Automatic XFA native graphics and images
Unit: HPDFXFAHeadless
HPDFXFAWriteHeadlessPDF consumes rectangle, line, arc and inline image draw callbacks directly from XDP, alongside the existing text and widget consumers
Each draw produces a native Form XObject placed on the actual layout page; image draws contain real Image XObjects and alpha soft masks when needed
Vector paths
Rectangles use native PDF rectangle paths, lines preserve slash or backslash slope, and elliptical arcs use cubic segments of at most 90 degrees
The first effective edge supplies RGB stroke colour and thickness, including subpoint widths; stroke appearance bounds include half the thickness so the form does not clip its own border
Arcs consume startAngle and signed sweepAngle, with a default full revolution and a maximum absolute sweep of 360 degrees
Draw rotation, scale and translation hints are applied once in a finite affine placement around the draw centre, with counterclockwise positive rotation; top-left XFA coordinates are converted once to the native PDF origin
PNG and JPEG
| Format | Actual native consumption |
|---|---|
| PNG grayscale | 1, 2, 4, 8 and 16-bit samples, with packed samples expanded to 8 bits and 16-bit precision retained |
| PNG RGB | 8 and 16-bit colour samples |
| PNG indexed colour | 1, 2, 4 and 8-bit indices expanded through the palette into RGB samples |
| PNG transparency | Grayscale and RGB transparent samples, palette alpha, and 8 or 16-bit grayscale-alpha and RGBA; actual /SMask streams preserve alpha precision |
| PNG filtering and interlace | All five PNG row filters and Adam7 passes, validated through the shared native zlib and predictor implementations |
| JPEG | 8-bit baseline and progressive grayscale, RGB and CMYK frames, validated with the shared native JPEG decoder before preserving the original /DCTDecode stream |
PNG chunk checksums, ordering, dimensions, palette indices, transparent sample ranges and exact decompressed length are checked before committing a candidate; truncated or excess compressed data fails preparation
JPEG frame dimensions must agree with real native decoding, and Adobe CMYK inversion is declared explicitly in the image dictionary
JPEG requires the existing native TurboJPEG runtime; HOTPDF_TURBOJPEG_LIBRARY can identify an absolute library path, with dependent libraries discoverable by the operating system
Image aspect supports fit, none, width, height and actual; PNG pHYs and JPEG JFIF density determine physical pixel size, with 96 dpi used when no usable density exists
Scaled images are centred and clipped to the XFA draw rectangle; alpha composition remains a native PDF reader operation
Bounds and transaction
Encoded image bytes, decoded working buffers and raw image or alpha streams are bounded before allocation; native object limits and aggregate image and font stream limits remain active
Cancellation is checked whilst parsing chunks, inflating data, decoding rows and placing native objects; the existing native JPEG decoder receives the caller token
Preparation failures preserve destination position, length and bytes; the final destination WriteBuffer follows the caller stream's own I/O semantics
Current boundaries
Filled shapes, corner styling, edge-specific borders, dash patterns, barcodes, external image transport, imageEdit widgets, alternate image formats, embedded color profiles and Exif orientation still require additional consumer integration
Animated PNG is consumed as its default static image; foreign XML exData remains a separate feature, and this draw adapter does not execute arbitrary embedded objects
Acceptance
Tests/Linux/Build-XFAHeadlessGraphicFixtures.py produces deterministic row-filter, Adam7, sample-precision, transparency, physical-aspect, JPEG and malformed protocol fixtures
Tests/Linux/XFAHeadlessGraphicsTests.pas writes and reopens actual native PDFs and verifies caller preservation on failures; Verify-XFAHeadlessGraphics.py independently checks decoded image samples, /SMask precision, vector operators, placement and rendered pixels with pypdf and MuPDF
Automatic XFA native PDF creation describes the existing field and text consumers and the public options and result APIs