HotXLS Release Notes

Version 2.384.120

Added

• Portable Classic XLS capture supports ordinary column, bar, line, scatter, pie, doughnut and single-series area charts with validated local stored values, core appearance and caller-declared plot, title and legend bounds
• Read-only chart appearance declarations retain worksheet and chart identities, explicit frame dimensions, effective legend point sizes and native font geometry for detached PDF, SVG and PNG output
• Bounded helpers inspect simple stored formula references and cached chart cells without formula evaluation or sparse-cell expansion

Fixed

• Classic chart capture preserves fractional stored font and marker sizes, source ownership and read leases, and rejects malformed records, unresolved appearance and exhausted resource budgets before output
• Chart double-valued property updates reject non-finite values before comparison or mutation
• Repeated native portable cell initialization releases prior managed strings during capture

Version 2.384.119

Fixed

• Classic chart inspection validates stored series text, trendline names and data-label separators before applying typed state, and releases partial models when malformed text raises an error
• Compressed Unicode data-label separators retain their complete stored characters

Version 2.384.118

Added

• Classic chart inspection exposes marker diameters in points, including fractional sizes, while preserving the original stored units

Fixed

• Supported native Classic chart fill tables decode properties after complex entries, and malformed tables preserve the preceding typed fill

Version 2.384.117

Added

• Explicit byte-encoding helpers preserve embedded nulls and reject text that cannot be represented by the selected code page

Changed

• Classic BIFF5 text, formulas and workbook metadata use their declared encoding on Windows, Linux and macOS, with encoded byte lengths and complete-character boundaries
• Supported legacy chart text retains its original decoding after codepage edits and copies, and is explicitly transcoded before saving under a different encoding

Fixed

• Compressed BIFF8 Unicode preserves exact code units, and legacy custom styles, chart text and double-byte formula-cache continuations use their actual record layouts
• Native Unicode text calculation, criteria, lookup keys, local formula bindings and Classic search and replacement preserve supported case matching
• Named-style lookup resolves the first stored entry without treating it as an absent name
• Repeated ZIP entry reads release prior header metadata while retaining concurrent stored and deflated stream behavior

Version 2.384.116

Added

• Local SUM calculated Pivot fields evaluate formulas from leaf, subtotal and grand-total context sums, including dependencies that are absent from displayed measures
• Cache-owned calculated items retain native member identities, hidden dependencies and global solve order across supported ungrouped string-member fields
• Supported calculated-item display calculations and aggregate filters retain visible member order, original filter order and unambiguous displayed-measure identities

Fixed

• Native Pivot formula date projection preserves the workbook date system through cache import, cloning and calculation
• Missing-item cache cleanup preserves validated materialized report ownership across both workbook date systems and legacy saved reports
• Unsupported calculated Pivot metadata and edited native output reject before mutating worksheet results

Version 2.384.115

Added

• Native portable PDF, SVG and PNG output supports rich text, modeled shapes, ordinary images, in-cell images, Boolean checkboxes and seven supported cached XLSX chart families
• Immutable workbook print plans apply complete print areas, repeated row and column titles, page breaks, paper setup, margins, orientation, centering, scaling and page order on Linux and macOS
• Explicit native font geometry supports generation-checked capture without changing the workbook's geometry provider

Fixed

• Fit-to-page mode survives workbook round trips and ignores stale percentage scaling, including unlimited fit dimensions
• Supported chart title fonts, line appearance and marker fill and borders retain explicit imported and authored values, including black
• Rich text preserves explicit false overrides and distinguishes effective run appearance during string reuse and semantic comparison

Version 2.384.114

Added

• Native Linux x64 and macOS ARM64 applications load, edit, calculate and save Classic BIFF8 workbooks, including supported document properties and RC4 encryption
• Portable compound storage supports nested scopes, large chained DIFAT containers, metadata and explicit native storage interface ownership

Fixed

• 128-bit CryptoAPI workbooks advertise a compatible cryptographic provider and retain readable plaintext document properties
• Native document timestamps use the correct FILETIME epoch and preserve UTC values across platforms
• Classic defined names share canonical Unicode identity across scope lookup, formulas and workbook reopening
• Escaped date-format separators and embedded quoted literals retain the selected complete date format

Version 2.384.113

Added

• Range clear and Unicode text paste support bounded atomic undo and redo with protection and structural preflight
• Typed local in-cell images and Checkbox formats support XLSX roundtrips, accessibility metadata and ownership-aware content edits
• Sparse row and column Checkbox formats preserve precedence, explicit clears and absent cells through edits and XLSX roundtrips
• Expanded XLSB support includes formulas and caches, names, merges, hyperlinks, panes, tables and filters
• Native Linux x64 and macOS ARM64 workbook APIs load, edit, calculate and save supported XLSX, XLSB and ODS files
• Native Pivot output supports numeric, date and manual grouping and all eleven custom subtotal functions

Fixed

• Scoped names use consistent Unicode normalization and case matching on Windows, Linux and macOS without discarding accents or symbols
• Modern aggregation formulas compile consistently with Free Pascal on Windows
• Opening another workbook replaces named styles instead of retaining style definitions from the previous file
• Modern image and control content survives range undo and saving without materializing absent cells
• Imported formula caches retain original value metadata, and ordinary metadata is distinguished from in-cell image content
• Semantic comparison distinguishes local image content, accessibility metadata and effective inherited control formats with bounded sparse traversal

Version 2.384.112

Added

• Regular expression testing, extraction and replacement support UTF-16 text, captures, array broadcasting and bounded evaluation on Windows, Linux x64 and macOS ARM64
• GROUPBY and PIVOTBY support scoped aggregate callbacks, multiple measures, headers, filters, hierarchical totals, sorting and relative percentages

Fixed

• Scalar SUM arguments retain Boolean and numeric-text coercion while range and array aggregation keep their existing ignoring rules
• Scoped Boolean literals and aggregate callback formulas survive future-function storage and workbook reopening

Version 2.384.111

Added

• Native local Pivot output supports base-item differences and percentages, running totals, row and column percentages, index, parent percentages, cumulative percentages and dense ranking

Fixed

• Display calculations retain native total, missing-value, zero-divisor and sorted-member semantics, including nonadditive averages
• Extended display modes and base-field identities survive saving, reopening and explicit Excel refresh

Version 2.384.110

Added

• Native local Pivot output supports compact, tabular and outline layouts, repeated ancestor labels, default subtotal placement and multiple measures on either data axis

Fixed

• Repeated-label settings survive imported namespace aliases and save/reopen, while native output distinguishes absent intersections from numeric zero
• Staged layout metadata keeps field and data captions consistent when Excel explicitly refreshes generated reports

Version 2.384.109

Added

• Workbook comparison distinguishes typed values and exact large integers, supports finite numeric tolerances and optional stored formula-cache, style and geometry comparison
• Bounded semantic comparison covers supported names, rules, tables, drawing order, blank-style regions, effective rich-text runs, local Pivot models, Slicer selections and Timeline state

Changed

• Comparison preserves packed and absent cells, resolves resource identities to effective values and retains the previous successful result on unsupported models or resource exhaustion

Fixed

• Typed database REAL parameter precision validation compiles with Delphi XE5 while retaining exact-value, overflow and null checks

Version 2.384.108

Added

• ODS round trips retain ordinary note text and authors, representable value and custom filters, explicit cell protection and equivalent passwordless worksheet protection
• Conversion reports retain precise metadata losses through worksheet copies and later saves, including unsupported filter expressions and non-equivalent protection

Fixed

• Cancelled or failed ODS stream saves preserve destination bytes, size and position through bounded temporary-file staging
• Successful ODS saves starting at a nonzero stream position preserve the prefix and correct ZIP entry offsets

Version 2.384.107

Added

• Stored-formula inspection and reference-aware Pivot aggregate formulas work in Classic XLS, XLSX, read-only and isolated evaluation contexts, including available array results and typed errors
• Pivot aggregate queries honor modeled visible members, report filters and supported subtotals through bounded detached cache evaluation without refreshing source data

Fixed

• Native Excel XLS Pivot caches retain successive inline numeric values and current cache-validity metadata instead of returning incorrect repeated-value totals
• Pivot references include the modeled report-filter area above the main table location

Version 2.384.106

Added

• Automatic database queries bind typed literal and stored-cell parameters in positional order, preserving Unicode, empty text, zero, Boolean, dates and nulls without substituting values into SQL text
• Cell parameters read available formula caches and compact stored values without recalculation or materialization, with bounded metadata validation and transactional result application

Fixed

• Jet and ACE date parameters preserve DATE, TIME and TIMESTAMP values through typed OLE DATE bindings independently of locale-dependent timestamp projections

Version 2.384.105

Added

• Formula metadata reads actual referenced file identity, supported base number formats, negative colors, positive parentheses, text alignment, locked state, column width and calculation mode across XLS and XLSX evaluation contexts
• Detached base-style and sparse axis-footprint inspection preserves stored and packed cells, worksheet defaults, theme colors and borrowed workbook ownership

Changed

• Column-width metadata returns its rounded character width and default-width flag as a two-column array; unavailable Excel application environment information and unclassifiable formats report explicit errors

Version 2.384.104

Added

• Import validation supports strict scalar types, numeric and date bounds, explicit Excel date systems, composite uniqueness and cross-column relationships
• Structured violations identify the worksheet, cell, rule, related column and first duplicate row, while stored-value validation preserves packed cells and reports unavailable formula caches without implicit recalculation

Version 2.384.103

Added

• Automatic text queries refresh fixed-width files using declared field starts, supported character encodings, typed field conversion, skipped fields and header or first-row selection
• Fixed-width refresh preserves literal quotes, trims field padding and rejects invalid boundaries, oversized input and cancellation before applying worksheet changes

Version 2.384.102

Changed

• Classic formula symbol audits scan faster while preserving function and name findings, case-insensitive deduplication, scope and finding order

Fixed

• Windows workbook saves, direct package patching, PDF output and page-image output can atomically replace an existing file held by delete-sharing metadata or reader handles on supported filesystems, while retaining lock checks, readonly protection and the legacy path on unsupported filesystems

Version 2.384.101

Added

• INDEX area-selector arrays broadcast with row and column selectors over calculated arrays and supported same-sheet reference unions, retaining orientation, selected value types and individual errors
• Defined names and LET bindings retain supported multi-area reference identity, and scalar CHOOSE retains the selected reference for downstream calculation

Fixed

• INDEX accepts an explicit area argument of 1 for calculated sources and distinguishes invalid, missing, fractional and out-of-range area selectors
• Cross-sheet unions and calculated-array unions reject with their supported error contract instead of acquiring incorrect reference identity
• Invalid references qualified by existing local worksheets retain their reference errors through direct formulas and defined names, including quoted sheet names, without changing formula strings or interpreting unknown qualifiers as local sheets

Version 2.384.100

Added

• Portable headless PDF, SVG and PNG output on Linux with explicit font files, native text shaping, bounded resources and staged exclusive file publication
• Read-only Windows capture of supported Classic XLS and XLSX rectangles into immutable transferable snapshots, including stored formula results, scoped Table and conditional appearance, physical dimensions and complete merges

Fixed

• Portable appearance validation rejects invalid enum representations consistently across optimized compiler builds

Version 2.384.99

Added

• Explicit automatic query refresh for supported Windows Text, ADO, ODBC and Web connections, with custom provider registration and bounded results

Fixed

• Native query result relationships and column identities persist through supported table-backed or standalone saving and reopening
• Invalid provider connection kinds reject before dispatch without changing worksheet results

Version 2.384.98

Added

• Local Pivot Timeline creation, shared-view connections, inclusive whole-day filtering and guarded selection-clearing or selection-preserving removal
• Explicit native Slicer no-data item refresh across connected views, with cumulative resource limits, cancellation and preserved selection and item identity

Fixed

• Native Timeline relationship and filter ownership handling preserves supported imported selections through materialization, saving and reopening
• Pivot Slicer cross-filter settings persist through their native tabular representation
• Timeline discovery retains UTF-16 workbook and linked-part decoding during Open XML loading

Version 2.384.97

Added

• Scalar and calculated-array defined names in supported numeric and counting aggregate calculations, with native coercion, dependency tracking and resource limits

Fixed

• Numeric aggregation propagates typed errors from local worksheet references, empty averages return division-by-zero errors, and failed named references retain their errors without restarting partial traversal

Version 2.384.96

Added

• Calculated-array arguments for text joining and supported median, order-statistic, percentile, quartile, percent-rank, multiple-mode and population covariance calculations

Changed

• Array and reference collection preserves native scalar coercion, typed errors, original covariance pairing and calculation resource limits

Fixed

• Finite median midpoint calculation avoids same-sign overflow, and exclusive percentiles accept valid endpoint and singleton ranks

Version 2.384.95

Added

• Transactional native compact Pivot report output with synchronized worksheet values, layout metadata and persisted cell ownership
• Read-only Slicer member availability across connected views under other active filters
• Explicit historical Pivot cache member cleanup with configurable retention and coordinated linked selection remapping

Changed

• Supported native Pivot output retains current manual filters and captions while rejecting unsupported layouts or edited owned cells before mutation

Version 2.384.94

Added

• Shared-cache PivotTable views and local Pivot Slicer creation with explicit worksheet bounds
• Transactional Slicer removal that either clears its restriction or preserves selection as ordinary Pivot filters, including future-member policy
• Fixed numeric, date and discrete grouped-cache reconstruction with stable member identities
• Detached native compact Pivot layout metadata previews using exact member and aggregate identities

Fixed

• Reopening a populated workbook now replaces previous typed Pivot caches, and authored native Slicer links remain recognized by Excel
• Grouped cache serialization preserves declared native member domains and official grand-total and data-axis attributes

Version 2.384.93

Added

• A validated XLSB reader and writer for typed cell values, basic styles, ordinary worksheet views and dimensions
• Explicit binary output from workbook file and stream saving, with automatic binary input detection and conversion fidelity reports

Changed

• XLSB conversion stages complete packages and rejects unsupported features or calculation settings before output, including unchecked or loss-allowed saves

Fixed

• Configured ZIP compression levels are applied before deflater initialization

Version 2.384.92

Added

• Transactional Pivot cache refresh and safe worksheet materialization for supported local defined-name and table sources
• Worksheet-scoped defined-name lookup and caller-resolved source bounds for cache refresh

Fixed

• Local source identity and defined-name scope survive worksheet moves and renames, including quoted and Unicode names
• Defined-name formula XML omits the optional leading equals sign required only by the in-memory formula syntax
• Table totals membership follows the actual totals-row count, preserving all data rows when only the totals UI hint is present

Version 2.384.91

Added

• Calculated-array inputs and typed results for transpose, matrix multiplication, determinants, and matrix inversion
• Ordered selector vectors, duplicates, and negative indices for row and column selection
• Calculated lookup vectors and array-valued lookup requests, including complete scalar lookup slices and typed fallback values

Changed

• Matrix work, lookup scanning, selector traversal, copying, and padding honor evaluation-step and array-memory limits across composition

Fixed

• Formula and array-result text caches display correctly in Excel before recalculation
• Modern rich-error caches include the workbook version and calculation-feature declarations required for native cached display
• Lookup and flattening options reject out-of-range values before numeric narrowing, while fractional options follow native truncation
• Empty take, drop, and flatten results retain their typed calculation errors

Version 2.384.90

Added

• Isolated Pivot result previews retain shared Slicer selections and caller filters without changing the live calculated cache
• Safe worksheet materialization for supported local Pivot sources stages results, rejects conflicting content and structures, and preserves existing cell objects, formatting, and comments
• Exact typed-value ownership survives XLSX saving and reopening, enabling safe result resizing, movement, and explicit adoption of existing scalar Pivot values

Fixed

• Deleting the last bound Pivot Slicer view rejects before removing its drawing, cache, reserved name, or selection
• Clearing Slicer views validates every affected cache before mutation and cleans permitted views, drawings, and unbound caches together

Version 2.384.89

Added

• INDEX accepts array-valued row and column selectors over calculated arrays and worksheet ranges, including paired selections and row-column cross products

Fixed

• Selector results preserve typed values and errors, pad missing selector positions with #N/A, and respect calculation memory and evaluation-step limits
• Fractional scalar selectors that truncate to zero produce complete row or column spills, while negative selectors return #VALUE! before truncation
• Scalar range selection retains reference behavior, and selector spills recover from obstructions, resize, and recalculate after saving and reopening

Version 2.384.88

Added

• Typed member inspection, shared-member selection, and clearing for supported imported non-OLAP Pivot Slicer caches
• Connected PivotTable calculations intersect Slicer selections with independent restrictions on other fields and caller-supplied record filters

Changed

• Selection changes validate every connection and member before committing, reject ambiguous imported same-field masks, and preserve unsupported cache representations
• Slicer cache collection changes respect workbook write protection when they affect connected PivotTable calculations
• Deleting bound non-OLAP Pivot Slicer caches is rejected atomically to preserve consistent runtime and saved filter state

Fixed

• Edited Pivot Slicer selections and connected view member masks save together while retaining imported namespaces and unknown extensions
• Supported Slicer connections follow saved sheet identities and resolve extended cache identifiers independently of ordinary workbook cache identifiers
• Imported PivotTable definitions remain associated with their original views when worksheets are reordered

Version 2.384.87

Changed

• Local rectangular Pivot cache refresh preserves existing field and member identities, appends new typed members, and keeps shared-view filters and selections bound to their original members after source reorder, removal, or reintroduction
• Source headers match database fields by name, allowing column reorder while calculated fields retain their formulas and receive fresh empty record placeholders for subsequent evaluation
• Refresh validates and stages the complete source before updating the cache, rejecting unsupported schemas, source types, derived grouping, and new unmapped discrete-group members without partial changes

Fixed

• Pivot error values retain their types and codes in shared items and records through XLSX saving and reopening
• Refreshed shared-item domains recompute type flags and representable bounds, and XLSX cache metadata accurately describes mixed values, dates, errors, Boolean values, and blanks
• Source refresh reads absent cells without creating worksheet cells and compares Unicode field headers without narrowing their text
• Saved refreshed PivotTable views include new cache members before subtotal items while retaining existing member order, hidden flags, and imported layout extensions
• Page selections remain bound to the correct member when view order differs from cache order, calculated fields no longer shift saved source records, and refresh timestamps use the numeric XLSX representation

Version 2.384.86

Added

• INDEX accepts computed arrays and constants with scalar selection, complete-array selection, and row or column extraction while retaining reference behavior for range sources
• SORT and UNIQUE accept calculated arrays and constants, allowing direct composition with FILTER, SEQUENCE, LET, and other array-producing expressions

Changed

• Composed array expressions retain automatic XLSX spill resizing, fixed-range array calculation, typed errors, and recalculation after saving and reopening

Fixed

• Array selection and sorting truncate fractional numeric options, invalid Boolean option text returns #VALUE!, and computed-array operations honor memory and evaluation-step limits

Version 2.384.85

Added

• XLSX dynamic arrays now calculate into automatically resized spill ranges, preserve user obstructions with #SPILL!, support spill references and explicit implicit intersection, and recover cached spill ownership after reopening
• Conversion preflight reports known target-format losses, and checked saving can reject those losses before writing an existing file or stream
• Explicit macro-enabled template output preserves existing VBA projects in XLTM packages
• Caller-provided query-table refresh supports bounded fixed-schema results, synchronized table and filter ranges, progress, cancellation, and transactional rollback

Changed

• Repeated report bands clone complete threaded conversations with remapped identities and preserve rollback when a target fails
• ODS custom Boolean validation, supported formula-valued bounds, list sources, and qualified references translate with base-cell rebasing; unsupported definitions and cell references are omitted together

Fixed

• Fixed-range array formulas evaluate raw range arithmetic element by element, multiple-array MAP accepts matching lambda parameters, and FILTER evaluates calculated arrays and masks
• Modern #SPILL! and #CALC! errors retain typed XLSX rich-value metadata while preserving unrelated imported metadata and rich-value buckets
• Newly authored XLSX formulas save with canonical argument separators and future-function prefixes for Excel interoperability; saving preserves the in-memory expression and reopening returns the canonical formula syntax
• Supported local chart series refresh numeric, text, date, error, bubble, and hierarchical category caches after edits and recalculation
• Native pagination renders merged spans through their original anchor, clips page fragments, and suppresses interior borders and gridlines

Version 2.384.84

Changed

• The Delphi and C++Builder getting-started examples now include a practical sales order with typed dates, prices, discounts and calculated totals, and verify saved XLS and XLSX workbooks after reopening
• The invoice example now connects a printable invoice to editable payment terms and a receipt ledger, calculating discounts, tax, payments and the outstanding balance
• The worksheet layout example now demonstrates merged heading creation, enumeration, removal and preservation after reopening
• Nine more focused API example families have moved into Miscellaneous, keeping the main catalog centered on complete business workflows and interactive tools

Removed

• The minimal standalone merged-cell and rich-text examples have been consolidated into the worksheet layout and dashboard examples

Version 2.384.83

Changed

• The demo catalog now separates complete experiences and substantial workflows from focused API recipes in each platform's Miscellaneous folder, with direct navigation and ready-to-build projects in both tiers

Removed

• Three redundant range and worksheet-layout demos have been removed; their complete scenarios remain available in RangeOps and SheetLayout

Version 2.384.82

Changed

• Delphi and C++Builder dashboard galleries open directly on a populated workbook with live totals, rendered charts, editable details, price scenarios, and exports of the current result
• The API tour provides direct dashboard, sales data, chart preview, and API recipe navigation with a repeatable dataset covering all twelve months
• Examples carry their required scenario units and manifest resources locally, allowing each project to be copied without a separate common directory

Fixed

• The API tour creates its sales data before entering dashboard formulas, allowing its live showcase to initialize successfully
• Live showcase charts refresh their cached values and labels after recalculation, and repeated C++Builder gallery exports preserve current edits without accumulating generated report sheets

Version 2.384.81

Changed

• Worksheet copying examples verify independent content, remapped formats, formulas, and embedded objects after reopening their output
• Date-system, search, grouping, array-formula, and picture examples verify their distinct behavior and report failed output through a nonzero exit code

Fixed

• Classic XLS AutoFilter ranges retain their complete row and column bounds after saving and reopening, including copied filters with active criteria
• Dashboard galleries create referenced worksheets before entering formulas, calculate totals before export, and generate all requested companion formats even when another output fails
• The C++Builder dashboard gallery produces the documented CSV, HTML, and RTF companion files
• Compiled reports preserve ordinary failure reporting in classic Win64 C++Builder static builds when a destination cannot be replaced
• Interactive API examples report PDF, HTML, and CSV export failures instead of claiming the files were written
• The feature showcase completes its supported semantic-hash demonstration and validates its saved workbooks without invoking unsupported modern drawing snapshots

Version 2.384.80

Changed

• The example catalog groups projects by learning task and identifies recommended entry points, format-specific lessons, and compatible project names
• Worksheet layout and range relocation examples share complete scenarios while preserving their existing project and output names
• Memory, advanced CSV, printing, document properties, text box, and worksheet protection examples demonstrate complete workflows with saved-data verification and explicit failure reporting

Fixed

• Reopened XLSX text boxes retain their drawing relationships when saved again, so annotations remain available through repeated save and reopen cycles without duplicate XML namespace declarations
• Batch conversion examples honor Skip existing before opening sources or invoking writers, and verify workbook success codes before reporting completion
• The invoice example includes payment terms and its complete title in the printable area
• The page setup console example starts with ordinary user privileges instead of triggering Windows installer detection

Version 2.384.79

Fixed

• ODS file saves with export options preserve the existing destination when serialization fails or saving is cancelled, and report destination replacement errors consistently
• ODS export options use the normal save lifecycle, including recalculation policy, progress cancellation, diagnostics, and workbook write protection for file and stream output
• XLSX stream saves honor the configured recalculation policy, preserving cached values in manual mode and refreshing dirty formulas when requested
• Completed XLSX and ODS save notifications do not turn an already committed output into a cancellation failure
• Failed streaming worksheet patch initialization and finalization release source file handles, allowing callers to clean up their temporary sources and reuse the writer after initialization failure

Version 2.384.78

Fixed

• Range copies and clones preserve intersecting virtual formatting and replace prior overlay formatting only inside the destination, including overlapping copies and cross-workbook formatting
• Explicit formatting resets remain effective after saving and reopening workbooks with row or column default formatting
• XLSX cells that explicitly use the workbook's default formatting retain its custom fonts, fills, and number formats

Version 2.384.77

Fixed

• Whole-row and whole-column conditional-format and data-validation targets follow row and column edits and range cloning when fully contained in the source, with bounded target geometry and preserved relative formula anchors
• Finite repeated blank styles imported from ODS retain their full used-range extent, and newly created or reset workbooks report the current library version in document properties
• Whole-row, whole-column and whole-sheet cell copies retain sparse storage, clear replaced destination contents without expanding blank coordinates, and preserve overlapping source data and cross-workbook formatting

Version 2.384.76

Fixed

• Conditional-format and data-validation rules now apply consistently to absolute, reversed, whole-row, whole-column and whitespace-separated targets during lookup, deletion, conditional calculation and validation checks
• Malformed and out-of-grid target areas no longer match cells, while valid areas in the same rule remain active; full-grid conditional statistics retain sparse cell storage
• ODS validation targets use valid cell anchors for whole rows and columns, and equivalent overlapping conditional targets no longer create duplicate rules when reopened

Version 2.384.75

Fixed

• ODS cell comparison and expression rules retain their full whole-row, whole-column and wide target ranges in Excel-compatible output; repeated blank carriers keep sparse ranges compact and preserve overlapping rules and worksheet metadata
• Reopening ODS repeated styled blank regions keeps their base formatting compact, and standard conditional rules retain overlapping precedence and coverage inside merged regions

Version 2.384.74

Fixed

• ODS conditional formatting retains supported font, background and number-format overrides when saved and reopened, including explicit resets; rules with different highlights or target ranges remain separate
• XLSX conditional formatting retains built-in and custom number formats, including formats used only by a conditional rule; saving no longer removes those custom formats or points rules at a different format
• ODS percentage styles include the percent symbol required by other spreadsheet applications and reopen without duplicating that symbol

Version 2.384.73

Fixed

• Copying text to the clipboard, reading clipboard text and pasting text or XML Spreadsheet data no longer fail at random on Windows 10 and 11. After every clipboard change, Explorer and the Clipboard User Service (clipboard history) open the clipboard for a few milliseconds to read it; text operations tried to open it only once and could fail with "access denied" during that moment
• All clipboard operations now use the same policy as the BIFF8 (Excel format) copy and paste: they keep trying to open the clipboard for about 500 ms. When another application holds the clipboard for longer, the operation still fails and the existing clipboard content is left unchanged

Version 2.384.72

Fixed

• Opening an ODS file reads the General number format correctly. Excel and LibreOffice write General as a number style without a decimal-place count, which was read as two fixed decimals, so a value such as 1.5 showed as 1.50; such cells now carry no number format, as in the source application
• Text around a General number is kept, for example General" kg"

Version 2.384.71

Fixed

• ODS files saved by Excel or LibreOffice no longer report a used range covering the whole sheet. Both give every row and column a size out to row 1048576 and column 16384; empty rows whose height is automatic now keep the default height, and the column sizes that run out to the last column become the sheet's standard column width. A 10-row Excel file opened and saved again shrinks from about 2.8 MB to a few KB
• Row heights set by hand, and automatic heights on rows with content (for example rows grown to fit wrapped text), are still kept
• Saving to ODS writes the sheet's standard column width (StandardWidth) on columns that have no width of their own; it used to be dropped, since ODS has no sheet-wide default column width

Version 2.384.70

Fixed

• ODS files saved by Excel or LibreOffice open in milliseconds again. Both pad every row out to column 16384 and add a padding block down to row 1048576; a 10-row file from Excel used to take more than 5 minutes to open, and one from LibreOffice did not finish either
• Empty padding cells that look exactly like the document's Default cell style are no longer turned into formatted blank cells; a 10-row Excel sheet used to gain about 180,000 of them. Empty cells with their own fill, border, font, alignment, number format or protection are still kept

Version 2.384.69

Fixed

• Conditional formats saved to ODS now show up in Excel. Value rules (equal, between, greater than and so on) and formula rules are also written as standard ODF style:map conditions on the cells they cover, the only form Excel reads; Excel used to open HotXLS ODS files without any of these rules
• LibreOffice now imports ODS value rules written by HotXLS with their real operator and bounds; it used to turn every one of them into "equal to 0"
• Opening an ODS file reads conditional formats stored as style:map conditions, as written by Excel and other ODF producers. When a file also carries the same rules in LibreOffice's own form, each rule is read once
• Opening an ODS file reads LibreOffice's spelling of value rules (=5, between(1,10), !="x"); files from earlier HotXLS versions still open as before

Version 2.384.68

Fixed

• Formulas that apply an operator to a multi-cell range or an inline array, such as =SUM(A1:B1*{10,100}), =SUM((A1:B2>2)*1) or =MAX(A1:B2-1), are stored the way Excel 365 stores them, so Excel shows the same result as HotXLS: in XLSX as single-cell dynamic-array formulas, in XLS as one-cell array formulas. Excel used to open them with implicit intersection (=SUM(@A1:B1*@{10,100})) and show #VALUE! or a wrong value
• XLS array formulas write their references and array constants in array form, as Excel does; Excel used to evaluate array formulas written by HotXLS with implicit intersection
• XLSX files with dynamic-array formulas (including those created with SetDynamicArrayFormula) open in Excel; Excel refused the whole file because the dynamic-array metadata used a differently cased extension identifier

Version 2.384.67

Fixed

• Text comparisons follow Excel's order in both engines: they ignore case and use the system locale's word sort, so hyphens and apostrophes only break ties ("a-b">"ab" is TRUE) and other punctuation sorts before digits and letters ("a~b">"ab" and "a_b">"ab" are FALSE). This applies to the comparison operators, >text / <text criteria in COUNTIF, SUMIF, the *IFS family and the database functions, array comparisons, lookups, SORT-style functions and range sorting
• VLOOKUP and HLOOKUP compare text case-insensitively, so VLOOKUP("ABC",...) finds "abc" as in Excel

Version 2.384.66

Fixed

• Formula-based conditional formatting written to ODS is now recognised by LibreOffice: the rule is stored as formula-is(...) with its base cell, the form LibreOffice itself writes. Before, LibreOffice read every such rule as a plain "cell value equals 0" rule
• Conditional formatting formulas written to ODS translate inline arrays, unions, whole-column and whole-row references and keep $ markers the same way as cell formulas (COUNTIF($A:$A,A1)>1 becomes COUNTIF([.$A:.$A];[.A1])>1)
• Opening an ODS file reads formula-based conditional formatting written by LibreOffice (formula-is) as well as files from earlier HotXLS versions

Version 2.384.65

Fixed

• Whole-column and whole-row references (A:A, $A:$B, 1:1, $1:$2) in formulas written to ODS use the OpenFormula reference form ([.A:.A]). Excel used to open such cells as =SUM(A:(A)) with #NAME?, or drop the formula; ODS files from HotXLS now show the same formulas and values in Excel and LibreOffice

Version 2.384.64

Fixed

• Database functions (DSUM, DCOUNT, DCOUNTA, DAVERAGE, DGET and the rest) treat a text criterion without an operator as "begins with", as Excel does: ab in the criteria range matches abc, ab and AB, and wildcards still apply (a? matches every entry of two or more characters starting with a). =ab and <>ab keep comparing the whole entry

Version 2.384.63

Added

• Formulas can contain inline array constants such as {1,2;3,4} (columns separated by commas, rows by semicolons, with numbers, text, TRUE / FALSE and error values) and reference unions such as (A1:B2,D4), in both the XLS and XLSX engines; entering them used to fail with "invalid formula syntax"
• Array constants are evaluated as arrays (=SUM({1,2}*{3,4}), =ROWS({1,2,3;4,5,6}), =SUMPRODUCT(A1:A2,{1;10})), and unions work in SUM, COUNT, AVERAGE, MIN, MAX and COUNTA, in AREAS, in ISREF and in the area_num argument of INDEX

Fixed

• XLS files store array constants so that Excel reads them: the constant used to be written as a reference, which Excel displayed as =#N/A; XLS formulas that contain array constants or unions also read back with their formula text

Version 2.384.62

Fixed

• The intersection operator (a space between two references) returns the cells both references cover, as in Excel: =SUM(A1:B2 B1:B2) is 6 instead of 2, references without a common cell give #NULL!, and an intersection can be passed to functions that take references such as ROWS, COLUMNS and INDEX. This applies to both engines, to calculated formulas and to formulas stored in cells
• XLS files keep intersection formulas usable in Excel: the referenced ranges are now written as references, so Excel no longer reads =SUM(A1:B2 B1:B2) as =SUM(@A1:B2 @B1:B2) and shows #VALUE!
• An intersection used as a single value with more than one common cell picks the cell in the formula's row or column, like any multi-cell reference

Version 2.384.61

Fixed

• SUMPRODUCT accepts array expressions such as (B1:B2>0)*1, --(B1:B2>0), B1:B2*C1:C2 and a single argument, in both the XLS and XLSX engines; these used to return #N/A. Arguments of different shapes return #VALUE!, an error inside any argument is returned, and text and logical values count as 0, as in Excel
• Operator expressions over ranges inside SUM, COUNT, AVERAGE, MIN, MAX and COUNTA are evaluated element by element, as under Excel's dynamic arrays: =SUM((B1:B2>0)*1) counts both rows instead of using only the first cell
• TRUE counts as 1 in arithmetic: =TRUE*1 returned -1
• Element-wise comparisons treat a blank element as 0, "" or FALSE to match the other side, and compare logical values as logical values; joining a logical value with & in an array gives TRUE / FALSE

Version 2.384.60

Fixed

• Wildcard Find and Replace with whole-cell matching (FindText, ReplaceText, FindTextIn, ReplaceTextIn in both engines) match the whole cell the way Excel does: a*b now finds abcb, which was rejected because matching stopped at the first b
• In wildcard Find and Replace, ~ makes whatever character follows it literal and a trailing ~ is ignored, as in Excel: a~b finds ab, and a~~b finds a~b

Version 2.384.59

Fixed

• XLSX document creation and modification times (Created, Modified) and date-valued custom properties are read correctly when the stamp has fractional seconds (2026-03-15T08:30:00.123Z) or a time zone offset (+08:00, -05:00): the time is converted to local time instead of being taken as local time unchanged
• Date-only stamps (2026-07-01) are read as that date; stamps with a time but no zone, which the W3CDTF format does not allow, are still read as local time

Version 2.384.58

Fixed

• XLSX AutoFit through TXLSAutoFit (FitXlsxSheet, FitXlsxSheetColumns, FitXlsxSheetRows, MarkSheet) measures each cell in the font it is displayed with; since 2.384.46 it picked the next font in the workbook's font list, so a bold heading could be measured in another cell's small font and come out too narrow
• AutoFit also uses the font a cell inherits from its row or column style

Version 2.384.57

Fixed

• "Precision as displayed" (FullPrecision = False) in the XLSX engine now rounds the way Excel and the XLS engine do: percent formats keep two extra decimals (0.1234 shown as 0.0% stores 0.123, not 0.1), negative and zero values use their own format section, and halves round away from zero (2.5 shown as 0 stores 3, not 2)
• Both engines honour thousands-scaling commas under "precision as displayed": 12345.678 shown as 0.0, (12.3) stores 12300, as in Excel
• A logical formula result in a cell with a number format is no longer turned into -1 by the XLS engine's "precision as displayed"

Version 2.384.56

Fixed

• ODS files written by HotXLS declare the namespaces of their formula prefixes, so LibreOffice evaluates the formulas instead of showing #VALUE! in every formula cell

Version 2.384.55

Fixed

• Formulas written to ODS keep their meaning: inline arrays use OpenFormula's row separator ({1,2;3,4} becomes {1;2|3;4} instead of a single row), a reference union such as (A1,B2) becomes the ~ operator instead of two function arguments, and absolute references keep their $ markers
• Reading such a union back from ODS no longer adds a second pair of parentheses, so these formulas survive an XLSX to ODS to XLSX round trip unchanged

Version 2.384.54

Fixed

• Password-protected XLS files saved in the Excel 5.0/95 format (XOR obfuscation) now open in Excel. The key stored in the file is derived from the password instead of being random, and the obfuscation array uses the bit rotation Excel expects; before, Excel reported a wrong password for every such file, and with the key alone corrected it reported the file as damaged
• XOR-obfuscated workbooks are read with the same corrected array, so files whose key matches the password are decrypted the way Excel encrypts them; XOR files written by earlier HotXLS versions still open

Version 2.384.53

Fixed

• Comparing a blank cell with a logical value treats the blank as FALSE in both the XLS and XLSX engines, as Excel does: with A1 empty, =A1=FALSE and =A1<TRUE now return TRUE and =A1=TRUE returns FALSE. Before, the blank was compared as 0, which ranks below every logical value

Version 2.384.52

Fixed

• COUNTIF, COUNTIFS, SUMIF, SUMIFS, AVERAGEIF, AVERAGEIFS and the database functions match text criteria with Excel's wildcard rules in both the XLS and XLSX engines: ~* and ~? match a literal asterisk or question mark, square brackets are ordinary characters, and a criterion without * or ? is a plain case-insensitive comparison. Before, a criterion such as "[x]" was read as a character set, "~*" did not match an asterisk, and a malformed pattern silently matched nothing
• "<>text" criteria count numbers, logical values and blank cells as not equal, and a bare "<>" counts every non-blank cell, as in Excel; text comparison under <> is now case-insensitive
• MATCH and XLOOKUP wildcard matching treat ~ as an escape for whatever character follows it and ignore a trailing ~, as Excel does

Version 2.384.51

Changed

• Read-only cell snapshots retain worksheet names in shared UnicodeString storage; SheetName is now a readable and writable WideString property with the same ordinary read and assignment syntax as before, but its raw field address and the previous record layout are unsupported; callers must recompile against this release
• Report output to an empty destination stream skips the unnecessary backup snapshot while preserving rollback of the destination contents and position if output fails
• Deep formula audits process numeric caches more efficiently while preserving cached Variant values, evaluation results, and reported findings

Fixed

• XLS and XLSX clipboard paste captures BIFF8 data and fallback text together under one clipboard lock, so both formats come from the same clipboard state; readable BIFF8 remains usable when its text companion cannot be retrieved, unavailable text alone inserts no cells, and the existing clipboard-open retry policy is unchanged
• Failed required text or BIFF8 clipboard publication preserves the original Win32 error code across cleanup so callers can retrieve the failure diagnostic immediately after the call
• The single-attempt clipboard text reader preserves native data-retrieval failures across clipboard cleanup without changing its retry behavior

Version 2.384.50

Changed

• Formula symbol audits run faster on sparse XLS workbooks while preserving the reported findings and their order

Version 2.384.49

Fixed

• WorkbookAudit demos (Delphi and C++Builder): AuditReport.json is written without a UTF-8 byte order mark so strict JSON parsers accept it; XLS sheets report their AutoFilter range; XLSX defined names that point at another workbook show the linked file instead of its index; command-line runs no longer overwrite the saved settings; an empty folder box falls back to the program folder instead of the drive root; a failed sample save is logged instead of reported as created
• ApiTour demos: rebuilding the showcase workbook keeps the previous one when the rebuild fails, so the export buttons no longer crash; workbooks are freed when building throws; the HTML export logs its errors; the log reports the five conditional formats that are actually created
• ExcelGraphics demos: an unsupported picture type is reported instead of saving an empty sheet, and the open dialog lists only supported formats
• OdsExport demos compare the font size in the round-trip check and return a non-zero exit code when the check or the save fails; the C++Builder OdsImport demo checks the save result and no longer creates empty cells while reporting
• DataExport / ExportWithColoring demos narrow only the first column as their comments say, and the C++Builder export demos open the save dialog in the program folder
• Demo readmes list the APIs and output files the demos actually use

Version 2.384.48

Fixed

• Document creation and last-saved times are now written in UTC, as the file formats require, and read back as local time: XLSX Created / Modified and custom date properties, and XLS CreatedDate / LastSavedDate. Before, the local time was stored as if it were UTC, so Excel and Windows Explorer showed the stamps shifted by the time-zone offset; values written and read by HotXLS itself are unchanged

Version 2.384.47

Fixed

• XLS files with XOR obfuscation (the password protection of Excel 5.0/95 workbooks, EncryptionType = xletXor): the key position used for each record now follows the file format specification, the same way other spreadsheet applications read and write it. Before, HotXLS could only read its own XOR files, and any record whose length was not a multiple of 16 decoded as garbage elsewhere

Version 2.384.46

Fixed

• XLSX AutoFit (FitXlsxSheet, FitXlsxSheetColumns, FitXlsxSheetRows, MarkSheet) measures each cell in its own font, size, bold and italic instead of always Arial 10, so Calibri 11 text, bold headers and large titles no longer come out too narrow
• XLSX AutoFit measures the text as displayed through the number format (dates, currency, percentages) rather than the raw stored value
• XLSX AutoFit keeps fitted column widths within MaxWidth; the cell margin used to be added after the limit was applied
• XLSX AutoFit sizes rows of wrapped cells in columns that have no explicit width against the sheet default width; before, such cells were measured as a single line and the row came out too short

Version 2.384.45

Fixed

• XLS workbooks: ApplyAutoFilter criteria such as '>=100' or '=2.5' are now saved as number conditions; they were saved as text, which Excel compares as text, so the filter hid every numeric row. A value without a comparison operator (such as 'North' or '100') is still saved as text, the way Excel stores a picked list item
• Criteria1.DataType reports 4 (number) for such criteria instead of 6 (text); Criteria1.Value still reads back as 100

Version 2.384.44

Fixed

• INDIRECT with an A1 reference text read the cell one row down and one column right (INDIRECT("D1") returned E2), in both XLS and XLSX workbooks; R1C1 text was already correct. A reference past the last row or column now returns #REF!

Version 2.384.43

Fixed

• XLSX workbooks: formulas can now reference the whole 1,048,576 × 16,384 sheet. References past column IV or row 65,536 (=XFD1, =A100000, =SUM(XFC1:XFD1), Data!C1048576) used to be rejected as unknown names on entry or evaluated to an error; they now parse, calculate, shift with inserted rows and columns, and save as written
• XLSX workbooks: whole-column and whole-row references (A:A, 1:1) cover the full sheet, so ROWS(A:A) returns 1048576 and SUM(A:A) includes values below row 65,536; ADDRESS returns three-letter columns such as $XFD$1
• XLS workbooks: references to the last row and column of the 65,536 × 256 sheet (=A65536, =IV1) are accepted; they were rejected as unknown names

Version 2.384.42

Fixed

• Legacy lxRows / lxCols units (kept for source compatibility; the workbook classes do not use them): a row index past 1,048,575 raises ERangeError instead of wrapping around to row 1, the ROW record no longer marks every outline-level change as collapsed, and writing COLINFO for a column without its own settings uses the defaults instead of crashing

Version 2.384.41

Fixed

• Lazarus / Free Pascal builds compiled in ANSI string mode: scratch files whose path contains characters outside the system code page are now removed at cleanup; before, their names were stored lossily and the files were left behind in the temp folder

Version 2.384.40

Fixed

• TlxCellMMU (the packed cell store) raises EInvalidOperation when a cell is written out of row order within a 256-row block or over an existing cell, and ERangeError for a negative position or a column past 65,534; before, such a write silently corrupted the neighboring row. The library's own bulk writes already keep row order and are not affected

Version 2.384.39

Changed

• XLSX workbooks: assigning TXLSXWorksheet.TabColor now also sets TabColorIsAuto to False, so the tab color shows without a second assignment, matching the font, fill and border color properties. Setting TabColorIsAuto to True afterwards still removes the tab color

Version 2.384.38

Changed

• XLS workbooks: TXLSWorksheet.AddPivotTable now treats DestRow and DestCol as 1-based, like Cells[Row, Col], the documentation and the XLSX engine; AddPivotTable('Data!A1:C9', 3, 1, ...) anchors the pivot table at A3. Code that passed 0-based positions must add 1, and an anchor of 0 or outside the sheet now returns nil
• XLS workbooks: the positions on TXLSPivotTable (FirstRow to FirstDataCol) and the source range on TXLSPivotCache (SourceFirstRow to SourceLastCol) are now 1-based for new and opened pivot tables, so the same code reads the same cells in both engines and a pivot table copied between them keeps its place

Version 2.384.37

Fixed

• XLSX workbooks: a pivot table placed below row 65,535, or one pushed there by inserting rows, kept its true position; previously the anchor was cut down to 16 bits and the saved pivot table landed on the wrong rows
• The pivot table position properties (FirstRow, FirstCol, LastRow, LastCol, FirstHeaderRow, FirstDataRow, FirstDataCol) are now Integer instead of Word
• XLSX workbooks: AddPivotTable and AddPivotTableCopy return nil when the anchor lies outside the sheet or the copied block would run past the last row or column, instead of writing a pivot table Excel rejects

Version 2.384.36

Fixed

• C++Builder demos XlsTables, XlsCondFormat12, XlsCreateChart and InteractiveFeatures write their text columns as text; passing a const wchar_t* to OleVariant picked the bool constructor and every header and label came out as TRUE
• The DefinedNames demo lists the XLS names from index 1 instead of crashing on index 0, WorkbookAudit counts external references in XLS workbook names, and FileMonitor writes Monitored.xlsx instead of the misspelled .xlsxx that made every sheet-name read fail
• The Pictures and OdsExport demos embed a valid PNG; the byte literal was corrupted by the string conversion and the Pictures chunk lengths were wrong
• ApiTour evaluates its DEMOADD step on the API Tour sheet instead of the active Dashboard sheet, where a text argument could abort start-up; the C++Builder classic tour now shows the step's result
• FreezePanes, QuickStart and ApiTour freeze exactly their header rows and columns; FreezePanes takes counts, not the first scrollable cell
• Conditional format colors in ApiTour, CondFormatGallery, OdsExport and WorkbookAudit are given in BGR as the API expects, so data bars and color scales show their intended colors; CondFormatGallery gives its top-N, duplicate and above-average rules a visible highlight
• MemoryDataExport and HtmlExport no longer report successful CSV, HTML and RTF exports as failures (these writers return 0 on success), and MemoryDataExport writes all three files even if one fails
• Demos that save and then report or open the file (QuickStart, ExcelGraphics, HotXLSInMemory, OrderCalc, PurchaseOrder and several console demos) check the save result instead of announcing success when the file is locked
• ConversionWorkbench treats any negative writer result as a failure and removes old output first, so a stale file no longer hides a failed conversion; the C++Builder version exports every sheet to HTML
• InvoiceTemplate, OrderCalc and DataExport no longer depend on the Windows number and date format: tax formulas, prices and date thresholds work on systems that use a comma decimal separator or a year-first date order
• Smaller corrections: QuickStart totals use dark text on their light fill, DateSystems shows the raw serials and a valid 1904 epoch date, CrossWorkbookCopy keeps its merged cells outside the table, CrossWorkbookCopyXls uses a comma-separated list and a real light blue, Hyperlinks shows the intended link text, SheetProtection describes and sets the allowed actions correctly, OdsImport no longer creates empty cells while scanning, InteractiveFeatures saves a workbook without a VBA payload as .xlsx, WorkbookAudit reports encrypted XLS files and their VBA projects, ChartTypes keeps its charts clear of the data, the C++Builder data export demos find Demos.mdb from the IDE output folder, and FormulaFunctions checks the reopened cell and counts reopen failures

Version 2.384.35

Fixed

• Assigning a color to an XLSX font, fill, border edge or rich text run now takes effect on its own. These objects start with the automatic color, and an explicit Color, ForegroundColor or BackgroundColor was silently dropped unless the matching ColorIsAuto / ForegroundIsAuto / BackgroundIsAuto flag was also cleared
• Assigning an RGB color also clears an earlier theme or indexed color source, which would otherwise win over it; assigning a theme or indexed slot of 0 or more clears the automatic flag
• Border colors pasted into an XLSX workbook from the clipboard, border colors of classic workbooks exported to XLSX, and the font color passed to TXLSXComment.SetFont are now written instead of being lost to the automatic color

Version 2.384.34

Fixed

• Subtotal, grand total and blank items in a pivot field's item list are no longer read as data items; copying such a pivot table used to write an item pointing at cache index -1
• Pivot fields from Excel files keep their default subtotal: Excel omits defaultSubtotal when it is on, and HotXLS treated the missing attribute as off
• Pivot cache fields whose shared items leave out containsString are read as text fields, following the schema default, instead of mixed fields

Version 2.384.33

Fixed

• Pivot tables written by the XLSX engine use the schema's axis values (axisRow, axisCol, axisPage); earlier versions wrote rowAxis / colAxis / pageAxis, which are not valid OOXML. HotXLS still reads files that contain the old values
• Pivot fields used in the values area are flagged with dataField="1", and fields without items no longer get an empty <items> list that the schema does not allow
• Hidden and collapsed pivot items keep their state when a pivot table is copied or rebuilt, and each field's item list ends with the subtotal items Excel expects for its subtotal functions
• Pivot cache fields carry a numeric numFmtId instead of the text General

Version 2.384.32

Fixed

• ODS import converts inline array rows correctly: {1;2|3;4} becomes {1,2;3,4} instead of a formula Excel cannot parse
• ODS import converts the reference union operator: SUM([.A1:.B2]~[.C3]) becomes SUM((A1:B2,C3))

Version 2.384.31

Fixed

• ODS import reads time cells whose duration carries a date part, such as P0Y0M0DT12H30M15S, and week durations such as P1W, instead of falling back to the display text
• An M before the T separator of an ODS duration is a month, not a minute; non-zero years and months, which have no fixed length, now fall back to the display text instead of turning P1M into one minute

Version 2.384.30

Fixed

• TGWmfEmfRead closes the metafile when the file is too short or damaged, instead of keeping it locked until the process exits
• TGWmfEmfRead treats a placeable WMF whose header gives zero units per inch as unsized instead of dividing by zero
• TGWmfEmfRead keeps 32-bit sizes and bounds in BSEWMF, matching the picture header it describes

Version 2.384.29

Fixed

• WMF and EMF pictures of 64 KB or more in classic XLS workbooks survive open and save; their stored and uncompressed sizes were cut to 16 bits, so Excel could not decode the picture in the saved file
• Metafile picture bounds outside the -32768..32767 range are kept instead of wrapping around

Version 2.384.28

Fixed

• TXMLWriter no longer produces malformed XML when text or attribute values contain control characters, unpaired surrogates or U+FFFE / U+FFFF; characters that XML 1.0 cannot represent are left out
• TXMLWriter writes a carriage return in element text as a character reference, so the text reads back unchanged instead of having CR turned into LF
• TXMLWriter raises an error when the target stream accepts fewer bytes than written, such as on a full disk, instead of leaving a silently truncated file

Version 2.384.27

Fixed

• Copying a range whose formulas point above or to the left of the copied block, such as =B4+1 copied from B5, now pastes with the reference still one row up; it used to stay pinned to the original cell, and a formula could end up referring to itself
• The Excel clipboard data HotXLS publishes keeps the copied cells at their sheet position, so applications that read it see the original addresses

Version 2.384.26

Fixed

• The zlib string helpers (ZCompressStr, ZDecompressStr and the in-place variants) now process the whole string; with Unicode strings only the first half of the text was compressed, so a round trip lost data

Version 2.384.25

Fixed

• Hyperlinks in classic workbooks whose display text is flagged without the site flag are read correctly, and hyperlinks with an unusual target type no longer fail the whole open or return a garbled sub-address

Version 2.384.24

Fixed

• An exception raised by the file monitor change handler no longer stops the monitor silently; it keeps watching and reports later changes

Version 2.384.23

Fixed

• Theme colors with a tint below -1 in a malformed file are clamped to black instead of producing random colors when exporting or printing

Version 2.384.22

Fixed

• Data connections created through the API are saved with the refresh version attribute that the file format requires
• A connection with a missing or repeated name or id no longer makes the whole workbook fail to open; the entry is kept as it was and saved back unchanged

Version 2.384.21

Fixed

• Changing the source code of a VBA module now takes effect in Excel: the project is saved so that Office compiles the new source instead of running the cached compiled code of the old one
• VBA module source is now saved in the compressed form the format defines; the last block of every module was stored in a way Office treats as damaged
• A malformed VBA directory stream no longer makes opening the workbook hang, and a module entry missing its stream information no longer leaves an empty module in the project

Version 2.384.20

Fixed

• Defined names loaded from a classic workbook keep their shifted references after rows or columns are inserted or deleted and the workbook is saved; the file previously still contained the original reference

Version 2.384.19

Fixed

• Fitting only the rows or only the columns of an XLSX sheet no longer sets the width of every measured column, or the height of every measured row, to zero when it had no explicit size

Version 2.384.18

Fixed

• AutoFilter criteria in classic workbooks are joined the way they were set: AND conditions were saved as OR and the other way round, so Excel showed the wrong rows
• Boolean AutoFilter criteria are saved as TRUE or FALSE; they were stored in a form Excel reads as an error value

Version 2.384.17

Fixed

• The created and last-saved dates of classic workbooks now match Excel; they were written one day late and read one day early
• The last-saved date is stored under its proper document property; it was saved in the page-count slot, where Excel and Windows did not show it as a date
• The total editing time is stored as a duration, so Excel and Windows show the real number of minutes instead of about 299 years; times written by earlier versions are still read correctly

Version 2.384.16

Fixed

• The unique-value rule of the import check now compares the cell values; every value after the first one in a column was previously reported as a duplicate
• Adding a second rule for the same column replaces the first one cleanly; the old rule was freed but still checked, and freeing the checker raised an invalid pointer error
• Checking a sheet no longer creates empty cells for the blanks it reads
• Keeping only the violating rows no longer writes past its row table when a violation lies below the used range, and the caption row is inserted above the data instead of overwriting the first violation

Version 2.384.15

Fixed

• Reading a ZIP entry in small pieces no longer loses the last bytes of the entry when the compressed data ends in the middle of a repeated sequence, which could make a valid XLSX fail with an entry size error
• Decompressing data that expands beyond 2 GB now raises an error instead of crashing the 64-bit process
• The decompression stream returns all remaining data after its input is exhausted instead of stopping early

Version 2.384.14

Fixed

• Interlaced PNG images with 1, 2 or 4 bits per pixel no longer write past the image buffer, which could crash or corrupt neighboring pixels when exporting or printing a workbook that contains them
• Interlaced PNG images decode correctly when a pass starts with the Average or Paeth filter
• 2-bit grayscale PNG images now render their true gray levels, and reading single pixels of 2-bit grayscale and palette images returns the right colors
• PNG images whose compressed data has trailing bytes no longer hang the decoder, and valid images with a large image data chunk are no longer rejected with a CRC error
• Images with an invalid bit depth or a chunk length beyond the end of the file are rejected cleanly instead of crashing or allocating gigabytes of memory
• Text chunks saved into PNG images are no longer truncated or corrupted

Version 2.384.13

Fixed

• Classic workbooks encrypted with RC4 now use the correct password hash for passwords of 28 characters or more, so they open in Excel; passwords of 32 characters or more previously produced files that even HotXLS could not open
• RC4 passwords of up to 255 characters are accepted when opening, and files with 28 to 31 character passwords saved by earlier versions still open
• A malformed RC4 CryptoAPI header with an oversized length is reported as a corrupt file instead of raising a generic error

Version 2.384.12

Fixed

• Agile-encrypted workbooks that declare SHA-256 now decrypt correctly; the data segments were always decrypted with SHA-512 initialization vectors, so the workbook failed to open even with the right password
• Agile files whose key data and password key encryptor use different hash algorithms are no longer reported as having a wrong password
• A spin count above the 10,000,000 limit of the format is rejected immediately instead of keeping the CPU busy for hours
• A truncated password verifier no longer reads past its buffer, and the integrity key written by SaveAsEncryptedAgile now has the length the format requires

Version 2.384.11

Fixed

• Workbooks saved with SaveAsEncrypted now open in Excel. The ECMA-376 Standard Encryption key was derived without its final step and the package size was encrypted together with the data, so Excel rejected the password; workbooks encrypted that way by Excel 2007 can now be opened as well
• Encrypted workbooks saved by earlier HotXLS versions still open with the right password
• A malformed encryption header with an oversized length no longer causes an access violation while opening

Version 2.384.10

Fixed

• Copying and pasting formulas between cells, and sorting with formulas copied along, no longer moves absolute references: $A$1 stays put, A$1 moves only its column and $A1 only its row, matching Excel
• Function names that look like cell addresses, such as LOG10(, are no longer shifted as references when a formula is pasted, so the pasted formula no longer turns into LOG11( and #NAME?
• References qualified with a sheet name, such as Sheet2!A1 or 'My Sheet'!B3, now shift on paste like unqualified ones
• Pasting text that starts with = or an apostrophe into a classic workbook keeps it as text; it was previously turned into a formula or lost its first character

Version 2.384.9

Fixed

• Elapsed-time number formats such as [h], [mm] and [ss].00 are now recognized as time formats instead of being mistaken for color names; with precision as displayed turned off (FullPrecision set to False), a duration formatted as [ss].00 is no longer rounded as if it were a plain two-decimal number, which previously turned values of a few seconds into zero
• Minute tokens separated from the hour or second by a literal, as in h:mm and mm:ss, are now read as minutes rather than months when a number format is analyzed

Version 2.384.8

Fixed

• Copying text, HTML, or spreadsheet data to the Windows clipboard now uses a valid temporary owner, including in console programs and worker threads without an application window; copied data remains available after that owner is destroyed
• Failed clipboard publication now releases every allocated data block that Windows did not accept, including companion text and HTML blocks when spreadsheet publication fails

Version 2.384.7

Fixed

• Compiling the library from source with RAD Studio XE5 through RAD Studio 12 Athens failed after the Free Pascal support and the DATE leap-day fix landed: those compilers still parse the Free Pascal version check while skipping it, and reject the numeric cast behind the DATE(1900,2,29) special case. Both spots now use forms every supported compiler accepts, so the full RAD Studio range builds again

Version 2.384.6

Fixed

• Copying cells that carry a comment in a classic .xls workbook, whether within a workbook, into another workbook, or by shifting rows and columns, kept the comment text but dropped its formatted runs, so every copy showed the whole note in one font; the runs now travel with the text, and across workbooks their fonts are added to the destination workbook
• Copying a whole sheet with Sheets.AddCopy in a classic .xls workbook dropped comments on cells that held no value or formatting; those comments are now copied too
• Comments in .xlsx workbooks reopened without the font name, strikethrough, and color of their formatted runs, and saved runs never carried underline or superscript/subscript; all of these now survive a save and reopen, including runs copied from another workbook with CloneRangeTo

Version 2.384.5

Fixed

• Formatted runs in comments and text boxes in classic .xls workbooks no longer lose their fonts on save. A font used only by such a run could be dropped from the saved font list, most often when HotXLS re-saved a workbook it had already saved once, and the run then showed the default font in Excel or picked up a font added in the meantime; those fonts are now kept, and the runs are renumbered together with the rest of the font list
• TextRuns.FontIndex on comments and text boxes follows the file's font numbering, where index 4 is never used and the fifth font is index 5; the palette reservation for indexed font colors now reads these indexes the same way

Version 2.384.4

Fixed

• The font-numbering change in 2.384.1 was itself the bug: classic .xls files number fonts the way Excel does ([MS-XLS] §2.5.129), where index 4 is never used and the fifth font is index 5, but 2.384.1 switched to a plain zero-based count; saved workbooks then showed every custom font one slot early in Excel (the first one fell back to the default font), and Excel-authored workbooks opened with every custom font shifted one slot late. Saving and opening follow Excel's numbering again, verified against workbooks written and read by Excel itself
• Rich-text cells keep the font of each formatted run through open, save, and cross-workbook sheet copies, including runs whose font no cell uses directly
• Workbooks saved by 2.384.1 through 2.384.3 with five or more fonts already carry the shifted numbering, which neither Excel nor HotXLS can tell apart from valid data; regenerate them from their source, or reapply the affected fonts and save again

Version 2.384.3

Fixed

• Dates from 1900-01-01 to 1900-02-28 were off by one day against Excel in both workbook engines, because Excel's 1900 calendar counts a phantom 1900-02-29; reading and writing now apply the documented correction, DATE(1900,2,29) answers with 1900-02-28 instead of #VALUE!, and the shared-string table survives embedded NUL characters and double-byte code pages across record boundaries without desynchronising
• Comparisons chain left to right as Excel evaluates them, so =1<2<3 is now read as (1<2)<3 and returns FALSE like Excel (TRUE outranks every number in a comparison), where it used to group as 1<(2<3) and return TRUE; a blank operand equals both 0 and an empty string in comparisons; wildcard criteria apply to "not equal" in COUNTIF and SUMIF exactly as they already did for "equal"; the sum range of SUMIF and its siblings resizes from its top-left cell to the criteria range's shape instead of clamping; and INDIRECT honours its second argument (absolute R1C1 text) and rejects trailing garbage after the reference
• YEARFRAC with basis 0 applies the NASD last-of-February adjustments that DAYS360 already implemented, so spans touching February's end count correctly
• Text reading "endstream" inside a cell can no longer truncate a PDF page during stream compression, the total-pages token patch keeps the stream length consistent when compression is skipped, object offsets beyond 32 bits fail loudly instead of corrupting the cross-reference table, and grayscale or CMYK JPEGs are tagged with their real colour space instead of /DeviceRGB
• A workbook whose pivot-cache storage was stripped or truncated opens with the affected caches marked incomplete instead of failing the whole load, and re-saves those caches through the legacy inline layout rather than erroring
• A shared-string table whose declared string count exceeds what the record chain delivers now fails the open instead of silently resolving leftover cells to empty strings

Version 2.384.2

Fixed

• Converting a classic .xls workbook to .xlsx silently dropped merged cells that sat to the left of an earlier merge in a lower row; the scan no longer reuses its used-range variable for the merge area
• An XML comment inside a workbook extension list was re-emitted with doubled <!-- markers on every save, producing a file Excel must repair; comment nodes now round-trip byte-for-byte
• Unrecognised worksheet records captured from an opened .xls were dropped from every subsequent save despite being kept for replay, and leaked memory on each open; they now persist and replay verbatim through the standard record writer
• Defined names longer than 255 characters and name comments longer than 65535 characters corrupted the NAME record framing on save; both are now rejected up front
• An aborted save leaked the workbook and worksheet record buffers when an encryption mismatch raised mid-store; the save now releases its buffers on every exit path
• The in-place patching writer deleted the original file before renaming the replacement into place, so a failed rename destroyed the workbook; the swap is now a single atomic replace that keeps both files on failure
• Direct-reader cell references enforce the 16384-column and 1048576-row bounds instead of accepting overflowed coordinates, malformed integer attributes fall back to defaults instead of aborting the whole read, and dates between 1900-01-01 and 1900-02-28 convert with Excel's phantom leap day accounted for in both directions
• Attribute values containing tab, line feed, or carriage return round-trip unchanged instead of collapsing to spaces, and an unterminated entity in element text is now rejected like the attribute branch
• Verbatim copies of ZIP entries of 4 GiB or more, which cannot be framed correctly by the copy path, are now refused like the streaming writer instead of producing a corrupt archive
• The corpus runner writes its results atomically so an interrupted run cannot destroy retained evidence, records build timeouts instead of crashing, and tolerates malformed stage-trace lines; the LibreOffice verifier writes its evidence atomically and the file-replacement probe actually collects the antivirus product list again

Version 2.384.1

Fixed

• Every font from the fifth onwards was written with a record index one past the end of the font list, so Excel repaired the workbook and custom fonts came back as the default; both saving and opening now use the documented zero-based index, so custom fonts survive a round trip and Excel-authored files keep their fonts when opened in HotXLS
• Evaluating =0^-1 or a negative base with a fractional exponent crashed recalculation with a floating-point exception instead of returning #DIV/0! or #NUM!
• Conditional-format expression rules read from an existing .xls were re-saved as an orphan CF12 record under the plain legacy header, which Excel discards; they now keep the legacy record form and their highlight formatting
• The reference intersection operator and the implicit intersection of resized OFFSET and multi-cell INDIRECT results followed the wrong axis or rejected valid single-row and single-column ranges, returning the wrong cell or #REF!
• The streaming row reader reported every cached formula boolean as FALSE and every cached error as #NULL! when reading classic .xls files
• Large .xls files whose compound-file container exceeds the 109 header-resident FAT sectors could not be opened by the self-contained compound-file reader because the DIFAT extension chain was never followed
• Custom table styles reloaded their fill and font colours with the red and blue channels swapped, so a saved style came back in the wrong colour
• Malformed PNG images embedded in a workbook could crash the process through an out-of-bounds write during PDF image embedding or an unbounded text-chunk scan; embedded PNGs are now dimension-checked and malformed chunks rejected
• Multi-rectangle EMF regions rendered as their bounding box, EMF pictures without an explicit viewport extent were scaled wrongly, and superscript text rendered below the baseline on canvas, print, and metafile output
• Auto-prioritised BIFF conditional-format rules all carried priority 1, which Excel flags as a repair; auto priorities are now unique per sheet like their XLSX counterparts
• The xlSlantDashDot border style was written as xlDashDotDot in classic .xls border output
• A truncated encrypted-workbook header record raised an exception through Open instead of returning the password-error code, and several error paths leaked memory while parsing malformed image or workbook files

Version 2.384.0

Changed

• Classic XLS pivot caches are saved the way Excel saves them: each cache's records go to their own storage stream, while the stream identifier, cache scope, and worksheet or named-range source definition are written once in the workbook globals; caches authored by older HotXLS versions keep their previous inline layout unless the opt-in framing upgrade is enabled
• Password-protected workbooks encrypt each pivot cache stream as its own record stream that starts fresh at the beginning of the stream instead of continuing the workbook stream's encryption position

Fixed

• A PivotTable could point at the wrong cache after saving a workbook that mixed caches read from an existing file with caches created in code; the view-to-cache reference now follows the actual save order
• Every HotXLS package still carried the 2.379.7 version resource, and rebuilding a package in C++Builder could strip the version resource entirely; all packages now carry the current version

Version 2.383.4

Fixed

• Deep formula audits now keep changed results available to dependent formulas even when the changes fall within the reporting tolerance, and explicit apply refreshes these unreported results
• Relative tolerance is now honored for double-precision cached results in both workbook engines, while unchanged results avoid a duplicate comparison

Version 2.383.3

Fixed

• File-based workbook saves now preserve the native Windows error when a completed XLSX, XLTX, or ODS output cannot replace its destination, including when cleanup or a progress callback changes the thread error state

Version 2.383.2

Fixed

• Repeated report-band and range copies reuse defined-name suffix searches instead of rescanning every earlier copy, while preserving the first available name, worksheet scope, and reuse of names freed by later edits

Version 2.383.1

Fixed

• Formula dependency rebuilding avoids repeatedly scanning preceding formula rows, reducing the cost of large calculation chains while preserving array-output references, cross-sheet dependencies, and lookup ordering

Version 2.383.0

Added

• Macro-free Excel templates can be saved as XLTX through an explicit output-format selection, using files or streams and buffered or streaming output
• Template output preserves formulas, defined names, print areas, and worksheet metadata while leaving the default workbook save behavior unchanged; workbooks containing VBA are rejected before the destination is written

Version 2.382.5

Fixed

• Phonetic settings retain the required default font index, preventing Excel from rejecting saved workbooks when the index is zero
• Dynamic print-area formulas and sheet-qualified range unions survive a round trip without adding a sheet prefix to a function or truncating references inside an expression
• Generated and preserved package parts share unique content-type declarations, and custom properties no longer collide with opaque root relationship identifiers

Version 2.382.4

Fixed

• Scalar defined-name references use the current row or column intersection while range arguments retain their full ranges, avoiding false dependency cycles in loan templates
• Mixed integer and floating-point formula arithmetic retains floating-point precision instead of being truncated to four decimal places
• Recalculation prioritizes available lookup precedents before reading their caches, avoiding stale payment counts and similar lookup results

Version 2.382.3

Fixed

• AGGREGATE handles uncalculated errors and nested aggregation with independent selection settings, preserves array calculations, and propagates original error types
• Classic XLS saves preserve existing formula caches until explicit recalculation, including shared and array formula roots; shared formulas translate relative references for their destination cells
• Structured references support backslashes and periods in table names, nested current-row syntax, and escaped column names without changing similar text inside formula strings
• XLSX conditional formatting retains RGB colors, explicit font flags, underline styles, font names, sizes, and schemes
• Unmodified charts preserve their original definitions and nonzero anchor offsets, avoiding changes to series colors, positions, and sizes
• Setting page margins no longer writes an unspecified page orientation, and Classic XLS preserves absent imported page margins rather than adding defaults that change print placement

Version 2.382.2

Fixed

• ODS workbooks containing large repeated blank row ranges open and save without expanding every row height and inherited blank-cell style into individual objects
• ODS output preserves explicit row heights and compact repeated styled rows, including single-row overrides and cleared heights; row-height ranges also survive worksheet copying, row insertion and deletion, tiling, and XLSX export

Version 2.382.1

Fixed

• Opening and saving XLSX workbooks with full-width column settings no longer repeatedly scans column metadata or copies the growing column XML, eliminating round-trip timeouts in date and template workbooks while retaining column settings
• ODS pivot definitions retain inherited XML namespace declarations, including producer extensions, so their saved XML remains well formed

Version 2.382.0

Fixed

• Saving a classic XLS workbook now keeps nested SUBTOTAL cells out of enclosing subtotal results, preserving the original cached totals instead of double-counting group rows
• XLSX charts are loaded only through the drawing relationship referenced by each worksheet, so an unlinked conventional drawing part can no longer duplicate a chart during a round trip
• OpenDocument data-pilot table definitions survive an ODS save, including their source ranges, fields, target ranges, buttons, and producer extension attributes
• AGGREGATE options 0-7 now follow the documented hidden-row, error, and nested SUBTOTAL/AGGREGATE selection rules across reference and array calculation forms

Version 2.381.0

Added

• Table slicers (Excel 2013+): Worksheet.Slicers.Add(Table, ColumnName) creates a slicer view for any table column end to end — the backing slicer cache named Slicer_<column>, the reserved #N/A defined name, the worksheet slicerList extension and a markup-compatibility drawing anchor with a fallback notice for older Excel versions are all written automatically, and the view can be positioned, captioned and styled through Caption, ColumnCount, StartItem, StyleName, RowHeight and SetBounds
• Slicer round-trip model: slicer views and caches parse back on open (shared caches are reused across views, deleting the last view drops its cache), unmodified slicer parts and anchors replay byte-for-byte, and a table cache keeps its table binding valid even when saving renumbers table ids

Fixed

• Slicer caches written by earlier versions referenced the table id the file carried at open time; when saving renumbered table ids workbook-globally, the stale tableId could send Excel into repair. The binding now tracks the table through the renumbering

Version 2.380.0

Added

• Waterfall and treemap charts: the Excel 2016 extended chart types can be created, saved and reopened. They are serialized as dedicated extended-chart parts with their own content type, worksheet data binding and markup-compatibility drawing anchor carrying a fallback notice for older readers
• Treemap charts support hierarchical categories: every category level of the series is written and read back as its own level inside the category dimension, preserving leaf and parent labels

Version 2.379.4

Fixed

• A corrupt or hostile classic XLS container no longer corrupts memory while being opened. The compound file reader copied a whole sector into the mini stream even when only part of it belonged there, followed the mini allocation table past the end of that stream, accepted sector sizes the format does not define, and sized its allocation table from an unchecked count. Sector and mini sector chains are now cycle-guarded, and a stream name that fills its field without a terminator is bounded
• A style record naming a cell format index beyond the format table no longer writes through that index. The reference counters in the shared format and key lists checked only the lower bound, so a 12-bit index from the file could reach past the array
• A local file header written with ZIP64 extended information no longer overruns its buffer: the declared header size now accounts for the 28-byte extra field the writer appends. Entry offsets and stream positions are tracked as 64-bit values, a ZIP64 entry is read using the length from its extra field rather than the sentinel, and an entry larger than 4 GiB is refused instead of being recorded as its remainder
• A ZIP entry name containing a Windows separator is normalised. A single backslash was never converted to a forward slash, so ..\..\name passed the directory traversal check that splits on forward slashes
• A mask range with no upper bound, such as [a-, no longer reads past the end of the mask, and building a mask that starts with a character set no longer leaks a set per mask. Both are reachable from COUNTIF and SUMIF criteria
• Whole-cell search with wildcards now requires the pattern to match the entire cell, matching both the documented behaviour and the non-wildcard whole-cell comparison. FindText with pattern abc previously also matched a cell containing abcdef
• Precision as displayed no longer rounds cells whose number format has no digit placeholder. A cell formatted General by name, as text with @, or with a literal-only format was rounded to a whole number
• The shared string table no longer reports more unique strings than it writes, no longer attaches a deleted string's rich text runs to whichever string reuses its slot, and no longer emits a continuation record holding nothing but its option byte when a string overflows only because of its formatting tail
• RTF import stops duplicating non-ASCII characters. A \\uN escape is followed by fallback characters written as \\'hh, which the skip logic did not recognise, so each one was also emitted as text. Skipped groups now honour \binN raw payloads, and a table wider or taller than the worksheet grid stops at the edge instead of raising
• A number reference outside the character range XML defines is rejected on read instead of producing an embedded NUL, a lone surrogate or a wrapped code point. Tab, line feed and carriage return survive an OPC round trip as character references rather than being normalised to spaces
• CSV import keeps the SkipRows value across parses; it was decremented in place, so the second file read through the same loader skipped nothing
• A xl/tables/ part is recognised when deciding which parts to preserve; the prefix comparison was one character short and never matched

Version 2.379.3

Fixed

• HTML import decodes decimal character references. é and every other &#NNN; form produced a NUL character instead of the character itself, because the conversion wrote its result and its error position into the same variable and the success code overwrote the value. Hexadecimal references were unaffected; both forms now reject values outside the representable range instead of emitting a stray character
• HTML export keeps the whole sub-address of an internal hyperlink when building its anchor target. A sub-address longer than 255 characters was truncated, so the generated anchor no longer matched the link

Version 2.379.2

Fixed

• A GIF inserted through AddPicture or Images.AddFromFile now takes its size from the logical screen descriptor. Every GIF previously reported 100 by 100 points regardless of its real dimensions, so the picture arrived stretched or squashed
• BMP sizing reads the bitmap header the file actually declares: a BITMAPCOREHEADER with its 16-bit dimensions is understood instead of being read as 32-bit fields, and a top-down BMP with a negative height no longer yields a negative picture height
• JPEG sizing accepts every SOFn frame marker rather than only baseline, extended sequential and progressive. Lossless, arithmetic-coded and hierarchical JPEGs previously fell through to the 100 by 100 default

Version 2.379.1

Fixed

• WMF pictures now decode as vector scenes for SVG, PDF and paginated output. Since 2.363.0 the decoder read each record as a 16-bit size followed by the function code, but a metafile record starts with a 32-bit size, so every WMF written by Windows or Office stopped at its first record and the picture fell back to a raster placeholder. Records are now framed per [MS-WMF] §2.3, and the pen, brush, font, text, shape and polygon parameters are read from their specified offsets
• META_STRETCHDIB expected the DIB before the destination rectangle and assumed a 40-byte bitmap header; the rectangle now precedes the DIB as the format specifies, BITMAPV4HEADER / BITMAPV5HEADER and colour tables are honoured when locating the pixel bits, and the record contributes an image command at XDest, YDest sized DestWidth by DestHeight
• META_EXTTEXTOUT skips its optional rectangle when either ETO_OPAQUE or ETO_CLIPPED is set, so text drawn with an opaque background is no longer read from the wrong offset
• Metafile images embedded in a page render read their pixels from the offset the bitmap header states rather than a fixed 40 bytes, so a BITMAPV4HEADER or BITMAPV5HEADER image no longer renders as shifted colour noise

Version 2.379.0

Added

• TGridToXLS is now part of the Free Pascal runtime package, so a Lazarus TDBGrid exports the way it does under the VCL: explicit columns keep their titles, fonts, colours, alignment, widths and visibility, a grid without explicit columns exports one column per dataset field, and the result saves as XLS, XLSX, HTML or RTF through the same SaveAs call

Version 2.378.4

Fixed

• TEXTBEFORE and TEXTAFTER read their argument count from a local that was never assigned, so the optional instance, match-mode, match-end and if-not-found arguments could be ignored, or a valid call rejected with #VALUE!, depending on leftover stack contents; the count now comes from the call itself
• A resized OFFSET, one with a height or width other than 1, evaluated in a single cell compared the formula cell against an anchor that had not been computed yet, so the implicit intersection could return #REF! or the wrong cell; the shifted anchor is now computed before the intersection test

Version 2.378.3

Fixed

• Under Free Pascal, matching or replacing an empty string raised the regular-expression error, so a report expression such as regexreplace(Code,"[^0-9]","") failed with REPORT_EXPRESSION_INVALID_REGEX whenever the field was blank. The FPC backend now answers the way TRegEx does under Delphi: a pattern that accepts an empty string, such as ^$ or .*, matches it and yields its replacement text, while [0-9]+ or \s+ reports no match

Version 2.378.0

Added

• HotXLS now builds with Free Pascal 3.2.2 and Lazarus/LCL on Windows, Win32 and Win64. The XLS and XLSX workbook APIs, formulas, styles, the direct reader and writer, dataset export and the rendering helpers all work from FPC code with Delphi Unicode string semantics; the Free Pascal and Lazarus support topic covers application setup and the backend differences
• The runtime package Lib/FPC/HotXLSLaz.lpk opens in Lazarus or builds per architecture from the command line through build-FPC-Lib.cmd, and the Lazarus smoke suites under Tests/Lazarus cover workbook round trips, compression, encryption, COM streams, graphics, text reading, regular expressions, dataset export and report expressions
• FPC builds use bundled Pascal backends for ZIP, compression streams and AES, so XLSX files, including password-protected ones, need neither a zlib DLL nor the prebuilt object files Delphi links; PNG keeps its alpha channel, EMF drawing and playback go through Windows metafiles, and multipage TIFF export uses the GDI+ runtime

Version 2.377.2

Added

• A classic XLS PivotCache written by HotXLS now carries the records the format defines the cache header against. The header's stream id and source-data type used to name an SXStreamID and an SXVS that were never written, so no reader could check either one; both records are now written and both header fields are taken from the same values, so they agree by construction
• Cache substreams also carry SXDBEx, the record that follows the cache header in every cache Excel writes. It records when the cache was last refreshed, so TXLSPivotCache.RefreshedDate now survives a classic XLS save and reload the way it already did in XLSX
• TXLSPivotCache.StreamId and TXLSPivotCache.SourceDataType expose the two values, read back from a file that states them and written into the pair of records plus the cache header. A cache built by AddPivotTable reports stream 0 and a worksheet-range source

Changed

• The opt-in rewrite of old cache framing (UpgradeLegacyPivotCacheFraming) now produces these three records as well, so an upgraded cache is not just correctly framed but complete in the same way freshly written ones are. A cache that already carried an SXStreamID, an SXVS or an SXDBEx keeps what it carried; one whose SXDBEx claims formula records HotXLS does not write stays on raw replay
• Reading is unaffected: files written before this release load exactly as before, and cache records HotXLS still has no model for — the DConRef or DConName naming a source range among them — go on being preserved byte for byte across a round trip

Version 2.377.1

Added

• A workbook can now ask for its old PivotCache substreams to be rewritten on save. TXLSWorkbook.UpgradeLegacyPivotCacheFraming is off by default; turned on, a classic XLS file whose caches were written by HotXLS before v2.377.0 is saved with those caches re-emitted through the corrected SXDB / SXFDB / SXDBB writers, so the run of pivot cache fixes in v2.376.1 through v2.376.7 reaches files that already exist instead of only new output
• The upgrade applies only where it provably costs nothing. A cache read from a file is re-emitted only when its whole substream held records HotXLS models and its framing is one HotXLS itself wrote wrong; anything else — every cache Excel produced, and any cache carrying a record such as SXDBEx that HotXLS keeps only as opaque bytes — goes on replaying its own bytes with the opt-in on. TXLSPivotCache.CanUpgradeFraming reports the verdict for one cache, and RawFramingIsLegacy / RawModelIsComplete report the two halves it rests on
• An upgraded cache is written where its substream sat rather than moved ahead of the others, so a PivotTable view still resolves to the cache it named before the save

Fixed

• A classic XLS PivotCache read from a file and then handed to the writer with its raw-replay flag cleared produced two cache substreams, the model's and the file's. It now produces one

Version 2.377.0

Added

• Classic XLS worksheets accept embedded chart objects: AddChartObject on a worksheet anchors a chart frame on the sheet grid (OBJ type 5 host-control shape plus the MSODRAWING client anchor), builds the chart substream through the shared chart builder, and rides the sheet's record list so the file layout matches what the reader reattaches on open; a round trip brings the frame back with its chart model — series, title, and range references intact

Fixed

• SeriesText records decode their string payload with the right width: the reader passed the character count where a byte count was expected, so 16-bit captions and series labels came back truncated to half their length (a five-character title read as two characters)

Version 2.376.7

Fixed

• The cached rows behind a PivotTable in a classic XLS file are written the way [MS-XLS] 2.4.276 describes them: one record per row of cached data, holding one index per cache field, a single byte wide, or two bytes for a field with more than 255 distinct values. HotXLS had been writing a scheme of its own instead, squeezing each field into just enough bits to count its values and packing many rows into one record, which no other program could decode and which contradicted the more-than-255-values flag the field definition already carried
• Excel can now read the values a HotXLS PivotCache holds instead of having to rebuild them from the source range when the workbook opens, which is what a workbook whose source data has since been moved, changed or deleted was relying on
• Cached rows are read back as well. HotXLS ignored the record entirely before, so a cache loaded from a file came back with every row pointing at the first value of every field, and anything computed from that cache followed those wrong values. This applies to a workbook written by Excel just as much as to one written by HotXLS
• A cache field carrying no values contributes nothing to a cached row, and a field with more than 255 values is stored and read at the wider size the format requires
• Workbooks written by HotXLS v2.376.6 and earlier still load with their cached rows intact

Version 2.376.6

Fixed

• The PivotCache header record in a classic XLS file is laid out the way [MS-XLS] 2.4.275 describes it. The record count is a four-byte field there and HotXLS was writing it in two bytes, so every field behind it went out one position early and the record ran four bytes long — the number of cache fields ended up inside the record count, the flag word landed on the stream identifier, and the field count landed on the length of the name of whoever last refreshed the cache, promising a name the record was too short to hold
• The header's flags now say what the cache actually is: whether its records are stored in the file, whether it refreshes when the workbook opens, whether refreshing is allowed, and whether the refresh query runs in the background. Those four settings already appeared in XLSX output and now travel with a classic XLS file as well, together with the name of whoever last refreshed the cache
• The source data type in the header is written as a worksheet range rather than left at a value the format does not define, and the field counts and the used record count carry the cache's real totals
• The refresher name is trimmed to the 255 characters the record allows, and a cache without one writes no name at all instead of an empty placeholder
• Workbooks written by HotXLS v2.376.5 and earlier still load, and still give up their cache record count. Nothing else was ever stored in the old header — the writer filled every field behind the record count with fixed values it never read from the cache — so the flags and the refresher name of such a file come back as defaults rather than as recovered values

Version 2.376.5

Fixed

• PivotCache field names survive a save and reload of a classic XLS file. The name length was being read from the wrong place in the cache field definition — an index that is always $FFFF — so the name was discarded every time and every cache field came back unnamed
• The cache field definition record is laid out the way [MS-XLS] 2.4.283 describes it: grbit, ifdbParent, ifdbBase, citmUnq, csxoper, cisxoper, catm, and then the field name. HotXLS had been writing three fields the record does not have — a subtype item index, a number format and a second item count — and putting $FFFF into indexes the specification bounds by the number of fields in the cache
• The record's flag word now describes the field it belongs to: whether the field carries cache items, whether those items include numbers, dates or text, whether a minimum and maximum can be computed, whether any item is not a date, and whether there are more than 255 of them. The item count in the record now matches the number of cache item records that actually follow it
• Cache field names longer than the 255 characters the record allows are trimmed to 255 instead of being written out at their full length
• Workbooks written by HotXLS v2.376.4 and earlier still load with their cache field names intact — the name occupied the same position in the old layout, so nothing that carried a name loses it

Version 2.376.4

Fixed

• PivotTable caches in classic XLS files now use the record numbers [MS-XLS] gives them. Seven records in the cache substream had been going out under some other record's number: the cache field definition as $00C8 (SXDBB), the field type as $00FB (SxFormat), the per-record item indices as $00C5 (SXDI), and the numeric, boolean, error, date and blank cache values as $00CE, $00CF, $00D0, $00CC and $00CB (SXDtr, SxNil, SXTbl, SXInt, SxErr). They are now SXFDB $00C7, SXFDBType $01BB, SXDBB $00C8, SXNum $00C9, SxBool $00CA, SxErr $00CB, SXDtr $00CE and SxNil $00CF, so Excel can read a pivot cache HotXLS wrote and HotXLS can read one Excel wrote
• Blank values in a pivot cache field survive a round trip. The writer had always emitted them but the reader had no case for them, so every blank was dropped on load and each value after it in that field shifted down one position, leaving the PivotTable's item references pointing at the wrong values
• Workbooks written by HotXLS v2.376.3 and earlier keep loading with their cache values intact. Each pivot cache substream is read in whichever numbering it was written with, decided from its own field-definition record, and a cache read out of an existing file is written back under the numbers it came in with, so reopening and saving an old workbook does not quietly change its bytes

Version 2.376.3

Fixed

• PivotTable views in classic XLS files now write their SXEx extension record under the number [MS-XLS] gives it, $00F1; the writer had been using $00C6, which is SXDB, the pivot cache header record — Excel looking for the view extension found nothing and a record claiming to be a cache header sat in the middle of a worksheet, so the view properties SXEx carries were lost even after the field-layout fix in v2.376.1
• Workbooks written by HotXLS v2.376.2 and earlier still open with their PivotTable views intact: inside a worksheet the old $00C6 is still accepted as the view extension it was meant to be, and inside a pivot cache substream $00C6 is still read as the cache header, so nothing that used to load stops loading

Version 2.376.2

Fixed

• Classic XLS 3-D chart sheets created without an explicit scene now carry the flags they are documented to carry: the default Chart3d flag word set a bit the record reserves instead of the automatic-scaling bit, so a default 3-D chart reopened with clustering on but automatic scaling off and never matched the scene it claimed to write; charts that pass an explicit 3-D scene were unaffected and their bytes are unchanged

Version 2.376.1

Fixed

• PivotTable views write an SXEx record whose fields sit where readers expect them: the fixed 24-byte body was shifted by two fields, so it announced two formatting records that never followed and left every optional-string length at zero, which [MS-XLS] 2.4.282 forbids — those lengths must be either the absent marker or a real length from 1 to 255; the body now starts at the correct field and marks every optional string absent, so Excel reads the view properties instead of a garbled block
• The SXEx record length is derived from the bytes actually written rather than a hardcoded constant, so the declared size and the payload can no longer drift apart the way they did before v2.376.0

Version 2.376.0

Added

• Classic XLS chart sheets can link to a PivotTable view: a new AddChartSheet overload accepts the source pivot view name and the hide-field-captions flag, which write into the SXViewLink ($0858, [MS-XLS] 2.4.316) and PivotChartBits ($0859, 2.4.196) records in the chart globals before Units per the Chart Sheet Substream ABNF; chart inspection decodes both records — including 8-bit and 16-bit string encodings — back into the typed model (pivot-chart marker, source view name, caption flag) and the canonical semantic snapshot, while charts without the link records stay unmarked

Fixed

• PivotTable views write a correctly sized SXEx record: the emitter declared a 24-byte body but wrote 26 bytes, so the two surplus bytes landed where the next record header was expected and every record after the view (including the worksheet EOF and any chart-sheet substream that followed a pivoted sheet) misparsed, leaving workbooks that combine a PivotTable with a chart sheet to reopen with the chart missing
• Trendline display names in classic charts decode 8-bit string payloads correctly; the previous wide-character reader returned an empty name for records Excel writes with 8-bit characters

Version 2.375.0

Added

• Classic XLS chart series accept theme-colored fills: the per-series style now carries a theme color index and tint which write into a GelFrame record after the AreaFormat per [MS-XLS] 2.4.131, using the OfficeArt theme color extension (fillType, fillColor with the flattened RGB, fillColorExt scheme index, fillColorExtMod tint payload); when no explicit fill color is supplied the flattened color resolves from the default Office theme with the tint applied, and chart inspection decodes the theme intent back into the typed model and the canonical semantic snapshot

Fixed

• The full multi-version library build is green again for every supported compiler: casts and calls that only newer compilers accept (an Extended type cast of Int64 values, a hard Double cast of an integer, a nil procedural argument in overload resolution, access to TPicture.LoadFromStream where it is protected, and the GdiplusStartup token type that changed between VCL generations) were made version-safe from XE5 through the current release

Version 2.374.0

Added

• 3-D chart scenes are configurable and inspectable: a new AddChartSheet overload accepts rotation, elevation, perspective distance, height, depth, gap, and the cluster, autoscale, right-angle-scaling, and two-dimensional-walls flags which are written into the Chart3d record per [MS-XLS] 2.4.46, and chart inspection decodes the scene back into the typed model (Scene3D) on read; charts that keep the previous overloads still emit the Excel-matching defaults byte-for-byte

Version 2.373.0

Added

• Classic XLS chart series accept per-series style overrides: line color and style, fill color, and marker shape, colors, and size write into the series formatting block with the automatic flag cleared so Excel honors the supplied values, and chart inspection decodes the LineFormat, AreaFormat, and MarkerFormat records back into the typed model and the canonical semantic snapshot

Version 2.372.0

Added

• Classic XLS chart series accept named and external references for categories and values: text that is not a plain A1 range compiles through the workbook formula engine, and the BRAI record carries a chart area item holding the compiled formula so the save pass re-serializes it with the current defined-name store indexes; chart inspection decompiles the reference back to its name on read, while unresolvable text keeps the empty placeholder BRAI

Version 2.371.0

Added

• Classic XLS chart series can bind to the secondary axis group: marking a series in AddChartSheet raises AXESUSED to two, emits a second AxisParent block with its own category and value axes plus its own chart group per the [MS-XLS] Chart Sheet Substream ABNF, and routes the series through SerToCrt to the second group; chart inspection decodes the per-series chart-group binding on read

Version 2.370.0

Added

• Classic XLS chart series accept data labels: AddChartSheet emits an AttachedLabel record with the show-value, show-percent, show-category-name, show-label-and-percent, show-bubble-size, and show-series-name flags ([MS-XLS] 2.4.5 bit layout), and chart inspection decodes both AttachedLabel and the DataLabExtContents extended-label record including its separator back into the typed model and the canonical semantic snapshot

Version 2.369.0

Added

• Classic XLS chart inspection now decodes the axis group structure: AXESUSED surfaces the declared axis group count and every axis records whether it belongs to the primary or secondary AxisParent group, with the group index carried into the canonical semantic snapshot

Version 2.368.0

Added

• BIFF chart trendlines and error bars now decode into the typed chart model on read: each Series block surfaces its Trendline records with regression kind, order or period, intercept (including automatic), equation and R-squared display flags, forecast and backcast periods, and display name, plus its SerErrBar records with direction side, value type, end caps, and magnitude
• The Classic semantic snapshot maps trendlines and error bars into the shared canonical containers instead of raising, so workbooks with advanced chart series compare deterministically and trendline changes alter the canonical stream

Version 2.367.0

Added

• RecalcAndVerifyEx runs a deep calculation audit on both workbook engines: every formula evaluates once through the dependency graph into an isolated result overlay so dependent cells read this pass instead of stale caches, evaluation failures classify into typed categories (unsupported arguments, circular references, denied or missing external references, skipped data tables, cancellation, internal failures) with result codes, and each failure carries a bounded call stack
• The audit honours an issue budget with a stable sheet, row, column, and category ordering, a progress sink that can cancel at node boundaries, configurable numeric tolerances, and an opt-in ApplyResults that commits recomputed values atomically after a fully successful pass over an unchanged workbook structure; the default stays strictly read-only

Changed

• The existing RecalcAndVerify and CalculateAndVerify entry points keep their previous shallow-compare semantics as compatibility shims while sharing the new report fields

Version 2.366.0

Changed

• PDF export now assembles documents through a sequential object writer that streams every object straight to the destination, so no whole-document intermediate copy is built; unsigned documents stream directly into forward-only streams, while signed documents assemble into a bounded spool that receives the ByteRange and signature contents patches before a single forward copy
• PDF attachment payloads copy from their provider spools to the final document in fixed chunks and reference their stream size through an indirect length object, keeping peak memory flat for multi-hundred-megabyte attachments

Fixed

• Signature digests are now computed incrementally over the two covered byte ranges of the assembled document instead of through a full document copy, and the reserved signature fields are located from the signature object itself rather than by scanning the assembled bytes

Version 2.365.0

Added

• Visible PDF signature text now renders through the embedded Unicode font pipeline with left, centered, or right alignment, configurable size, color, and bold or italic styling, covering any script in signer names, reasons, and locations
• PDF signature appearance images accept JPEG payloads with dimensions derived from the frame marker, while oversized or corrupt images fail deterministically before the digest is computed
• PDF attachments expose an explicit duplicate-name policy, spool large provider payloads to a temporary file beyond a configurable byte threshold, and reject leaf-hostile names containing path separators or control characters

Changed

• PDF attachment name trees are now sorted by their encoded name bytes per ISO 32000-1 7.9.6, and the attachment collection stays frozen while providers run so reentrant additions fail atomically

Version 2.364.0

Added

• Embedded EMF and WMF pictures replay path brackets with fill, stroke, and combined stroke-and-fill passes, rounded rectangles, arcs, chords, pies, and angle arcs through exact arc-to-Bézier conversion, region fills and frames resolved from bounded RGNDATA rectangles, and geometric or cosmetic extended pens, all flattened into the immutable shared scene with the same fail-closed boundaries against clipping, escape, and private records

Version 2.363.0

Added

• Embedded EMF and WMF pictures now replay through an expanded safe record whitelist: device-context state stacks with save and restore, window and viewport mapping with world transforms, pen and brush object tables including stock objects, rectangles, ellipses, line chains, polygons, polylines, and Bézier curves with dash styles, bounded text runs with fonts, colors, and alignment, and restricted stretch-DIB images re-encoded as PNG, while unknown state records still fail closed into the bounded raster fallback instead of invoking platform metafile playback

Fixed

• WMF pictures now parse the standard two-byte record header layout, so real placeable and standard metafiles decode instead of only synthetic fixtures

Version 2.362.0

Added

• Classic XLS and XLSX now preserve table-formula array state, typed data-validation IME modes, calculation-thread hints, force-full-calculation presence, refresh-all hints, and workbook code names through their native binary or OOXML representations
• RecalcAndVerify returns a bounded structured calculation audit with stable cache, parse, missing-function, and missing-name categories while leaving formula caches and workbook semantics unchanged
• TXLSPageImageExporter.ExportRange renders a requested cell rectangle as one compact PNG or JPEG without paper, margins, headers, footers, print titles, or page breaks, and page-image before and after events can draw on the actual encoding canvas
• A dedicated cell-image exporter measures, renders, and saves any classic or XLSX cell rectangle as a single compact image with configurable resolution, scale, format, and JPEG quality; transparent or colored backgrounds, independent object, gridline, and cell-border switches, pixel and byte budgets, and file saving through an atomic same-folder replacement that leaves an existing target untouched when anything fails
• Page-image paint events now receive a complete read-only context carrying the live canvas, pixel bounds, page size in points, resolution, scale, document and in-sheet page numbers, total page count, and the originating worksheet, fire exactly once per rendered frame including multipage TIFF frames, stay silent during measurement, and never trigger their trailing event or write partial image bytes when the leading event raises
• TXLSSvgExport supports complete per-page callbacks, deterministic document and page ID namespaces, and opt-in safe hyperlink overlays with ScreenTips and secure new-window attributes
• Classic and XLSX HTML exporters provide an opt-in spreadsheet view with semantic row and column headings plus bounded pure-CSS hot tracking, and both reject active or local URI schemes while retaining link display text
• Visible PDF signatures can use contained, covered, or stretched PNG appearance images, and PDF attachments can obtain their payload lazily from a single-call source interface during each save
• A bounded EMF and WMF vector decoder validates headers, record sizes, totals, exact EOF placement, and escape restrictions, then replays its safe rectangle command subset through the shared Canvas and SVG backends without invoking driver metafile playback

Changed

• Direct XLSX writing can author table formula array attributes and typed data-validation IME modes, while array table formulas are no longer silently expanded as scalar per-row formulas

Version 2.361.99

Changed

• Automatic tagged PDF now emits informative XLSX pictures as independent Figure structure elements with Unicode alternative text, dense page-local marked-content identifiers, exact parent-tree entries, and one semantic element per visible instance while equal image payloads still share one XObject; pictures without alternative text remain decorative artifacts

Version 2.361.98

Fixed

• Lookup formulas that scan a range covering their own cell (VLOOKUP, HLOOKUP, LOOKUP, XLOOKUP, XMATCH, MATCH) no longer report false circular-reference errors during recalculation, while edits inside the scanned range still trigger recalculation of the formula
• AutoFilter text criteria longer than 255 characters are capped to the filter record limit instead of writing a corrupted autofilter condition
• Consecutive spaces used by the implicit intersection operator are preserved when formula text is regenerated instead of collapsing to a single space
• Built-in defined names beyond print areas, print titles and filter databases (consolidate areas, extract, database, criteria, recorder, sheet titles and Auto_Open-family markers) are preserved verbatim when classic XLS workbooks are opened and saved

Version 2.361.97

Added

• Classic XLS and XLSX workbooks now expose a shared three-state object display policy with lossless untouched source preservation and cross-format mapping; hidden objects no longer expand paginated or HTML output or enter drawing and media rendering paths
• The workbook viewer renders lightweight named placeholders for placeholder-mode objects, while PDF, SVG, bitmap, print and HTML exporters retain full objects by default and offer an explicit placeholder option that avoids image decoding and chart-scene construction

Version 2.361.96

Changed

• Manual chart layouts now preserve independent source fields and invalid tokens, resolve factor and edge coordinates with bounded size-preserving placement, honor inner and outer plot targets plus title and legend overlay, convert supported ODF geometry through absolute chart distances, and render consistent title, legend, plot-area and axis-title placement across pagination, SVG, PDF, image, preview, print, viewer and HTML output

Version 2.361.95

Changed

• XLSX cell-sourced chart data labels now belong to each series with complete formulas, indexed caches, explicit shared and per-point visibility, live worksheet refresh, external-cache fallback, formula translation during copies and report expansion, and consistent custom-first label composition across pagination, SVG, PDF, image, preview, print, viewer and HTML output

Version 2.361.94

Changed

• Multi-level chart category axes now preserve sparse hierarchical caches and live two-dimensional ranges, build parent-aware spans in linear time, honor single-level, reversed, right-to-left, hidden and blank-category behavior, and render consistent hierarchy labels and separators across pagination, SVG, PDF, image, preview, print, viewer and HTML output; ODS export uses the leaf category range

Version 2.361.93

Changed

• Date axes now infer calendar units once across all visible series, choose bounded automatic intervals from the visible span, axis width, measured label font, number format, and rotation, and derive labels, coordinates, ticks, and gridlines from shared major and minor tick collections across pagination, SVG, PDF, image, preview, print, viewer, and HTML output

Version 2.361.92

Changed

• XLSX date axes now preserve axis number formats and source-link state, render formatted labels with text styling and rotation, and honor reversed direction while retaining continuous spacing for duplicate and sparse dates across pagination, SVG, PDF, image, preview, print, viewer, and HTML output

Version 2.361.91

Changed

• XLSX date axes now preserve whether base, major, and minor calendar units and automatic-category detection were omitted or explicitly supplied, including explicit day units and false values; Classic date-axis extension records expose typed fields, retain masked or invalid raw values, and can map active settings into the XLSX axis model

Version 2.361.90

Changed

• XLSX date-axis charts now preserve continuous date spacing and render day, month, or year major intervals with the workbook's 1900 or 1904 date system across paginated, SVG, PDF, image, preview, print, viewer, and HTML output; pagination-based output also honors configured minor ticks and gridlines

Version 2.361.89

Added

• The workbook viewer now tiles valid worksheet backgrounds behind unfilled cells, while print, PDF, SVG, bitmap, metafile, and Classic or XLSX HTML output expose an explicit ExportSheetBackgroundImages opt-in that is disabled by default

Changed

• Paginated background rendering now preserves worksheet-relative tile phase across pages, decodes each background once per render session, reuses one PDF image XObject or SVG resource, and skips corrupt payloads with one diagnostic

Version 2.361.88

Added

• Classic worksheets now expose typed background-image authoring and a shared read-only background resource, read and write BIFF8 background records with continuous payload splitting, reconstruct bitmap payloads as standard BMP data, preserve unsupported native and unknown formats byte-for-byte with diagnostics, deep-copy background ownership, and convert known backgrounds in either direction between XLS and XLSX

Version 2.361.87

Fixed

• XLSX worksheet copies now retain sheet background image bytes and media type in same-workbook and cross-workbook copies, while same-workbook copies reuse preserved source media and cross-workbook copies rebase media ownership before save

Version 2.361.86

Added

• Classic XLS and XLSX workbooks expose AllowUnsafeFormulaCallbacks, while contextual evaluation provides the matching per-call option and the structured xlfeUnsafeFunctionDenied status

Security

• Dangerous external and macro-style formula names no longer reach workbook-local, process-wide, resolver, or event user-function handlers by default, and denial occurs before formula arguments are evaluated

Version 2.361.85

Added

• Classic XLS and XLSX workbooks expose AuditFormulaSymbols for structured unknown-function, unknown-name, and syntax diagnostics across formula cells and defined names, with per-operation unique-formula and cache-hit statistics

Changed

• Preserved unknown functions and names now evaluate to Excel #NAME? errors through direct, dependent, contextual, serial, and parallel calculation paths
• Workbook-local or process-wide user-function registration changes and defined-name formula changes now invalidate affected calculation state before the next recalculation

Fixed

• Formula errors read from dependent cached cells now remain typed errors instead of entering scalar conversion and raising a Variant conversion exception

Version 2.361.84

Added

• Normal Classic XLS and XLSX formula assignment now applies the workbook formula-entry policy atomically, so strict rejection leaves the cell value, formula cache, and dependencies unchanged while preservation mode stores unknown functions and names

Changed

• Classic XLS writes preserved unknown calls through BIFF add-in user-function tokens and retains preserved unknown-name text across save, reopen, and XLS/XLSX conversion
• Formula entry validation recognizes configured callback user functions, LAMBDA and LET callables, defined-name lambdas, report placeholders, registered external links, and valid structured references without invoking handlers

Fixed

• XLSX validation now preserves valid nested and current-row structured references and no longer reuses stale column text while scanning an unresolved table reference

Version 2.361.83

Changed

• Classic XLS and XLSX value assignment now applies FormulaEntryPolicy.EmptyStringMode consistently: the default clears exact empty strings, preservation mode stores a constant empty string, and Null or Unassigned always clears the cell
• Native XLS, XLSX, and ODS loading plus XLS-to-XLSX and XLSX-to-XLS conversion now preserve trusted constant empty strings without confusing them with absent cells or formulas such as =""
• XLSX string and Variant row or column writers retain packed-storage fast paths while leaving cleared entries absent and storing exact empty strings only in preservation mode

Version 2.361.82

Added

• Classic XLS and XLSX workbooks now expose a shared FormulaEntryPolicy, while both worksheet engines provide mutation-free ValidateFormulaEntry methods that return a structured issue, action, and unknown symbol before a caller changes a cell

Changed

• Formula entry validation now distinguishes exact empty input from formulas returning an empty string, separates unknown functions, unknown names, syntax failures, invalid targets, and target-format limitations, and recognizes registered user functions without invoking their handlers

Version 2.361.81

Changed

• Whole-row and whole-column insertions or deletions now remap every pane's ordered selection areas and active area consistently in XLS and XLSX, preserving surviving order and creating a boundary-cell fallback when every area is removed
• Large pane selection normalization, structural editing, and queries now use amortized-linear builders and capacity growth instead of repeated full-string or array copies

Fixed

• Classic XLS now preserves malformed or contradictory Selection record bodies byte-for-byte while reporting xlsDiagnosticSelectionRecordInvalid instead of exposing partially trusted typed state

Version 2.361.80

Added

• TXLSWorksheet and TXLSXWorksheet now expose matching TryGetWindowScroll and ScrollWindow methods for reading and changing the first visible cell of any existing logical pane with 1-based coordinates

Changed

• Four-pane scroll state now preserves the shared upper, lower, left, and right axes independently through XLS and XLSX save, open, and worksheet copy, while missing panes and invalid coordinates fail without changing view state

Version 2.361.79

Changed

• Classic XLS worksheets retain independent ordered selection groups for every pane, aggregate consecutive BIFF Selection records into one logical sequence, preserve untouched record bodies, and copy the complete selection state with the worksheet
• Classic selections with more than 1369 areas are written as consecutive standards-compliant chunks while keeping one aggregate active-area index across every chunk

Version 2.361.78

Added

• TXLSWorksheet and TXLSXWorksheet now expose matching pane-aware GetSelectedAreas and SelectAreas methods through the shared TXLSPanePosition type

Changed

• Pane selection queries return caller-owned ordered snapshots with the active area index, preserve overlapping areas, and leave worksheet state unchanged when a pane is absent or an update is invalid

Version 2.361.77

Added

• IXLSWorkbook, TXLSWorkbook, and TXLSXWorkbook now expose matching streaming SaveSemanticSnapshot and incremental SHA-256 ComputeSemanticHash methods with Content and SecurityComplete profiles plus explicit viewer-state, volatile-metadata, and formula-cache options

Changed

• Semantic snapshots traverse sparse workbook models without materializing a complete snapshot, and both snapshot and hash operations hold a read lease so concurrent public mutations fail before changing workbook state

Version 2.361.76

Added

• The typed Classic XLS chart model now exposes chart, category-axis, and value-axis titles

Fixed

• Classic XLS chart model extraction now identifies series, legends, chart groups, axes, 3-D state, series labels, and series source formulas from their defined BIFF records
• Legend positions and pie-versus-doughnut classification are decoded from their defined record fields

Version 2.361.75

Added

• The Delphi FeatureShowcase and C++Builder XlsxFeatureGallery now demonstrate the same compiled report data context, memory table, strict data access, recoverable-error continuation, diagnostics worksheet, immutable run result, and atomic file-output workflow
• The compiled reporting topic now provides matching Delphi and C++Builder examples, including the C++Builder smart-interface acquisition required for interfaced report tables

Fixed

• The C++Builder XLSX feature gallery links the complete version-matched HotXLS static package so compiled-report implementation dependencies resolve without fragile per-unit linker directives

Version 2.361.74

Changed

• RunToStream, RunToFile, and RunToEncryptedFile now share one output transaction that publishes a serialized workbook only after report generation and saving complete
• Unencrypted report output preserves the template workbook format, including ODS output from ODS templates, while encrypted output remains an encrypted Office container

Fixed

• Serialization, replacement, cancellation, and fatal report failures leave existing destination streams, files, and encrypted files unchanged and release per-run data sessions and temporary storage before returning

Version 2.361.73

Added

• Named range and Excel-table report bands now expose ErrorMode, StrictDataAccess, RuntimeLimits, and immutable LastResult snapshots, with a distinct CompletedWithErrors band state

Changed

• WriteAndContinue can write an explicitly recoverable row error as literal diagnostic text, record the source cursor row and actual output cell, then continue later targets in that row and later input rows
• Report expressions consume typed band snapshot values before falling back to text, preserving numeric, date, Boolean, and other supported Variant values during expression evaluation

Fixed

• Fatal target, cancellation, and runtime-limit failures during band expansion restore source targets and inserted structure, clear rolled-back diagnostic flags, and publish zero applied targets

Version 2.361.72

Added

• TXLSReportRuntimeLimits bounds retained runtime issue count, sanitized message length, and issue memory, while RuntimeDiagnosticsSheetName optionally writes one structured diagnostics worksheet
• Explicit resolver data errors, expression division by zero, and invalid regular expressions now carry stable provider or expression issue categories and error codes

Changed

• WriteAndContinue preserves safe non-cell targets and records their issue, while worksheet names and defined-name formulas remain fail-fast structural targets
• Resolver data-error results remain recoverable when served from the per-run alias cache, while exceptions thrown by a resolver remain fatal

Security

• Runtime messages stop at the first line, redact common credential and bearer-token forms, and enforce count, length, and memory limits before mutating the failing target

Version 2.361.71

Added

• Exact {{#usertable(alias, logicalName, providerKey, name=expression)}} cells compile into typed lazy table declarations, and TXLSCompiledReportTemplate.UserTableCount reports the declaration count separately from output targets
• TXLSReportTemplateChecker.DataContext preflights registered provider keys, typed parameter bounds, malformed declarations, and case-insensitive alias conflicts without invoking a resolver

Changed

• Report templates and generators evaluate declaration parameters before target writes, register them in the current run session, and resolve a declared table only when dbvalue or dbcount first references its alias
• Repeated equal declarations share one request, while declarations that reuse an alias with a different logical name, provider, parameter order, type, or value fail with the declaring cell location before workbook mutation

Security

• User-table declarations accept only a registered provider key and bounded typed data parameters, reject data-table access inside parameter expressions, and expose no template field for SQL, paths, URLs, or command text

Version 2.361.70

Added

• TXLSReportDataRequest carries named or user-table identity, typed parameters, source location, and read-only run control to IXLSReportDataResolver, whose explicit result states distinguish unhandled, resolved, not-found, and data-error outcomes
• TXLSReportDataContext.RegisterResolver, RemoveResolver, and ClearResolvers manage a case-insensitive named resolver registry, while TXLSReportDataLimits bounds aliases, resolvers, parameters, names, messages, and returned table schemas

Changed

• Each report data session snapshots the resolver registry into an O(1) lookup, keeps explicit tables authoritative, adapts legacy table providers, caches typed requests by alias, and rejects a reused alias whose logical name, provider, parameter order, type, or value differs
• Report runners and bands now pass one thread-safe cancellation control through data resolution, stop after cancellation without committing the active target, and leave a canceled alias available for a clean retry instead of caching cancellation as a data error

Security

• Typed data requests reject control characters, negative source coordinates, duplicate parameter names, unsupported Variant arrays or references, excessive strings, excessive aliases or providers, and oversized table schemas before unsafe work or workbook mutation

Version 2.361.69

Added

• TXLSXWorksheet.ClearRowHeight and ClearColWidth remove one explicit dimension override without disturbing other row or column metadata

Changed

• Compiled report targets now finish expression evaluation before invoking dynamic formatting and commit cell content, number formats, row heights, and column widths through one rollback boundary
• Recoverable cell diagnostics derive red-font and yellow-fill styles from the cell's existing font, preserve the remaining cell style, and reuse pooled styles across equal base fonts and repeated runner executions

Fixed

• A recoverable evaluation failure no longer leaves dynamic number formats or explicit row and column dimensions behind, and diagnostic writes now clear any formula before storing literal text

Version 2.361.68

Added

• Compiled report runners and generators expose ErrorMode, StrictDataAccess, and an immutable LastResult with terminal state, applied count, failure details, and ordered runtime issues
• WriteAndContinue can replace explicitly recoverable cell data failures with literal diagnostics and finish as CompletedWithErrors, while the default Stop mode remains compatible

Changed

• Strict data access reports missing tables, unavailable random access, missing rows, and missing fields through stable error codes when no explicit template fallback exists
• Only explicitly marked recoverable data failures may continue; structural, callback, cancellation, resource, security, serialization, and other fatal failures still stop the report

Fixed

• File and encrypted-file output failures now publish a failed run result instead of retaining a completed result from the preceding workbook generation phase

Version 2.361.67

Changed

• Compiled templates, runners, range bands, and Excel-table bands now resolve dbvalue and dbcount through one isolated data session per logical run
• Repeated aliases are loaded once within a run, later runs reload provider data, concurrent runs keep independent caches, and both passes of PreprocessAndRun share one consistent table snapshot
• Report-band expressions bind only their row or column variables, so literal table names and built-in location variables are no longer mistaken for source columns

Fixed

• Report data sessions and their provider snapshots are released after successful, canceled, or failed execution, including provider exceptions and reusable runner retries

Version 2.361.66

Added

• TXLSReportDataContext.BeginRun returns an isolated IXLSReportDataSession snapshot with state inspection and resolved or cached alias counts

Changed

• Report data sessions use a case-insensitive hash cache for explicit tables, lazy provider results, missing aliases, and provider data errors
• Each distinct alias reaches the provider chain at most once per session, provider results stay bound to the requested alias, and same-alias recursion fails immediately
• Direct TXLSReportDataContext.TryGetTable and RequireTable calls retain a separate compatibility cache that is invalidated when registrations change

Version 2.361.65

Added

• IXLSXImageBlob and TXLSXImageBlob provide immutable raster and SVG payloads with one-time SHA-256 digests, byte size, and an optional cache key
• TXLSReportImageResult.CacheKey lets a V2 image provider explicitly identify reusable content while conflicting content for the same key is rejected before worksheet mutation

Changed

• Repeated report images share one immutable payload by explicit cache key or exact digest match while retaining independent anchors and transforms
• XLSX and ODS save each distinct image payload once, reuse its media relationship across sheets, and restore shared payloads when the workbook is reopened
• XLSX reopening now restores SVG companion data from DrawingML extension relationships and pairs it with the raster fallback

Version 2.361.64

Added

• TXLSReportImageResult can declare source pixel dimensions and DPI, while image requests and results expose anchor-relative target offsets in EMU
• TXLSXImage.OffsetXEMU and OffsetYEMU preserve picture placement inside an anchor through XLSX and ODS save/open cycles

Changed

• V2 image binding derives physical geometry from PNG, JPEG, GIF, BMP, EMF, and WMF headers, applies Natural, Contain, Cover, or Stretch sizing, and centres contained images without decoding their pixels
• Raster geometry is rejected before worksheet mutation when headers are truncated, declared dimensions disagree, a side exceeds 100,000 pixels, or total decoded size exceeds 100 million pixels

Version 2.361.63

Added

• IXLSReportImageDataProviderV2 can return image bytes or caller-borrowed and provider-transferred streams with workbook, worksheet, target, band, table, row, anchor, and cancellation context
• OnTransformImage can replace or resize an image before binding, while compiled range and Excel-table bands expose a read-only snapshot of each actual data row

Changed

• V2 image streams are read in bounded blocks with a configurable 64 MiB default limit; borrowed seekable streams restore their position and transferred streams are released exactly once
• V2 payloads reject empty data, mismatched format signatures, inconsistent dimensions, invalid anchors, and out-of-grid offsets before mutating the worksheet
• Compiled range and Excel-table bands bind image targets once per generated row from materialized bound columns without scanning generated worksheet cells

Version 2.361.62

Added

• TXLSCompiledReportRunner and TXLSReportGenerator now accept an existing IXLSReportImageProvider with missing-image and directive-cell policies
• TXLSReportImageBinder.ApplyTarget binds one precompiled image directive without scanning unrelated worksheet cells

Changed

• Exact {{#image Key}} cells compile into typed report targets, remain unchanged when no provider is configured, and reuse their compiled key and coordinates across runner sessions

Version 2.361.61

Added

• TXLSCompiledReportRunner now exposes paired workbook and worksheet generation events with stable worksheet identity, original and current coordinates, read-only report data, target counts, cancellation, and failure details
• TXLSReportGenerator runs workbook, file, stream, or encrypted-file templates through template-read and generation lifecycle events with selectable in-place or atomic-clone modification

Changed

• Report generation freezes the original worksheet set, recompiles after callback-driven structure changes, rejects concurrent mutation of the same workbook, pairs entered events during cancellation and failure, and preserves the first exception
• Report streams remain caller-owned, seekable template streams restore their position, forward-only streams are consumed in bounded blocks, and stream, file, and encrypted-file destinations change only after successful completion
• Compiled reports without lifecycle handlers retain the direct target loop without worksheet scans, event contexts, or per-target lifecycle allocations

Version 2.361.60

Added

• XLSX workbooks now expose HasWebAddinTaskPanes, WebAddinTaskPaneCount, and RemoveWebAddinTaskPanes for exact package-root task pane inspection and removal

Changed

• Web Add-in task pane removal now validates a bounded internal relationship graph before mutation, preserves shared and unrelated package parts, removes exclusive descendants and owned relationship parts, rebuilds content types, prevents source-copy resurrection, and is idempotent
• Classic XLS and ODS workbooks report Web Add-in task pane removal as unsupported without scanning package content

Version 2.361.59

Added

• Classic XLS and XLSX workbooks now expose matching SaveExternalLinkValues and UpdateLinks properties for cache persistence and file-declared link-update preference

Changed

• Disabling external-link cached values now omits only Classic XCT/CRN records or XLSX sheetDataSet content while retaining link definitions, defined names, relationships, unknown XML, and in-memory caches
• External-link policies now round-trip reserved source values losslessly until edited, convert between XLS and XLSX, preserve unchanged compressed package parts, and never initiate external file or network access

Version 2.361.58

Added

• TXLSHTMLExport and TXLSXHtmlExportSession now expose the immutable RenderGeometry snapshot used by the current HTML operation

Changed

• Classic XLS and XLSX HTML exporters now convert Excel character widths to fixed CSS lengths through backend-specific maximum-digit-width geometry instead of hard-coded conversion factors or CSS character units
• Workbook, worksheet, range, and incremental multi-sheet HTML exports capture one operation-local calibration with a bounded metric cache and never modify stored row heights or column widths

Version 2.361.57

Changed

• PDF, SVG, bitmap, print-preview, and workbook-viewer operations now capture distinct immutable render-geometry snapshots with backend-aware bounded metric caches
• Shared page rendering and SVG wrapping now use operation-local font measurement instead of process-global GDI bitmap and string-list state, while preserving stored row heights and column widths

Version 2.361.56

Changed

• Classic XLS and XLSX AutoFit now capture one immutable workbook render-geometry service per operation, including calibrated text measurement, wrapping, line spacing, cell margins, and actual XLSX visible fonts
• Pagination now captures workbook-local geometry for each build, removes process-global normal-font and maximum-digit-width state, and applies calibrated implicit column widths without modifying stored row heights or column widths

Version 2.361.55

Added

• Classic XLS and XLSX workbooks now create immutable operation-owned geometry and font-measurement services with bounded local caches, shared width conversions, and no AutoFit process-global DPI or base-width cache

Version 2.361.54

Added

• Classic XLS and XLSX workbooks now expose validated TXLSRenderCalibration values plus independent CaptureRenderCalibration snapshots and ResetRenderCalibration defaults for future rendering operations

Version 2.361.53

Added

• XLSX loading now reports xlsDiagnosticPrinterSettingsRelationshipInvalid when a worksheet printer-settings relationship is missing, mistyped, or dangling

Changed

• SaveXLSWorkbookAsXLSX and SaveXLSXWorkbookAsXLS now carry typed page-setup fields and opaque printer-driver bytes independently, with standard fields taking priority over duplicate private defaults

Fixed

• XLSX loading no longer binds an unrelated printer-settings part when the relationship named by pageSetup is invalid
• XLSX-to-BIFF8 conversion now rejects oversized or unrepresentable printer settings before writing any destination bytes

Version 2.361.52

Added

• Classic XLS and XLSX worksheets now expose copy-isolated PrinterDriverData and HasPrinterDriverData properties plus ClearPrinterDriverData for direct opaque printer-driver payload control

Fixed

• Clearing XLSX printer-driver data now removes its worksheet relationship and binary package part even when unsupported-part preservation is enabled

Version 2.361.51

Changed

• XLSX loading and saving now preserves each worksheet's binary printer settings through the pageSetup relationship, content-type declaration, and printerSettingsN.bin package part
• Printer settings survive same-workbook and cross-workbook worksheet copies with byte-exact independent ownership, while unchanged source parts can retain their original compressed payload

Version 2.361.50

Changed

• Classic XLS loading and saving now preserves the complete BIFF PLS printer-driver payload together with every immediately following Continue segment
• Printer-driver payloads are deep-copied with worksheets, emitted before Setup in exact BIFF-sized chunks, and rejected when structurally incomplete or larger than the 1 MiB safety limit

Version 2.361.49

Added

• Classic XLS page setup now exposes read-only HasSetupRecord and PrinterSettingsInitialized properties so callers can distinguish missing or uninitialized printer settings from defined values

Changed

• Saving an untouched classic worksheet now preserves a missing BIFF Setup record and retains every byte of an existing 34-byte record, including reserved and unmodeled flag bits
• Page-setup setters now create a missing Setup record only when needed and initialize printer-dependent or orientation fields according to the affected property

Version 2.361.48

Added

• SaveXLSXWorkbookAsXLS now converts an XLSX-model workbook to BIFF8 through file and stream overloads, preserving loaded values, formulas, core styles, merged ranges, row and column dimensions, workbook date mode, and active sheet state

Changed

• XLSX-to-XLS conversion builds one deterministic 56-color palette before writing styles, preserves indexed slots already in use, and applies weighted OKLab quantization when direct colors exceed BIFF8 capacity
• The conversion path traverses loaded sparse or packed cells without materializing the source grid, rejects coordinates outside BIFF8 limits before touching the destination, and is now available in the Delphi and C++Builder ConversionWorkbench examples

Version 2.361.47

Added

• Classic XLS and XLSX workbooks now expose matching ApplyBiffPalettePlan methods that validate and atomically apply a current 56-slot palette plan
• Classic workbooks now expose BiffPaletteSavePolicy, with an explicit BIFF8 optimization option and a default that preserves historical palette output

Changed

• Palette-plan application now validates source generation, source hash, all slot identities and colors, aggregate counts, preservation locks, change flags, and opaque targets before the first mutation, then rebuilds the cached OKLab matcher once
• Optimized BIFF8 saves leave the applied palette in the workbook after success but restore all 56 source colors and the original generation if output fails or is cancelled; BIFF5 and non-BIFF output paths remain unchanged

Version 2.361.46

Added

• Classic XLS and XLSX workbooks now expose matching BuildBiffPalettePlan methods that return a mutation-free 56-slot plan with source palette identity, locked slots, proposed colors, aggregated true-color mappings, exact coverage, and perceptual error statistics

Changed

• BIFF palette planning now preserves indexed consumers and existing exact colors before assigning free slots, then uses deterministic frequency-weighted OKLab clustering only when direct and resolved theme colors exceed the remaining capacity
• Palette plan construction reads stored style and conditional-format metadata without changing palette entries, custom-color counts, cells, save behavior, or sparse storage

Version 2.361.45

Added

• Classic XLS and XLSX workbooks now expose matching ScanIndexedColorUsage methods that return a deterministic 56-slot inventory with public and physical indexes, normalized ARGB values, definition and reference counts, consumer kinds, and preservation locks

Changed

• Indexed-color usage scanning now covers workbook style pools, sparse cell references, named styles, range overlays, worksheet gridlines, shapes, comments, rich text, chart fonts, differential styles, data bars, and color scales without materializing unused cells or changing workbook state
• Indexed conditional-format colors now round-trip through classic BIFF and OpenXML differential styles, data bars, and color scales, including streaming saves that emit differential-style identifiers before worksheet parts

Version 2.361.44

Added

• Classic XLS and XLSX workbooks now expose matching NearestIndexedColor methods that return the closest public palette index, physical slot, normalized ARGB value, squared perceptual distance, and exact-match status

Changed

• Nearest indexed-color matching now shares one cached OKLab implementation across both workbook engines and the legacy color manager, with deterministic lowest-index tie breaking and cache updates limited to effective palette mutations
• Nearest-color queries scan only the 56 physical BIFF palette entries, reject partially transparent inputs, and never modify palette or workbook state

Version 2.361.43

Added

• Classic XLS and XLSX workbooks now expose matching ResolveIndexedColor overloads that classify public color indexes, BIFF icv values, and OOXML indexed colors while returning palette, automatic, system, or invalid metadata
• The typed indexed-color result reports the raw index, palette slot, system-color role, normalized ARGB value, and whether a concrete color was resolved

Changed

• Automatic and system color tokens no longer query the current Windows desktop implicitly; callers can supply an explicit resolver when a concrete environment color is required
• Theme system colors now use their stored fallback RGB value, keeping load, save, and headless processing deterministic across machines

Fixed

• Invalid OOXML indexed-color values no longer resolve to opaque black, and the legacy color-manager defaults now use the exact 128-channel palette values

Version 2.361.42

Added

• TXLSWorksheet and TXLSXWorksheet now expose matching TryGetOutlineNode, SetOutlineNodeCollapsed, and SetOutlineDisplayLevel methods for one-based row or column summary nodes

Changed

• Worksheet outline queries and updates now adapt sparse engine metadata to compressed plans, preserve nested child state by default, and apply hidden and collapsed changes without scanning unused worksheet coordinates
• OpenDocument row and column groups now preserve native nested display state across direct metadata changes, structural edits, ODS round trips, and conversion to XLSX summary-node markers
• TXLSXWorksheet.ToggleRowGrouped and ToggleColGrouped remain available as direction-aware compatibility wrappers around the summary-node API

Version 2.361.41

Added

• TXLSOutlineGeometry can now build compressed TXLSOutlineMutationPlan instances for idempotent node collapse or expansion and one-based outline display-level changes
• Outline mutation plans expose explicit preserve, collapse, and expand child modes while keeping nested collapse markers independent by default

Changed

• Expanding a nested outline node now accounts for collapsed ancestors and collapsed descendants when deriving visibility, and bulk node discovery traverses only materialized spans with a linear monotonic-stack pass

Version 2.361.40

Added

• The new engine-neutral lxOutline unit builds row or column outline geometry from sorted sparse spans and discovers summary-before or summary-after nodes, nested detail ranges, summary levels, hidden state, and explicit collapse markers without scanning unused worksheet coordinates

Version 2.361.39

Added

• Classic XLS column metadata now exposes the independent outline collapse marker through TXLSColumnInfo.Collapsed and TXLSColumnInfoList.Collapsed

Fixed

• Classic XLS column outline collapse markers now survive BIFF round trips, column-format copies, worksheet copies, and adjacent-record coalescing while defined width, best-fit, and phonetic flags remain intact

Version 2.361.38

Fixed

• Classic XLS outline summary columns now preserve their logical left or right position across left-to-right and right-to-left worksheet views instead of writing the opposite WSBOOL direction bit

Version 2.361.37

Added

• Classic XLS and XLSX worksheets now expose EvaluateFormulaAt for read-only A1 or R1C1 formula evaluation in an explicit target-cell context
• IXLSFormulaEvaluationTemplate compiles an immutable formula anchor once and evaluates it repeatedly at other cells through EvaluateAt or EvaluateAtWithOptions
• Formula evaluation options now include limits for formula length, syntax-item steps, recursion depth, and generated array bytes, with exhaustion reported as xlfeResourceLimit

Changed

• One-shot and reusable evaluation now share worksheet-local names, structured references, uncached formula dependencies, array roots, volatile status, cancellation, and explicitly authorized external references without writing workbook state
• The existing workbook and worksheet Calculate methods remain compatibility facades over the typed contextual evaluator

Fixed

• Contextual evaluation now preserves formula error variants and distinguishes parse, unsupported-function, circular-reference, external-reference, cancellation, worksheet-lifetime, and resource-limit failures
• Reusable R1C1 templates now translate relative row and column offsets from their compile anchor when evaluated at another target cell

Version 2.361.36

Added

• ConvertExternalDefinedNamesToRefErrors is now available on Classic XLS and XLSX workbooks to detach confirmed external-workbook defined names without rewriting ordinary cell formulas
• Skipped malformed or unsupported definitions are reported through xlsDiagnosticDefinedNameConversionSkipped

Changed

• The conversion replaces selected definitions atomically with native #REF! while retaining workbook or worksheet scope, visibility, comments, macro and built-in flags, unknown attributes, and external-link parts

Fixed

• External defined-name slot syntax such as =[1]!RemoteName now resolves through the declared workbook link, and dependent formulas retain the resulting Excel error value during recalculation

Version 2.361.35

Added

• TXLSWorkbookWorkspace.Recalculate now evaluates registered Classic XLS, XLSX, and ODS formula nodes in one cross-workbook topological pass with strongly connected component detection and dirty-generation propagation
• TXLSWorkspaceRecalcResult reports graph reuse, dirty and evaluated nodes, cycles, blocked dependents, invalidations, and runtime cache hits and misses

Changed

• The workspace retains its dependency graph until registrations or formula topology change, while ordinary value edits dirty only the affected workbook nodes and their transitive dependents
• Repeated external-cell reads in one calculation pass share an exact workbook, worksheet, row, and column cache that is invalidated before each scheduled formula update

Fixed

• Bidirectional links between workbooks are no longer treated as circular unless the formula-cell graph contains a real cycle; cycle members and their blocked dependents are invalidated instead of returning stale cached values
• A failed or cycle-invalidated workspace session now retains the engine-local dirty state so a later workbook recalculation retries unfinished formulas instead of treating a partial pass as complete

Version 2.361.34

Added

• TXLSWorkbookWorkspace.BuildDependencyGraph now creates one dependency graph across registered Classic XLS, XLSX, and ODS workbooks, retaining workbook, worksheet, cell, rectangular range, and scoped defined-name identities
• TXLSWorkspaceDepGraph exposes formula-node, reference, edge, cell-dependent, and defined-name-dependent queries plus candidate-check diagnostics for performance verification

Changed

• Large formula ranges remain compact intervals instead of expanding into individual cells, while row-interval trees prune both dependent lookup and edge construction without materializing packed worksheet values
• Local defined names retain both their symbolic dependency and statically resolved cell or range dependencies, while external defined names preserve their declared external-workbook slot, name, and worksheet scope

Fixed

• Ordinary defined names are no longer cleared while the formula parser checks for an optional external-workbook prefix

Version 2.361.33

Added

• TXLSWorkbookWorkspace.OnLoadWorkbook can now request an external workbook through caller-controlled policy at most once per normalized identity, while ResolveWithLoader reports typed not-found, error, cycle, and resource-limit diagnostics
• MaxLoadDepth, MaxWorkbookCount, and ResetLoadAttempts provide explicit bounds and retry control for external workbook loading

Changed

• Concurrent external-workbook requests now share one in-flight load result, and nested re-entry is rejected without waiting in a cross-thread load cycle; the default workspace still performs no file-system or network access

Version 2.361.32

Added

• Classic XLS workbooks now expose CreateWorkspaceWorkbook, RegisterExternalWorkbook, UnregisterExternalWorkbook, and ExternalWorkspace, allowing Classic, XLSX, and ODS workbooks to serve external formula values through one engine-neutral workspace
• Classic XLS external-link caches now decode number, string, Boolean, error, and blank values from typed XCT and CRN records into a sparse constant-time lookup without materializing worksheets

Changed

• External formula evaluation now resolves worksheet names before positional fallbacks and consistently reads an explicitly registered live workbook first, then a typed file cache, then returns #REF!, while destroyed registrations are revoked without implicit file or network access
• Quoted external workbook identities can contain URI schemes, Windows drive letters, and other colons without being mistaken for a three-dimensional worksheet range separator, so ODF external-source IRIs remain intact during formula import

Version 2.361.31

Added

• TXLSWorkbookWorkspace, IXLSWorkspaceWorkbook, and TXLSXWorkbook.CreateWorkspaceWorkbook provide a thread-safe external-workbook identity and registration foundation with explicit resolution status, generation tracking, and revocable workbook adapters

Changed

• TXLSXWorkbook.RegisterExternalWorkbook now preserves paths and extensions, resolves relative relationship targets against the owning workbook, rejects ambiguous aliases or same-basename targets instead of silently choosing one, revokes registrations when a target workbook is destroyed, and retains typed external-link cache fallback without implicit file or network access

Version 2.361.30

Added

• Classic and XLSX read-only workbook views can now evaluate formulas on demand through xlsvEvaluateFormulas and report the independent result through FormulaEvaluation without replacing stored formula caches

Changed

• Concurrent read-only views now isolate formula recursion flags, scalar results, compiled XLSX formulas, and array matrices, while XLSX external formulas use only file-local external-link caches and never dereference live-workbook registrations

Fixed

• XLSX external-link cached values now resolve the one-based worksheet index carried by external XTI records to the correct zero-based cached sheet entry

Version 2.361.29

Added

• XLSX workbooks now expose IXLSXWorkbookOwner.CreateReadOnlyView with view-local active sheets, by-value materialized and packed cell snapshots, cached formula results, opt-in formula text, safe owner or view release order, and fail-fast write exclusion

Changed

• XLSX read-only views use only locally stored external-formula caches and never dereference weak live-workbook registrations

Version 2.361.28

Added

• Classic workbooks now expose CreateReadOnlyView and IXLSReadOnlyWorkbookView for constant-time shared-model views with independent active sheets, by-value cell snapshots, cached formula results, opt-in formula text, reference-counted owner lifetime, and fail-fast write exclusion

Version 2.361.27

Added

• Classic, XLSX, and ODS workbooks now expose CreateFormulaCacheReader and the shared IXLSFormulaCacheReader interface for engine-neutral formula-cache inspection with safe lifetime invalidation

Changed

• Shared-formula cache inspection now preserves each member's independent stored result, so a follower without a cached value reports missing instead of inheriting the root member's value

Version 2.361.26

Added

• TXLSXWorkbook.TryGetCachedFormulaValue now reports XLSX and ODS formula-cache presence, provenance, and value kind without calculation or workbook mutation

Fixed

• XLSX and ODS formula caches now preserve missing and explicit blank results, empty strings, Boolean false, numeric zero, typed errors, and explicit dates across save and reopen, while changed formulas no longer serialize stale cached results

Version 2.361.25

Added

• Classic workbooks now expose TryGetCachedFormulaValue with explicit cache presence, provenance, and value kind, allowing a stored formula result to be read without calculation, user-function callbacks, token decompilation, or workbook mutation

Fixed

• Classic XLS formula Boolean and error caches now use the specified BIFF payload byte in ordinary and shared formulas, and loaded error results remain typed varError values across save and reopen

Version 2.361.24

Added

• Classic and XLSX workbooks now provide a CreateLoadedCellCursor(Boolean) overload that can include independently styled stored blanks as xlckBlank snapshots

Changed

• Opt-in blank traversal preserves worksheet, row, and column order, keeps packed XLSX storage compressed, and excludes inherited-format coordinates plus unstyled Empty and Null placeholders

Version 2.361.23

Added

• XLSX workbooks now expose TXLSXWorkbook.CreateLoadedCellCursor for constant-state row-major traversal of stored value and formula cells across worksheets
• TXLSLoadedCellKind now includes xlckError so stored XLSX error values retain their distinct value kind

Changed

• XLSX cursors traverse materialized sparse row blocks or packed storage directly without sorting cells or materializing packed values, and formula snapshots return the currently stored cached value without recalculation
• Workbook-wide topology changes and workbook release invalidate XLSX cursors through a shared lifetime lease, while edits to existing cell content keep traversal valid and captured snapshots owned

Version 2.361.22

Added

• Classic workbooks now expose IXLSWorkbook.CreateLoadedCellCursor for constant-state row-major traversal of stored value and formula cells across worksheets
• TXLSLoadedCellSnapshot returns owned sheet identity, one-based coordinates, style, value kind, value, and formula text without exposing internal cell objects

Changed

• Loaded-cell cursors fail fast after cell topology, worksheet collection, or workbook model replacement changes, while edits to existing cell content keep traversal valid and previously captured snapshots unchanged

Version 2.361.21

Added

• Classic and XLSX worksheet cell stores and worksheet collections now expose independent structural generation counters for safe long-running traversal
• XlsValidateStructureGeneration and EXLSLoadedCellCursorInvalidated provide one fail-fast validation contract for loaded-cell readers

Changed

• Cell creation, removal, clearing, coordinate reordering, packed-storage transitions, index rebuilds, and worksheet add, delete, or move operations advance the relevant generation, while existing-cell content edits, worksheet rename or activation, and no-op mutations keep it stable

Version 2.361.20

Added

• TXLSForwardReader now accepts ODS input through the same callback surface as XLS, XLSX, CSV, and TSV
• PeakRowBufferedBytes and SheetPassesStarted expose the shared row backend's peak logical-row working set and successful worksheet pass count

Changed

• TXLSForwardReader now adapts TXLSRowCursor for every supported format, sharing row bounds, column projection, formula state, rich text, password, skipped-record, and stream-position behavior instead of maintaining separate scan paths
• Forward-reader sheet indexes are consistently one-based, and worksheets skipped by OnSheet are not opened or decompressed

Version 2.361.19

Added

• TXLSRowCursor now exposes explicit closed, before-first, active, EOF, cancelled, and faulted states plus an idempotent Cancel method that releases the active worksheet stream without closing the source
• CurrentRowView provides a generation-checked borrowed view of the current row without duplicating its cell array; advancing, selecting a sheet, cancelling, closing, or destroying the cursor invalidates prior views with EXLSRowCursorViewInvalidated
• Stream opens now accept xsoBorrowed or xsoOwned, transferring ownership only after a successful open, and PeakRowBufferedBytes reports the estimated peak logical-row buffer since the last open

Changed

• Worksheet-part and pass-start failures now preserve their original read exception and place the cursor in the faulted state instead of appearing as cancellation or normal EOF; explicitly selecting a sheet resets the pass for retry

Version 2.361.18

Added

• TXLSDirectCell and TXLSRowCursorCell now expose HasFormula and FormulaTextAvailable, distinguishing plain values, formulas with available text, and formulas whose text is suppressed or unavailable
• TXLSDirectReader, TXLSForwardReader, and TXLSRowCursor now provide IncludeFormulaText; disabling it preserves cached values, coordinates, styles, and formula presence while avoiding formula-text accumulation in XLSX and ODS streams

Changed

• Row bounds and column projection now push into XLS, XLSX, ODS, CSV, and TSV backends before value, formula, and rich-text allocation; repeated ODS column spans jump directly between selected columns and repeated row blocks are clipped to the requested range
• Row-cursor pass options are frozen while a worksheet pass is active and become mutable again after EOF, sheet selection, or close, preventing a pass from mixing incompatible filter or decoding settings

Version 2.361.17

Added

• TXLSRowCursor now opens ODS files and caller-owned streams, auto-detects ODS packages by their mimetype or content.xml, preloads worksheet names, and streams the selected table through FindFirst and FindNext without materializing a workbook

Changed

• ODS row traversal expands repeated rows and columns lazily, omits covered cells as independent values, preserves styled blanks and cached formula values, restores borrowed stream positions, and reports invalid or grid-limited repeat counts through OnSkippedRecord

Version 2.361.16

Added

• TXLSRowCursor now streams CSV and TSV files or caller-owned streams with UTF-8 and UTF-16 BOM detection, multiline quoted records, escaped quotes, comma, semicolon, tab, or pipe delimiter sniffing, optional delimiter override, and optional number, ISO date, Boolean, and error type recognition

Changed

• TXLSForwardReader and TXLSRowCursor now share one pausable text-row backend, preserving identical callback and pull results while releasing Variant, formula, and rich-text payloads from the previous row before advancing

Version 2.361.15

Added

• TXLSRowCursor now opens classic .xls workbooks from files or streams, auto-detects BIFF compound containers, and returns cached numeric or string formula results plus stored Blank and MulBlank cells while retaining only the current logical row

Changed

• Row-cursor Open overloads accept an optional TXLSForwardFormat; file opens own their internal BIFF stream, stream opens borrow the caller's stream and restore its initial position on Close, and repeated FindFirst passes seek directly to the selected worksheet

Version 2.361.14

Added

• Formula calculation now supports ISFORMULA for individual cells, cross-sheet references, and rectangular ranges, returning a same-shaped Boolean array for range inputs without evaluating the referenced formulas

Changed

• Formula-presence checks distinguish ordinary values and blanks, shared-formula members, classic and XLSX array-formula members, dynamic-array anchors, and dynamic spill children; generated arrays are bounded by FormulaArrayMemoryLimit
• XLSX stores ISFORMULA with the _xlfn. prefix, ODS emits the standard OpenFormula spelling, and Classic XLS saves reject the unsupported future function without committing a partial stream

Fixed

• Formula text parsing now recognizes the seven standard Excel error literals, allowing #REF! to propagate through reference-only functions, while a defined name that resolves to a scalar reports #VALUE! instead of #NAME?

Version 2.361.13

Added

• Formula calculation now supports UNICHAR across the Unicode scalar range, including supplementary and private-use characters, with Excel-compatible errors for zero, out-of-range values, surrogate code points, and Unicode noncharacters
• Formula calculation now supports BAHTTEXT with Thai baht and satang text, two-decimal midpoint rounding away from zero, negative values, repeated million groups, and finite scientific-notation inputs

Changed

• UNICHAR and BAHTTEXT map ranges and generated arrays element by element while preserving local errors; XLSX stores UNICHAR with its future-function prefix while BAHTTEXT remains a standard bare function in both XLS and XLSX

Version 2.361.12

Added

• Formula calculation now supports COMBINA and PERMUTATIONA with truncated non-negative arguments, checked overflow, singleton array broadcasting, and per-element errors, plus MUNIT identity matrices without a fixed dimension ceiling
• Classic and XLSX workbooks expose FormulaArrayMemoryLimit with a 64 MiB default to bound generated formula-array payloads, including parallel XLSX recalculation workers

Changed

• PERMUTATIONA uses logarithmic exponentiation, COMBINA shortens its multiplicative recurrence, and MUNIT allocates a compact numeric matrix while writing only its diagonal

Fixed

• Numeric formula arrays are no longer classified as scalar numbers, and newly created classic workbooks apply BIFF97 grid limits during formula calculation before their first save

Version 2.361.11

Added

• Formula calculation now supports COT, ACOT, COTH, ACOTH, CSC, CSCH, SEC, and SECH with principal-value, domain, singularity, finite-input, and large-argument rules across scalar and broadcast-array operands

Fixed

• Mapped formula functions now preserve direct scalar failures as spreadsheet error values, use cancellation-resistant reciprocal and exponential-tail branches for large valid inputs, write exact _xlfn. names without _xlws., and reject unsupported Classic BIFF formulas cleanly

Version 2.361.10

Added

• Formula calculation now supports CEILING.MATH and FLOOR.MATH with omitted significance and mode arguments, sign-independent significance, precise decimal-multiple boundaries, and the specified negative-number mode directions

Changed

• Formula-level array functions can now share a linear element mapper with singleton-row and singleton-column broadcasting, one-time range snapshots, and per-element conversion or calculation errors; XLSX uses the _xlfn. prefix for both new functions while Classic BIFF saves reject unsupported formulas cleanly

Version 2.361.9

Added

• Formula calculation now supports SORTBY with stable multi-key row or column ordering, independent ascending or descending directions, omitted-direction defaults, range or generated-array keys, and one-time source and key snapshots followed by O(n log n) sorting

Fixed

• Classic BIFF saves now fail cleanly when a formula cannot be represented instead of reporting success after an internal future-function identifier is rejected, while XLSX writes SORTBY with _xlfn. and never adds _xlws.

Version 2.361.8

Changed

• Worksheet zero-display mode now uses one raw-value decision across the workbook viewer, preview, print, bitmap, EMF, SVG, PDF, Classic HTML, and XLSX HTML paths; actual numeric zero is hidden before number formatting and text measurement, while formula-display mode retains precedence

Fixed

• XLSX worksheet views now parse, write, and copy zero-display state through one setting, correctly handle the 1904 date base, preserve conditional-format visuals and hyperlink geometry for hidden zero cells, and leave stored values, public formatted text, CSV, and ODS output unchanged

Version 2.361.7

Changed

• Worksheet formula-display mode now uses one protected-cell-aware text decision across the workbook viewer, preview, print, bitmap, EMF, SVG, PDF, Classic HTML, and XLSX HTML paths; formula text uses text alignment under General alignment, preserves explicit alignment, and receives a non-persistent twofold effective layout width without overflowing into neighbouring cells

Fixed

• XLSX worksheet views now accept both showFormulas="1" and showFormulas="true", preserve the setting during same-workbook and cross-workbook worksheet copies, and emit the required sheetView state without changing formula caches, CSV output, ODS persistence, or the public formatted-value contract

Version 2.361.6

Changed

• XLSX external-workbook formulas now share one grammar-aware book-prefix lexer across OnlyExternal conversion, formula diagnostics, calculation binding, cross-workbook copying, and generated formula serialization; numeric link slots are canonical, while structured references, string literals, malformed indexes, and ambiguous filename prefixes fail closed
• TXLSXExternalLink now exposes its source RelationshipId, one-based SourceSlot, and typed LinkKind so workbook, DDE, OLE, and unresolved external-reference entries can be audited without inferring identity from collection order or target text

Fixed

• Changing an XLSX external-link target or sheet list now invalidates compiled external-formula bindings by revision instead of rescanning every cell, and unchanged DDE or OLE external-link parts without relationship files remain byte-preserved during save

Version 2.361.5

Fixed

• XLSX external-workbook formula slots now follow the document order of workbook.xml external references rather than relationship-file order; missing, unresolved, and wrong-type relationships retain empty slots so later [N] references do not shift

Version 2.361.4

Changed

• Classic XLS formula freezing with OnlyExternal now classifies preserved BIFF references and converts only confirmed external-workbook formulas, while internal, add-in, DDE/OLE, and malformed references remain formulas

Fixed

• Formula results beginning with an equals sign or apostrophe now remain literal strings during value conversion, and array-formula groups are snapshotted before any root or spill formula is released

Version 2.361.3

Changed

• Classic XLS supporting links and external-sheet indexes now use bounded typed classification for self, same-sheet, add-in, external-workbook, DDE/OLE, unused, and unknown records

Fixed

• Malformed supporting-link payloads are preserved byte-for-byte instead of being reconstructed as a different link type, and truncated external-sheet tables no longer read beyond their record data

Version 2.361.2

Fixed

• Classic XLS formulas now resolve the zero-based XTI index in external defined-name tokens to the correct internal external-sheet entry, including the first entry

Version 2.361.1

Fixed

• The direct XLSX writer now emits the base SpreadsheetML showDropDown suppression flag only when ShowDropDown is disabled, preserving the public display-oriented API semantics

Version 2.361.0

Added

• Chart series gained CategoriesMultiLevel: set it to emit the categories as a multiLvlStrRef (hierarchical axis labels) instead of a flat string reference, with round-trip recovery from saved charts

Version 2.360.0

Fixed

• Classic XLS name comments now use the correct NAMECMT record identifier ($0894); the identifier previously used ($0019) belongs to WINDOWPROTECT per the BIFF specifications

Added

• WindowProtected on classic workbooks round-trips the workbook window-protection flag through the WINDOWPROTECT record ($0019), with raw captured bodies winning unless the property is explicitly written

Version 2.359.0

Added

• Chart data tables expose independent showHorzBorder / showVertBorder / showOutline switches instead of always drawing every border
• Line charts gained drop lines (HasDropLines) and stacked bar charts gained series lines (HasSeriesLines), both round-tripped; series can be removed from the legend individually through LegendDeleted while staying plotted

Version 2.358.0

Added

• Report bands gained fixed and empty-dataset semantics: EmptyBandPolicy keeps, clears or deletes the template rows when the data set is empty (no more unresolved placeholders left behind), FixedBand writes the copies over the rows below without inserting anything for overlay-style form bands, and DeleteLastRow collapses the final copy so trailing content joins flush (X-range behaviour)

Version 2.357.0

Added

• PDF export can embed standalone OpenType fonts with CFF outlines through a FontFile3 OpenType stream and CIDFontType0 descendant; automatic and full modes retain the complete program, while an unavailable required subset fails before output is committed
• OnGetFontDirectories supplies semicolon-separated directories for a requested font family and style; accepted trees are searched recursively once and indexed by internal font metadata for cached fallback resolution

Changed

• Embedded fonts carrying OpenType GPOS or legacy kern data now use positioned glyph advances for simple text as well as complex scripts, keeping measurement, wrapping, alignment, and PDF glyph placement consistent

Version 2.356.0

Added

• PDF font embedding became a configurable four-mode policy: FontEmbedding selects automatic historical behaviour, required subsetting, full-program embedding, or map-only fonts whose programs are stripped while the CID mapping and ToUnicode tables survive; PDF/A rejects map-only output
• The OnFontEmbed event fires once per Unicode font before metrics and glyph mappings are resolved, receives its name and file path, and can accept a per-font mode or replacement font bytes through the explicit Handled contract

Version 2.355.0

Added

• Classic XLS recalculation now uses a lazily built dependency graph, propagates input changes only to transitive dependants, recalculates volatile formulas on every pass, and evaluates formulas in dependency order instead of rescanning the entire sparse cell heap
• Circular references in classic XLS now follow EnableIteration, MaxIterations, and MaxIterationChange; strongly connected components isolate independent cycles, iterate them deterministically, and then calculate downstream formulas
• PDF/A-3 associated files gained a detailed AddAttachment overload with Unicode display names, descriptions, modification times, escaped MIME names, byte-size metadata, and Source, Data, Alternative, Supplement, or Unspecified relationships
• TXLSPDFWriter creates standalone PDF documents without a workbook, with explicit page sizes, multi-page output, rectangles, lines, ellipses, measured styled text, coordinate transforms, rectangular clipping, and graphics-state control

Changed

• Disabling UseFullPrecision on classic XLS now rounds formula and What-If data-table cached results to the effective displayed numeric precision before dependant formulas consume them, including percentage formats and half-away-from-zero midpoint rounding

Fixed

• Classic formula recalculation now uses the native zero-based cell coordinates, keeps calculation failures out of the success cache, and preserves error propagation through dependent array functions

Version 2.354.0

Added

• Chart axes gained the crossing-mode enumeration (Crosses: autoZero / max / min, emitted as crosses when no explicit CrossesAt is set) and per-axis major / minor tick-mark styles (in, out, cross), all round-tripped
• AutoTitleDeleted on charts replaces the hardcoded automatic-title flag and suppresses the series-derived title when set

Version 2.353.0

Added

• Chart series gained Smooth/SmoothSet for curved line rendering and MarkerFillColor/MarkerBorderColor for explicit marker colours; both persist through save and open, and the shared renderer draws smoothed segments as quadratic curves and consumes the marker colours

Version 2.352.0

Added

• Chart space formatting round-trips: RoundedCorners, explicit chart background and border colours, and the plot-area background colour persist through save and open, and the shared chart renderer consumes the background and border colours instead of always drawing white on grey
• Chart protection switches round-trip behind a single master: ProtectionEnabled plus ProtectedChartObjects, ProtectedDataEditing, ProtectedFormatting, ProtectedSelection and ProtectedUserInterface emit and parse the OOXML protection element

Version 2.351.0

Changed

• Bar and column charts in the shared pagination renderer now follow the plot group's layout options: GapWidth sizes the cluster, Overlap widens and pulls neighbouring bars together, and stacked / percent-stacked groupings accumulate one full-cluster bar per category instead of drawing series side by side

Version 2.350.0

Added

• Cross-workbook range copy now remaps every cell style by value into the destination workbook's font, fill, border, number-format, alignment and protection pools, so CopyRangeTo and CloneRangeTo between two open workbooks keep their formatting independent of pool order
• Classic XLS workbooks round-trip the numeric and date built-in document properties: CreatedDate, LastSavedDate, RevisionNumber and EditTimeMinutes persist through the OLE SummaryInformation stream, with zero values skipped and a full clear removing them

Version 2.349.0

Added

• Classic XLS workbooks expose the shared-workbook revision protection state behind structured properties: SharedRevisionProtected and SharedRevisionPasswordHash round-trip, while raw record bodies captured from parsed files still win unless the properties are explicitly written
• Report encryption awareness: TXLSCompiledReportTemplate.CreateEncrypted opens a password-protected template directly, and RunToEncryptedFile runs the report and writes an ECMA-376 encrypted result in one call

Version 2.348.0

Added

• Clipboard paste options for both engines: PasteFromClipboard accepts values-only (keep cached results, skip formulas, formats and merges) and insert-rows / insert-columns modes that shift existing content before the block lands
• Forward readers report skipped content: the direct reader, row cursor, and forward reader expose OnSkippedRecord, fired when a malformed cell reference is skipped or a BIFF sheet scan aborts on truncated or mismatched records

Version 2.347.0

Added

• Report table shaping combinators: Filter streams matching rows lazily, Sort performs a stable multi-key sort, and Distinct keeps the first row per key tuple
• Data source adapters: FromDataSet snapshots an open TDataSet into a report table (field names become columns, the dataset position is restored), and FromObjects maps published properties of arbitrary object lists and arrays to columns
• Bidirectional report bands: RunBidirectional lets each master row drive its own columns from a related child table, producing jagged column layouts where shorter rows leave the extra clones blank
• PreprocessAndRun runs a template twice: the first pass fills values and executes structural directives so the template adapts to the data, the second pass recompiles the reshaped workbook and fills the remaining placeholders

Version 2.346.0

Added

• Report placeholders accept a default value: {{name;fallback}} writes the fallback text when the key is missing or empty
• Report expressions gained dbvalue(table, row, field[, default]) and dbcount(table) to read any registered data-context table by row and column, plus tonumber(text) for string-to-number conversion and ref(rowOffset, colOffset) relative references that follow band copies; templates, runners, and bands expose a DataContext property
• A bare {{=formula(...)}} cell now writes a live formula into the cell instead of plain text, and other bare expressions keep their evaluated type (numbers stay numeric)
• New structural report directives: {{#imgsize zoom=150}} rescales the image anchored at the directive cell, {{#comment text}} attaches a cell comment, {{#html <b>bold</b>}} feeds a fragment into rich text, and {{#sheetvisible Sheet false}} / {{#deletesheet Sheet}} control worksheet visibility and removal

Version 2.345.0

Added

• XLSX workbooks gained a save-time recalculation policy: RecalcBeforeSave selects manual (default), recalculate-if-dirty, or always-recalculate behavior when a file is saved
• RecalcNeeded reports whether any formula cell or captured edit is currently dirty, and RecalcCell recalculates the workbook and returns the freshly computed value of a single cell by sheet, row and column

Version 2.344.0

Added

• Report templates support structural directives in cell text: {{#delete A2}} clears the named range, {{#format fmt=0.00 bold fill=FFFF00}} applies number formats, bold, italic, font and fill colors to the column above the directive cell, and {{#aggregate sum A2:A99}} writes =SUM(A2:A99) style formulas — all executed after value placeholders fill, with directive cells consumed afterwards and unknown directive verbs still reported as template diagnostics

Version 2.343.0

Added

• The ZIP writer produces ZIP64 archives: entries at or beyond the 32-bit size limits reserve a ZIP64 extra field in their local header with sentinel classic fields, central directory records emit 64-bit sizes and offsets with a version-needed bump, and the end-of-central-directory record is preceded by the ZIP64 EOCD record and locator whenever any classic field saturates
• Write-side entry and archive size ceilings rise to 16 GB and 64 GB respectively; read-side decompression guards are unchanged

Version 2.342.0

Added

• Embedded pictures carry accessibility metadata: AltText and AltTextTitle on every image serialize as the drawing descr and title attributes and load back into the typed model for workbooks that do not replay the raw anchor XML

Changed

• Slicer and timeline parts already round-trip losslessly through the opaque package layer (verified end to end this release); the round-trip contract is now pinned by a regression test covering open and re-save of slicer-bearing packages

Version 2.341.0

Added

• Report runners can autofit the finished output: AutofitMode selects row, column, or both passes over every sheet after all targets applied and before serialization, and AutofitAdjustment adds optional headroom above the measured sizes
• Wrapped rows size from the measured line count against the column width, so multi-line cell text produced by templates grows the row the way Excel's own autofit would

Version 2.340.0

Added

• PDF exports can set the initial viewer mode — bookmarks, thumbnails, full screen, optional content, or attachments — through the catalog /PageMode entry
• PDF exports can define an open action that jumps to a chosen page at an explicit zoom factor or opens the bookmark panel when the document loads
• XMP metadata now carries DocumentID and InstanceID identifiers, an xmpMM:History event, modify and metadata dates, and the dc:format declaration expected by archival validators

Version 2.339.0

Added

• A physical print pipeline: the new TXLSPrintDocument prints worksheets, workbooks, or all visible sheets through the shared pagination model with page windows, copy collation, a document title, and before-new-page / after-generate-page events
• The workbook viewer preview gains zoom modes — fit page, fit width, or an explicit 25% to 400% factor — with Ctrl+wheel zoom, wheel page turning, and horizontal panning when the zoomed page exceeds the window

Version 2.338.0

Added

• Chart rendering draws the value axis: tick marks and right-aligned numeric labels along the left edge, major gridlines when enabled, display-unit scaling, reversed orientation, and logarithmic clamping
• Series assigned to the secondary value axis map against their own scale, so combo charts keep independent primary and secondary magnitudes
• Bubble charts render area-scaled bubbles derived from the bubble size column instead of fixed marker dots
• In-memory charts resolve their value, category, and bubble ranges against the owning sheet when caches are empty, so freshly created charts render identically to loaded ones
• The chart model gains editable doughnut hole size, first slice angle, radar style, and of-pie type, split position, split mode, secondary plot size, and gap width, all persisted through save and load

Fixed

• Category charts whose points carry values but no numeric categories no longer abort rendering with a placeholder, which previously suppressed tick marks, gridlines, and series output for every column, bar, and line chart

Version 2.337.0

Added

• Paginated output can render cell comments as an at-end listing through the new PrintComments mode on the SVG exporter and pagination model: cell address, author, and text appear below the sheet's last data row
• Named drawing objects export as PNG images at the configured DPI and scale through ExportObjectAsPNG, with a measure-only GetObjectSize counterpart returning the object's point dimensions without rendering
• Text-shape anchors outside the used range now extend the printed surface like charts and pictures already did

Changed

• The pagination renderer caches resolved cell fonts per pass and memoizes repeated text measurements, removing redundant font-pool reads and GDI measure calls from shrink-to-fit loops and word wrap

Version 2.336.0

Added

• Cell text rotation now renders in page image, SVG, print, and preview output: 1-90 degree counter-clockwise and clockwise angles rotate around the cell center and the 255 stacked form emits one letter per line
• Subscript and superscript cells render with a shifted baseline and reduced glyph size instead of losing their vertical alignment
• Header and footer sections honor their &nn point-size and &K color format codes in every pagination-driven output
• Column widths in paginated output derive from the measured maximum digit width of the workbook's Normal font, so workbooks based on fonts other than Calibri 11 paginate with Excel-accurate column boundaries
• Worksheets with a custom classic grid color render their grid lines in that palette entry instead of the fixed gray default

Version 2.335.0

Added

• About seventy classic worksheet functions now evaluate in addition to being parsed: the math and trigonometry group ATAN2 ASIN ACOS SINH COSH TANH ASINH ACOSH ATANH EVEN ODD, the date and time group TIME HOUR MINUTE SECOND DATEVALUE TIMEVALUE DAYS360, the finance group MIRR IRR IPMT PPMT DB, TRANSPOSE TYPE N ADDRESS ROMAN HYPERLINK MAXA MINA, and the pre-2010 spellings of the distribution and statistics family such as NORMSDIST TDIST CHIDIST FDIST POISSON QUARTILE PERCENTILE PERCENTRANK COVAR CHITEST GAMMALN that forward to the dotted implementations sharing their argument lists

Fixed

• Legacy one-argument normal CDF forms such as NORMSDIST computed an incorrectly scaled error-function approximation
• MIRR compounded positive cash flows with an off-by-one period exponent, shifting results away from Excel's values

Version 2.334.0

Added

• Report band placeholders that consist of a single value field now write the raw data type into the target cell: numbers stay numeric, dates stay date cells and receive the default date format when the cell is General, and booleans keep their cell type, so downstream SUM, sorting, and chart references work on generated reports
• Custom table styles gained the full section set including pivot subtotal, subheading, blank-row, and page-field sections, per-section font name, size, italic, stripe size, and four-edge borders through SetSectionEx, plus pivot/table flags, DefaultTableStyleName / DefaultPivotStyleName workbook properties, and full load-side round-trip of styles, sections, and dxf attributes
• Opening an encrypted XLSX workbook without a password now fires TXLSXWorkbook.OnPassword with the same retry contract as the classic engine, enabling GUI password prompts without knowing the file is encrypted in advance
• Saving classic BIFF5 workbooks with a password now applies XOR obfuscation and BIFF8 workbooks can select standard RC4 through EncryptionType, completing the classic encryption matrix beside the existing RC4 CryptoAPI default
• Worksheet text rendering in page image, SVG, print, and preview output now honors vertical alignment and WrapText word wrap with greedy line breaking, instead of clipping wrapped cells to a single centered line

Fixed

• Error cells now display their Excel error literal such as #DIV/0! or #REF! in every rendering path instead of an undefined empty text
• Custom table-style sections are written in the schema tableStyleElement form so Excel recognizes the style, and a section font color no longer discards its bold setting
• Report band snapshots keep every bound column value when a band grows past its initial capacity

Version 2.333.0

Added

• TXLSPDFSignature.CertificationPolicy can create a DocMDP certification signature that permits no changes, form filling and signing, or form filling, signing, and annotations; certified catalogs reference the signature through /Perms, ordinary certified output uses PDF 1.5, and incompatible PDF/A-1 certification is rejected before the destination stream is changed

Version 2.332.0

Changed

• Automatic tagged PDF now emits a complete table structure with rows, semantic header and data cells, dense per-page marked-content identifiers, exact parent-tree mappings, merged-cell row and column spans, and artifact isolation for page decoration, charts, images, headers, and footers

Fixed

• Reusing one PDF exporter now clears binary object slots between documents, preventing an ICC profile, image, or other stream from an earlier export from replacing a newly allocated object when object numbering changes

Version 2.331.0

Added

• A unified XLSX drawing facade exposes mixed image, chart, text-box, preserved-anchor, and group order, with forward, backward, front, back, transform-aware group, and save-reopen-ungroup operations that retain drawing relationships
• Named styles support automatic or explicit linked format parts while preserving local overrides, and all 60 built-in light, medium, and dark table styles resolve through the shared HTML, PDF, image, preview, and viewer formatting paths
• TXLSPDFBookmark supports hierarchical Unicode bookmarks, sibling and child relationships, colors, bold or italic titles, and expanded or collapsed descendants
• TXLSPageImageExporter.ExportAllAsTiff writes selected pages to a stream or file as one LZW-compressed multipage TIFF with progress, cancellation, and atomic file replacement
• Classic XLS opens can stop after global metadata or seek directly to selected worksheets by validated BOUNDSHEET offsets, including encrypted-stream repositioning
• TXLSFormatSnapshot captures a complete classic cell format independently of its source workbook and IXLSRange.ApplyFormat applies it to cell, row, or column ranges
• Compatible row-major typed XLSX bulk writes use packed RLE storage, with public packed-cell and byte statistics plus transparent one-time materialization for stable object handles

Changed

• Style-only blank XLSX cells share reference-counted payloads with copy-on-write local edits, automatic compaction across bulk formatting, named-style updates, copy, sort, format-patch, save, and load paths, and measurable avoided payload bytes
• Dynamic arrays stream directly into scalar aggregates, repeated exact lookups reuse typed incremental range caches, and immutable stateless formula tokens are shared across compilations
• Elliptical arcs use flatness-driven adaptive cubic Bezier subdivision shared by PDF, SVG, and bitmap output, increasing precision only where scale or eccentricity requires it
• XLSX package output now stabilizes shared strings, hyperlinks, relationships, and part ordering so reverse construction produces byte-identical deterministic packages
• The unified XLS, XLSX, CSV, and TSV forward reader keeps memory bounded; BIFF8 shared-string continuations spill beyond a configurable threshold while retaining projection, selected-sheet, encryption, and malformed-input safeguards

Fixed

• C++Builder Win64 package linking no longer emits duplicate Delphi generic object-list vtables from workbook comparison and import-check units

Version 2.330.0

Added

• TXLSXWorksheet.Subtotal inserts grouped subtotal rows, a configurable grand-total row, formulas, inherited visible formatting, and Excel-compatible outline levels for sorted data ranges
• GetCellVisibleFormat resolves worksheet, row, column, range-overlay, and explicit cell formatting without materializing an empty cell
• CheckDataValidation tests a candidate value against the rules covering one cell and can commit it atomically only when valid

Changed

• Lookup scan memos use compact exact binary keys and a single payload lookup per cache hit, reducing recalculation allocation and hash-table work without changing lookup comparisons
• Fixed-width CSV export pads each field in one operation, keeping padding time linear in the configured output width

Fixed

• Backward ReplaceTextIn scans now invoke replacement callbacks in the requested reverse scan order
• Styled cells with no value or formula no longer block neighbouring text overflow in SVG, image, and preview rendering
• Classic XLS numeric output normalizes negative zero in addition to non-finite and subnormal values before writing BIFF numeric records

Version 2.329.0

Added

• A1ToR1C1 converts A1-style formulas to R1C1 notation (relative offsets, absolute $ forms, ranges, and round-trips through R1C1ToA1), completing both directions of reference-style conversion

Changed

• The recalculation lookup memo skips scan vectors of fewer than five cells in VLOOKUP, XLOOKUP/XMATCH, and MATCH, where a raw scan is cheaper than the dictionary round-trip

Version 2.328.0

Added

• The shared page renderer behind SVG, image, and preview output now paints XLSX cell styles: solid fills (with alpha), per-cell fonts and colors, horizontal alignment with Excel's numeric-right default, Excel-style border collapsing where the heavier shared edge wins, shrink-to-fit that scales the font to the column, and overflow of long text into consecutive empty neighbour cells

Version 2.327.0

Added

• The CSV/TSV loader gains per-column import directives (cctText, cctNumber, cctDate, cctBoolean, cctSkip) mirroring Excel's Text Import wizard, so individual columns can be forced to text to keep leading zeros, coerced to a type, or dropped from the import
• Fixed-width text import: assigning field widths switches the loader from separator parsing to character-position slicing, and the saver can emit space-padded fixed-width columns with truncation

Version 2.326.0

Added

• XLSX worksheet search gains windowed FindTextIn and ReplaceTextIn methods that confine the scan to a rectangular cell range, plus a column-major scan order flag that mirrors the Search: By Columns mode of Excel's Find and Replace dialog; the backward scan now accepts a zero anchor to start from the bottom-right corner
• Saving now guards against formulas longer than Excel's 8192-character limit, failing fast with the offending cell identified instead of writing a workbook Excel would open in repair mode

Fixed

• Non-finite numeric cell values (NaN and positive or negative infinity) can no longer leak unparseable tokens into saved XLSX parts; they now serialize as #NUM! error cells, and subnormal doubles flush to zero in both the XLSX and BIFF8 writers, matching what Excel can store and display

Version 2.325.0

Added

• Workbook-format APIs now include editable XML Maps and schema bindings, ordered multiple workbook windows, multi-area and per-pane selections, and caller-owned lossless formula-token inspection, reconstruction, reference replacement, and atomic writeback for XLS and XLSX
• Workbook processing now includes stable horizontal or vertical sorting with custom comparers and explicit formula movement, BIFF2 through BIFF4 style and formula conversion, one projected forward-reader facade for XLS, XLSX, CSV, and TSV, typed custom views, and editable OLAP, web-table, fixed-text, and parameter connection submodels
• Calculation and inspection now include structured missing-function, external-reference, and circular-dependency diagnostics, global and workbook-local UDF registries, Unicode bidirectional visual ordering, merged-cell AutoFit strategies, typed Excel 4 macro-sheet preservation and command scanning, and the ISOMITTED, ARRAYTOTEXT, VALUETOTEXT, PDURATION, and RRI functions
• Compiled reporting now covers whole-workbook placeholders, named row, column, and cross-tab bands, expandable Excel-table bands, one table and cursor abstraction, exact compound-key relationship indexes, lazy data contexts, streaming table combinators, controlled named query providers, per-record sheet bands, overflow continuation sheets, and immutable recursive includes
• Report authoring now includes compiled conditions, lookup, regular expressions, dynamic formulas and formats, image fields, chart-series rebinding, reusable runners with stream or atomic file output, dependency-aware range cloning, fixed-footer and balanced pagination planning, addressed diagnostics, strict execution, template checking, and safe migration of supported legacy placeholders
• PDF output now includes automatic tagged structure trees, visible or invisible host-signed fields, explicit PDF/A-1, PDF/A-2, and PDF/A-3 conformance profiles at levels a and b, atomic progress and cancellation, style-aware cached font fallback, contextual complex-text shaping, native multi-stop gradients, per-stop alpha, cell fill patterns, shadows, and positioned comment annotations
• Shared rendering now includes native range, page, workbook, and named-object SVG, incremental PNG and JPEG pages, backend-neutral layout geometry, inline vector charts and shapes in HTML, stateful partial HTML sessions with shared CSS, rendered picture effects, vector preset-shape and WordArt geometry, advanced scientific and financial chart details, and deterministic data-label collision and leader-line layout

Changed

• All 55 audited capability items F01 through F18, R01 through R19, and V01 through V18 now have completed supported implementations with explicit ownership, format-safety, cancellation, and output boundaries
• Large report templates compile object targets and expressions once, materialize only bound axis fields, consume data cursors once, hash relationship and join keys, freeze source collection prefixes for repeated disjoint cloning, and avoid rescanning generated workbook objects
• Rendering paths share pagination, geometry, object placement, effect state, and document resource caches while PDF, SVG, HTML, bitmap, metafile, preview, and print outputs retain only their format-specific encoders

Fixed

• XLSX cell-write events now serialize their effective replacement values without retaining stale rich-text runs, BIFF external-name tokens preserve the complete 4-byte name index, and top-level connection sources round-trip independently from text-import sources while retaining the legacy source fallback
• Legacy BIFF style and formula meaning, workbook-window and selection remapping, mixed absolute formula references, formula-token ownership, and invalid token writeback now retain deterministic lossless behavior across load, edit, and save
• Report expansion now rejects deterministic table, filter, name, overlap, partial-object, grid, and query-backed conflicts before structural writes, preserves zero-row table cardinality where representable, keeps exact integer join keys beyond 2^53, and reports deterministic applied counts after cancellation or failure
• PDF and vector export now preserve fallback-font Unicode mappings, grapheme and shaping clusters, archival profile restrictions, gradient alpha, image effects, chart decorations, dense label placement, and output atomicity without resource growth from repeated content

Version 2.324.0

Added

• Lossless formula-token inspection and guarded writeback for XLS and XLSX through TXLSFormulaTokenStream, ParseFormulaTokens, GetFormulaTokens, FormulaTokensToText, and SetFormulaTokens expose RPN token kinds and classes, raw BIFF token and extra payloads, independent reference flags, cloning, reference replacement, context provenance, and exact compatible BIFF reuse while rejecting malformed, incompatible, array-owned, or unsupported streams before cell mutation
• Expandable Excel-table report bands through TXLSCompiledTableReportBand use one bounded cursor snapshot to expand a single data-template row while preserving headers, totals, table and AutoFilter ranges, filter and sort state, column metadata, calculated and totals formulas, styles, structured references, and dependent worksheet objects; representable zero-row results remove the template data row, while query-backed, conflicting, invalid, and overflowing layouts fail before writes
• Style-aware PDF font fallback chains through SetFontFallbackChain, AddFontFallback, and ClearFontFallbackChains resolve source-family and style tiers by glyph coverage, cache loaded faces and selections, preserve supplementary scalars and cross-run grapheme clusters, assign distinct CIDs to shared-glyph Unicode sequences, embed standalone TTC faces, and reject PDF/A output atomically when no embedded glyph is available

Version 2.323.0

Added

• Reusable workbook report templates through TXLSCompiledReportTemplate: placeholders are compiled once and can be applied repeatedly without rescanning the workbook, using a sorted value lookup and pre-split text fragments across cells, rich text, comments, threaded comments, hyperlinks, text boxes, chart titles and series names, headers and footers, worksheet names, defined names, and document properties
• Report data binding through IXLSReportTable, IXLSReportCursor, lazy IXLSReportTableProvider loading, random-access memory tables, and TXLSReportRelationIndex, which builds a compound-key child hash once and retains row numbers instead of copying row payloads for repeated master-detail lookup
• Formula calculation now evaluates PDURATION for the number of periods needed to reach a target investment value and RRI for the equivalent periodic growth rate, including Excel-compatible numeric error boundaries and the required OOXML _xlfn. storage prefix
• Formula calculation now evaluates ISOMITTED, ARRAYTOTEXT, and VALUETOTEXT, including a dedicated omitted-argument value that remains distinct from a blank cell, BIFF missing-argument token round trips, one- and two-dimensional arrays, strict string escaping, localized separators, error conversion rules, and OOXML future-function prefixes
• Structured formula preflight through TXLSXWorkbook.AnalyzeFormulaIssues reports missing functions, external workbook references, and exact circular dependency components with cell coordinates and bounded member lists, while string literals and local structured references are excluded from false matches
• Dependency-aware range cloning through TXLSXWorksheet.CloneRangeTo copies selected cells, merged ranges, conditional formats, data validations, hyperlinks, comments, images, and charts in one operation, relocates applied ranges and anchors, and translates formula payloads with mixed relative and absolute reference semantics; report bands use the same path for dependent objects
• PDF export progress and cancellation through TXLSPDFExport.OnProgress, OnProgressEx, and Cancel: begin, page, and end notifications add negligible callback overhead, cancellation is checked between row passes and page objects, cancelled streams receive no partial PDF, and the exporter remains reusable
• Compiled report diagnostics through TXLSCompiledReportTemplate.Analyze, strict execution, and WriteDiagnostics identify missing values, empty placeholders, and unclosed placeholders with target type plus worksheet and cell coordinates before mutation, or write the same findings to a diagnostic worksheet
• Reusable report execution through TXLSCompiledReportRunner provides begin, target, and end progress with completed, cancelled, and failed states, supports cancellation before the first write or between compiled targets, reports the exact applied count, and remains reusable after cancellation or exceptions without rescanning the workbook
• Merged-cell AutoFit strategies for classic XLS and XLSX allow the first or last row or column to absorb the required growth, or distribute it evenly across the merged span, while preserving legacy overload behavior, hidden states, explicit bounds, and deterministic repeated calls
• Optional PDF comment annotations through TXLSPDFExport.ExportCommentsAsAnnotations place Unicode author and comment text on the originating cell rectangle while preserving the existing comments-at-end page; per-sheet comment indexes build in approximately O(C log C) time and support O(log C) cell lookup
• Hierarchical user-defined function registries let XLS and XLSX workbooks share process-wide handlers while retaining workbook-local overrides, replacement, removal, clearing, and the existing workbook event fallback; average lookup is O(1), local matches mask global names, and handlers run outside registry locks for safe parallel recalculation and self-removal
• Lazy report-table combinators through TXLSReportTableFactory provide top rows, column projection, row count, minimum and multiple aggregate definitions, schema-mapped union, contiguous split partitions, and duplicate-preserving inner or left hash joins without materializing complete intermediate row payloads
• Incremental PNG and JPEG page export through TXLSPageImageExporter supports XLS and XLSX ranges, worksheets, and workbooks with page selection, DPI, scale, JPEG quality, progress, cancellation, and reusable sessions; one pagination model is built per session and bitmap memory is released after every page
• Multiple XLSX workbook windows through TXLSXWorkbook.WorkbookViews preserve independent active tabs, first visible sheets, visibility, minimized state, geometry, tab display, scrolling, unknown attributes, extensions, and correctly remapped worksheet view bindings while the existing single-window properties continue to address the first view
• Named row, column, and cross-tab report bands through TXLSCompiledReportBand expand a defined-name template structurally from one-pass bounded data snapshots, preserve formulas and dependent worksheet objects, tile explicit row and column metadata without creating zero-sized defaults, and reuse fixed source collection boundaries so repeated same-sheet cloning does not rescan generated payloads
• PDF export now writes two-stop and multi-stop axial or radial gradients as native PDF shadings, reuses identical function, shading, and transparency resources, clips each fill to its cell, and applies solid or uniform gradient alpha without the former banded rectangle approximation

Fixed

• Custom point-based paper width and height overrides now participate in the pagination band calculation as well as the final page object, so page breaks and rendered dimensions use the same available area

Version 2.322.0

Added

• Formula calculation now evaluates the Bessel functions of the first kind BESSELJ (oscillating) and BESSELI (modified), computed via the power series with term recurrence for numerical stability across the practical argument range

Version 2.321.0

Fixed

• XLSX RTF export now builds a real font table instead of hardcoding Calibri: the pre-scan collects every distinct cell font name, the \fonttbl lists each with its own \fN index, and each cell’s text references its actual font — workbooks styled in Times New Roman, Courier New, or any other font no longer visually re-flow to Calibri when opened in Word or WordPad

Version 2.320.0

Fixed

• PDF export now renders text rotation on the classic XLS engine path: cells with 90-degree or 270-degree orientation draw their text vertically through a PDF text matrix pivoting around the cell centre, matching the existing XLSX-path behaviour; previously vertical headers and rotated labels rendered horizontally in PDFs produced from .xls workbooks

Version 2.319.0

Added

• Formula calculation now evaluates the Excel 365 LAMBDA-dependent array combinators MAP (element-wise transform), REDUCE (fold to a single value), SCAN (fold spilling each intermediate step), MAKEARRAY (build from row and column indices), BYROW (row-wise aggregate spilling a column), and BYCOL (column-wise aggregate spilling a row), completing the LAMBDA functional-programming family alongside the existing LAMBDA and LET support

Version 2.318.0

Added

• Formula calculation now evaluates the bond and treasury finance functions: COUPDAYBS, COUPDAYS, COUPDAYSNC, COUPNCD, COUPPCD, COUPNUM (coupon day-count primitives), PRICE / YIELD (standard bond pricing with bisection yield solver), PRICEDISC / YIELDDISC (discounted securities), PRICEMAT / YIELDMAT (maturity-based), DURATION / MDURATION (Macaulay and modified), TBILLEQ / TBILLPRICE / TBILLYIELD (treasury bills), DISC / INTRATE / RECEIVED (discount and interest), ACCRINTM (accrued interest at maturity), DOLLARDE / DOLLARFR (fractional-dollar conversion), EFFECT / NOMINAL (effective and nominal rates), CUMIPMT / CUMPRINC (cumulative interest and principal), FVSCHEDULE (variable-rate future value), and AMORDEGRC / AMORLINC (French depreciation); all share a self-contained nested coupon-date engine with the five standard day-count bases

Version 2.317.0

Added

• Encrypted-workbook loading now accepts Agile Encryption files that declare the SHA-256 hash algorithm (produced by some third-party writers): the password key derivation, spin loop, password verifier, and the data-integrity key/value IVs and HMAC all follow the declared hash, while files declaring mismatched hashes between the key data and key encryptor are still rejected; saving continues to write Excel’s default SHA-512 variant

Version 2.316.0

Added

• SortRange preserves formulas on both workbook engines: formula cells travel with their rows and their relative references re-anchor with Excel's sort semantics ($-absolute parts stay, references shifted off the grid clamp in place), so sorted tables keep computing instead of collapsing to values
• Sheet background images: SetBackgroundImage tiles a picture behind the cells of an XLSX worksheet (display-only, like Excel), round-tripping through the worksheet picture element, its media relationship, and the package content types

Fixed

• Recalculating an arithmetic formula that references a never-touched empty cell no longer aborts with a conversion error — empty operands contribute 0 like Excel

Version 2.315.0

Added

• CopyToClipboard publishes the native Biff8 clipboard format alongside text and HTML on both workbook engines, so ranges pasted into Excel or other HotXLS workbooks carry values, formulas, number formats, fonts, fills, borders, alignment, and merged cells instead of plain text
• PasteFromClipboard prefers the Biff8 payload when present: text pasted from Excel keeps its original values without locale-dependent date and number re-parsing, formulas re-anchor their relative references to the paste target, and merged areas rebuild at the destination; both engines also paste each other's blocks with full formatting

Changed

• Formulas pasted into Excel arrive as their computed values (a limitation of Excel's third-party Biff8 clipboard parsing); everything else — text, numbers with formats, fills, borders, and merges — pastes losslessly into Excel, and formulas survive fully in Excel-to-HotXLS and HotXLS-to-HotXLS directions

Version 2.314.0

Added

• Camera objects on XLSX worksheets: AddCamera places a live snapshot picture bound to a cell range — the Excel camera tool's linked picture — with the range binding written to the documented drawing extension and a placeholder image sized to the anchor; camera objects created by Excel reopen with their range binding preserved

Version 2.313.0

Added

• In-place worksheet patching for XLSX packages: BeginPatch on the streaming writer opens an existing workbook and rewrites exactly one worksheet — every other part (styles, drawings, charts, pivot tables, custom parts) is copied across with its compressed bytes verbatim, without decompressing or re-authoring anything, so a large template gets its data sheet refreshed in a fraction of a full rebuild
• Patch mode can also append a new worksheet: the sheet is registered in the workbook part list, relationships and content types with numbering that never collides with existing parts, following the package's own strict or transitional relationship convention
• Patched packages drop the stale calculation chain automatically (Excel rebuilds it on load), and features that would need new package parts (tables, charts, images, comments, defined names) report a clear error instead of writing an invalid package

Version 2.312.0

Added

• Strict Open XML output: setting StrictOOXML on an XLSX workbook saves the package in ISO/IEC 29500 strict conformance — every namespace declaration and relationship type switches from the transitional vocabularies to the strict purl.oclc.org ones across worksheets, shared strings, styles, theme, drawings, charts, comments, document properties, and preserved parts, while content types, part names, and the VML vocabularies stay as both conformance classes share them
• Reading strict-conformance workbooks is now complete: relationship types written in strict form resolve to the same targets as their transitional equivalents, complementing the strict element-namespace acceptance so files saved in Excel's "Strict Open XML Spreadsheet" mode open like any other workbook
• Saving with strict output enabled reports a clear error for features that exist only as Microsoft extensions (VBA projects, form controls, threaded comments, dynamic-array metadata) instead of writing a package that would not validate

Fixed

• Custom document properties declared a misspelled value-types namespace, which could make the properties part fail validation in external tools

Version 2.311.0

Added

• Reading an XLS workbook preserves unrecognized worksheet records: any record the typed parser does not model is captured verbatim and replayed on save, so files carrying vendor-specific or newer-spec records round-trip through HotXLS without losing them

Version 2.310.0

Added

• Preset-geometry shapes on XLSX worksheets: AddPresetShape writes any of the 187 OOXML ST_ShapeType preset tokens (rightArrow, flowChartDecision, star5, …) into the drawing's <a:prstGeom> verbatim, and the PrstToken property lets existing shapes switch geometry — tokens round-trip through save and reopen

Version 2.309.0

Added

• Template processing on worksheets: ApplyTemplate replaces {{key}} placeholders across all cells — values keep their native types, formula placeholders substitute before parsing — and FillTemplateBand clones a template block once per record with per-clone formula row translation, merged-range cloning, and a #skip field for row-level conditionals, turning any hand-crafted sheet into a data-driven report generator

Version 2.308.0

Added

• PDF export can attach files: AddAttachment carries raw documents into the exported PDF as /EmbeddedFiles entries (file specification, embedded stream with optional MIME subtype, and the catalog name tree), so reports ship together with their supporting data in one self-contained file

Changed

• Reading an XLS workbook with an unrecognized BIFF version word no longer rejects the whole stream: the reader falls back to the BIFF8 layout so files from writers that emit unusual version numbers still load when their records parse

Version 2.307.0

Added

• Central scratch-file management: every temporary file the library creates is registered and swept, so an aborted export or an exception between create and delete no longer leaves orphaned temp files behind
• VerifyEncryptedSave re-opens a just-written encrypted container through the full decrypt pipeline and fails the save when the password cannot unlock it — broken containers surface immediately instead of at the user's next open
• Dataset export raises an OnExportLimit event when row-splitting kicks in, and workbook grid limits per format version are exposed as constants (65536×256 for BIFF8, 1048576×16384 for the 2007+ formats) for capability negotiation

Version 2.306.0

Added

• XLSX worksheets gain form-control authoring: AddFormButton, AddFormCheckBox, AddFormRadioButton, AddFormComboBox, and AddFormListBox write button / check box / radio / drop-down / list controls with captions, linked cells, checked state, and macro binding — the save path emits the VML shapes, the per-control ctrlProps parts, their relationships, and the content-type entries, and files carrying controls round-trip their parts verbatim on reopen

Version 2.305.0

Added

• PDF export gains an opt-in UsePDFA mode producing PDF/A-2b identified documents: the sRGB IEC61966-2.1 profile is embedded as the GTS_PDFA1 output intent, the XMP metadata carries the pdfaid part/conformance identification, and the header announces PDF 1.7 — pair it with an embedded TrueType font (UnicodeFontFile or the auto-selected installed font) for archivable output

Version 2.304.0

Added

• Live cross-workbook evaluation: RegisterExternalWorkbook pairs a workbook object with a book name so formulas like [Book.xlsx]Sheet!A1 compile (the [Book]Sheet!Ref prefix now parses into a real external reference) and read the registered object's current state on every recalculation; unregistering drops the live link and falls back to the external-link cache

Version 2.303.0

Added

• Workbook comparison through TXLSXWorkbookCompare: pairs sheets by name and reports added/removed sheets, added/removed cells, changed values, changed formulas, and added/removed merged ranges — as a structured diff list for tooling and as one human-readable line per difference through Report

Version 2.302.0

Added

• The streaming direct writer gains a pull-mode row supply: OnNeedRow fires once per prospective row and the supplier writes that row straight through the writer, declaring end-of-data to stop — data pipelines can drive an export without pre-buffering row lists

Version 2.301.0

Added

• ODS export writes cell formulas: same-sheet references translate to the standard of:= OpenFormula spelling (bracketed refs, semicolon argument separators, string literals untouched), while sheet-qualified and structured references ride the msoxl: passthrough Excel itself uses for cross-suite round trips — saved packages reopen with the original Excel formula text

Version 2.300.0

Added

• Right-to-left worksheets now render mirrored everywhere: PDF export flips the column geometry across the printable width (column A on the right, cell content untouched) and both HTML exporters emit dir="rtl" on the table so the browser flips the column order — the sheet flag already round-trips through XLSX and BIFF

Version 2.299.0

Added

• Dataset export (TDataToXLS) gains a title band and a footer band: Title renders one merged, boldly styled row above the header (own TitleFont and optional TitleColor), and Footer appends a line under the last data or summary row
• ForEachFormattedCell on a worksheet walks every populated cell in row/column order and delivers its display text — number format applied — through a callback, ready for indexing and text extraction

Version 2.298.0

Added

• The ApiTour example now opens straight into a live showcase: an in-memory sales dashboard with KPI tiles, cross-sheet SUMIF/COUNTIF formulas, data bars, a color scale, an icon set, and two charts renders inside the embedded workbook viewer the moment the sample starts, every grid cell stays editable with instant recalculation and refreshed conditional formats, and one-click actions export the showcase to PDF, HTML, and CSV or hand it to Excel with live build and export timings
• The C++Builder ApiTour sample mirrors the new Delphi showcase, embedded viewer, formula bar, export actions, and activity log

Changed

• The ApiTour examples replace modal result dialogs with an in-window activity log while the classic create, modify, and export walkthrough keeps its behavior

Version 2.297.0

Added

• PDF export gains an opt-in UseXRefStream mode producing PDF 1.5 files with compressed object streams and a binary cross-reference stream: small dictionary objects travel inside one Flate-compressed container and the classic xref table and trailer are replaced by a compact stream, shrinking exported documents while page content stays byte-identical

Version 2.296.0

Added

• The streaming direct writer gains an opt-in AutoSplitRows mode: a row past the worksheet limit (1048576) finishes the current sheet and continues on a freshly created <name>_2, <name>_3… sheet, so unbounded exports never emit an illegal row index
• FillTableCalculatedColumn on a worksheet pushes a table column's calculated-column formula down into every data row, row-translated per cell the way Excel's own authoring does, then recalculates naturally

Version 2.295.0

Changed

• XLSX HTML export centralizes cell styling: identical style combinations are emitted once as td.cN classes in a style sheet instead of a repeated inline style attribute on every cell, shrinking HTML output for uniformly-styled tables; conditional-format visual overlays stay inline so they keep their per-cell character

Version 2.294.0

Fixed

• Structured references in current-row form (Table[@Col]) now evaluate against the formula cell's own row during recalculation, in both the serial and the parallel evaluation paths; previously every such formula read row 1 of the column — typically the header text — and recalculation failed outright with a type-cast error when that row held text

Version 2.293.0

Changed

• The opt-in RecalcAggregateCache memo now also covers unconditional single-range aggregates — SUM, AVERAGE, MIN, MAX, COUNT, and COUNTA over one plain single-sheet range — so repeated totals over the same block reuse the first pass's result under the same reference-stability gate as the conditional family

Version 2.292.0

Added

• Workbook-to-code generator (TXLSCodeGen): walks a workbook and emits the source statements that rebuild it — sheet creation, cell values, formulas, number formats, column widths, and merged ranges — in Delphi or C++Builder dialects, so a hand-crafted file becomes a ready-made API example
• Column-rule import validator (TXLSImportCheck): per-column required, text-length, regex-pattern, and uniqueness rules plus a caller callback run over a worksheet's data rows; violations can be annotated back with a red font and a reason comment, and KeepOnlyViolations compacts the sheet to the header plus failing rows for a "please fix and return" workbook

Version 2.291.0

Added

• CSV import accepts an options record: OpenCSV overloads take leading-line skipping (SkipRows, for title banners above the grid) and whitespace trimming (AutoTrim) alongside the separator and type-detection switches
• FillMergedCellValues on a worksheet copies each merged range's anchor value into every covered member cell, so report-style sheets with merged labels read as one flat grid
• AddListValidation gains an array-of-strings overload that joins the items into one dropdown list in a single call
• Dataset export (TDataToXLS) gains zebra striping (AlternatingColor, painted on odd detail rows) and a totals row (SummaryRow) that appends real SUM formulas for the exported numeric fields or a caller-named field list, with a bold caption cell

Fixed

• Dataset export now applies the configured DetailColor background to detail cells; previously the property was accepted but never painted

Version 2.290.0

Changed

• The opt-in RecalcLookupCache memo now also covers VLOOKUP exact and approximate scans, HLOOKUP, and MATCH match types 0 and 1, so workbooks that repeat these classic lookups over the same range reuse the first scan's outcome during a recalculation pass under the same reference-stability safety gate as XLOOKUP/XMATCH

Version 2.289.0

Added

• The streaming direct writer gains an opt-in TrackColumnWidths mode: the worksheet is staged in memory while rows stream so the saved file carries the <dimension> used-range hint and per-column widths sized to the longest display text seen (characters plus padding, capped at 255), approximating autofit for exports whose data is only known as it streams; explicitly set column widths always win

Version 2.288.0

Added

• Opening XLSX packages tolerates non-canonical namespace prefixes: worksheets, charts, drawings, and document-property parts that bind their elements to a differently prefixed (or ISO 29500 strict) namespace are normalized onto the standard vocabulary while parsing, so files from third-party generators and strict-mode savers load correctly
• Worksheets whose cells omit the r position attribute place each cell after its predecessor in scan order, and rows without an r attribute continue the row sequence — the convention used by several streaming writers

Version 2.287.0

Added

• Open reads Excel 2.0/3.0/4.0 workbooks (BIFF2-4): raw BIFF streams without an OLE wrapper load with their numbers, text, booleans, errors, and formula cached results; string results of formulas arrive through the BIFF2-4 String record, and the stream's code page record drives text decoding

Changed

• BIFF2-4 documents are read for values — number formats and styles of those generations are not modeled, so dates surface as raw serial numbers and formulas as their last computed values

Version 2.286.0

Added

• Number formats with comparison sections ([>=100], [<0], …) select the section by value everywhere display text renders, matching Excel's conditional-format behaviour
• Number-format text sections ("ID:"@) template string values instead of ignoring the format
• SetCellFromHtml fills a cell from a light HTML fragment: <b>/<strong>, <i>/<em>, <u>, <br>, and <font color=...> become rich runs; other tags drop while their text is kept
• The TaggedText markup accepts <br> line breaks, <strong>/<em> aliases, the <font color=...> attribute form, and " /   entities
• Column keep-together blocks (KeepColumnsTogether) stop vertical page breaks inside a column group, mirroring the row-level feature in pagination, preview, and PDF output
• Classic Spinner and ScrollBar form controls can be authored into XLS files with Min/Max/Inc/Page settings and a linked cell, emitted through the OBJ FtSbs subrecord
• WriteAutoValue recognizes m/d/yyyy and d.m.yyyy dates with optional times (AM/PM included) beyond ISO form, rejecting impossible calendar dates
• ConvertFormulasToValues gains an OnlyExternal mode that freezes only formulas referencing other workbooks, and the classic XLS facade gains the whole-sheet method
• Row auto-fit measures merged cells against their combined column width and distributes the needed height across the merged rows

Version 2.285.0

Added

• PDF export draws worksheet charts as native vector graphics: clustered, stacked, and percent-stacked columns and bars, line charts with markers, area charts, pie and doughnut charts with percentage labels, scatter plots, and radar charts — each with a titled frame, auto-scaled axes, tick labels, gridlines, a legend, optional per-point data labels, per-series solid fills, and Excel's default accent palette; 3D chart types render as their 2D projection
• Charts anchored beyond the last cell now extend the printed surface, so a chart placed below the data rows still appears in PDF and preview output instead of being clipped away

Version 2.284.0

Added

• PDF headers and footers render Excel's style codes: &B bold, &I italic, &U/&E single/double underline, &S strikethrough, &X/&Y super/subscript, &nn font size, &K colour, and &"Font,Style" weight keywords
• PDF export honors the worksheet's black-and-white print mode by collapsing fills, fonts, borders, data bars, and icons to luminance grey
• Printed row and column headings (row numbers and column letters) render beside the grid when the sheet requests them
• Print centering: CenterHorizontally / CenterVertically translate the printed grid between the margins while header and footer bands stay in place
• Cell comments print either as a red corner indicator (as-displayed mode) or as a comment listing page after the sheet through the new PrintCommentsAtEnd worksheet property, which round-trips Excel's cellComments="atEnd" page-setup value
• Internal hyperlinks (Sheet!Cell targets) become clickable in-document GoTo links that jump to the target sheet's page
• The exporter exposes a viewer PageLayout preference and writes an XMP metadata packet alongside the document information dictionary
• TrueType collection (.ttc) fonts can serve as the embedded Unicode font; the bold and italic cuts of the embedded face are synthesized when only a regular face exists

Changed

• Column widths convert through Excel's maximum-digit-width formula (Calibri 11 metrics) instead of a fixed per-character estimate, so PDF and preview widths match Excel's stored pixel widths
• Wrapped lines re-measure from each line's own first character, fixing wrapped lines that ignored the width of words moved down by an earlier break

Version 2.283.0

Added

• PDF export wraps WrapText cells with real word breaking, line-by-line horizontal alignment, vertical block alignment, and justified or distributed spacing
• PDF export renders rotated cell text through PDF text matrices, including vertical stacked text, and honors ShrinkToFit by stepping the font size down until the value fits
• PDF text that overflows a cell now extends across empty neighbouring cells and clips at the first occupied cell, matching Excel print behaviour; general alignment right-aligns numbers like Excel
• PDF export draws rich-text runs with per-run font style, size, colour, underline, strikethrough, superscript, and subscript; wrapped lines keep the per-run styling
• PDF export draws diagonal cell borders and true double-line borders
• Text spanning merged cells now draws, aligns, wraps, and anchors hyperlinks across the whole merged rectangle, and covered merge members no longer paint their own background

Changed

• PDF text widths come from calibrated per-character advance tables instead of a fixed average character width, so alignment, wrapping, and shrinking match the printed page
• Cell indentation shifts text like Excel, and icon-set cells indent their value next to the icon

Version 2.282.0

Added

• PDF export preserves PNG transparency: images with an alpha channel now emit a soft-mask plane instead of losing transparency, including palette images with a tRNS chunk

Changed

• PDF page content streams are now Flate-compressed, shrinking exported documents substantially with no viewer requirements beyond PDF 1.4
• PDF export batches consecutive text operations into shared text blocks and skips redundant font, size, and colour selection, reducing both output size and generation time
• Adjacent cells with the same solid fill render as one merged rectangle, and collinear border segments with the same style stroke as a single run instead of per-cell edges
• PNG images are embedded in PDF without decoding and re-encoding, keeping the original compressed stream and exact pixels
• Images with identical byte content are embedded once and shared, regardless of how many image objects reference them

Fixed

• Hyperlink annotations are now also created for link cells whose text renders through the embedded Unicode font path
• PDF content bytes no longer depend on the active system code page, so Western text renders identically on any Windows locale

Version 2.281.0

Fixed

• Clipboard readers no longer truncate the data handle to 32 bits, which made XLSGetClipboardBiff8 and the XML Spreadsheet paste path return empty on 64-bit applications
• Clipboard operations retry opening the clipboard for up to half a second when another application holds it, instead of failing on the first contention

Changed

• Conditional-format rule lookups (FindAt / IndexOfCell paths) parse each rule’s cell range once instead of on every queried cell, matching the data-validation cache
• PDF export builds a per-sheet hyperlink index, replacing the linear per-cell scan with a binary search

Version 2.280.0

Added

• User-defined functions can be registered by name: RegisterUserFunction dispatches formulas calling that name straight to its handler (no global if-chain in one OnUserFunction), with the workbook event still serving unregistered names
• ScanUnknownFunctions audits a workbook for formula functions the calculation engine cannot evaluate, reporting each cell and name — ideal before enabling recalculation on foreign files
• RecoveryMode: a structurally damaged worksheet part is skipped with a warning diagnostic instead of aborting the whole open
• Native BIFF8 clipboard format: XLSSetClipboardBiff8 / XLSGetClipboardBiff8 publish and read the Biff8 stream Excel itself uses, so clipboard exchange keeps values, formulas, and formatting

Changed

• Data-validation auditing parses each rule’s cell range once instead of on every cell, speeding CheckDataValidations over large sheets

Version 2.279.0

Added

• PDF export embeds worksheet images: PNG/GIF/BMP bitmaps are converted to Flate-compressed RGB XObjects and JPEG files pass through untouched via DCTDecode (no recompression); identical images on several pages share one object, and images added without an explicit size now default to their pixel dimensions at 96 DPI
• Keep-together blocks: KeepRowsTogether stops print/PDF pagination from breaking inside a row block that fits one page, and XlsxAutoPageBreaks turns the natural (keep-aware) page boundaries into explicit row breaks so Excel paginates like the export
• Configurable XLSX compression: ZipCompressionLevel selects the deflate level for saves (-1 default through 9)
• Reproducible saves: DeterministicOutput pins zip entry timestamps so saving the same workbook twice produces byte-identical files

Fixed

• Corrupt compound files with cyclic FAT chains no longer hang the classic .xls loader — directory and stream chain walks carry cycle guards

Version 2.278.0

Added

• PDF export emits clickable link annotations for worksheet hyperlinks: cells carrying a URL get an invisible /Link rectangle with a URI action over the cell area
• PDF export writes a document outline (bookmarks): each exported worksheet becomes a bookmark pointing at its first rendered page

Version 2.277.0

Added

• Data validation gets an audit API: CheckDataValidations scans populated cells against the sheet’s rules (list membership, whole/decimal/date/time bounds, text length, custom formulas through the calculator) and reports every violating cell with the rule index and a reason
• Typed custom document properties round-trip through docProps/custom.xml: text, number, date, and boolean entries on Workbook.CustomProperties
• Custom XML data parts round-trip through Workbook.CustomXmlParts with workbook-level relationships and content types
• Images accept an SVG companion: AddImagePair stores the SVG bytes next to the raster fallback, writes the svgBlip extension and the extra media part Excel 2016+ consumes
• Charts can source data-label text from a cell range (“Value From Cells”) through DataLabelsRange, emitted as the chart extension and recovered on open
• Custom table styles: Workbook.CustomTableStyles defines named styles with colored/bold sections (header row, total row, stripes, edge columns), written to tableStyles with their own differential styles so tables can reference them by name

Version 2.276.0

Added

• Cell gradient fills: linear sweeps by degree and path gradients with per-stop colors round-trip through gradientFill, render as CSS gradients in HTML export, band-interpolated fills in PDF export and the workbook viewer, and are offered through the TXLSXFills.AddLinearGradient convenience
• Elapsed-time number formats ([h]:mm:ss, [mm]:ss, [s]) now render running totals instead of wrapping like a clock, in every display path (FormattedText, HTML, PDF, conditional-format values)
• Fraction number formats (# ?/?, # ??/??, fixed denominators like 0/8) render with Excel’s best rational approximation, including zero-integer hiding and alignment spacing
• [Red]-style color brackets now surface through TXLSXCell.GetFormattedTextColor and repaint the cell text in PDF export
• TXLSXWorkbook.ValidateNumberFormat checks a custom format string the way Excel’s Custom Format dialog does — section count, quote balance, and bracket content (colors, conditions, elapsed units, locale directives)
• Frozen pane state can be read back through GetFrozenPanes, and the bottom-right pane scroll position round-trips through SetPaneScroll / GetPaneScroll (written as topLeftCell)
• The row/column heading display toggle DisplayRowColHeaders round-trips through showRowColHeaders
• Charts expose DispBlanksAs (gap / span / zero) and PlotVisibleOnly instead of hardcoding them, and recover both from preserved ChartML on open
• Comments carry editable box geometry (WidthPt / HeightPt, VML points) plus a text-driven AutoSizeFromText heuristic

Version 2.275.0

Added

• Find & replace understands Excel wildcards: * (any run of characters), ? (exactly one character), and ~ as the escape for a literal *, ?, or ~ — the same semantics Excel itself applies in Find & Replace, COUNTIF-family criteria, and autofilters
• The XLSX worksheet FindText gains an options overload: match case, match entire cell contents, search formula text as well as values, wildcard mode, and backward scanning, with a start anchor so repeated calls walk through matches
• ReplaceText gains the same option set plus an optional per-match callback that can accept, rewrite, skip, or abort each replacement cell by cell
• The classic XLS worksheet FindText / ReplaceText gain wildcard and entire-cell overloads, so both facades share Excel search semantics
• GetRangeStatistics scans a rectangle and reports minimum, maximum, sum, average, numeric value count, and populated cell count in one pass
• WriteAutoValue writes text the way Excel's own cell entry interprets it: TRUE/FALSE become booleans, error literals such as #N/A become error values, unambiguous ISO dates become dates, plain invariant-format numbers become numbers, and everything else — including leading zeros and thousands separators — stays text (both facades)
• GetHtmlFromCell renders one cell as a styled HTML fragment with escaped markup, complementing the whole-sheet HTML export
• SortRange accepts a case-insensitive flag so text keys order alphabetically regardless of case (XLSX facade)

Version 2.274.0

Added

• Excel tables now round-trip header-less tables and per-table comments through the new HasHeaderRow and Comment table properties, written as headerRowCount="0" and the comment attribute
• Charts can select an Excel gallery style through the new ChartStyle property (1–48 or the extended 101–148 range), round-tripped through the chart part’s style element
• Combination charts can control plot-area layering through a new ZOrder property on each plot group, so a column series can render behind or in front of a line series
• Worksheets and single cells can freeze formulas into their calculated values through ConvertFormulasToValues and ConvertFormulaToValue, keeping the cached result while dropping the formula, shared-formula link, and array metadata
• Defined names now carry an optional comment through the new Comment name property, round-tripped through the classic name-comment record and the XLSX defined-name comment attribute
• PDF export accepts custom paper dimensions in points through the new PaperWidth and PaperHeight properties, overriding the paper-size enum for every rendered worksheet
• PDF export can suppress the header or footer band through the new HidePrintObjects set
• PDF header and footer field codes &F (file name) and &Z (path) now expand using the source workbook path captured by file-based exports
• Data bars support an independent negative-value border colour through the new NegativeBorderColor and NegativeBarBorderColorSameAsPositive settings, emitted in the Excel 2010+ data-bar extension
• Built-in cell styles whose identifiers have no legacy record slot (Hyperlink, Title, Headings, Accents, Good/Bad/Neutral, …) now resolve their canonical display names on save, so the Excel Styles gallery survives a round-trip
• Agile-encrypted workbooks now carry a real dataIntegrity block: the writer emits a spec-conformant HMAC over the complete encrypted package, and opening verifies it, detecting a wrong password or a tampered package; files from writers without the block still open unchanged
• Classic workbooks support Excel's “password to modify” write reservation through SetModifyPassword (recommend read-only flag, legacy password hash, and reserving user name emitted as the WRITEPROT + FILESHARING record pair); IsWriteReserved reports the state and an empty password clears it
• An opt-in RecalcLookupCache memo accelerates lookup-heavy recalculation: repeated XLOOKUP/XMATCH scans over the same range reuse the first scan's result. The memo engages only when the dependency graph is provably reference-stable (no OFFSET/INDIRECT or unresolvable references, no circular chains, no What-If data tables) and stays silently off otherwise, so results are always identical
• A companion RecalcAggregateCache memo accelerates conditional-aggregate recalculation: repeated SUMIF/COUNTIF/AVERAGEIF scans over the same range and criteria reuse the first result, under the same reference-stability gate and with identical results
• LAMBDA expressions evaluate with full closures: a defined name whose RefersTo is a LAMBDA can be called by name with arguments (=MyFunc(5)), closures bind lexically inside LET (=LET(f, LAMBDA(x, x*2), f(21))), parameters shadow same-spelling column references, and the formula text round-trips verbatim; a bare LAMBDA stored as a cell result keeps the skip-without-value behaviour
• LET now has complete semantics: every binding is visible to later bindings and to the body expression, including nested LET composition — replacing the former partial implementation
• Formula text now accepts the comma as an argument separator alongside the classic semicolon, matching Excel's canonical syntax for pasted and imported formulas
• Structured references in formulas now evaluate: Table[Column], Table[[Column]], column spans, and the [#Data] / [#All] / [#Headers] / [#Totals] item forms resolve against the workbook's Excel tables at parse time, so =SUM(Table1[Amount]) calculates while the original formula text still round-trips verbatim; unresolvable references keep the previous skip-without-value behaviour and current-row forms ([@Column]) remain unsupported

Changed

• When a workbook disables full precision, recalculated results are now stored rounded to the number of decimal places shown by each cell’s number format, matching Excel’s “precision as displayed” behaviour for HotXLS-computed values
• Conditional-formatting rules written to XLSX now receive sheet-unique priorities, eliminating duplicate-priority files that made Excel prompt for repair
• Conditional formats flagged as PivotTable-scoped through the new IsPivot property are routed to the extension bucket instead of the worksheet’s main conditional-formatting block, where Excel would drop or misplace them

Version 2.273.0

Changed

• TXLSXWorkbook.PreserveRawParts now copies unchanged worksheet images, header and footer images, and chart parts with their original compressed payloads, while image data edits and chart model edits take the serialization path
• A successful file SaveAs now accepts the committed package as the next raw-copy baseline, allowing subsequent saves to reuse newly serialized opaque data, VBA, connections, external links, images, charts, themes, and pivot payloads without reopening the previous source package

Fixed

• Unknown worksheet and drawing relationships from packages with nonstandard part locations are now rebased against their generated canonical owners, including drawing-only packages whose only anchors are preserved unsupported shapes
• A failed compressed-member copy now rolls back its partial local header, payload bytes, parent folder entries, and central-directory count before safe serialization fallback, preventing unreachable data from accumulating in the output archive

Version 2.272.0

Changed

• XLSX SaveAs now opens one validated source archive for the complete save operation and copies every unchanged relationship-reachable opaque part with its original compressed payload, avoiding repeated central-directory scans plus decompression and recompression while modified payloads still take the serialization path

Fixed

• Unknown workbook relationships from packages whose workbook uses a nonstandard part name are now rebased against the generated workbook location, preventing their preserved target parts from becoming unreachable after SaveAs
• Projected direct reads now retain shared-formula master definitions outside the selected row or column set, so selected follower cells still receive their correctly translated formulas without rebuilding unselected values or rich text
• Editing a chart no longer removes an unmodelled axis number format merely because data-label number-format authoring uses an element with the same local name

Version 2.271.0

Fixed

• The classic HTML exporter now writes UTF-8 and declares it with a <meta charset> plus an HTML5 doctype; previously the stream was written in the system ANSI code page, corrupting any non-ASCII cell value, and the missing charset declaration left the page at the browser’s mercy

Version 2.270.0

Added

• Chart data labels can now carry an explicit number format and a placement through the new LabelNumFmt and LabelPos series properties (with their *Set flags), emitted as <c:numFmt> and <c:dLblPos> in both the standard and combination-chart writers — enabling percentage or currency label text and label positions such as outside-end or center; either setting alone also turns on the label block, and unset series stay byte-identical

Version 2.269.0

Added

• Line and scatter chart series can now carry an explicit marker symbol and size through the new MarkerSymbol, MarkerSize, and MarkerSet properties, emitted as the series’ <c:marker> block in both the standard and combination-chart writers; unset series keep automatic markers
• Value axes with display units can now show the unit label (“Thousands”, “Millions”, …) via the new ShowDispUnitsLabel axis property

Version 2.268.0

Added

• XLSX chart trendlines now support a fixed intercept plus forward and backward forecast periods through the new Intercept/InterceptSet, Forward, and Backward properties, emitted as <c:intercept>/<c:forward>/<c:backward>; previously only the classic BIFF chart builder offered these controls, and unset trendlines stay byte-identical

Version 2.267.0

Added

• Formula calculation now evaluates the complex-number functions IMCOS, IMSIN, IMTAN, IMSEC, IMCSC, IMCOT, IMLOG2, and IMLOG10, completing the IM* engineering family with trigonometric, reciprocal, and base-2/base-10 logarithm coverage over complex string operands

Version 2.266.0

Changed

• The streaming direct reader now applies column projection before accumulating cell values: when a projected column set is configured, cells outside it skip the value, formula, and rich-text string building that previously ran for every cell and was discarded at cell end, speeding up wide-sheet reads that project only a few columns

Version 2.265.0

Added

• Bar and column chart plot groups now expose a configurable grouping mode (clustered, stacked, percentStacked), a cluster gap width, and a series overlap through the new BarGrouping, GapWidth, and Overlap properties; previously every bar/column chart was hard-coded to clustered with no gap or overlap control, and unset groups stay byte-identical while the edit-after-open merge list treats the new elements as owned

Version 2.264.0

Fixed

• PDF cell borders now reflect the worksheet’s border style and weight instead of every border drawing as an identical 0.75 pt solid line: XLSX cells render thin, medium, thick, hair, and double weights and dashed, dotted, dash-dot, dash-dot-dot, and slant-dash-dot patterns, and BIFF cells render thin, medium, and dashed/dotted styles

Version 2.263.0

Added

• PDF export can now embed document metadata through the new DocumentTitle, DocumentAuthor, DocumentSubject, and DocumentKeywords properties; when any is set, the PDF carries a /Info dictionary with those fields plus a Creator/Producer identity and a CreationDate, while default exports with no metadata stay byte-identical

Version 2.262.0

Fixed

• CSV export no longer over-quotes fields: an internal space no longer forces quoting, so values like John Smith are written unquoted, matching RFC 4180 and Excel; fields with the delimiter, a double quote, a line break, or leading or trailing whitespace are still quoted as before

Version 2.261.0

Added

• Each chart series can now carry a solid fill color through the new SolidFillHex property, emitted as the series’ <c:spPr> shape properties; assigning a hex color stamps the series with a brand or palette color instead of Excel’s automatic theme color, while unset series stay byte-identical and the edit-after-open merge list treats spPr as owned so it replaces rather than duplicates

Version 2.260.0

Changed

• The sparse cell-storage back end now grows each 256-row block buffer geometrically (doubling) instead of reallocating on every appended cell, removing the quadratic copy cost when loading dense worksheets

Version 2.259.0

Fixed

• BIFF8 worksheet object-protection (OBJECTPROTECT) and scenario-protection (SCENARIOPROTECT) records are now read back into the sheet protection state, so the lock flags survive a round-trip instead of being written but lost on reload
• BIFF8 workbook-globals MMS, RefreshAll, and UseSelFS records are now preserved through round-trip instead of being silently dropped

Version 2.258.0

Added

• Formatted display text now recognizes [Cyan] and [Magenta] colour directives and the indexed palette directives [Color1] through [Color8], so HTML-style formatted output applies those colours instead of dropping them

Version 2.257.0

Added

• Formula calculation now evaluates the engineering and math functions FACTDOUBLE, SQRTPI, DELTA, GESTEP, SERIESSUM, and CONVERT, with CONVERT covering mass, distance, time, pressure, force, energy, power, magnetism, temperature, and volume units and rejecting mismatched unit categories

Version 2.256.0

Fixed

• Formatted display text now renders minutes correctly in time masks such as hh:mm, mm:ss, and [h]:mm:ss; previously Delphi’s mm (month) was used, so minute fields showed the month number
• Locale and currency directives such as [$-409] and [$$-1009] are now stripped from formatted output instead of leaking as a literal $-409 prefix, while a currency symbol present in the directive is preserved

Version 2.255.0

Added

• ZIP package reading now understands ZIP64 archives, parsing the extended-information extra field and the ZIP64 end-of-central-directory record and locator so workbooks whose parts or offsets exceed the 4 GB / 65535-entry classic limits open correctly instead of failing the end-of-central-directory sanity check

Version 2.254.0

Security

• Encrypted-workbook password verification now compares the verifier hash in constant time for both Agile and Standard Encryption, closing a timing side-channel that could reveal how many leading bytes of the password hash match
• Agile encryption no longer falls back to a time-seeded pseudorandom generator for salts, verifiers, and intermediate keys when the system CSPRNG is unavailable; it now fails closed instead of emitting predictable key material

Version 2.253.0

Added

• Formula calculation now evaluates the Excel 365 array-shaping functions VSTACK, HSTACK, TOROW, TOCOL, WRAPROWS, WRAPCOLS, TAKE, DROP, CHOOSECOLS, CHOOSEROWS, EXPAND, RANDARRAY, and TEXTSPLIT, spilling their results through the existing dynamic-array machinery
• Formula calculation now evaluates the modern text functions TEXTBEFORE and TEXTAFTER with instance-number, case-insensitivity, match-end, and not-found handling
• Formula calculation now evaluates RANDBETWEEN, ISEVEN, ISODD, and the FORECAST.LINEAR alias of FORECAST

Version 2.252.0

Added

• Defined names carry an optional comment, round-tripped as the BIFF8 NAMECMT record and the OOXML definedName comment attribute
• Local-file hyperlink targets are normalized (backslash-to-slash conversion, file:// and file:/// prefixes, percent-encoding of reserved characters) so Excel stores them verbatim
• Shape, picture, and text-box fill kinds (solid, picture, gradient, none, automatic) are now typed over the raw fill-type property and round-trip faithfully
• Conditional-format data bars support Excel 2010+ independent negative-value bar border colors
• Pivot cache fields round-trip multi-level date and numeric grouping linkage
• User-function callbacks receive an argument-kind descriptor so handlers can distinguish literal values, cell or range references, and array constants
• Formula compilation accepts localized function names (German, French, Spanish), rewritten to canonical English before lexing
• Password and encryption helpers expose named return-code constants

Changed

• Package and XLSX saves now flow through larger deflate buffers with a higher compression memory level and skip a redundant checksum pass over already-compressed bytes, reducing save CPU and wall-clock time
• Internal string-builder, list, and hash-table growth now uses geometric (doubling) allocation together with an earlier hash resize point, lowering allocation churn and lookup cost when a workbook carries many styles, names, defined ranges, or shared strings
• Shared-string table lookups now compare the stored bytes directly before materializing a full string, cutting lookup allocations on large shared-string tables

Version 2.251.0

Fixed

• XLSX open and save now retain foreign workbook-root namespace declarations, markup-compatibility attributes, complete workbook-level <extLst> payloads, and their unknown relationships while regenerating typed workbook state; preserved r:id references follow relationship-ID reassignment instead of becoming stale when they conflict with generated worksheet or workbook relationships

Version 2.250.0

Changed

• XLSX sparse row entries now store only their column and stable cell handle, eliminating unused payload tags plus redundant classification and a second row-index lookup whenever a cell value, formula, or rich-text object changes

Version 2.249.0

Added

• TXLSRTFImport.LoadFromRTF and TXLSWorksheet.OpenRTF overloads can import tables or paragraph lists at an explicit one-based target row and column while the original overloads continue to start at A1

Changed

• XLSX typed row writers now append strictly forward batches directly to pre-sized sparse row vectors, and sparse cell lookup resolves first-column and last-column hits without binary search while overlap writes retain their existing replacement path and stable cell handles

Fixed

• File-backed XLSX raw-part copying and selective sheet loading now use a private source-archive identity with size and timestamp validation, so a later stream open or a public FileName assignment cannot redirect preserved parts to another workbook
• RefreshPivotCache now switches refreshed caches to typed serialization so rebuilt records cannot be overwritten by stale raw package bytes during save
• Raw compressed ZIP copying now requires the source stream to contain exactly the declared compressed byte count, rejecting both truncated and overlong entries before central-directory registration
• Reusing one RTF importer now resets code-page and character-format parser state between documents so bold, italic, underline, strikeout, and font size cannot leak into the next import
• MetadataOnly XLSX opens now avoid shared-string, person, worksheet-relationship, comment, drawing, chart, table, query-table, pivot, and worksheet-reachable opaque payloads while retaining workbook properties, names, sheet states, styles, theme, connections, and external links

Version 2.248.0

Changed

• XLSX SaveAs now raw-copies unchanged connections.xml parts from the source package when the parsed connection model is unchanged, preserving source XML fidelity and avoiding unnecessary decompression and recompression while still rewriting edited connections from the typed model

Version 2.247.0

Changed

• XLSX SaveAs now raw-copies unchanged external-link XML and relationship parts from the source package when the parsed model is unchanged, preserving unknown extension XML and avoiding unnecessary decompression and recompression while still rewriting edited links from the typed model

Version 2.246.0

Added

• TXLSHTMLImport.LoadFromHTML overloads can import at a one-based target row and column, select one zero-based top-level table or all tables, and parse numbers with an explicit decimal separator without changing global format settings

Changed

• XLSX style loading now resolves each cell style through a shared immutable cellXf array instead of repeated string-key construction and linear map searches, reducing styled-workbook open time and temporary allocations

Fixed

• Modified VBA project bytes now bypass preserved raw-part copying so saving an edited macro-enabled workbook cannot restore the original vbaProject.bin payload
• Raw ZIP-member copying now rejects compression methods other than STORE and DEFLATE instead of emitting a mismatched compression header, allowing callers to use safe decompression and rewrite fallback paths
• HTML import now ignores row and cell tags outside top-level tables and preserves the established all-table behavior of the original overloads

Version 2.245.0

Added

• Editable TXLSWorkbookViewer cells with protected-cell-aware input, formula entry and recalculation, rectangular range selection, formula-reference picking, and bounded circular undo and redo history
• TXLSPivotDesigner provides a reusable non-visual field-list model with row, column, filter, and value zones, deterministic field movement, aggregation selection, layout validation, and result application
• Classic XLS workbooks now provide CalculateAndVerify diagnostics with shared mismatch records containing worksheet identity, one-based cell coordinates, cached and freshly calculated values, and formula text
• Classic XLS and XLSX worksheets now provide typed numeric, string, and variant row and column writers with one-based coordinates, grid-boundary clipping, direct cell-model writes, and a returned write count

Changed

• XLSX viewer painting now preserves sparse worksheets by avoiding blank-cell materialization during repaint and style resolution
• TXLSXWorksheet.ReplaceText now visits stored cells directly in linear time without copying and sorting the complete cell collection

Fixed

• Classic XLS RC4 and RC4 CryptoAPI read and write now preserve every plaintext record body required by [MS-XLS], keep BoundSheet8.lbPlyPos unencrypted, consume matching keystream positions, and activate decryption before reading the first encrypted record after FILEPASS
• SpreadsheetML serialization now applies the symmetric _xHHHH_ codec only to Xstring fields, while formulas and reference syntax use ordinary XML escaping so token-shaped literals retain their exact formula meaning

Version 2.244.0

Added

• Multi-format clipboard import: XLSPasteFromClipboardRich prefers the XML Spreadsheet clipboard format (carrying formulas, styles, and merges) over plain Unicode text, with XLSClipboardHasXmlSpreadsheet, XLSGetClipboardXmlSpreadsheet, and XLSParseXmlSpreadsheet for format detection, extraction, and SpreadsheetML 2003 parsing

Version 2.243.0

Added

• External defined names: TXLSXExternalLink now parses, stores, and round-trips <definedNames> entries (name + refersTo) from external workbook links, with AddDefinedName, FindDefinedName, and indexed DefinedNames[] access
• Per-sheet refresh failure state: each TXLSXExternalSheetData carries a RefreshError flag that records whether the last refresh attempt failed, parsed from and emitted to the refreshError attribute

Version 2.242.0

Added

• TXLSXWorkbook.PreserveRawParts copies the compressed representation of unchanged parts (theme, VBA project, pivot tables, and pivot caches) directly from the source archive during SaveAs, avoiding decompression and recompression of preserved opaque entries

Version 2.241.0

Added

• TXLSXWorkbook.SelectedSheets property enables selective worksheet loading: Open parses cell XML only for sheets named in the list, while other sheets keep their names and visibility states with zero cells. Call LoadSheet(Index) to materialize a skipped sheet on demand from the original archive

Version 2.240.0

Added

• TXLSXWorkbook.MetadataOnly open mode parses workbook properties, defined names, sheet names and visibility states, styles, theme, connections, and external links, but skips all worksheet cell XML and satellite parts for fast property and name discovery from large workbooks

Version 2.239.0

Added

• Compact 256-row block cell storage with stable public handles, row-major cache-friendly serialization, and style compaction across all workbook pools before every save
• Interval style overlays let row, column, and rectangle formatting stay compact without materializing empty cell objects, resolve lazily on cell access, survive structural edits, and round-trip as compact column, row, and style-only cell entries

Changed

• Normal workbook save now streams each worksheet row directly into the package deflate stream, so peak memory follows the largest single row rather than the complete worksheet XML
• Shared strings and auxiliary XML parts now stream one entry at a time through a single reusable UTF-8 serializer, bounding peak memory by the largest entry instead of the whole part

Version 2.238.0

Added

• TXLSPaginationModel builds reusable logical pages for XLS and XLSX ranges, worksheets, and workbooks with paper geometry, margins, headers, footers, row and column bands, repeated titles, page order, gridline policy, and drawing placements
• TXLSWorkbookViewer now provides shared-model print preview, keyboard and scroll-wheel page navigation, cached PaginationModel access, explicit RefreshPagination, and selected-page EMF output
• TXLSPDFExport.CreatePaginationModel exposes the exact owned page model used by PDF counting and rendering for all six XLS and XLSX input types

Changed

• PDF export and page counting now consume one pagination engine instead of maintaining separate XLS and XLSX page loops
• Fit-to-width and fit-to-height scaling now solves against actual page bands, including manual breaks and repeated titles, within Excel's 10% through 400% scale range

Version 2.237.0

Added

• TXLSFormulaDebugger creates an owned interactive session containing the parsed expression tree, source spans, operators, operand nodes and values, intermediate results, deterministic evaluation order, and navigable current-step state
• Formula-debug sessions expose cell and range dependencies with resolved sheet coordinates, report selected and skipped branches for IF, IFERROR, IFNA, CHOOSE, IFS, and SWITCH, and preserve recovered inner errors with their originating node

Changed

• Workbook and worksheet Calculate now return Excel formula failures as standard Variant error values instead of converting them to Null
• TXLSFormulaTracer now records exact before-and-after operand stacks for push and reduction steps

Version 2.236.0

Added

• TXLSXWorksheet.StructuralEditRegistry now lets application-owned worksheet metadata participate in row and column insertions or deletions through IXLSXStructuralEditParticipant

Changed

• XLSX structural edits now complete their reference-sensitive work in one guarded transaction covering formulas, names, validations, conditional formats, comments, hyperlinks, drawings, tables, queries, scenarios, data tables, print settings, sparklines, pivot caches and views, and worksheet extensions

Fixed

• Inserted or deleted rows and columns now update protected ranges, threaded-comment anchors, reopened classic-comment VML, local and cross-sheet sparkline references, typed pivot sources and anchors, calculated pivot formulas, and recognizable references inside preserved worksheet extension XML

Version 2.235.0

Added

• TXLSDirectWriter now adds classic comments, anchored worksheet images, formula-backed column, bar, line, area, pie, doughnut, and scatter charts, and complete caller-supplied ChartML without constructing a full workbook object model
• Images can be supplied as bytes, seekable streams with optional ownership transfer, or file paths and can use PNG, JPEG, GIF, BMP, EMF, or WMF content

Changed

• Rich direct-writer parts use current-sheet side-channel descriptors, assign worksheet relationship identifiers before the sheet closes, and stream media and related XML parts after worksheet XML to keep row memory independent of sheet size

Version 2.234.0

Added

• TXLSDirectReader now exposes opt-in dimension, sheet-format, column, row, pane, and merged-range events with typed layout records
• Column and row events report widths or heights, hidden and collapsed state, outline levels, and source style indexes without constructing worksheet cell objects

Changed

• Assigned layout callbacks use a bounded-memory metadata pass before cell delivery, ensuring merged ranges and other layout facts arrive before cells that depend on them while ordinary cell-only scans retain the original single-pass path
• Parallel direct reads keep layout callbacks on the calling thread and finish selected-sheet metadata before draining ordered worker cell queues

Version 2.233.0

Added

• TXLSXChartTextStyle now provides typed chart-title, legend, axis-title, and axis-label font names, point sizes, RGB colours, emphasis, and text rotation
• TXLSXChartLayout now provides normalized manual plot-area, chart-title, axis-title, and legend positioning with explicit edge or factor modes and inner or outer plot targets
• Charts now expose typed legend visibility, position, overlay, text style, and layout controls together with chart-title overlay settings

Changed

• Manual layouts always serialize complete x, y, width, and height groups using Office-compatible default modes so Excel does not discard partial or conflicting coordinates

Fixed

• Editing typed chart fonts or layouts preserves unrelated extension lists and custom namespace content while clearing a known colour or emphasis property no longer restores the imported value during structural merge

Version 2.232.0

Added

• TXLSXImage now exposes percentage-based crop edges, transparency, brightness, contrast, rotation in degrees, and stretch or tile fill modes
• TXLSXImage.HasPreservedXml reports whether imported DrawingML is available, while ClearPreservedXml opts one picture into clean typed-model regeneration

Changed

• Image effect values use compact fixed-point storage and write native DrawingML units without allocating per-effect helper objects

Fixed

• Editing a supported property on an imported image now preserves unrelated extensions, markup-compatibility choices, custom attributes, shape effects, and namespace-qualified content

Version 2.231.0

Added

• Imported XLSX charts now preserve their original ChartML while TXLSXWorkbook.PreserveUnsupportedParts is enabled, retaining unknown namespaces, markup-compatibility alternatives, extension lists, custom styles, and unmodelled formatting
• TXLSXChart.HasPreservedXml reports whether original ChartML is available, and ClearPreservedXml lets applications opt one chart into clean typed-model regeneration

Changed

• Saving an untouched imported chart replays its original ChartML, while edits to supported properties structurally overlay only the modelled chart nodes and keep unrelated content in place

Fixed

• Editing a chart title, series, axis, or other supported property no longer discards unrelated Excel-authored extensions, alternate content, styles, or formatting from the chart part

Version 2.230.0

Added

• TXLSXChart.AddPlotGroup now builds ordered combination charts whose chart families and series retain independent primary or secondary axis assignments
• Charts now expose primary and secondary category and value axes plus a series axis, with typed category, value, date, and series-axis settings including positions, identifiers, crossings, date units, label intervals, and visibility
• Column, bar, line, pie, and area charts now have dedicated 3D chart types with configurable rotation, height, depth, perspective, right-angle axes, and gap depth
• Series now expose typed string and numeric caches for names, categories, values, and bubble sizes, including sparse point indexes and numeric format codes

Changed

• The compatibility ChartType, AddSeries, and flat Series APIs continue to address the primary plot group while the chart model retains every ordered plot group

Fixed

• Opening and resaving combination charts no longer collapses them into one chart family or loses secondary, date, or series axes
• Chart formula references and cached series values now survive workbook copies and open/save cycles without being flattened or discarded
• Chart anchors are now matched to their chart parts by relationship identifier, so applications that reorder relationship entries no longer swap chart contents between worksheet positions

Version 2.229.0

Added

• TXLSXWorksheet.HeaderFooterImages now creates, finds, replaces, and removes typed image slots for every left, centre, and right header or footer position on odd, even, and first pages
• Header and footer images expose their raw data, format, title, point dimensions, and point margins, with direct loading from PNG, JPEG, GIF, BMP, EMF, and WMF files

Changed

• Even-page and first-page images automatically enable the matching page variants and maintain the required picture field codes when header or footer text is edited

Fixed

• Worksheet comments and header or footer images now retain independent VML relationships when they share one worksheet
• Excel-authored header and footer image slots now preserve their image bytes, formats, titles, dimensions, margins, and relationships through open and save cycles

Version 2.228.0

Added

• Classic XLS worksheets and shape collections now create and discover typed TXLSButton and TXLSRadioButton controls alongside check boxes, combo boxes, and list boxes
• Form controls now expose macro names, linked cells, input ranges, check states, flat or 3D appearance, and radio-group links through typed properties
• Defined names expose IsMacro so macro bindings retain the required workbook-name metadata

Fixed

• Editing and resaving classic form controls no longer discards check states, macro bindings, linked-cell formulas, source ranges, or supported radio-group metadata
• Clearing a combo-box or list-box source range now removes the serialized input formula instead of restoring the previous range on the next open

Version 2.227.0

Added

• LINEST and LOGEST now return complete multivariable coefficient arrays and optional five-row statistics matrices
• TREND and GROWTH now return shaped prediction arrays for single-variable and multivariable inputs

Changed

• Array arithmetic, comparisons, concatenation, unary negation, and percent operations now broadcast scalars and singleton rows or columns while retaining per-cell results
• Calculation dependencies now expand nested workbook and worksheet names, formula-defined names, and multi-area references with scope-aware cycle protection

Fixed

• An error in one array element no longer collapses the complete result or incorrectly falls back to scalar evaluation
• Formula-defined names whose dependencies can be resolved statically no longer force unnecessary volatile full recalculation

Version 2.226.0

Added

• Existing XLSX pivot caches and pivot-table views now load into the typed model through package relationships, including cache records, fields, shared items, grouping, filters, formats, calculated definitions, and extensions
• TXLSPivotTable.Make now evaluates calculated cache fields, calculated items, and data calculated members with dependency-aware name binding
• Extended ShowDataAs modes add parent percentages, running percentages, and ascending or descending rank with valid XLSX extension markup

Changed

• Pivot result generation now applies item visibility, page selections, caption, date, aggregate, and value filters before building grouped domains
• Numeric, date, and discrete groups, label or data-value sorting, configured subtotals, and optional materialized grand-total rows or columns now participate in result generation

Fixed

• Pivot filters now use schema-valid root-level filter markup and symbolic filter types instead of ambiguous numeric values
• Relationship-driven loading no longer assumes conventional or sequential pivot cache, cache-record, or pivot-table part names

Version 2.225.0

Added

• TXLSXWorkbook.Connections and TXLSXWorksheet.QueryTables now provide typed authoring and editing for external database, web, text, and query-result metadata
• Connection parameters preserve their exact Boolean, integer, floating-point, string, or cell-reference kind, including explicit zero, false, and empty-string values

Changed

• XLSX open and save now resolve and write connection and query-table parts through their owning relationships, synchronize result tables and defined names, and preserve unknown attributes, child nodes, and extension lists
• Row and column edits keep query-result ranges aligned, while same-workbook and cross-workbook worksheet copies preserve query ownership and remap connection identifiers and conflicting names

Fixed

• Query tables can coexist with standard tables, threaded comments, and pivot tables without worksheet or workbook relationship identifier collisions

Version 2.224.0

Added

• Classic XLS and modern XLSX worksheets now expose AddDataTable and DataTables for one-variable row, one-variable column, and two-variable What-If data tables

Changed

• Workbook recalculation evaluates each data-table result with an isolated input context, including indirect formula dependencies, and preserves definitions plus cached results through open, save, worksheet copy, and structural row or column edits

Fixed

• Deleted What-If input cells now produce #REF! results, and classic workbook recalculation visits every worksheet instead of omitting the first sheet
• Inserting rows at the top of a classic worksheet no longer risks an access violation on Win64

Version 2.223.0

Changed

• Parallel direct-reader cancellation now propagates through active ZIP member reads and every decompression step instead of waiting for the next XML pull

Fixed

• Aborting a parallel worksheet scan no longer waits for a large compressed XML read to complete before blocked workers are released

Version 2.222.0

Added

• TXLSDirectReader.ParallelBufferMemoryLimit bounds aggregate raw-cell queue payload for parallel worksheet scanning, while ParallelPeakBufferedBytes exposes the measured peak for diagnostics and capacity tests

Changed

• Parallel direct reads now stream worksheet ZIP entries into worker-local XML readers and bounded producer-consumer queues instead of retaining complete inflated worksheet XML and complete per-sheet cell buffers
• Active worksheets receive independent queue budgets and are processed in worker-sized windows, preserving callback order without allowing later worksheets to starve the worksheet currently being emitted

Fixed

• Cell aborts, callback failures, and worker errors now wake blocked producers and consumers, stop XML pulls cooperatively, join all workers, and leave the direct reader reusable

Version 2.221.0

Changed

• Multi-sheet XLSX opening now streams each worksheet ZIP entry directly into its worker-local inflater and XML reader instead of retaining inflated worksheet XML before parsing
• Concurrent worksheet readers serialize only short positioned reads from the package source, so decompression and XML parsing remain parallel while transient worksheet input memory is bounded by the active worker count

Version 2.220.0

Added

• PivotTable result generation now executes all nine classic ShowDataAs modes, including difference, percent of a base item, percent difference, running total, percent of row, percent of column, percent of total, and index calculations
• xlPivotBaseItemPrevious and xlPivotBaseItemNext provide typed selectors for adjacent base items in comparison calculations

Changed

• Pivot percentages and index values use row, column, and grand totals aggregated directly from source records, preserving correct Average, Product, Count, variance, and standard-deviation semantics without rescanning the source
• Running totals follow pivot-item order, reset independently for every peer group, and retain sparse result storage for high-cardinality pivots

Fixed

• BIFF8 SXDI records now use the standard 14-byte fixed layout and preserve ShowDataAs, base field, base item, number format, and display name through typed reads, while legacy HotXLS records remain readable
• XLSX data fields now write pivot-field indexes, map previous and next selectors to the required OOXML values, and emit base attributes only for modes that use them
• Missing comparison cells remain blank, while present zero denominators return #DIV/0! instead of being confused with missing data

Version 2.219.0

Fixed

• VDB now applies variable declining-balance depreciation across fractional periods, switches to straight-line depreciation when beneficial, and honours no_switch
• Zero factors, salvage boundaries, fractional starting periods, and invalid period ranges now follow Excel-compatible results instead of rounding periods and returning a simplified declining-balance approximation

Version 2.218.0

Added

• TXLSXWorksheet.Scenarios now provides typed What-If Scenario Manager authoring with named scenarios, active and shown indexes, changing-cell ranges, user and comment metadata, locked and hidden states, and owned input-cell values
• Scenario definitions now follow worksheet copies and row or column structural edits, while Excel compatibility limits are validated before invalid metadata can be written

Fixed

• Opening and saving Excel-authored scenarios now preserves root attributes and nested input cells instead of flattening the hierarchy into lossy raw XML
• Scenario, AutoFilter, data-consolidation, merged-cell, and phonetic worksheet elements now use schema order so strict spreadsheet consumers do not require a repair pass

Version 2.217.0

Added

• TXLSXWorksheet.IgnoredErrors now provides typed authoring for every standard ignored-error category with multi-area references, worksheet-copy support, structural-edit updates, and open/save round trips
• TXLSXWorkbook now exposes reference mode, recalculate-on-save, full-precision, concurrent-calculation, and manual concurrency settings alongside its existing calculation options

Fixed

• Every non-default workbook calculation setting now independently produces and round-trips a schema-ordered calcPr element, including iteration limits configured before iteration is enabled
• Calculation modes, reference modes, iteration bounds, calculation identifiers, and manual concurrency counts now reject or normalize invalid values instead of writing ambiguous workbook state

Version 2.216.0

Changed

• XMATCH and XLOOKUP now perform logarithmic ascending or descending binary searches when requested, including Excel-compatible duplicate and approximate-match boundaries
• XIRR now honours its guess with a safeguarded solver that can find high and guess-sensitive roots, while XNPV uses compensated summation and validated date and rate inputs
• High-cardinality XLSX defined-name lookup now uses a lazy scope-aware hash index while retaining serialized order, local-name precedence, and first-match behaviour
• Preserved XLSX package parts now allocate their declared uncompressed size once while retaining complete length and CRC verification

Fixed

• Invalid lookup modes, wildcard binary searches, two-dimensional lookup ranges, and mismatched XLOOKUP return vectors now report formula errors instead of producing ambiguous results
• XIRR and XNPV now reject invalid cash-flow signs, rates, guesses, and dates, truncate fractional date serials, and report non-convergence instead of returning an unverified midpoint

Version 2.215.0

Added

• FREQUENCY now returns a vertical array with one result per bin plus the overflow bucket

Changed

• SORT now uses a stable scalable ordering for rows or columns, preserving the source order of equal keys
• UNIQUE now uses stable hash-indexed matching for rows or columns, including exactly_once results and case-insensitive text
• PERCENTRANK.INC, PERCENTRANK.EXC, and MODE.MULT now scale efficiently on high-cardinality input instead of relying on quadratic scans

Fixed

• The by_col option of SORT and UNIQUE now operates on columns instead of falling back to row behaviour

Version 2.214.0

Added

• XLSX opening now accepts streaming UTF-16 little-endian and big-endian XML parts with a byte-order mark or the required endian-specific declaration, including characters outside the Basic Multilingual Plane

Changed

• High-cardinality PivotTable cache refreshes now use deterministic hash indexes for shared values and allocate the record matrix once, preserving first-seen item order while avoiding repeated linear searches and row-by-row growth
• Projected direct reads discard cells outside the selected rows and columns before copying raw values, formulas, and rich text into per-sheet buffers

Fixed

• Malformed XML with duplicate byte-order marks, unsupported or contradictory encoding declarations, UTF-32 signatures, odd UTF-16 byte lengths, or invalid surrogate pairs is now rejected
• Parallel direct-reader workers now initialize all job state before starting, stop taking new sheet jobs after a failure, re-raise the lowest-index job error deterministically, and clean up every created worker if pool construction fails

Version 2.213.0

Added

• T.TEST now evaluates independent samples with either pooled equal variance or Welch unequal variance in addition to paired samples, including one-tailed and two-tailed probabilities

Changed

• XLSX opening now follows workbook, worksheet, and drawing relationship targets for worksheets, shared strings, styles, themes, macros, external links, comments, threaded comments, VML, drawings, images, charts, tables, and PivotTables, including valid packages that use renamed or non-contiguous part paths

Fixed

• Malformed packages with duplicate relationship identifiers no longer allow a later unsupported relationship to replace the first usable target during compatibility recovery

Version 2.212.0

Changed

• Formula dependency tracking now scales better on large worksheets by indexing cell and range dependencies before rebuilds and dirty-cell invalidation
• Large statistical and regression formulas now evaluate every numeric value in the supplied ranges instead of stopping at the previous fixed-size buffer

Fixed

• XLSX cells outside the worksheet grid are rejected consistently so invalid coordinates can no longer alias valid cells
• Editing formulas no longer resets workbook calculation mode, iteration limits, full-calculation flags, or other calculation settings
• Blob-list reads now validate requested ranges explicitly and allow valid empty slices at the end of the stream

Security

• Classic XLS encryption now covers pivot cache streams along with worksheet data
• Malformed SST extension tails and ZIP central-directory records are rejected before oversized allocation, truncated reads, or comment-embedded fake directory records can affect loading

Version 2.211.0

Added

• TXLSPivotTable.Make now materializes every configured data field with its own aggregation and expands the values on rows or columns according to DataFieldAxisIsRow

Changed

• Pivot result generation reuses hash-indexed row and column key domains across all data fields, selects dense or sparse accumulator storage from the result shape, preserves cache item order beyond nine keys, and emits readable combined labels for multi-field axes

Fixed

• Count now includes non-empty text, Boolean, error, date, and numeric cache values while CountNums includes only numeric and date values, Product returns zero for intersections without numeric values, blanks are no longer counted as entries, and values-only pivots can materialize their grand totals without row or column fields

Version 2.210.0

Added

• TXLSWorkbookViewer now renders XLSX conditional fills, font color and emphasis, number formats, directional positive and negative data bars, gradients, borders, axes, hidden values, and all built-in icon-set families
• TXLSWorkbookViewer.RefreshConditionalFormats rebuilds prepared aggregate and threshold state after callers modify an attached XLSX workbook in place

Changed

• The viewer caches one prepared conditional-format evaluator per active worksheet, reuses it across scrolling, selection, and repaint operations, and releases it when the workbook or worksheet changes

Fixed

• Data-bar axes at the rightmost boundary are now clipped inside the owning cell instead of drawing into the adjacent cell

Version 2.209.0

Added

• XLSX worksheets now expose a prepared conditional-format evaluator that resolves rule priority, StopIfTrue, relative formulas, differential styles, Top/Bottom ranking, above/below-average tests, duplicate and unique values, relative date periods, two- and three-color scales, data bars, and icon sets for any cell

Changed

• HTML and PDF exports now render matching conditional fills, font emphasis and color, number formats, color scales, directional positive and negative data bars, explicit axes, icon-family glyphs, hidden values, and theme colors with tint resolution
• Large conditional-format domains now reuse compiled relative formulas without per-cell syntax-tree allocation, stream mean, deviation, minimum, and maximum statistics, retain ordered values only for ranking or percentile rules, batch-sort Unicode duplicate keys, and use row-band candidate indexes to avoid scanning unrelated rules

Fixed

• Classic conditional rules now load expression types correctly and evaluate reversed between bounds, case-insensitive text predicates, blank cells, and error predicates consistently

Version 2.208.0

Changed

• The packed cell storage engine now writes native tagged string, Boolean, error, and blank payloads, iterates cells in row-major order, and run-length encodes consecutive equal-typed cells for lower memory use and faster sequential output
• Classic workbook shared strings now keep character bytes in one allocation and create decoded string values lazily, reducing allocation overhead for large unique-string tables

Added

• Pivot tables can now compute a complete result matrix from cache records with row and column key grouping and all 11 supported aggregate functions
• Classic workbook charts now expose typed series, axis, legend, and chart-group information for conversion and inspection workflows

Version 2.207.0

Added

• TXLSPivotCache.RefreshFromSource builds a pivot cache from a source worksheet range: it reads the header row to name the fields, distils each data row’s distinct values into the per-field shared-item domain, locks the field type from the first non-empty value, and records the shared-item index for every (record, field) — so a pivot table can be created programmatically from a data range instead of relying on Excel’s refreshOnLoad. A callback-based value reader keeps the pivot unit free of a direct worksheet dependency

Version 2.206.0

Added

• New TlxCellMMU (lxCellMMU unit) implements a packed, row-major cell storage engine with run-length encoding of equal-typed and empty cell runs, and a tagged-byte type code that doubles as the data width selector — so a column of small integers costs one byte per cell instead of eight. The engine encodes numbers in the smallest width that preserves the value (int8 / int16 / int32 / 2-decimal / 4-decimal scaled integers / IEEE double), stores string-index, boolean, error, and blank cells compactly, and exposes a GetValue / SetNumber / SetString API plus a forward iterator. This is the next-generation backend behind a feature flag for the existing TXLSXCells API, targeting a 10–25× working-set reduction on dense sheets

Version 2.205.0

Added

• TXLSCustomChart.GetChartInfo now decodes a semantic summary from the raw BIFF chart record stream — chart type family (bar, line, pie, area, scatter, radar, doughnut, surface, bubble), the series count, and legend presence — so a BIFF chart’s type can be queried programmatically and a BIGH-to-XLSX chart conversion can pick the right target family. The byte-passthrough storage that guarantees round-trip fidelity is unchanged

Version 2.204.0

Changed

• The classic .xls shared-string table now writes ASCII-only strings in the compressed Latin-1 encoding ([MS-XLS] §2.5.294 fHighByte=0, one byte per character) instead of always using UTF-16, roughly halving the SST size — and therefore the .xls file — for ASCII-heavy workbooks

Version 2.203.0

Added

• New TlxCompoundFile reads and writes the Compound File Binary (OLE2) container format directly in pure Pascal, with no ole32.dll / COM IStorage dependency — opening the door to in-memory and cross-platform .xls handling. The reader covers the v3 layout HotXLS meets in the wild (header + in-header DIFAT + FAT chain walk + MiniFAT and the mini-stream for streams under 4096 bytes + the directory tree); the writer emits a minimal-but-valid v3 container sufficient for HotXLS’s output

Version 2.202.0

Added

• The autofilter sort state now has a typed model (TXLSXSortState with TXLSXSortCondition children) covering sort-by value / cell color / font color / icon, sort method (none / stroke / pinYin), case sensitivity, column sort, and per-condition custom lists — so applications can build sort states programmatically instead of editing raw XML, while the legacy opaque SortStateXml passthrough still wins for files HotXLS did not author
• Dynamic filters are now read and written as a typed TXLSXDynamicFilterType (above/below average and std-dev, today/yesterday/tomorrow, this/last/next week, month, quarter and year, year-to-date, Q1–Q4, M1–M12) with optional val / maxVal bounds — opening a workbook that uses a dynamic filter no longer drops it on save

Version 2.201.0

Added

• Surface and 3-D surface charts (xlsxChartSurface / xlsxChartSurface3D) are now first-class chart types — written, read, and round-tripped like the other chart families
• The autoshape authoring API covers a practical subset of the OOXML preset-geometry list — triangles, diamonds, parallelograms, hexagons, octagons, arrows, chevrons, callouts, flowchart nodes, stars, hearts, sun, moon, brackets and braces — each mapping to its correct prst token and [MS-ODRAW] shape id, so authored shapes render correctly in Excel (the full 187-preset list was already preserved on round-trip via raw-anchor passthrough)

Version 2.200.0

Fixed

• Opening a workbook whose autofilter uses a color filter or an icon filter no longer drops that filter on save — colorFilter and iconFilter are now read into typed columns (TXLSXAutoFilterColumn.SetColorFilter / SetIconFilter) and re-emitted verbatim, instead of being lost as unrecognised filter-column children

Version 2.199.0

Added

• TCondFormatRule.Evaluate answers whether a conditional-formatting rule fires for a given cell, covering cell-value comparisons, formula expressions, text contains/begins/ends, blanks, and error rules — the foundation for querying which cells a rule currently highlights and for rendering CF results outside Excel

Changed

• FACT and the internal factorial helpers now use a precomputed 0!..170! table instead of recomputing an O(n) product on every call, speeding up FACT, multinomial coefficients, and the Poisson/Binomial/chi-square distribution paths

Version 2.198.0

Changed

• Saving a worksheet now groups cells into rows with a single count-then-fill pass instead of growing each row’s index with a realloc-per-cell, removing a hidden O(cells × cells-per-row) copy on large sheets
• The formula dependency graph now indexes its reference ranges by (sheet, row) so seeding a dirty cell binary-searches the candidate band instead of scanning every formula on every changed cell — the bigger the workbook and the more sheets it has, the larger the speedup on recalculation
• Row-major, column-sorted cell traversal is now exposed as TXLSXWorksheet.BuildRowMajorCellOrder so save, clipboard, autofilter and other row-ordered consumers share one traversal instead of re-deriving it

Version 2.197.0

Fixed

• Radar, stock, and pie-of-pie/bar-of-pie charts are no longer lost when an .xlsx workbook that contains them is opened and re-saved — the chart reader now recognises the c:radarChart, c:stockChart, and c:ofPieChart container elements it already writes, restoring symmetric read/write coverage
• SUBTOTAL with function codes 101–111 now excludes values on hidden rows, and AGGREGATE now honours its ignore-hidden-row option codes (2, 3, 6, 7), matching Excel instead of treating the include-hidden and ignore-hidden variants identically
• Calling a fixed-arity built-in function with more arguments than declared (for example =SIN(1,2)) now raises #VALUE! centrally instead of being accepted by the dispatcher

Version 2.196.0

Fixed

• Row and column insertion or deletion now partitions cell-value and expression conditional formats plus data validations when edited target areas require different relative formula anchors, while preserving rule settings, assigning unique conditional-format priorities, and keeping aggregate conditional-format domains intact

Version 2.195.2

Changed

• PDF export now subsets the embedded TrueType font via the Windows fontsub.dll CreateFontPackage API, keeping only the glyphs the document actually rendered instead of embedding the whole .ttf — typically shrinking the PDF 8–20× for Unicode-heavy sheets; falls back to full embedding when subsetting is unavailable

Version 2.195.1

Fixed

• Inserting or deleting rows and columns on the classic .xls engine now rewrites every affected formula reference (cell formulas and defined names) to match Excel semantics — references below an insertion shift, references fully inside a deletion become #REF!, and partially overlapping areas shrink — instead of leaving stale coordinates

Version 2.195.0

Changed

• Deleting table columns now converts every affected structured-reference token to #REF! across worksheet formulas, table formulas, defined names, and chart series while preserving unrelated references and string literals

Fixed

• Deleting an entire table no longer leaves qualified structured references pointing to a table that no longer exists
• Cloned data validations now retain their IME input mode

Version 2.194.0

Changed

• Row and column edits now keep Excel table column identifiers, names, calculated formulas, totals formulas, filter columns, filter ranges, and sort conditions aligned with the edited table

Fixed

• Conditional-format and data-validation formulas now rebase relative and mixed references when deletion removes their first target area
• Worksheet and table sort states now write valid attributes and discard conditions whose ranges were deleted

Version 2.193.0

Changed

• Row and column edits now update complex and cross-scope defined-name formulas, bubble-chart data ranges, and formula-based visual conditional-format thresholds
• Bubble and scatter charts now restore their X-axis, Y-axis, and bubble-size series ranges when reopened

Fixed

• Multi-area conditional-format and data-validation targets now shift each surviving area independently instead of leaving the complete target list unchanged
• Text box and WordArt anchors now move or resize with inserted and deleted rows and columns

Version 2.192.0

Added

• Streaming direct reads and row-cursor traversal can opt into structured rich strings with complete run formatting, phonetic runs, and phonetic properties while retaining bounded shared-string storage
• TXLSDirectWriter.WriteRichString writes formatted and phonetic text as inline or deduplicated shared strings

Changed

• Rich shared-string deduplication now distinguishes run formatting and phonetic metadata while keeping ordinary string interning on its existing fast path

Fixed

• Phonetic text in inline strings is no longer included in the displayed cell value

Version 2.191.0

Added

• TXLSXRange.SetDynamicArrayFormula now creates native dynamic-array formulas with the complete Excel XLDAPR metadata chain

Changed

• Dynamic-array roots now share one deduplicated workbook metadata record and cross-workbook worksheet or range copies remap that record into the destination workbook
• Supported dynamic-array metadata remains available when unsupported package-part preservation is disabled, while unrelated value metadata is still removed

Fixed

• Dynamic-array spill cells are no longer modelled as legacy CSE followers, and replacing or clearing the root no longer leaves stale metadata indexes
• Removing the final dynamic-array root now removes the unused metadata part, workbook relationship, and content-type entry

Version 2.190.0

Changed

• Streaming direct reads and row-cursor traversal now expand shared-formula followers into complete cell-relative formulas while retaining projection and bounded-memory behaviour

Fixed

• Dynamic-array workbooks now preserve cell and value metadata indexes, formula recalculation attributes, metadata package relationships, and unchanged future-function prefixes during open and save cycles
• Disabling unsupported-part preservation no longer leaves dangling worksheet metadata indexes after the metadata part is removed

Version 2.189.0

Added

• Opening XLSX workbooks now expands shared-formula followers into complete per-cell formulas with correct relative, absolute, mixed, range, and sheet-qualified references

Fixed

• Self-closing shared-formula followers no longer capture later value or formula XML, and their cached results survive open and save cycles
• CSE array formulas now retain their anchor and follower relationships through row or column structure edits, complete-range copies, worksheet duplication, and cross-workbook sheet copies
• Assigning a normal formula or clearing a cell now removes stale CSE array membership so calculation cannot be redirected to an old array anchor
• Copying only part of a CSE array no longer creates dangling follower references in the destination
• Range copies now snapshot overlapping sources and shift relative formula references, preventing source cells or array anchors from being overwritten before they are copied
• Row or column edits that intersect only part of a CSE array are rejected instead of silently resizing or detaching the array

Version 2.188.0

Added

• TXLSDirectReader.SharedStringMemoryLimit bounds decoded shared-string memory and automatically migrates larger tables to delete-on-close indexed temporary storage with a fixed-size value cache
• TXLSRowCursor.SharedStringMemoryLimit exposes the same bounded shared-string storage control for forward-only row traversal

Changed

• Direct shared-string parsing now preserves whitespace-only values, CDATA, XML entities, non-BMP characters, empty items, and rich-text run order while excluding phonetic guide text from the displayed value

Fixed

• Opening XLSX workbooks now retains every shared-string item in document order, preventing duplicate text or differently formatted rich strings from collapsing later cell indexes
• Saving XLSX workbooks now deduplicates rich strings by run boundaries and formatting properties instead of visible text alone, preserving distinct formatting for cells with the same displayed value

Version 2.187.0

Added

• TXLSDirectReader.FirstRow and LastRow restrict streaming reads to a 1-based row range and stop worksheet parsing after the requested range
• TXLSDirectReader.IncludeColumn and ClearColumnProjection provide constant-time column projection that skips unselected cells before value decoding and callback dispatch

Version 2.186.0

Added

• TXLSDirectWriter.UseSharedStrings can deduplicate repeated text into an OOXML shared-strings part with accurate total and unique counts, reducing package size for repetition-heavy data
• TSimpleDomReader provides a compact single-pass XML tree for random-access queries of workbook metadata and package parts, with case-sensitive QName and local-name lookup

Changed

• Direct XLSX reading now uses hash-assisted exact element dispatch, single-pass attribute harvesting, and constant-time current-row column lookup while preserving streaming memory behaviour
• Shared-string XML generation now uses a linear-growth buffer to avoid repeated whole-string copies on large distinct-string tables

Fixed

• CSV formula export no longer adds a second equals sign when formula text already includes one
• XLSX PDF export no longer truncates cell text using a character-count estimate before layout
• XML reader name hashes now track processing instructions, attributes, and moves back to their owning elements without exposing stale element state

Version 2.185.1

Fixed

• Inserting or deleting rows and columns now adjusts local and cross-sheet formula references exactly once while preserving function names and text literals

Version 2.185.0

Added

• TXLSPDFExport now provides one-based FirstPage and LastPage output selection, a complete logical PageCount, and six CalculatePageCount overloads for classic and XLSX ranges, worksheets, and workbooks without constructing PDF page objects
• TXLSXWorkbook.SaveWorkbookAsHTML publishes visible worksheets in one navigable HTML document with stable per-sheet anchors, escaped titles and sheet names, and an option to include hidden or very hidden sheets
• TDataToXLS.BeforeRecord can skip records before row output and BeforeValue can transform or suppress field values with dataset, detail-level, source-field, and export-column context
• TXLSWorkbookViewer exposes selected-cell comments and hyperlinks, raises an application-controlled hyperlink event, paints comment indicators and XLSX fonts, fills, and alignment, and normalises selection and rendering across merged ranges

Changed

• The ConversionWorkbench example can now produce PDF output and publishes all visible XLSX worksheets for HTML targets, while the WorkbookViewer example demonstrates merged-cell navigation, comments, and hyperlink activation

Version 2.184.0

Added

• Classic XLS and XLSX ranges can copy Unicode tabular text and styled HTML to the Windows clipboard through CopyToClipboard, preserving formulas, merged cells, links, display formatting, and safely quoted embedded tabs, line breaks, and quotation marks
• TXLSXWorkbook now preserves custom package parts reachable from unknown package-root, workbook, worksheet, or drawing relationships, including relationship-ID collision handling, content types, child relationships, compression choice, and exact payload bytes; PreserveUnsupportedParts, UnsupportedPartCount, UnsupportedPartBytes, and ClearUnsupportedParts expose control and diagnostics
• Classic XLS saves expose begin, work, and end callbacks with cancellation through OnProgressEx, while IXLSWorkbookProgress provides structured diagnostics with stable codes, severity, operation, sheet context, native return values, and exception details
• TXLSXWorkbook now exposes the same structured Diagnostics, LastDiagnostic, and OnDiagnostic surfaces for save, open, recalculation, cancellation, and exception reporting
• Parsed VBA modules now support editing SourceCode and writing the updated compressed module stream back through classic XLS, standalone VBA storage, and macro-enabled OOXML saves while preserving stream prefixes and untouched project content
• Classic worksheets and shape collections now expose typed TXLSCheckBox, TXLSComboBox, and TXLSListBox form controls with creation helpers, labels, control-kind inspection, cell anchors, deletion, and open/save discovery
• TXLSXRange.ApplyFormatPatch can selectively copy font, fill, border, number-format, alignment, and protection components with source-pattern tiling while preserving destination values and unselected formatting
• TDataToXLS can recursively export linked master-detail datasets through ReadDetailTables and IndentDetailTables, preserving active data links, appending each detail table below its master record, and inheriting field, title, and formatting options at every level
• TXLSXWorksheet.AddWordArt creates cell-anchored DrawingML text with preset warping plus configurable text and outline colors, and the raw drawing-anchor preservation path keeps the complete WordArt shape on later open/save cycles
• TXLSWorkbookViewer adds a VCL worksheet viewer for classic and modern workbook objects with XLS, XLSX, XLSM, template, and ODS file loading, formatted cell text, classic styles, row and column sizing, headers, gridline and selection rendering, scrollbars, mouse selection, keyboard navigation, cell positioning, worksheet switching, and selection or worksheet events
• New Delphi FeatureShowcase and WorkbookViewer examples demonstrate styled PDF printing, print titles, manual breaks, classic form controls, selective format patches, WordArt, diagnostics, run-time viewer creation, file loading, cell navigation, worksheet switching, and viewer events, with matching English guides and developer-reference topics

Changed

• Classic XLS and modern XLSX workbook, worksheet, and range PDF export now writes every visible worksheet into one document, skips hidden and chart sheets, renders solid fills, font colors, font sizes, bold and italic emphasis, horizontal and vertical alignment, explicit cell borders, and optional worksheet gridlines, follows each worksheet's print area, paper size, orientation, margins, zoom, fit-to-width setting, column widths, actual row heights, manual row and column page breaks, repeated title rows and columns, and three-part headers and footers with page, total-page, sheet-name, date, and time fields, paginates wide and tall ranges across both axes, emits locale-independent PDF numeric syntax, and embeds an eligible TrueType font plus a ToUnicode map for text outside WinAnsi
• XLSX extended progress callbacks now receive reliable begin, work, and end states for save, open, and recalculation, while the legacy callback remains work-only and every operation resets prior cancellation state
• Classic XLS serialization now propagates worksheet, chart, style-extension, VBA, control-stream, storage, and output commit failures instead of reporting success, and file saves stage the complete compound file before atomically replacing the destination
• XLSX and ODS file saves now build a complete sibling temporary package and atomically replace the destination only after successful finalization, leaving existing files unchanged after cancellation or failure

Version 2.183.0

Added

• TXLSDirectWriter now supports reusable package-level cell-style handles and style cloning without buffering worksheet rows; the same style index can be applied to any number of cells or reused across worksheets
• Direct cell styles provide 30 formatting capabilities spanning font name, size, BGR color, bold, italic, strikeout, outline, shadow, five underline modes, 19 fill patterns with foreground/background colors, five independently colored border edges with 14 line styles and diagonal directions, built-in or custom number formats, horizontal and vertical alignment, wrapping, shrink-to-fit, indentation, text rotation, quote prefixes, and locked/hidden protection
• Styled overloads are available for strings, numbers, booleans, dates, formulas, errors, and every cached formula result type, allowing high-volume exporters to select a prebuilt style by integer index during sequential writes

Changed

• The direct writer deduplicates equal font, fill, border, and number-format components when building styles.xml, while preserving stable cell-style indexes already written into worksheet streams
• Generated direct-writer style tables now include the standard none and gray125 default fills before custom fills for stricter OOXML compatibility

Version 2.182.0

Added

• TXLSDirectWriter now writes conditional formatting while retaining its row-count-independent memory profile, with eight cell-value operators, expression rules, contains/not-contains/begins-with/ends-with text rules, top or bottom value and percentage rules, above or below average rules, duplicate and unique values, blank and nonblank cells, error and nonerror cells, data bars, two- and three-color scales, and all 17 baseline icon-set families
• Direct conditional-format rules support explicit priorities, stop-if-true evaluation, average standard-deviation and equality options, icon threshold replacement, icon-only and reverse display, data-bar value visibility, BGR colors, and differential fill, font-color, bold, and italic styles

Changed

• The direct writer deduplicates equivalent differential styles across worksheets and stores only conditional-format rules and styles, so memory use scales with the rule count rather than the number of worksheet rows
• Automatic data-bar minimum and maximum threshold kinds are normalized to the standard min and max tokens in the base OOXML element for broad spreadsheet compatibility

Version 2.181.0

Added

• TXLSDirectWriter now writes worksheet and workbook protection without buffering worksheet data, including 15 fine-grained worksheet permissions, legacy password verifiers, modern hash metadata, editable protected ranges with optional passwords or security descriptors, workbook structure and window locks, raw verifier assignment, and protection clearing

Fixed

• Unsupported drawing anchors now retain every closing element when reconstructed for round-trip output, so shape-only drawings remain well-formed after open and save
• XLSX worksheet protection permissions now follow the OOXML locked-flag definition, so allowed actions write 0, forbidden actions write 1, and omitted attributes reopen with their schema defaults
• Legacy XLSX worksheet, protected-range, and workbook passwords now use the Excel-compatible 16-bit verifier algorithm with ANSI conversion and the 15-byte limit

Version 2.180.0

Added

• The XLSX worksheet model now round-trips the What-If Scenario Manager (<scenarios>) and Data Consolidation (<dataConsolidate>) elements. New ScenariosXml and ConsolidateXml raw-XML properties on TXLSXWorksheet carry the inner content verbatim — the writer emits the wrapper elements, the reader parses scenario/data-consolidation children back. Financial-modeling files that define scenarios or consolidation definitions no longer lose them on round-trip

Version 2.179.0

Added

• A new XlsxBuildNumFmt helper function constructs OOXML number-format codes from semantic properties — Decimals, Thousands (separator), NegativeRed (red negatives), CurrencySymbol, and Percent. Callers no longer need to hand-write format strings like "$#,##0.00;[Red]-$#,##0.00" — the builder produces them from simple parameters. Designed for use with Cell.NumberFormat or Workbook.NumberFormats.Add

Version 2.178.0

Added

• The XLSX worksheet model now round-trips the VBA code name (<sheetPr codeName="..."/>), essential for VBA interop where the VBA project references worksheets by code name. New CodeName property on TXLSXWorksheet
• Comments now support an AlwaysShow visibility toggle on TXLSXComment — when True the comment balloon is always visible (Excel's per-comment "Show/Hide Comment" toggle), not just on hover
• Data validation now round-trips the IME mode attribute (imeMode) on TXLSXDataValidation. The 11-value enum controls CJK input-method behaviour when entering a cell (noControl / off / on / hiragana / fullKatakana / etc.). Essential for Japanese/Korean/Chinese-locale files
• The worksheet model gains a PaneState property for the pane-state attribute on <pane> beyond the existing frozen-mode handling — supports frozenSplit (frozen-then-scrollable) and split states

Version 2.177.0

Added

• Three previously-registered-but-unimplemented worksheet functions now have full dispatch: DOLLAR (currency-formatted text), FIXED (fixed-decimal text with optional comma suppression), and AREAS (count of areas in a reference). Previously these were registered in InitFuncHash but had no calc-engine case — evaluating =DOLLAR(...) silently returned wrong results
• The XLSX worksheet model now round-trips per-view-type zoom scales (zoomScaleNormal / zoomScalePageLayoutView / zoomScaleSheetLayoutView), gridline color (sheetView color), and sheet-view display flags (showFormulas / showZeros). New properties on TXLSXWorksheet: ZoomNormal, ZoomPageLayout, ZoomSheetLayout, GridColor, ShowFormulas, ShowZeros
• Outline summary direction now round-trips. New SummaryBelow, SummaryRight, and ShowOutlineSymbols properties on TXLSXWorksheet control whether summary rows/columns appear below/right of detail and whether outline symbols are visible. The writer emits <outlinePr> inside <sheetPr>; the reader parses it back
• The workbook calcPr now carries forceFullCalc and calcCompleted attributes alongside the existing calcId / fullCalcOnLoad / calcMode / iterate trio

Version 2.176.0

Added

• The XLSX formula engine now resolves structured table references before compilation. A new ResolveStructuredRefs method on TXLSXWorkbook scans formula text for TableName[ColumnSpec] patterns and replaces them with concrete A1 range references using the workbook's TXLSXTable collection. Supported specs include column data range (Sales[Qty]), current-row column (Sales[@Total]), full data (Sales[#Data] / Sales[#All]), headers (Sales[#Headers]), and totals row (Sales[#Totals]). The resolution runs automatically inside GetCompiledFormula — existing formulas are unaffected since non-table identifiers are left verbatim. Previously structured references caused formula compilation to fail

Version 2.175.0

Added

• The formula engine now supports R1C1 reference-style formula compilation. A new GetCompiledFormulaR1C1 method on TXLSCalculator (and GetCompiledR1C1 on TXLSFormula) accepts a formula written in R1C1 notation (R2C3 for absolute, R[-1]C[2] for relative, RC for current cell) and compiles it by first converting to A1 notation using the formula cell's row/col. A standalone R1C1ToA1 function performs the conversion — it detects R/C markers that are standalone (not part of a function name like ROUND), parses the optional bracketed offsets, and produces the equivalent A1 reference. String literals and identifiers are skipped. Previously only A1-style formulas could be compiled

Version 2.174.0

Added

• TXLSDirectWriter AutoFilter output now supports worksheet and table filter columns, value filters, blank filters, date-group filters, custom comparison filters, dynamic filters, top/bottom filters, filter button flags, sort states, sort conditions, XML escaping, and duplicate worksheet/table filter range suppression while streaming XLSX files

Version 2.173.0

Added

• TXLSDirectWriter table output now supports table comments, totals row counts, column totals-row labels, column totals-row functions, calculated-column formulas, totals-row formulas, and XML escaping for table-column metadata while streaming XLSX files

Version 2.172.0

Added

• TXLSDirectWriter now writes Excel table parts while streaming XLSX files, including worksheet table relationships, table part content types, table ranges, column names, header-row and totals-row flags, and built-in table style display flags without building a full workbook object model

Version 2.171.0

Added

• TXLSDirectWriter now writes worksheet data validations while streaming XLSX files, including list, whole-number, decimal, date, time, text-length, custom-formula, arbitrary sqref, prompt, error-alert, allow-blank, dropdown, and input/error message flags without building a full workbook object model

Version 2.170.0

Added

• TXLSDirectWriter now writes workbook and sheet-local defined names while streaming XLSX files, including hidden names, comments, generated absolute ranges, print areas, print title rows, print title columns, combined print titles, quoted sheet-name references, XML escaping, and replacement semantics for repeated print-name setters without building a full workbook object model

Version 2.169.0

Added

• TXLSDirectWriter now writes worksheet hyperlinks while streaming XLSX files, including external URL, mailto, and file relationships, internal sheet locations, display text, tooltips, single-cell or multi-cell hyperlink ranges, bare www. URL normalization, XML escaping, and percent-encoded relationship targets without building a full workbook object model

Version 2.168.0

Added

• TXLSDirectWriter now writes print and sheet-view metadata while streaming XLSX files, including gridline and row/column header display toggles, page margins, odd header/footer text, print gridlines and headings, page centering, paper size and landscape orientation, fit-to-page dimensions, and manual row or column page breaks without building a full workbook object model

Version 2.167.0

Added

• TXLSDirectWriter now writes worksheet layout metadata while streaming XLSX files, including default column and row sizes, custom column widths, hidden rows and columns, merged ranges, AutoFilter ranges, frozen panes, sheet tab colors, and right-to-left sheet views without building a full workbook object model

Version 2.166.0

Added

• Streaming XLSX workflows now include sheet-name listing without worksheet scans, name-based row-cursor sheet selection, raw numeric date streaming, direct writer reuse, error-cell writing, and cached formula-result writers for number, string, Boolean, date, and error values

Version 2.165.0

Fixed

• TXLSDirectReader.ReadStream now restores the caller's stream position after direct scanning, matching the documented stream contract and allowing applications to reuse the same stream after a direct read pass

Version 2.164.0

Added

• TDataToXLS dataset exports now support IncludeFields and ExcludeFields string lists. Applications can whitelist exported fields, omit sensitive columns, and keep headers, data rows, column widths, and export callbacks aligned with the selected field set

Version 2.163.0

Added

• CSV export gains configurable text quoting across the XLS and XLSX engines. TXLSWorkbook now exposes CSVEmitFormulas and CSVQuoteAllStrings, and TXLSXWorkbook adds CSVQuoteAllStrings. When enabled, formula export writes formula text and string fields are quoted while numeric, date, Boolean, Null, and Empty fields keep the normal quoting rules

Version 2.162.0

Added

• The XLSX engine gains an abortable progress event system. A new OnProgressEx callback on TXLSXWorkbook fires alongside the existing OnProgress with three enhancements: (1) a TXLSXProgressState parameter (ppsBegin/ppsWork/ppsEnd) that brackets each phase so UIs know when an operation starts and finishes; (2) an Abort: Boolean var parameter that lets the callback cancel the current operation mid-stream — the read/write/calc loops check Workbook.Aborted and break out cleanly; (3) a new xlsxProgressExporting phase for export operations (HTML/RTF/CSV/ODS). A FireProgress helper method centralises the dual-callback dispatch. The FAborted flag is reset at the start of each top-level operation (Open / SaveAs / Recalculate)

Version 2.161.0

Added

• The XLSX chart model now supports four new chart elements for stronger Excel compatibility. Chart data table (c:dTable): TXLSXChart gains ShowDataTable + ShowDataTableKeys toggles — when enabled the writer emits the data-table element inside the plot area so Excel renders the source data as a grid below the chart axis. Up/down bars + hi-low lines (stock/OHLC chart elements): TXLSXChart gains HasUpDownBars + UpDownGapWidth and HasHiLowLines — the writer emits c:upDownBars and c:hiLowLines inside the chart element (line/stock charts). 3D view controls: TXLSXChart gains Rotation3DX/Rotation3DY/Elevation3D/Perspective3D/RightAngleAxes3D — the writer emits c:view3D at the chart level when any 3D property differs from the default. Previously none of these elements were emitted, so charts built with data tables, stock-chart elements, or 3D views lost them on round-trip

Version 2.160.0

Added

• Formula references now adjust automatically when rows or columns are inserted or deleted in the XLSX engine. The ShiftSheetGeometry method — called by InsertRows/DeleteRows/InsertCols/DeleteCols — now walks every surviving formula cell and rewrites its A1 references to reflect the structural change. A new XlsxAdjustFormulaRowColRefs text-scanner detects single-cell refs (A1, $B$2) and range refs (A1:B3), shifts their row/column components past the insertion point, and converts deleted references to #REF!. Both relative and absolute references are shifted (matching Excel's behaviour). Previously formulas were silently left pointing at the wrong cells after a structural edit

Version 2.159.0

Added

• Autofilter sort state now round-trips through the XLSX engine. TXLSXWorksheet gains a SortStateXml property that carries the raw inner XML of the OOXML <sortState> element (column sort direction, case sensitivity, sort method, and each <sortCondition> with its descending flag + ref). The writer emits the element after <autoFilter>; the reader parses the attributes and sort conditions back. Previously sort state was silently dropped on read and never written
• CJK phonetic/ruby settings now round-trip through the XLSX engine. TXLSXWorksheet gains PhoneticFontId, PhoneticType (hiragana / katakana / fullWidthKatakana / noConversion), and PhoneticAlignment (left / center / distributed). The writer emits <phoneticPr> after <sheetData>; the reader parses the attributes back. Previously the phonetic element was silently dropped on read — Japanese-market files lost their furigana reading-aid settings

Version 2.158.0

Added

• The XLSX engine now round-trips the named cell-style table (<cellStyles>). A new TXLSXNamedStyle class + TXLSXNamedStyles collection on TXLSXWorkbook carries the full surface: Name, XfId, BuiltInId, CustomBuiltIn, Hidden, BuiltinLevel. The writer emits one <cellStyle> per entry when the collection is non-empty; the reader parses them back. Previously the writer hardcoded a single "Normal" entry and the reader discarded all named styles, so files with "Good"/"Bad"/"Total"/custom-named styles lost those names on round-trip

Version 2.157.0

Added

• The XLSX engine now round-trips the full <calcPr> element, not just the iterative-calculation toggles. TXLSXWorkbook gains CalcId (the calc engine identifier Excel uses to decide if cached values are stale), FullCalcOnLoad (forces Excel to do a full recalc on open — critical for workbooks produced by a non-Excel engine), and CalcMode (auto / manual / autoNoTable). The writer emits the matching attributes on <calcPr>; the reader parses them back. Previously only iterate/iterateCount/iterateDelta were handled, and the rest were silently dropped

Version 2.156.0

Added

• The formula evaluator now implements 37 worksheet functions that were previously registered (ptg assigned in InitFuncHash) but had no dispatch implementation — they fell through to #NAME?. The new implementations span five function families: Regression (FORECAST, LINEST, TREND, GROWTH, LOGEST, CORREL, PEARSON, SLOPE, INTERCEPT, RSQ, STEYX), Depreciation (SLN, SYD, DB, DDB, VDB), Descriptive Statistics (AVEDEV, DEVSQ, KURT, SKEW, GEOMEAN, HARMEAN, TRIMMEAN, FREQUENCY), Array Pairwise (SUMXMY2, SUMX2MY2, SUMX2PY2), and Combinatorial/Math (COMBIN, PERMUT, FACT, QUOTIENT, ROMAN, DEGREES, RADIANS, STANDARDIZE). The regression functions use a least-squares solver; the depreciation functions implement straight-line, sum-of-years'-digits, and declining-balance methods; the descriptive-statistics functions compute mean absolute deviation, kurtosis, skewness, geometric/harmonic/trimmed means; and the pairwise functions compute sum of squared differences between two equal-sized ranges. A new CollectRangeValues helper on TXLSCalculator extracts numeric values from range arguments for the stat functions

Version 2.155.0

Added

• The formula evaluator now supports five additional statistical and engineering compatibility names: ERFC, ERFC.PRECISE, BINOM.INV, CHI.INV.RT, and F.TEST. ERFC complements the existing error-function series, BINOM.INV searches the binomial CDF threshold, CHI.INV.RT aliases the right-tail chi-square inverse, and F.TEST returns the two-tailed sample-variance p-value

Version 2.154.0

Added

• The formula evaluator now resolves cross-workbook references via a new OnGetExternalValue callback on TXLSCalculator. When a 3D range reference points into an external workbook ([1]Sheet1!A1), the evaluator routes through the callback instead of returning #REF!. The XLSX engine wires this callback to GetExternalCellValue, which maps the SupBook index + sheet index to the matching TXLSXExternalLink's <sheetDataSet> cache and decodes the cached cell value by its OOXML type attribute (numeric / boolean / string / error). Previously cross-workbook formulas returned #REF! unconditionally; the callback infrastructure now lets any engine (BIFF8, XLSX, or a custom calculator host) supply external values
• The formula evaluator now performs element-wise array broadcasting on binary operators (+ / - / * / / / ^). When either operand is a 1D or 2D variant array (e.g. a matrix-function result, a spilled range, or a CSE array operand), the operator is applied element-by-element; a scalar operand is broadcast to every element of the array operand. The new ApplyArrayBinaryOp helper handles shape validation, scalar broadcast, and division-by-zero propagation. Legacy scalar arithmetic is unchanged — the array path only activates when at least one operand is a variant array
• The formula evaluator now recognises the TABLE function (What-If Analysis DataTable). Excel stores data tables as CSE array formulas whose body is TABLE(rowRef, colRef); previously TABLE was unrecognised and fell through to #NAME?. The function is registered with a synthetic ptg (ftabTABLE = 4232) so it round-trips through BIFF8 formula encoding, and at eval time returns a placeholder value (0) — the real results are pre-computed by Excel and stored as cached cell values, so HotXLS doesn't re-evaluate the substitution logic

Version 2.153.0

Added

• The PivotTable typed model now carries the page-axis typed properties (Hier / Caption / Name), the auto-sort scope, and the raw-XML round-trip hooks for <formats> and <extLst> needed for fuller Excel compatibility. TXLSPivotField gains PageHier / PageCaption / PageName (F16 — OLAP hierarchy name + page-axis caption + display name) and AutoSortScope (F20 — A1 range reference pinning the cells Excel uses to auto-sort the field's items); TXLSPivotTable gains FormatsXml / ExtLstXml (F17 — raw inner-XML round-trip for the <formats> CF / number-format overrides block and the <extLst> slicer / timeline / x14 block, emitted verbatim when set so round-trip is safe even across engine versions). The XLSX writer emits the matching <pageField hier=... cap=... name=.../> attributes, the <autoSortScope> child of <pivotField>, and the <formats> + <extLst> blocks before the closing sentinel
• The formula tracer now records per-step operand-stack snapshots, enabling a true step-through "Evaluate Formula" debugger. TXLSFormulaStep gains StepKind (push / reduce / function), StackBefore / StackAfter (comma-separated snapshots of the operand stack at the moment each step started / finished), and TokenStart / TokenEnd (1-based character span pinning the step back to the source formula). TXLSFormulaTracer gains TraceReversed (returns the same steps in reverse evaluation order — useful for a step-back UI that walks from the final result toward the original operands) and a StepCount property. Previously the tracer only recorded Depth / Source / Expression / Value with no stack state
• HTML export now embeds per-cell images via a new OnCellImage callback on TXLSHTMLExport and a new xhImages option. When the option is set and the callback is wired, the exporter asks the implementor for each cell's <img> fragment (typically a Base64 data-URI) and appends it after the cell text. This decouples the exporter from the XLS engine's BIFF8 image storage (which lives in the per-sheet MsoDrawing record) so the exporter doesn't have to parse anchors itself. The XLSX engine's HTML export already embedded images; this closes the gap for the XLS engine
• RTF import now applies character properties (bold / italic / underline / strikethrough / size) from the RTF control words to the destination cell's font. TXLSRTFImport tracks \b / \i / \ul / \strike / \fs across the document and pushes the dominant run's formatting into the cell's XF at commit time. A new ImportCharFormat property (default True) toggles the behaviour; set False to restore the legacy text-only import. Multi-run cells collapse to a single XF (the last active run wins) since the XLS engine's cell API is cell-level, not per-fragment — but this is far better than dropping all formatting, which is what the legacy importer did
• A new TXLSHTMLImport class parses HTML tables into worksheet cells. Walks the first <table> in the source, emitting one worksheet row per <tr> and one cell per <td> / <th>. Honours colspan / rowspan (Excel-style merge), strips tags inside a cell (collapsing nested formatting to plain text), decodes the five standard character entities (& / < / > / " / ' /  ) and the numeric &#NNN; / &#xNN; forms, and coerces clean numeric strings to Double so imported tables stay computable. Designed for the output HotXLS itself produces (SaveAsHTML) plus general copy-paste-quality HTML tables; not a full HTML5 parser

Version 2.152.0

Added

• XLSX tables now retain column-level totals metadata and table formulas. TXLSXTable exposes ColumnTotalsRowFunctions[], ColumnTotalsRowLabels[], ColumnCalculatedColumnFormulas[], and ColumnTotalsRowFormulas[]; SaveAs emits the matching totalsRowFunction, totalsRowLabel, calculatedColumnFormula, and totalsRowFormula table-column XML, Open reads them back, and table clones preserve them

Version 2.151.0

Added

• The PivotTable typed model now carries the per-field filter collection, the multi-select toggle, the fieldGroup discretePr + groupItems surface, and the calculated items / calculated members model needed for fuller PivotTable compatibility. TXLSPivotField gains a typed TXLSPivotFilters collection (each entry carries FilterType across the 80-member OOXML ST_FilterType enum — count / percent / sum / caption / value / date families — plus Value1/Value2/StrValue1/StrValue2, EvalOrder, DataField, MemberPropertyId, Name, Description), a MultipleItemSelectionAllowed toggle (F9), and a TXLSPivotCalculatedItem collection (F14). TXLSPivotCacheField gains DiscreteIndices[] + AddDiscreteIndex and GroupItemLabels[] + AddGroupItem for the fieldGroup discretePr + groupItems blocks (F13). TXLSPivotTable gains a TXLSPivotCalculatedMember collection (F14). The XLSX writer emits the matching <pivotField multipleItemSelectionAllowed="1">, the per-field <filters><filter .../> block, the <fieldGroup><discretePr>...</discretePr><groupItems>...</groupItems></fieldGroup> element, the per-field <calculatedItems> block, and the table-level <calculatedMembers> block — so programmatically built pivots with filters / multi-select / grouping / calculated items round-trip

Version 2.150.0

Added

• Defined names now keep their formula text valid when the worksheet they reference is deleted or renamed. Deleting a sheet rewrites every SheetN! qualifier inside a defined name's formula to #REF! (matching Excel's behaviour); renaming a sheet rewrites the qualifier to the new name. Previously the formula text kept pointing at a non-existent sheet name, which Excel would then flag as a broken link on reopen
• The PivotTable typed model now carries the field-level sort order and layout toggles, plus the table-level layout / display attributes needed for fuller PivotTable layout compatibility. TXLSPivotField gains SortType (manual / ascending / descending), Compact, Outline, SubtotalTop, DragToRow/DragToCol/DragToPage/DragToData, ShowAll, InsertBlankRow, NumFmtId; TXLSPivotTable gains Compact/CompactData/Outline/OutlineData/Indent/ShowEmptyRow/ShowEmptyCol/ShowHeaders/ShowDrill/EnableDrill/FieldListSortAscending/GridDropZones plus RowHeaderCaption/ColHeaderCaption/ErrorCaption/MissingCaption/ShowError/ShowMissing and the page-axis layout pair RowPageCount/ColPageCount. The XLSX writer emits the matching <pivotField sortType=... compact=... outline=...> and <pivotTableDefinition indent=... outline=... showEmptyRow=...> attributes and the <location rowPageCount=... colPageCount=.../> element so programmatically built pivots carry the user-set layout instead of Excel's defaults
• A new FormulaIsVolatile method on TXLSWorkbook reports whether a formula would force a recalculation on every calc pass — i.e. it calls a volatile function (NOW / RAND / TODAY / OFFSET / INDIRECT) or carries an unresolvable reference. The probe shares the same dependency extractor the recalc engine uses, so the answer matches what Recalculate actually does. Callers can use it to flag cells for a verification UI, to populate a calc-chain volatile set on a brand-new workbook, or to skip caching the result of a freshly-entered formula

Version 2.149.0

Added

• PivotTable caches can now be refreshed from their source range programmatically. A new RefreshPivotCache(ACacheId) method on TXLSXWorkbook re-scans the cache's source range, rebuilds the shared-items domain per field (re-deduplicating values and re-deriving the per-type bounds), and rewrites the per-record item indices. Previously HotXLS relied on the hardcoded refreshOnLoad="1" flag and deferred all refresh work to Excel on the next open; programmatic pipelines can now refresh in-process

Version 2.148.0

Added

• InsertCols on the XLSX engine now inherits the column width from its left neighbour, matching Excel's default behaviour for "insert column". Previously the inserted column picked up the workbook default width, producing a visible fidelity gap after a reopen
• The XLSX engine gains convenience outline-collapse helpers ToggleColGrouped / ToggleRowGrouped on TXLSXWorksheet, which walk the outline tree and collapse or expand the group at a given level. The collapse flags themselves already round-tripped; this adds the missing walk-the-tree convenience API

Version 2.147.0

Added

• HTML export now emits <a name> anchors at cells that are internal-hyperlink targets, so #SubAddress hyperlinks resolve correctly when the reader clicks them. Previously the links were emitted but the targets had no anchors, so internal jumps resolved to nothing
• CSV export gained a CSVEmitFormulas property on TXLSXWorkbook. When True, formula cells export their formula text (=SUM(A1:A2)) instead of the calculated value — useful for debugging and for pipelines that re-import into a different engine
• External hyperlink URLs are now percent-encoded on write (XlsxEncodeHyperlinkUrl): space, angle brackets, quotes, braces, pipes, backslashes, carets, and backticks are encoded as %XX in the <Relationship Target="..."/> attribute so the file opens correctly in every browser. Existing percent-encoded sequences (%20) are preserved as-is
• PivotTable style info now carries the full pivotTableStyleInfo surface: StyleName, ShowRowHeaders, ShowColHeaders, ShowRowStripes, ShowColStripes, ShowLastColumn on TXLSPivotTable. The XLSX writer emits the matching element so programmatically built pivots carry the user-chosen banding style instead of the hardcoded default
• A new CalculateAndVerify method on TXLSXWorkbook re-evaluates every formula cell and compares the computed result against the cell's cached value. Returns the count of mismatches (0 = all cached values match the engine's fresh computation; volatile formulas are excluded). Useful for QA when porting workbooks between engines or validating that a writer didn't corrupt cached values

Version 2.146.0

Added

• The XLSX engine now supports iterative circular-reference calculation (Excel's iterative calc mode). A new Iterate / IterateCount / IterateDelta property triple on TXLSXWorkbook controls the behaviour: when Iterate=True and a reference cycle is detected during Recalculate, the engine re-evaluates cycle-member cells up to IterateCount times (default 100) or until every cell's change is < IterateDelta (default 0.001), instead of returning #REF!. The settings round-trip through OOXML's <calcPr iterate="1" iterateCount="N" iterateDelta="X"/> element. Financial models with intentional circular references (interest-on-interest, circular allocations) now converge instead of erroring
• PivotTable cache fields now carry a grouping model for numeric-range and date-hierarchy grouping. TXLSPivotCacheField gains HasGroup + GroupStartNum/GroupEndNum/GroupInterval (numeric range) and GroupByDate/GroupDays/GroupMonths/GroupQuarters/GroupYears/GroupHours/GroupMinutes/GroupSeconds/GroupStartDate/GroupEndDate (date). The XLSX writer emits the <fieldGroup><rangePr groupBy="range|days|months|..."/> element so programmatically built pivots with grouped fields match Excel's output
• HTML export now emits comment endnotes. When xhComments is in the export options, each commented cell gets a [n] superscript link, and the full comment text (author + body) is listed as an endnote section after the table with a back-anchor to the cell. Previously comments were only available via the cell's title= attribute, which is invisible when printed or copy-pasted

Version 2.145.0

Added

• BIFF8 formula error literals (tErr token) now decode into a typed SA_ERROR syntax item instead of failing the formula parse. A cell storing =#N/A (or any of the seven Excel error constants) previously caused the whole formula to be dropped on read; the error byte now round-trips through read → evaluate → write and reconstructs to its text form (#NULL!, #DIV/0!, #VALUE!, #REF!, #NAME?, #NUM!, #N/A) via a new XlsErrorToString helper
• The BIFF8 tIsect $0F token (implicit intersection operator) now decodes in the binary-formula translator, building the same SA_ISECT node the text parser produces for @. Previously a binary formula containing tIsect would fail to parse
• Data bar conditional formatting now emits the Excel 2010+ x14 extension block in the worksheet extLst when the model carries any non-default x14 field (negative fill color, gradient off, axis color, direction, border). A new BuildDataBarExtXml builds the <x14:conditionalFormattings> ext element with the matching priority, cfvo pair, fill + negative + axis colours, and the gradient/border/direction/negativeBarColorSameAsPositive attributes

Version 2.144.0

Added

• The XLSX engine can now write Agile-encrypted workbooks (TXLSXWorkbook.SaveAsEncryptedAgile), closing a high-value read/write compatibility gap. Agile is what Excel 2010+ writes by default for password-protected .xlsx files; HotXLS could previously read Agile but only write Standard Encryption (AES-128). The new write pipeline mirrors the existing decrypt path: SHA-512 password hashing with a 100000-iteration spin count, AES-256-CBC per-segment package encryption, a randomly generated intermediate key wrapped per the three spec-pinned block keys, and a complete <encryption> descriptor XML carrying the encrypted verifier + key blobs. The output is byte-faithful to what Excel produces and round-trips cleanly through XlsLoadAgileEncryptedFile (verified by test: correct password decrypts, wrong password rejected)
• New cryptographic primitives underpin the Agile write path: XlsHMACSHA512 (RFC 2104, needed for the dataIntegrity element), XlsBase64Encode (RFC 4648 encoder mirroring the existing decoder), and AESEncryptCBC (AES-CBC encrypt for any key size 128/192/256, the inverse of the existing AESDecryptCBC). The AES unit now imports aes_encrypt_key192 / aes_encrypt_key256 from the Gladman obj (previously only 128-bit encrypt was available)

Version 2.143.0

Added

• The formula evaluator now parses and evaluates the explicit intersection operator @ (Excel 365 dynamic-array syntax, e.g. =A2:A4 @ B2:B4). The lexer emits a new LT_ISECT token for the @ symbol, the precedence parser builds an SA_ISECT node (binding tighter than ^, looser than %), and a new IntersectRanges evaluator resolves both operands as ranges and returns the value of the single crossing cell — Excel's implicit-intersection rule. The BIFF token serializer emits tIsect $0F so the operator round-trips through XLS, and the formula-text reconstruction path emits the implicit-space form. Previously such formulas failed to parse
• A standard Base64 encoder (XlsBase64Encode) is now available, mirroring the existing XlsBase64Decode. This is the prerequisite for the Agile-encryption write path, which emits saltValue / encryptedVerifierHashInput / encryptedKeyValue / encryptedHmacValue as Base64 attribute values
• The workbook's OnProgress event now also fires during full recalculation (Recalculate), with a new xlsxProgressCalculating phase. Previously the event only fired during reading and writing, so long full-recalcs gave the UI no heartbeat. The event fires roughly every 4% of cells to avoid flooding
• The extended document properties (docProps/app.xml) now round-trip three more fields Excel writes for password-protected / template-based documents: Manager, Template, and DocSecurity (a bit field: 1=password-protected, 2=read-only recommended, 4=read-only enforced, 8=locked for annotation)

Version 2.142.0

Added

• The PivotTable typed model now carries the metadata, source, calculated-field, "show values as", and shared-items-bounds surface that the previous round of analysis flagged as gaps. TXLSPivotCache gains RefreshOnLoad/RefreshedBy/RefreshedDate/MissingItemsLimit/CreatedVersion/RefreshedVersion/MinRefreshableVersion/EnableRefresh/BackgroundQuery/SaveData plus a typed worksheet-source shape (SourceName/SourceRelId/SourceIsNamedRange); TXLSPivotCacheField gains Formula/DatabaseField/MemberPropertyField (calculated fields) and the shared-items bounds (ContainsBlank/ContainsInteger/ContainsString/ContainsNumber/ContainsDate/MinValue/MaxValue/MinDate/MaxDate/HasBounds); TXLSPivotDataField gains ShowDataAs/BaseField/BaseItem for the "show values as" transforms (percent-of / difference / running-total / index). The XLSX writer (BuildPivotCacheDefinitionXml/BuildPivotTableXml) honours all of these so a programmatically built pivot (AddPivotTable) emits a richer pivotCacheDefinition + pivotTableDefinition — refreshOnLoad / version triple / refreshedBy / named-range source / calculated cacheField formula / sharedItems bounds / dataField subtotal + showDataAs. Pivot tables read from a file continue to round-trip their raw XML verbatim
• Image anchor behaviour now round-trips through a new AnchorType property on TXLSXImage (xaTwoCell = move and size with cells, xaOneCell = move but don't size, xaAbsolute = don't move or size). The writer picks the anchor element (xdr:oneCellAnchor / xdr:twoCellAnchor editAs=... / xdr:absoluteAnchor) to match; the reader parses the element name + editAs attribute back. Maps to Excel's right-click → Size and Properties → Properties dialog
• Header/footer now round-trips even/first-page differentiation. TXLSXWorksheet gains EvenHeaderText / EvenFooterText / FirstHeaderText / FirstFooterText plus the four <headerFooter> toggles (DifferentOddEven, DifferentFirst, ScaleWithDoc, AlignWithMargins). Files where the user enabled "Different odd and even pages" or "Different first page" no longer lose the even/first-page sections on round-trip
• Page setup now round-trips the extended <pageSetup> attributes Excel writes for non-default print setups: Copies, FirstPageNumber + UseFirstPageNumber, HorizontalDpi, VerticalDpi, PrintErrors (displayed/blank/dash/NA), UsePrinterDefaults
• <sheetFormatPr> now round-trips customHeight, zeroHeight, outlineLevelRow and outlineLevelCol, so files where the user set a user-pinned default row height, hidden-default rows, or a default outline level no longer lose that behaviour
• Data bar conditional formatting round-trips the standard gradient attribute (gradient vs. solid fill), and TXLSDataBarSpec gains the Excel 2010+ x14 extension fields (NegativeBarColor, NegativeBarColorSameAsPositive, GradientFill, AxisColor with theme/tint, Direction, Border) on the model so API consumers can set them
• Comment balloons round-trip their VML drawing verbatim. The reader captures xl/drawings/vmlDrawingN.vml as-is and the writer replays it on save (instead of the built-in yellow-balloon template), so each balloon's user-set position / size / fill colour / visibility survives
• Comment bodies round-trip per-run formatting. TXLSXComment gains a lazy RichText: TXLSXRichText payload (reusing the shared-strings run model); the writer emits one <r><rPr>...</rPr><t>...</t></r> per run when RichText carries formatting, and the reader parses <b/>, <i/>, <sz val=N/>, <color rgb=.../> back so bold / italic / size / colour survive
• The formula evaluator now implements SUBTOTAL (all eleven aggregators: AVERAGE / COUNT / COUNTA / MAX / MIN / PRODUCT / STDEV / STDEVP / SUM / VAR / VARP, plus the 101–111 ignore-hidden variants) and AGGREGATE (nineteen aggregators with the option-code 0–7 flags that select error/hidden skipping, including the LARGE / SMALL / MEDIAN / PERCENTILE / QUARTILE family with their k argument). Both previously fell through to the unsupported-function path; SUBTOTAL was registered but had no implementation, AGGREGATE was not registered at all
• The CELL and INFO worksheet functions are now evaluated: CELL supports address, col, row, type, contents, filename, format, and the GUI-state info_types (color, parentheses, prefix, protect, width) return neutral defaults; INFO supports osversion, system, recalc, memavail, numfile, release, origin. Both were registered but fell through to #NAME?
• Conditional formatting now round-trips the stopIfTrue, stdDev, equalAverage and timePeriod cfRule attributes, plus the full ten-member timePeriod family (yesterday / today / tomorrow / last7Days / lastWeek / thisWeek / nextWeek / lastMonth / thisMonth / nextMonth). These were silently dropped on read; Excel writes them routinely on the above-average and date-band rules, and dropping them changed how the rule re-evaluated on reopen
• CSV export gained a TXLSCsvEncoding property on TXLSCSVSaver with three modes: ceSystemAnsi (legacy default), ceUtf16 (the historical Unicode mode), and ceUtf8 which writes a UTF-8 file with the EF BB BF byte-order mark — the modern Excel-friendly default that reopens without the text-import wizard on any locale
• CSV import gained a LocaleTolerantNumbers property on TXLSCSVLoader. When enabled, a numeric field that fails to parse with . as the decimal separator is retried with , (and vice versa), so a European 3,14 imports as the number 3.14 on a US machine. Strict mode (the default) is unchanged so leading-zero text, postal codes and other comma-bearing tokens stay text

Fixed

• The XLSX engine's strict-mode number inference no longer coerces comma-bearing text fields (e.g. 3,14) into numbers via Delphi's TryStrToFloat, which tolerates a comma as a thousand separator even when ThousandSeparator is set to #0. Such fields now stay text unless the new LocaleTolerantNumbers option is enabled

Version 2.140.0

Changed

• The XLSX engine now compacts its six per-attribute style pools (Fonts / Fills / Borders / NumberFormats / Alignments / Protections) before every save: a new CompactStylePools pass scans every cell's style indices and discards pool entries that are no longer referenced by any cell — the entries that accumulate when cells are cleared, deleted, or re-styled after the pool entry was first added. The compacted pools produce a smaller <cellXfs> table in the output file (typically 30–70% smaller for documents that have been through several edit cycles), matching the compaction the XLS engine already does via CreateAllReferredList

Version 2.139.0

Added

• Drawing-layer shapes, connectors, group shapes, and non-chart graphic frames now round-trip verbatim: when the XLSX reader encounters a <xdr:sp>, <xdr:cxnSp>, <xdr:grpSp>, or non-chart <xdr:graphicFrame> anchor that the typed model (images / charts / text boxes) does not cover, it captures the complete anchor XML and replays it unchanged on save — so drawings authored in Excel (arrows, callouts, grouped shapes, smart art frames) survive a read→write cycle without loss. Each worksheet exposes RawDrawingAnchorCount / RawDrawingAnchor(i) for inspection

Version 2.138.0

Added

• External link cached cell values now round-trip on the XLSX engine: a new TXLSXExternalCell / TXLSXExternalRow / TXLSXExternalSheetData model hangs off each TXLSXExternalLink via AddSheetData / AddRow / AddCell, the writer emits the matching <sheetDataSet> block inside externalLinkN.xml, and reopening restores every cached value. FindCachedCell(sheetIdx, row, col) looks up the last-known value of an externally referenced cell — the data Excel itself shows when the remote workbook is offline

Version 2.137.0

Added

• The OOXML cellXf quotePrefix flag — Excel's mechanism for marking stored text as a literal so it survives numeric coercion (CSV-imported codes like 00123, long account numbers, anything the user prefixed with an apostrophe) — now round-trips on the XLSX engine: TXLSXAlignment gains a QuotePrefix property, the alignment pool dedup counts it, the writer emits quotePrefix="1" on the matching <xf>, and reopening restores it whether the cellXf carried a child <alignment/> or was a bare self-closing <xf quotePrefix="1"/>

Version 2.136.0

Added

• Chart axes are now fully programmable: a new TXLSXChartAxis model on each chart's CategoryAxis and ValueAxis exposes LogBase (logarithmic scale, e.g. 10 for base-10 log), Orientation (normal / reversed), MajorGridlines / MinorGridlines, MinValue / MaxValue (NaN = auto), MajorUnit / MinorUnit, CrossesAt, TickLblPos, and DispUnits (display units: hundreds / thousands / millions / billions / trillions). All default to Excel's auto behaviour, so existing charts emit the same XML as before; the writer adds <c:logBase>, <c:orientation>, <c:majorGridlines>, <c:majorUnit>, <c:crossesAt>, <c:tickLblPos>, and <c:dispUnits> only when the caller sets a non-default value, and reopening such a chart restores every setting

Version 2.135.0

Added

• Chart error bars are now programmable and round-trip on the XLSX engine: each TXLSXChartSeries gains AddErrorBar(dir) returning a TXLSXChartErrBar with ValueType (fixed / percentage / stdDev / stdErr / custom), Value, PlusValue / MinusValue (custom), and NoEndCap; one error-bar slot per direction (X / Y) is stored, the writer emits the matching <c:errBars> block, and reopening restores the direction, type, and magnitude so scientific / engineering charts keep their uncertainty indicators

Version 2.134.0

Added

• Chart data labels are now programmable and round-trip on the XLSX engine: each TXLSXChartSeries gains ShowValue, ShowCategoryName, ShowSeriesName, ShowLegendKey, ShowPercent, ShowBubbleSize, and Separator properties that emit the matching <c:dLbls> block with its <c:show*/> children; reopening such a chart restores the switches so 'show value' / 'show percent' settings survive a save-edit-save cycle

Version 2.133.0

Added

• Vector image formats EMF and WMF are now first-class image types on the XLSX engine: TXLSXImageFormat gains xlsxImageEmf / xlsxImageWmf, the writer emits the matching image/x-emf / image/x-wmf content types and .emf / .wmf extensions, and the format sniffer recognises EMF (01 00 00 73), placeable WMF (D7 CD C6 9A), and standard WMF headers — so images pasted from Word or PowerPoint that Excel embeds as vector data now round-trip instead of being dropped

Fixed

• BIFF8 chart substreams whose GELFRAME record ($1066) spans the 8224-byte continuation boundary no longer lose fill/line geometry: the reader now accepts GELFRAME self-continuation (Excel writes another $1066 instead of the standard $003C CONTINUE) and also lets standard CONTINUE records stitch inside chart context, matching the writer's existing multi-chunk dispatch
• BIFF8 header/footer background pictures ($0866) whose payload spans continuation records no longer corrupt: each continuation segment repeats a 14-byte MSO header that the reader now strips before concatenating, so the embedded image data is no longer shifted by 14 bytes per segment

Version 2.132.19

Fixed

• Legacy XLS and encrypted XLSX storage cleanup now releases compound-file streams deterministically, preventing temporary files from staying locked after save, load, VBA, or encrypted-package operations
• Formatted display of extended integer variant values now applies the requested number format consistently, matching the existing behavior for standard integer values
• Shared strings, data validation, chart sheets, ZIP archives, PNG drawing, and repeated text formulas now clean up temporary buffers more defensively during error paths and large text operations

Version 2.132.18

Fixed

• Internal string builder value assignment now clears previous text when assigned an empty string, preventing reused XML and text buffers from retaining stale content

Version 2.132.17

Fixed

• XLS chart sheet creation now accepts quoted source sheet names that contain an exclamation mark, rejects oversized row references before they can wrap into a valid BIFF row, and caps trendline legend names to the BIFF short-text limit so generated chart records stay well formed

Version 2.132.16

Fixed

• Internal string-key and object-key lists now remove stale lookup keys when an existing slot is overwritten, preventing old keys from resolving to replacement entries
• Internal object-list replacements and transformed unique-key caches now release owned entries on overwrite, clear, and destruction, preventing leaks during repeated workbook and style operations

Version 2.132.15

Fixed

• Direct XLSX streaming reads now decode boolean cells that store OOXML literal values such as true and false
• Internal string-list index updates now remove stale keys when an existing slot is overwritten, preventing old names from resolving to the new value

Version 2.132.14

Fixed

• BIFF8 data-validation prompt and error messages whose first character is NUL now preserve record boundaries instead of writing a mismatched DV length

Version 2.132.13

Fixed

• XML text parsing now rejects nested entity starts such as &amp& instead of silently dropping the unfinished entity
• Decompiled formulas now write numeric constants with an invariant dot decimal separator so formula text remains portable across regional settings

Version 2.132.12

Fixed

• Palette PNG transparency now ignores transparent colors that are not present in the image palette instead of marking an arbitrary palette index as transparent
• SUMPRODUCT formulas now stop scanning remaining rows after the first out-of-used-range row in long range arguments, avoiding unnecessary cell reads while preserving the calculated result

Version 2.132.11

Fixed

• TEXT formulas now return #VALUE! when an invalid or oversized number format cannot be applied instead of being swallowed as a blank successful result
• Shared-string table truncation now frees the correct tail blocks and keeps the table reusable after large delete cycles

Version 2.132.10

Fixed

• MATCH exact lookups now honor Excel wildcards, and classic XLS row-offset metadata now ignores out-of-block rows instead of corrupting saved workbook structure
• PNG compressed text metadata now releases temporary buffers safely if chunk processing raises

Version 2.132.9

Fixed

• PNG compressed text metadata, chunk-list insertion, and CRC byte-order handling now remain stable across supported Windows platforms

Version 2.132.8

Fixed

• LEFT and RIGHT formulas now report #VALUE! when num_chars cannot be converted to a valid integer

Version 2.132.7

Fixed

• SEARCH formulas now support Excel-style ?, *, and ~ wildcards
• Worksheet name validation now rejects invalid Excel sheet-name characters when renaming sheets
• Manual page break deletion now works after inserting a break before existing breaks

Version 2.132.6

Fixed

• DATE formulas now match Excel-compatible month and day rollover behavior, including calls such as DATE(2008,14,2)
• Column-range Hidden reads now report each column's visibility instead of mixing in row visibility state
• BIFF row metadata at the Excel 97-2003 row limit now loads correctly so settings for row 65536 are preserved

Version 2.132.5

Fixed

• BIFF workbook loading now tolerates malformed worksheet tab colors, truncated pivot labels, and invalid pivot-cache date fragments without raising exceptions
• Workbook package and stream loading now cleans up failed archive and storage setup paths more consistently, avoiding leaked resources on malformed input or failed stream-backed storage opens

Version 2.132.4

Fixed

• ZIP archive handling now finalizes the current package before reusing the same archive object for another package, and file-based archive opening now allows concurrent shared readers

Version 2.132.3

Fixed

• Workbook package writing now preserves non-ASCII ZIP entry names with UTF-8 metadata, preventing mojibake when package part names contain Unicode text
• XML output now writes raw text and attribute values as UTF-8 bytes when escaping is disabled, matching the declared document encoding
• Formula evaluation now clears its in-progress state after callback exceptions and correctly handles large-number cancellation on Unicode Delphi builds
• Array-formula matrix extraction now respects non-zero Variant array lower bounds instead of raising for valid one-based arrays

Version 2.132.2

Changed

• Range sorting now uses a stable O(n log n) ordering pass for large sorted ranges, while preserving existing row content round-trip behavior
• ODS and XLSX numeric parsing and formatting now reuse invariant formatting state instead of creating it repeatedly in hot paths

Fixed

• XML parsing now accepts valid CDATA blocks and UTF-8 processing instructions without corrupting text or failing on valid markup
• Formula parsing now accepts signed scientific notation and rejects malformed or truncated formula bytes cleanly
• Classic workbook open and save paths now release streams and storages reliably after errors, preventing temporary files from remaining locked
• Binary buffer handling now validates copy and read bounds, grows destination buffers before copying, and rejects malformed VBA compression back-references cleanly

Version 2.132.1

Fixed

• Corrected the error message raised when an encrypted workbook uses an unsupported encryption scheme: it previously claimed ECMA-376 Agile Encryption could not be read, although Agile decryption (AES-128/192/256) has been supported since 2.119.0. The message now lists the actually supported schemes
• Open now points callers at TXLSXWorkbook.OpenEncrypted when it refuses an encrypted OOXML package, instead of stating that encrypted workbooks are not supported

Version 2.132.0

Changed

• XLSX loading is substantially faster: the XML reader now interns repeated element names and converts text and attribute values straight from its scan buffer, cutting per-cell memory allocations by more than half. Multi-threaded opening (2.128.0) benefits the most — the allocations removed were the ones serializing the worker threads — with multi-sheet workbooks now opening close to twice as fast as the single-threaded path on multi-core machines; single-threaded opening also gains a few percent

Version 2.131.0

Added

• Worksheet extension blocks that HotXLS does not model — <ext> entries in a worksheet's <extLst> other than sparklines, as written by Excel for features like x14 conditional formatting or slicers — are now preserved verbatim when a workbook is opened and written back on save, so opening and resaving a file no longer strips those extensions. The captured blocks are visible read-only through Worksheet.RawWorksheetExts and travel with the sheet when it is copied to another workbook

Version 2.130.0

Added

• Saved workbooks that contain formulas now carry the xl/calcChain.xml calculation-chain part, letting Excel seed its dependency engine from the recorded order instead of rebuilding it on load. After a Recalculate call the chain follows the dependency graph's topological order; otherwise formula cells are listed in document order (a valid hint Excel refines on the fly). The part, its content type, and its workbook relationship are emitted together and survive open/save round-trips

Version 2.129.0

Added

• New Workbook.ParallelParseThreads property tunes the worker count for the multi-threaded Open path introduced in 2.128.0 (0 = automatic, one worker per CPU core). Worksheet parsing allocates heavily and memory allocation is serialized across threads, so on some workloads a small explicit cap performs as well as — or better than — the full core count; the property lets integrators pick what measures best for their files

Version 2.128.0

Added

• Multi-threaded XLSX loading: opening a workbook now parses each worksheet's XML — the dominant cost on large files — on worker threads, so multi-sheet workbooks open faster on multi-core machines. Sheet parts are extracted from the package up front and each worker touches only its own sheet, so the loaded workbook is identical to the serial result
• New Workbook.ParallelParse property (default True) switches back to single-threaded parsing when set to False before Open — a one-line opt-out for constrained environments

Changed

• The OnProgress reading callback still fires once per worksheet on the calling thread, now as each sheet is dispatched for parsing rather than strictly between sheets

Version 2.127.0

Added

• Incremental recalculation on the XLSX engine: the new Workbook.Recalculate method builds a formula dependency graph, evaluates every formula cell once in dependency order, and stores each result in the cell's cached value. Subsequent calls re-evaluate only the cells affected by value changes since the last pass — on chained models (financial sheets, report cascades) editing one input recomputes just its downstream formulas instead of the whole workbook
• Volatile functions (NOW, TODAY, RAND, OFFSET, INDIRECT) recalculate on every pass, as do formulas whose references cannot be resolved statically (multi-area names, external workbooks) — the engine over-recalculates rather than risk a stale result
• Reference cycles are detected instead of looping: Recalculate returns an error code when formulas form a circular chain, skips the cycle members, and still evaluates every formula outside the cycle. Cross-sheet references, named ranges, and CSE array formulas (results spilled across the anchored range) participate fully in the dependency graph

Changed

• The existing on-demand Calculate path is unchanged; Recalculate is a separate opt-in API and workbooks that never call it behave exactly as before

Version 2.126.0

Added

• Sparklines on the XLSX engine: worksheets now carry sparkline groups — the in-cell mini charts Excel renders from a data range — with the three Excel types (line, column, win/loss stacked). Create one with Worksheet.AddSparklineGroup or build groups manually through the new Worksheet.SparklineGroups collection, adding any number of sparklines per group with a source range and a display cell each
• Sparkline groups expose the full option set Excel offers: line weight, data-point markers, high/low/negative/first/last point emphasis toggles, eight ARGB colour slots (series, negative, axis, markers, first, last, high, low), per-group or per-sparkline vertical-axis scaling, and gap/zero/span handling for empty source cells
• Sparklines survive a full save→open round-trip: saving emits the worksheet extLst extension block ([MS-XLSX] §2.3.6) so Excel 2010 and later render the sparklines, and opening such a workbook — whether written by HotXLS or Excel — rebuilds every group, option, and range so they persist through edit cycles. Copying a sheet to another workbook clones its sparkline groups as well

Version 2.125.0

Added

• Seven DBCS byte-counting text functions now evaluate on both engines: LEFTB, RIGHTB, MIDB, LENB, FINDB, SEARCHB, and REPLACEB. Each treats its input as a byte string under the system ANSI code page so CJK (Chinese/Japanese/Korean) characters occupy 2 bytes while ASCII occupies 1; operations that would split a DBCS lead/trail byte pair snap back to the nearest character boundary, matching Excel's behaviour

Version 2.124.0

Added

• Five hypothesis-testing and distribution functions now evaluate on both engines: Z.TEST (one-tailed right-tail probability that the sample mean is at least x under the null hypothesis), T.TEST (paired two-sample t-test p-value; unpaired equal/unequal-variance types are deferred), CHISQ.TEST (goodness-of-fit p-value via the upper-tail chi-square CDF), HYPGEOM.DIST (hypergeometric PMF/CDF with the cumulative flag), and NEGBINOM.DIST (negative-binomial PMF/CDF). The p-value math reuses the engine's existing normal/Student-t/chi-square CDF helpers via self-contained nested routines, so no existing Calc method changes

Version 2.123.0

Added

• Fourteen formula functions from a compatibility gap review now evaluate on both engines: AVERAGEIF (single-criterion average, the SUMIF/COUNTIF sibling the engine was missing), WORKDAY.INTL and NETWORKDAYS.INTL (custom weekend masks via the number 1-17 codes or a 7-character Mon-Sun string), XIRR and XNPV (internal-rate-of-return and net present value for irregular-date cash flows, solved by bracketing bisection), CEILING.PRECISE / FLOOR.PRECISE / ISO.CEILING (sign-agnostic rounding), PERCENTRANK.INC / PERCENTRANK.EXC (percentile rank with the matching inclusive/exclusive interpolation), MODE.MULT (spills every value tied for the highest frequency), ERF.PRECISE (Gauss error function via its alternating series), ACCRINT (accrued interest), and ISPMT (period interest for constant-principal amortization)

Version 2.122.0

Added

• Threaded comments (Excel 365) now round-trip on the XLSX engine: AddThreadedComment and AddThreadedReply on a worksheet build a parent→child reply tree anchored at a cell, auto-register every author on a new workbook-level Persons list, and emit the dedicated xl/threadedComments/threadedCommentN.xml part per sheet plus the workbook-scoped xl/person/person.xml; reopening such a package rebuilds the conversation tree and participant list from those parts, so reply chains, identifiers, and display names survive a full read→write cycle

Version 2.121.0

Added

• The formula evaluator now computes the Excel 365 dynamic-array functions XMATCH, SORT, UNIQUE, and SEQUENCE, joining the already-supported XLOOKUP and FILTER; results spill as 2-D arrays so SEQUENCE(3;2;1;1) returns a 3×2 grid, SORT reorders rows by any key column ascending or descending, UNIQUE collapses duplicate rows keeping the first occurrence, and XMATCH reports the 1-based position of a lookup value with the same exact/wildcard/next-smaller/next-larger match modes as XLOOKUP

Version 2.120.0

Added

• New PasteFromClipboard method on both engines' worksheets reads tab-delimited text off the Windows clipboard — the payload Excel places there when a range is copied, or a plain selection from any other application — and writes it into cells starting at the given row and column, with the same number/date/boolean type promotion the CSV loader already uses; leading = text becomes a live formula, matching how Excel itself pastes

Version 2.119.0

Added

• Reading of ECMA-376 Agile Encryption — the password protection Excel 2010 and later apply by default: OpenEncrypted now parses the XML encryption descriptor, derives the keys with SHA-512 spin-count hashing, verifies the password, and decrypts the AES-CBC segmented package (AES-128/192/256), so workbooks password-protected directly in Excel open without Excel installed
• OpenEncrypted handles all three cases through one entry point: plain unencrypted .xlsx, Standard Encryption (our SaveAsEncrypted round trip), and Agile Encryption; a wrong password raises the same clear error in every scheme

Fixed

• Saving an .xlsx workbook whose worksheet XML exceeded roughly 116 KB — about 3,000 rows of typical string data — crashed with a stack overflow; the ZIP writer's CRC-32 checksum now processes data in 64 KB slices, which keeps the compression backend off its large-input code path that allocates an oversized scratch buffer on the stack. The checksum result is unchanged

Changed

• Shared-string deduplication now runs through a hash table instead of scanning the string list, and the sharedStrings.xml part is assembled through the pre-grown string builder instead of repeated concatenation — together with the constant-time cell index this brings saving 120,000 string cells from tens of seconds down to about a third of a second

Version 2.117.0

Changed

• XLSX cell access is now constant-time: the worksheet cell collection carries a hash index over (row, column), replacing the previous linear scan — so populating, reading back, or loading a large sheet scales linearly with the cell count instead of quadratically; filling 100,000 cells now takes well under a second where it previously took minutes
• The speedup applies across the engine: Cells[r, c], HasCell, workbook loading, saving, cross-workbook copies, auto-fit, and every helper that touches cells by position all benefit; row and column insertion or deletion rebuilds the index in a single pass

Version 2.116.0

Added

• PDF export in the classic XLS engine: new SaveAsPDF methods on the worksheet and workbook render the used range (or a caller-supplied range through TXLSPDFExport) as a paginated grid — cell display text through the number format, gridlines, column widths derived from the Excel widths and scaled to fit the page, and automatic multi-page pagination
• Output is a self-contained PDF 1.4 file with no external dependencies, using the built-in Helvetica font with WinAnsi text encoding; portrait or landscape A4 orientation, page margin, and font size are configurable on the exporter
• This is a data-presentation export alongside HTML and RTF: it reproduces tabular content and layout for reports and document handoff rather than full per-cell styling

Version 2.115.0

Added

• One-call chart creation from a data range in the XLSX engine: new AddChartFromRange builds a complete chart and derives its series automatically from a rectangular data block — the equivalent of Excel's "insert chart from selection" — with a series-in-columns or transposed series-in-rows layout
• Named shortcuts AddColumnChart, AddBarChart, AddLineChart, AddPieChart, and AddAreaChart pick a chart type by name over the same one-call flow, reading series names from the header row and category labels from the first column
• Series references are qualified with the source sheet's quoted name, so a generated chart keeps working when placed on any sheet and survives save and reload

Version 2.114.0

Added

• RTF import in the classic XLS engine, the inverse of the existing SaveAsRTF export: new OpenRTF methods on the worksheet and workbook read an RTF document's table into worksheet cells — \cell advances the column, \row advances the row — starting at A1
• Recovers tabular data from files produced by SaveAsRTF, Word, or WordPad: numeric text is stored as numbers so formulas keep working, hex (\'hh) and Unicode (\uN) escapes decode correctly, and header groups (font table, color table, stylesheet, info) are skipped
• Documents with no table fall back to paragraph mode, importing each \par as the next row in column A

Version 2.113.0

Added

• Reading of encrypted .xlsx workbooks: OpenEncrypted now decrypts ECMA-376 Standard Encryption (AES-128) packages, completing the round trip with the existing SaveAsEncrypted — it reads the compound-file container, derives the key from the password, verifies it, and loads the recovered workbook
• A wrong password raises a clear "password is incorrect" error, and OpenEncrypted on a plain unencrypted .xlsx continues to load it directly, so callers can route every file through one entry point

Changed

• The unsupported-encryption message now names ECMA-376 Agile Encryption specifically as the remaining unsupported scheme, since Standard Encryption reading and writing are both available

Version 2.112.0

Added

• New TaggedText property on XLSX cells converts between rich text and lightweight markup: <b>, <i>, <u>, <s>, <sub>, <sup>, <color=#RRGGBB>, <font=Name>, and <size=N>, with < / > / & escapes — so multi-format cell content can be produced from a template string instead of assembling run objects by hand
• Writing markup builds the run list and keeps Value as the concatenated plain text; markup without tags stores a plain string with no rich payload; reading serializes each run's attributes as complete tag sets; unknown tags pass through as literal text

Fixed

• A rich-text run consisting entirely of whitespace (for example a single styled space between words) now survives save and reload; previously the space was dropped when the file was reopened

Version 2.111.0

Added

• New OnProgress event on the XLSX workbook reports load and save progress: it fires once per worksheet as it is parsed during Open or emitted during SaveAs, carrying the phase (reading or writing), the current 1-based sheet number, and the total — so applications can drive a progress bar over long multi-sheet operations at negligible callback cost

Version 2.110.0

Added

• New TXLSRowCursor dataset-style row cursor for .xlsx / .xlsm files: Open, SelectSheet, then FindFirst / FindNext walk a worksheet one row at a time with only the current row's cells in memory — a pull-driven alternative to TXLSDirectReader's callbacks for sequentially processing multi-million-row sheets
• Each row exposes its 1-based worksheet row number, the populated cells in document order (typed value, formula text, style index), and a ValueByCol lookup that returns Null for columns not populated on that row; rows with no populated cells are skipped
• Sheet names and count are available after Open (file and stream overloads), with shared strings and date-format detection handled the same way as the streaming reader

Version 2.109.0

Added

• New TXLSXCellFormat "format brush" for the XLSX engine: describe font, fill, border, alignment, number format, and protection once, then stamp any number of targets with Apply (single cell), ApplyRange, ApplyRows, ApplyCols, or ApplySheet
• Only the aspects actually touched are applied — a brush that sets just Font.Bold leaves the targets' fills, borders, and number formats untouched — so brushes compose with existing formatting like Excel's format painter
• Each apply resolves the touched aspects into the workbook style pools once and then writes only pool indices per cell, so formatting large ranges costs one pool lookup per aspect rather than per cell; ApplyRange materializes missing cells while the row/column/sheet forms stamp existing cells only

Version 2.108.0

Added

• Localizable boolean display text: the new XLSBooleanTrueText / XLSBooleanFalseText variables control how boolean cell values render wherever display text is produced — FormattedText, HTML export, and auto-fit measurement — so localized applications can show for example WAHR/FALSCH or VRAI/FAUX

Changed

• Boolean cells now display as TRUE / FALSE by default, matching Excel's grid, instead of the runtime's True / False capitalization

Version 2.107.0

Added

• New OnPassword event on the classic XLS workbook: when Open meets an encrypted workbook and the supplied password is missing or wrong, the event asks the application for a password and retries the open (up to three attempts) — so interactive tools can prompt the user instead of pre-loading passwords or inspecting error codes
• The event receives the current password by reference and a Retry flag; leaving Retry at False gives up and returns the original password error

Version 2.106.0

Added

• New TXLSFileMonitor class watches a workbook file on disk and fires OnFileChanged when another process — typically Excel — saves over it, so applications can reload and refresh their view automatically
• Polling-based (last-write time plus size, IntervalMs configurable, 1 s default), so it works on local disks and network shares alike without a message loop; SynchronizeEvent chooses between main-thread delivery for UI handlers (default) and direct watcher-thread delivery for console or service code
• A file that briefly disappears during a save cycle does not fire a false event; the change is reported once the new content is in place

Version 2.105.0

Added

• Bounded auto-fit in the classic XLS engine, matching the limits the XLSX engine already offers: new AutoFit(Columns, AMin, AMax), ColumnsAutoFit(AMinWidth, AMaxWidth), and RowsAutoFit(AMinHeight, AMaxHeight) overloads clamp every fitted column width (Excel width units) or row height (points) to the given range — 0 leaves a side unbounded, and an all-empty column or row still receives the requested minimum

Version 2.104.0

Added

• New read-only FormattedText property returns a cell's display text — the stored value rendered through its number format, the way Excel shows it in the grid — covering thousands separators, decimals, percentages, date and time masks, multi-section positive/negative/zero masks, and boolean formats
• Available in both engines: on IXLSRange in the classic XLS engine (multi-cell ranges render their anchor cell) and on TXLSXCell in the XLSX engine (rich-text cells return their concatenated run text)

Version 2.103.0

Added

• New DetectExcelContentFormat function (file name and stream overloads) identifies a workbook's container from its leading bytes rather than the file extension — reporting a BIFF/OLE compound file, an OOXML .xlsx package, an OpenDocument .ods package, or plain delimited text — so applications can route renamed or extension-less files to the right loader

Changed

• TXLSWorkbook.Open now sniffs the content signature and rejects ZIP-based packages (an .xlsx or .ods renamed to .xls) with a clear message naming the right loader, instead of failing with an opaque error code
• TXLSXWorkbook.Open now recognizes OLE compound-file content (a legacy .xls renamed to .xlsx, or an encrypted package) and reports what the file actually is, instead of surfacing a ZIP parser error

Version 2.102.0

Added

• Cross-workbook worksheet copy in the classic XLS (BIFF8) engine now brings the workbook-level defined names a formula uses over to the destination — the same thing Excel does when a sheet moves between workbooks: a name the destination already has is bound as-is (no duplicate is created), a missing name is registered with its original definition and visibility so the formula recompiles against it, and a name whose definition cannot resolve in the destination (for example, it points at a sheet that does not exist there) leaves the formula on its computed value as before

Version 2.101.0

Added

• Cross-workbook worksheet copy in the classic XLS (BIFF8) engine now re-points formulas at the source workbook as an external link when the destination has no sheet of the referenced name and the source has been saved to a file: the copied cell stays a live formula (its references switch to the [SourceBook]Sheet form), the source's computed value is kept as the cached result so reads keep working until the link is refreshed, and the external workbook reference round-trips through save and reload; formulas carrying defined-name references, and copies from unsaved source workbooks, keep the computed value as before

Version 2.100.0

Added

• Array formulas now survive worksheet copies in the classic XLS (BIFF8) engine: each array range is rebuilt as a whole in the copy — within the same workbook and across workbooks alike — with the root formula recompiled against the destination's sheet names; an array the destination cannot recompile keeps its computed values instead

Fixed

• Copying a range that contains array-formula cells between classic XLS worksheets no longer risks a crash on the spill cells (which carry no formula tree of their own); such cells now transfer their computed value

Version 2.99.0

Added

• Chart sheets can now be copied between classic XLS (BIFF8) workbooks: the chart record stream is cloned into the destination, series category / value / name references are recompiled against the destination's sheet names (a series whose sheet has no same-name counterpart falls back to an empty reference so the file still opens), chart font references migrate by value into the destination font table, and the page setup, tab color, zoom, and visibility of the chart sheet come along; this removes the last documented limitation of the cross-workbook worksheet copy

Version 2.98.0

Added

• Cross-workbook worksheet copy in the classic XLS (BIFF8) engine now preserves rich-text formatting inside cells: each formatting run's font is migrated by value into the destination workbook's font table and the run is rewritten to point at it, so mixed-font strings keep their appearance instead of falling back to plain text; phonetic (Far-East) auxiliary data travels with the string as well

Fixed

• Fonts referenced only by rich-text runs (and not by any cell format) are no longer dropped from the FONT table when saving a classic XLS workbook, and the runs now follow the same font reordering as cell formats on save, so rich-text strings keep the correct fonts after a save/reload cycle

Version 2.97.0

Added

• Cross-workbook worksheet copy in the classic XLS (BIFF8) engine now also clones embedded pictures: the image bytes migrate into the destination workbook's picture store (identical images are shared rather than duplicated), and each picture frame keeps its cell anchor, size, visibility, and transparent color; a new CrossWorkbookCopyXls sample project demonstrates the classic-engine copy end to end

Changed

• Classic XLS cross-workbook copy limitations are now documented: chart sheets are not copied (AddCopy returns nil for them), and a formula whose sheet reference has no same-name sheet in the destination keeps the source's computed value instead of becoming an external reference

Version 2.96.0

Added

• Cross-workbook worksheet copy in the classic XLS (BIFF8) engine now also clones conditional formats (cell-is and expression rules, data bars, color scales, icon sets, and their differential styles), data validations (list, numeric, date, text-length, and custom rules with prompt and error boxes), the AutoFilter definition with its per-column criteria (the dropdown buttons are rebuilt in the copy), and tables (ids and names are re-issued so they stay unique in the destination workbook); rule formulas are recompiled against the destination's sheet names, matching the behavior of cell formulas

Version 2.95.0

Added

• Classic XLS (BIFF8) worksheets can now be copied from one workbook into another: TXLSWorksheets.AddCopy adds a new sheet to the destination workbook and copies the source sheet's cell values, formulas (re-encoded as text so the destination recompiles them against its own sheet names, keeping same-name 3D references live), merged cells, comments, hyperlinks, row heights, column widths, outline levels, frozen panes, sheet protection, page setup, manual page breaks, tab color, and view settings; cell, row, and column formats are migrated by value into the destination workbook's format table, so a styled source produces an independently styled copy
• Cross-workbook format migration in the classic XLS engine now carries the full-precision RGB and theme colors alongside each format entry, so fills, fonts, and borders keep their exact colors instead of falling back to palette approximations; this also applies to range-level paste between workbooks

Version 2.94.0

Added

• Cross-workbook worksheet copy now clones conditional formats (with their differential styles and the dataBar / colorScale / iconSet visual payloads), tables, data validations, and charts (with every series); each sub-object deep-copies its own state into the destination worksheet
• Cross-workbook formula references are rewritten: a reference whose sheet name exists in the destination resolves as-is, otherwise it becomes an external reference back to the source workbook (which is registered as an external link) or #REF! when the source has no file identity; the rewrite applies to cell formulas, conditional-format rules, data-validation formulas, and chart series ranges alike

Version 2.93.0

Added

• A worksheet can now be copied from one workbook into another: TXLSXSheets.AddCopy adds a new sheet to the destination workbook and copies the source sheet's cells, formulas, merged ranges, column/row sizing, hidden/collapsed/outline flags, freeze panes, view state, protection, page setup, AutoFilter range, and tab colour onto it; the six per-cell style indices are remapped into the destination workbook's own Font/Fill/Border/NumberFormat/Alignment/Protection pools via value-equality lookup, so a styled source produces an independently-styled copy rather than indices pointing at the wrong pool
• The six per-attribute style pools (Fonts, Fills, Borders, NumberFormats, Alignments, Protections) each expose a 1-based LookupOrAdd that deduplicates by value; a new CrossWorkbookCopy sample project demonstrates the cross-workbook copy with style preservation

Version 2.92.0

Added

• XLSX workbooks now support pivot tables: TXLSXWorksheet.AddPivotTable scans a flat source range, auto-builds a pivot cache, and returns a typed pivot table for row / column / page / data-axis wiring, mirroring the long-standing BIFF8 (XLS) engine API so the same code targets either format; PivotSetDataFieldFormat assigns a number format to each aggregated value field
• Excel-created pivot tables round-trip without loss through Open then Save: the XLSX engine preserves each pivot cache definition, cache records, and pivot table part verbatim and replays the original bytes on save, so re-saving a workbook never drops or alters its pivots even when they use features the typed model does not expose

Version 2.91.0

Added

• A worksheet AutoFilter can now be applied to the data so the saved workbook opens with the non-matching rows already hidden, mirroring what Excel does when a value is picked from the filter dropdown
• Embedded images expose their geometry in real units (centimetres, inches, points) alongside the EMU values, and can be resized by a scale factor with optional aspect-ratio locking
• Page-setup margins can be set in centimetres directly, as a metric convenience over the inch-based margins
• Protected XLSX worksheets now expose fifteen fine-grained permission toggles (format cells, insert/delete rows and columns, sort, AutoFilter, select locked/unlocked cells, and more) via a new AllowOption property, matching Excel's Review > Protect Sheet checkbox list; the per-action settings round-trip through the file
• XLSX worksheets can now be duplicated (TXLSXSheets.Duplicate): a new sheet is added and the source's cells, merged ranges, column/row sizing, freeze panes, view state, protection, page setup, and tab colour are copied onto it, mirroring Excel's right-click > Move or Copy > Create a copy
• Twenty-two new sample projects cover capabilities the library supports: workbook encryption, cell comments, hyperlinks, defined names, AutoFilter evaluation, embedded pictures, an invoice template, advanced XLSX conditional formatting, merged cells, row/column outlining, freeze panes, multi-sheet layout, array formulas, range copy/move, find/replace, date systems, rich text, advanced CSV, HTML export options, custom number formats, hidden rows/columns, and worksheet protection

Version 2.90.29

Changed

• The streaming direct reader now counts every populated cell even when no per-cell handler is attached, so a plain ReadFile call can be used to size a worksheet quickly; the returned count and the CellCount property report the number of populated cells scanned

Version 2.90.28

Added

• Formulas can now be evaluated step by step, the equivalent of Excel's Evaluate Formula command: a new tracer reduces an expression one sub-step at a time, innermost first, and reports each reduction with its source sub-expression, the same sub-expression after its operands are substituted by their values, and the resulting value, so applications can show how a result is built up; it respects operator precedence, keeps range references intact for functions such as SUM, and reuses the worksheet calculation engine so every step agrees with a full recalculation

Version 2.90.27

Added

• XLSX workbooks can now be produced cell by cell through a new streaming direct writer that emits each worksheet straight into the package as rows arrive, without building a full in-memory cell tree, so very large files can be generated with a near-constant memory footprint; the writer handles inline strings, numbers, booleans, dates, and formulas across multiple sheets and round-trips with the streaming direct reader

Version 2.90.26

Added

• XLSX workbooks can now be scanned cell by cell through a new streaming direct reader that raises per-cell and per-sheet events without building a full in-memory cell tree, keeping the memory footprint near constant for very large files; each reported cell carries its value, kind (number, string, boolean, date, or error), formula text, and style index, with optional date detection and early-abort support

Version 2.90.25

Added

• Classic XLS worksheets can now add hyperlinks from A1-style cell references while preserving URL normalization, workbook-target subaddresses, display text, and screen tips

Version 2.90.24

Added

• Classic XLS page breaks can now be added from one-based row or column positions and A1-style cell references through page-break collection and worksheet helper overloads

Version 2.90.23

Added

• Classic XLS worksheets and shape collections can now create rectangle, oval, line, and typed autoshape drawing objects from A1-style range strings

Version 2.90.22

Added

• Classic XLS shape collections can now create text boxes from A1-style range strings through an AddTextBox overload

Version 2.90.21

Added

• Classic XLS text boxes can now be repositioned from A1-style range strings through a new SetArea overload

Version 2.90.20

Added

• Classic XLS worksheet text boxes can now be created from A1-style range strings through an AddTextBox overload

Version 2.90.19

Added

• Classic XLS worksheets can now create drawing text boxes directly from row and column anchors through a new AddTextBox method

Version 2.90.18

Added

• Classic XLS worksheets can now create cell comments directly from row and column coordinates through new AddComment overloads