Compression Overview

Overview

HotPDF uses several compression and decompression paths depending on whether it is generating new PDF content or reading existing streams and images from external files.

Main Compression Paths

FlateDecode: Used for generated PDF streams through THotPDF.Compression and backed by zlib-ng in zlib-compatible mode, with 64-bit runtime SIMD dispatch where available
LZWDecode: Supported for legacy PDF and TIFF-style data through HPDFLZW.pas
DCT/JPEG: Used for standard JPEG image workflows handled by HotPDF image support and backed by libjpeg-turbo 3.1.90 with NASM SIMD acceleration objects
JBIG2 and JPEG2000: Available for specialized image workflows where the source format already uses those codecs

PDF 1.5+ structural size reduction

For PDF 1.5+ output HotPDF can layer two additional structural compression paths on top of FlateDecode-compressed streams. Both are opt-in and pair naturally:

UseXRefStream: Replace the textual xref ... trailer cross-reference table with a compressed /Type /XRef stream (ISO 32000-1 7.5.8). Cuts xref overhead by 4-10x on large documents.
UseObjectStreams: Pack small indirect dictionaries (multi-page documents, annotations, AcroForm widgets, structure-tree elements, Optional Content layers) into /Type /ObjStm container streams (ISO 32000-1 7.5.7) referenced through type-2 xref entries. Requires UseXRefStream to also be on. A 30-page graphics-only smoke shrinks from 14502 to 8488 bytes (-41.5%) with the two toggles combined.

Related Units

HPDFZLib.pas for FlateDecode and predictor-aware inflate helpers
HPDFLZW.pas for LZW decompression
HPDFJBIG2.pas for JBIG2 decoding support
HPDFJpeg2000.pas and HPDFOpenJpeg.pas for JPEG2000-related workflows

See Also