pdf_cos 1.0.0

COS object layer for PDF: tokenizer, parser, filters (incl. CCITT, JBIG2, JPEG 2000), xref machinery, encryption, and serializer. Pure Dart, web-ready.

Changelog

1.0.0

First stable release. Changes since 0.1.0:

• JBIG2: Huffman-coded symbol dictionaries and text regions, generic refinement regions, and pattern dictionaries with halftone regions.
• JPEG 2000: reset-probabilities (RESET) code-block style support.
• Inline images: correct data-length detection for DCT-filtered streams.

0.1.0

Initial release.

• COS object model: dictionaries, arrays, names, strings, streams, references.
• Lenient tokenizer/parser for real-world PDFs (broken /Length, missing endobj, junk before the header, broken xref chains with object-scan recovery).
• Cross-reference tables and streams; lazy object loading; incremental updates.
• Filters: Flate, LZW, RunLength, ASCIIHex, ASCII85, DCT passthrough, CCITT G3/G4, JBIG2 (embedded profile), JPEG 2000.
• Encryption: RC4, AES-128, AES-256 decryption and encrypt-on-write.
• Crypto primitives for signing/validation: ASN.1, RSA, ECDSA, CMS, X.509 chain verification.
• Content-stream tokenizer and a from-scratch document builder.