Skip to content

CHANGELOG.md

Latest commit

 

History

History
232 lines (208 loc) · 12.3 KB

CHANGELOG.md

File metadata and controls

232 lines (208 loc) · 12.3 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.5.0] - Unreleased

Added

Changed

Fixed

[0.4.0] - 2026-04-26

Added

  • Phase A-1: Image embedding pipeline — file path and data URI images are read, base64-decoded or loaded from disk, and embedded as BinData/ entries in the HWPX ZIP; binaryItemIDRef references are emitted on <hp:pic> elements; HTTP/HTTPS URLs are preserved as external references.
  • Phase A-2: List writer with bullet/numbering — <hh:numbering> definitions (BULLET id=1, DIGIT id=2) emitted in header.xml; numPrIDRef set on list-item paragraphs; nested list support via write_list_items() with depth-based paraPrIDRef (id=2 depth-0, id=3 depth-1+); 23 new list tests.
  • Phase A-3: BlockQuote visual indentation — paraPr id=1 with left margin 800 HWP units (~20mm); quote_depth parameter threaded through write_block(); 6 new blockquote tests.
  • Phase A-4: Footnote OWPML structure — <hp:fn noteId> wrapping around footnote content blocks; 7 new footnote tests.
  • Phase B-1: HWPX nested list reader — StagedBlock enum for list-paragraph grouping; group_list_paragraphs collapses flat sequences into nested Block::List; paraPrIDRef/numPrIDRef parsing from <hp:p>; 13 new reader list tests.
  • Phase B-3: HWPX lenient XML error recovery — malformed section XML parsing continues with partial results; missing section files skipped with warning; missing attributes use defaults.
  • Phase C-3: Code block language preservation — language hint stored as <!-- hwp2md:lang:X --> XML comment in HWPX; reader parses it back; MD→HWPX→MD roundtrip preserves language info; --> injection sanitized.
  • Phase B-2: HWP binary list recognition — two-tier detection: numbering_id from ParaShape records, then text-prefix heuristics (●■▶•-* bullets, 1./2)/a. ordered); year-like prefix rejection; StagedBlock grouping; 32 new HWP list tests.
  • Phase B-4: Page layout parsing — PageLayout IR struct (width, height, landscape, margins); HWPX reader parses <hp:secPr> → <hp:pagePr> → <hp:pageSize>/<hp:margin>; writer emits <hp:secPr> with A4 portrait defaults; 11 new page layout tests.
  • Phase C-1: Heading paraPr (id=4) with 180% line spacing; ParaPrConfig struct replacing bare (id, left_margin) parameters; 14 new paraPr tests.
  • Phase D-1: Comprehensive roundtrip integration tests — 37 tests covering all block types (paragraph, H1-H6, ordered/unordered/nested list, table, code block, horizontal rule, blockquote, image, footnote, inline formatting, combined document).

Changed

  • paraPr table expanded from 2 to 5 entries (normal, blockquote, list-depth-0, list-depth-1+, heading).
  • writer_tests_roundtrip.rs split: golden tests → writer_tests_golden.rs, code language tests → writer_tests_code_lang.rs.
  • ir::Section now carries page_layout: Option<PageLayout>.
  • Headings use paraPrIDRef="4" (180% line spacing) outside blockquotes.

Fixed

  • Image filename collision: counter suffix dedup (photo_2.png) instead of silent drop; unique_entry_name bounded to 10,000 iterations.
  • XML comment injection: code language --> sanitized via -- collapse.
  • flush_paragraph marked #[cfg(test)] instead of #[allow(dead_code)].

[0.3.1] - 2026-04-26

Added

  • Phase 16: Golden file test (golden_comprehensive_document_structure): validates internal ZIP XML structure (section0.xml, header.xml, content.hpf, mimetype) of generated HWPX archives; OWPML schema validation re-verified with polaris DVC after inline charPr changes.
  • Phase 15: faceNameIDRef attribute emitted on section-level inline <hp:charPr> when the inline carries a font_name, completing the font name write-to-read roundtrip; 4 new font name roundtrip tests.
  • Phase 14: Section-level inline <hp:charPr> emission in write_inline_charpr(): bold, italic, underline, strikeout, superscript, subscript, and color attributes are now written inside <hp:run> elements for OWPML conformance; 16 new bold/italic/underline/strike/color roundtrip tests; 3 new xml_escape_content tests (apostrophe, all special chars, passthrough).
  • Phase 13: Font name reader: parse_face_names() from header.xml <hh:fontface> entries; faceNameIDRef / hangulIDRef resolution in apply_charpr_attrs; with_font_name() builder method on Inline; 9 new font resolution tests; README updated with Phase 9-12 features (hyperlinks, ruby, footnote_ref, inline code, metadata).
  • Phase 12: writer_tests.rs split into 8 topic-based test modules (writer_tests_charpr, writer_tests_section, writer_tests_metadata, writer_tests_hyperlink, writer_tests_ruby, writer_tests_footnote, writer_tests_roundtrip); ruby + hyperlink combination test added.
  • xml_escape_content now covers the complete set of XML 1.0 predefined entities (&, <, >, ", ').

Changed

  • ..Default::default() struct update syntax eliminated across the entire codebase (md/parser.rs, hwp/convert.rs) — all fields are now set explicitly.

Fixed

  • trim_start_matches('#') replaced with strip_prefix('#') in color attribute emission to prevent stripping multiple # characters.
  • Font name propagation in flush paths (flush_paragraph, flush_cell_paragraph, flush_list_item_paragraph, flush_footnote_paragraph) now chains .with_font_name().
  • Replaced .unwrap() with if let pattern in write_inlines to prevent potential panic on link URL access.
  • Empty text inline run guard in writer_section.rs: zero-length text runs are skipped, preventing emission of empty <hp:t/> elements.
  • Dead code removed: InlineStyle.code field (superseded by the CharPrKey path introduced in Phase 8).
  • Broken intra-doc link in md/mod.rs corrected.

[0.3.0] - 2026-04-26

Added

  • Phase 11: Crate-level //! documentation on lib.rs; /// doc comments on all public types in ir.rs, error.rs, and all public functions in convert.rs; #![warn(missing_docs)] lint enabled; xml_escape_content extended with " → &quot; escaping for defense-in-depth.
  • Phase 10: Ruby annotation writer (hp:ruby / baseText / rubyText); footnote_ref writer emitting hp:noteRef; Inline builder pattern with with_formatting / with_link / with_ruby constructors; ruby formatting propagation fix ensuring annotation runs inherit the base run's charPr.
  • Phase 9: Superscript/subscript writer using the supscript charPr attribute; hyperlink reader and writer using fieldBegin / fieldEnd controls; writer module split into three focused submodules; reader run-start reset fix preventing stale formatting from leaking across paragraphs.
  • Phase 3: charPr / paraPr / fontface reference tables in header.xml with IDRef linking between section paragraph runs and the header table entries.
  • Phase 4: Style table (hh:styles) with Normal + Heading1-6, numeric styleIDRef and charPrIDRef values replacing string-form references for OWPML schema compliance.
  • Phase 5: Sequential paragraph IDs (id attribute on <hp:p>), table block wrapping in <hp:p>/<hp:run>/<hp:tbl> hierarchy, heading-specific charPr entries with level-differentiated font heights.
  • Phase 6: OWPML schema validation pass (enable_schema=true) with polaris DVC; fixed breakSetting attribute set, align horizontal/vertical attrs, margin child elements, heading / border / autoSpacing / lineSpacing required children in paraPr.
  • Phase 7: hh:borderFills table with default entry (id=1), slash / backSlash / border / diagonal children; borderFillIDRef on every charPr entry; polaris_dvc rev pinning; schema validation expansion covering all writer-emitted elements.
  • Phase 8: Inline code (code: true) mapped to distinct charPr entry with Courier New monospace font; metadata preservation (hp:docInfo with title and author in content.hpf); HWPX structural roundtrip tests; dead code audit; version bump to 0.3.0.

Changed

  • CharPrKey struct gains a code field; from_inline() now forces Courier New font for inline code spans, producing a distinct charPr ID.
  • generate_content_hpf() emits <hp:docInfo> when the document carries title or author metadata.
  • README library usage example updated to hwp2md = "0.3".

Fixed

  • Font registration for inline code spans: the monospace font is now registered from the resolved CharPrKey (which overrides to Courier New for code) rather than from the raw IR inline's font_name field, ensuring the font table entry always exists when referenced by a charPr.

[0.2.0] - 2026-04-23

Added

  • Ruby text control parsing (ruby ctrl_id): base text and phonetic annotation are extracted and emitted as <ruby> HTML in Markdown output.
  • Lenient CFB fallback: corrupted or partially-written HWP files are now partially recovered instead of returning a hard error. Successfully read sections are returned with a warning.
  • Distributed-document decryption: HWP files with the distributed flag set are now decrypted with AES-128 ECB before parsing.
  • EQEDIT → LaTeX converter: HWPTAG_EQEDIT blocks are converted to fenced $$ display-math blocks in Markdown output.
  • Image embedding: gso (GShapeObject) controls are resolved to their BinData streams and written as ![](data:image/...) inline Markdown.
  • Footnote and endnote parsing: fn and en controls are extracted and rendered as Markdown footnote references with collected definitions.
  • Hyperlink extraction: hyln controls produce [text](url) Markdown links with URL sanitisation (scheme allow-list: http, https, mailto).
  • Superscript and subscript character shapes are now mapped to <sup> / <sub>.
  • Heading type detection from HWPTAG_PARA_SHAPE outlineLv attribute.
  • HWPX colspan and rowspan extraction for table cells.
  • Table parser now enforces a row-count cap (4 096) to prevent allocation DoS.
  • Decompression-bomb guard: deflate output is capped at 256 MB; exceeding this limit returns Hwp2MdError::DecompressionBomb.
  • CFB stream reads are capped at 256 MB (MAX_CFB_STREAM).
  • GitHub Actions CI workflow (cargo test, cargo clippy -- -D warnings).
  • 614 unit and integration tests (82%+ line coverage).

Fixed

  • 6 CRITICAL security issues resolved in the initial rewrite phase, including unbounded allocation on untrusted record sizes and integer overflow in dimension calculations.
  • Infinite-loop in EQEDIT tokeniser on malformed input (recursion depth guard).
  • Offset clamp in read_utf16le_str to prevent out-of-bounds reads.
  • Heading level clamped to 1–6 (H7+ demoted to H6).
  • URL deduplication: identical hyperlinks within the same paragraph are collapsed to a single link.
  • parse_records extended-size field handling (0xFFF sentinel + follow-up u32).
  • Zlib fallback path for deflate-wrapped CFB streams.
  • Dead CTRL_EQUATION, CTRL_HEADER, CTRL_FOOTER constants removed.

Changed

  • Replaced all HWP-specific third-party crate dependencies with a self-contained parser based on the cfb (OLE2) and zip (HWPX) crates.
  • control.rs split into category modules (table, image, hyperlink, ruby, dispatcher, common) for maintainability.
  • reader.rs split into focused submodules.
  • tracing::debug! demoted to tracing::trace! for high-frequency paths (table dims, image shapes, unhandled ctrl_id, list-item hints).
  • MAX_CFB_STREAM deduplicated: defined once as pub(crate) in reader.rs and imported in summary.rs.
  • Error types converted to thiserror-derived enum (Hwp2MdError) with structured variants instead of stringly-typed errors.
  • Markdown inline escaping hardened: characters that start a Markdown construct at line start are escaped.

[0.1.0] - 2026-04-02

Added

  • Initial project scaffolding with HWP 5.0 (CFB-based) and HWPX (ZIP/XML) parsing skeletons.
  • Basic paragraph text extraction.
  • CLI binary (hwp2md) with --input, --output, and --style flags.
  • Bidirectional conversion: HWP → Markdown and HWPX → Markdown.
  • Markdown → IR → Markdown roundtrip support.