pywpdoc 0.44.0rc1__tar.gz

pywpdoc-0.44.0rc1/CHANGELOG.md

@@ -0,0 +1,577 @@
+## 0.44.0rc1 — release structure candidate (not published)
+
+- Consolidate product implementation under `src/pywpdoc/`, with internal `_engine`; pre-release legacy import shims are deliberately omitted.
+- Consolidate deterministic business QA, Windows COM validation tools, and historical synthetic fixtures under `tests/`.
+- Introduce authentic native-rendered README image assets and unified README metadata for GitHub/PyPI.
+- Prepare CI and manual-approval PyPI Trusted Publishing workflows.
+- No intentional changes to business document algorithms. The 0.44.0rc1 packaged wheel passed focused Windows Word/Hancom native tests: 11/11 core cases and 7/7 extended QA stages (no new PDF/pixel-layout run).
+- Add explicit portable `pywpdoc.Document` methods, optional image-format sniffing, and a documented no-argument `add_section` contract without changing the backend engines.
+- Introduce bilingual MkDocs API manual, tested DOCX/HWPX examples, API contract tests, and CI docs-build gate; [API manual deployed to GitHub Pages](https://rydolabs.github.io/PyWPDoc/).
+
+# 0.43.1rc2 — Untracked-format safety and CI candidate (2026-10-15)
+
+- `compare(everywhere=False)` now rejects untracked DOCX table-border, cell-border, shading, merge and geometry changes rather than silently accepting them.
+- Body-only compare also checks DOCX drawing appearance/payload, and native HWPX table geometry and image data for out-of-scope changes.
+- Public source's default `pytest` skips the explicitly declared private fixture cases when the licensed source fixture is absent; public Actions still use the narrower published CI selector.
+- Windows native Actions and QA launchers select and validate the current wheel version instead of assuming 0.43.0.
+- Native Microsoft Word / Hancom 2020 verification of this release candidate is pending.
+
+# 0.43.1rc1 — Safety stabilization candidate (2026-10-15)
+
+- Restore native DOCX package when rebuilding DOCX to DOCX; reject missing payloads.
+- Reject unsupported DOCX to HWPX rebuild explicitly.
+- Reject body-only comparisons when excluded table/story content has changed.
+- Preflight strict template rendering before any changes.
+- Cache native run-to-paragraph linkage to remove quadratic HWPX text replacement traversal.
+- Avoid exact sub-line-height Word table rows when rebuilding simple HWPX tables.
+- 709 existing tests + 16 new defect regressions passed locally (725 total). Windows GUI revalidation pending.
+
+## v0.43.0 — 보수적 정규식 검색·치환 (2026-10-13)
+
+- Added `Document.find_regex(pattern, *, everywhere=True, flags=0, count=-1)` and `Document.replace_regex(pattern, replacement, *, everywhere=True, flags=0, count=-1)` for DOCX/HWPX.
+- Reused core `TextMap`, safe-boundary checks, paragraph traversal and structure-preserving `replace_text_range` without changing existing literal APIs or native Track Changes writers.
+- Python `re` capture groups/backreference expansion; unsupported zero-width matches and invalid replacements are rejected before mutation, opaque-boundary matches are skipped.
+- Regression tests: 623/623 (588 preexisting + 35 new). See `REGEX_FIND_REPLACE_V0430_KO.md` and `QA_REGEX_V0430.md`.
+
+## v0.42.0 — 검색·치환 미리보기 및 JSON 변경 보고 (2026-10-13)
+
+- Added `Document.preview_replace(search, replacement, *, everywhere=True, case_sensitive=True, count=-1)` to inspect literal replacement candidates without modifying DOCX/HWPX.
+- Uses exactly the existing replace matcher and text traversal; supports split runs, safe hyperlink joins, global replacement limits, case folding and body-only filtering.
+- Added immutable `ReplacementPreviewReport` / `ReplacementPreviewMatch`, `to_dict()`, and `to_json()` (Korean Unicode preserved).
+- Reports original zero-based logical paragraph offsets and before/after logical text; opaque/native non-text elements are not reconstituted in this preview.
+- Matching across unsafe field/revision/content-control boundaries remains excluded. Successful preview does not promise future writer success if document or backend changes.
+- Existing `replace_text()`, `find()`, `render_template()`, and Track Changes writers unchanged.
+- QA: `REPLACE_PREVIEW_V0420_KO.md` and `tests/test_replace_preview_v0420.py`.
+
+## v0.41.0 — 템플릿 사전 검사 (2026-10-13)
+
+- `Document.check_template()` 추가 (DOCX/HWPX 공통, 읽기 전용).
+- 누락된 활성 변수·섹션/반복 문맥·템플릿 문법·불안전 텍스트 경계 검사.
+- 구조적 반복 행 문맥은 기존 렌더러의 행 탐지 로직 재사용.
+- `TemplateCheckReport`/`TemplateCheckIssue` 및 `to_dict()` 제공.
+- 기존 `render_template()` 구현 및 인수는 변경 없음.
+- 테스트/제한 범위: `TEMPLATE_PREFLIGHT_V0410_KO.md`.
+
+# Changelog
+
+## 0.40.0 (QA batch diagnostics add-on; no product version bump)
+
+- Added a single-file Hancom 2020 native batch probe with ten isolated text/format/structure cases, one-edit-session instructions, per-case validation and issue reports.
+- Added `quality/run_batch_checks.py` to run all pytest cases and the native diagnostic gate independently; pytest failure never stops native analysis.
+- Linked text marker TcIds and format anchor IDs to matching header change IDs to prevent other scenarios' revisions from producing false passes.
+- Kept HWPX structural Track Changes fail-closed; case-level structural observations are never a native writer enablement signal.
+
+
+- Record Hancom 2020 confirmation for the v0.39.0 semantic named-style identity QA.
+- Keep HWPX table-row structural Track Changes fail-closed because no verified `hp:tr` revision anchor exists in current native evidence.
+- Add a read-only HWPX row-tracking analyzer that compares package entries, header TrackChange records, inline marker scope, paragraph/run anchors, table row matrices, and possible history/version-log parts.
+- Add isolated native Hancom 2020 row-insert and row-delete probe fixtures plus expected-final references and machine-readable manifests.
+- Enrich `compare-hwpx-table-row-track-unsupported` context with the verified TrackChange type set and public marker scope so callers can distinguish an intentional evidence boundary from a generic unsupported feature.
+
+## 0.39.0
+
+- Record Hancom 2020 GUI confirmation for v0.38.0 embedded image-bullet Accept/Reject behavior.
+- Compare HWPX paragraph named-style identity semantically instead of comparing raw header-local `styleIDRef` numbers.
+- Allow equivalent named styles with different local ids across unrelated documents to participate in ordinary ParaShape tracking.
+- Detect the inverse collision where equal numeric style ids resolve to different style identities and fail closed rather than silently missing the change.
+- Resolve `nextStyleIDRef` by target style identity so self/next-style local id differences do not create false revisions.
+- Keep actual named-style identity changes fail-closed; no unsupported style-identity TrackChange record is synthesized.
+- Add `qa_hwpx_style_identity_v0390` and release-QA hash/structure checks.
+
+## 0.38.0
+
+- Added cross-document HWPX image-bullet import/remap for ParaShape Track Changes.
+- Image bullets now carry their OPF manifest entry and embedded `BinData` payload into the NEW-based review clone.
+- Resolved bullet comparison includes binary payload identity, preventing equal local `binaryItemIDRef` values from masking different images.
+- Added payload deduplication when the OLD image already exists in the target package.
+- Extended validation to reject dangling image-bullet manifest and BinData references.
+- Kept external/non-embedded image dependencies and named-style identity changes fail-closed.
+
+## 0.37.0
+
+- Hancom 2020 GUI confirmation recorded for v0.36.0 HWPX ParaShape tracking.
+- Added cross-document ParaShape resource import/remap for tabPr, paragraph borderFill, numbering, text bullets, and nested paraHead charPr dependencies.
+- Paragraph-format comparison now resolves header-local resource meaning instead of comparing numeric ids across documents.
+- Added validator checks for ParaShape dependency references.
+- Image bullets and named-style identity changes remain fail-closed.
+
+## 0.36.0
+
+- HWPX ParaShape Track Changes candidate: `trackChange@type="ParaShape"` + OLD `parashapeID` + changed `hp:p@paraTcId`.
+- Grounded the paragraph anchor/header contract in Hancom's public OWPML model.
+- Fixed ParaShape-only revision plans being skipped by an early-return condition.
+- Added alignment, line-spacing, mixed-format, overlap, named-style, and missing-OLD-paraPr safety coverage.
+- Added `qa_hwpx_parashape_v0360` and `QA_INDEX_V0360.json`; Hancom 2020 GUI Accept/Reject remains pending.
+
+## 0.35.0 - HWPX cross-document CharShape resource import/remap
+
+- Extend native-backed HWPX CharShape Track Changes across independently-authored documents whose OLD character resources are absent from NEW.
+- Resolve `charPr` semantics through referenced font definitions and border fills instead of comparing raw numeric IDs only.
+- Import/deduplicate OLD fonts per language block and self-contained border fills, remap `fontRef`/`borderFillIDRef`, then author `trackChange@charshapeID` against the remapped OLD shape.
+- Keep the NEW input immutable; imports are applied only to the NEW-derived review clone.
+- Add validator checks for dangling `charPr` font and border-fill references.
+- Fail closed for embedded-font payloads or unverified external border-fill dependencies.
+- Preserve the Hancom 2020-native CharShape anchor contract (`run@paraTcId`) and keep ParaShape tracking fail-closed.
+- v0.34.4 Hancom 2020 GUI display + Accept/Reject QA was user-confirmed before this increment.
+
+## 0.34.4 - Hancom 2020-native CharShape tracking candidate
+
+- Re-enabled HWPX **character-format** Track Changes from a real Hancom 2020-authored oracle supplied during GUI QA.
+- Match Hancom 2020 native authoring: `trackChange(type="CharShape", charshapeID=<old charPr>)` plus `hp:run@paraTcId=<change id>`.
+- Preserve `trackchageConfig` tracking bit behavior (`flags |= 1`) and native author metadata.
+- Character-format-only and mixed text + character-format comparisons are enabled again.
+- Keep HWPX paragraph-format (`ParaShape`) Track Changes fail-closed pending an independent native ParaShape oracle.
+- Add a machine-readable native-oracle contract receipt and release QA with identical OLD/NEW/TRACKED text covering font, bold, and size changes.
+- Full regression: 452/452 passed before packaging.
+
+## 0.34.3
+
+- Correct release QA packaging so unrelated HWPX formatting OLD/NEW and text TRACKED artifacts cannot be presented as one scenario.
+- Add `scenario_id` manifests, SHA-256 checks, and `QA_INDEX_V0343.json`.
+- Keep HWPX formatting Track Changes fail-closed pending a native Hancom-authored fixture.
+- Preserve supported HWPX text Track Changes, paragraph structural tracking, DOCX formatting revisions, and v0.34.1 Hancom 2020 compatibility fixes.
+
+## 0.34.2 - HWPX formatting Track Changes safety rollback
+
+- Retracted HWPX CharShape/ParaShape Track Changes from the supported surface after Hancom 2020 GUI QA showed that the synthesized revisions were not displayed as formatting changes.
+- HWPX character/paragraph formatting differences now fail closed with `compare-hwpx-format-track-unsupported` before any review metadata is authored.
+- Mixed HWPX text + formatting comparisons also fail before partial review output is produced.
+- HWPX text revisions and whole-paragraph structural revisions remain supported; DOCX formatting revisions remain supported.
+- Kept the v0.34.1 Hancom 2020 creation profile and font-size unit fixes.
+- Full regression: 439/439 passed.
+
+## 0.34.1 - Hancom 2020 creation compatibility
+
+- Fixed HWPX font-size normalization for foreign EMU-backed length objects such as `docx.shared.Pt`; 14pt now serializes as `charPr@height="1400"`.
+- Added validation for invalid/non-positive HWPX font-size inputs.
+- Added a Hancom 2020-targeted real-authored blank HWPX skeleton and made it the default for new HWPX documents.
+- Added `compatibility="latest"` to explicitly retain the previous 13.x / xmlVersion 1.5 creation skeleton.
+- Preserved `version.xml` byte-for-byte when opening/editing existing HWPX documents and when producing compare output from NEW.
+- Preserved the existing header/footer style-ID contract while using the 2020-targeted skeleton.
+- Full regression: 439/439 passed.
+
+## 0.34.0 - Tracked HWPX formatting comparison
+
+- Added HWPX `CharShape` Track Changes anchored by `hp:run@charTcId`.
+- Added HWPX `ParaShape` Track Changes anchored by `hp:p@paraTcId`.
+- Kept the current/new shape references in body XML and linked revision metadata to the prior shape definition for reject semantics.
+- Added format-only Accept/Reject projection tests and mixed text + format coverage.
+- Added fail-closed boundaries for missing prior header definitions, named-style identity changes, and pre-existing format revision anchors.
+- Preserved the v0.33.0 baseline; full suite: 431/431 passed.
+
+## 0.33.0 - Tracked DOCX formatting comparison
+
+- Added DOCX run-format Track Changes using native `w:rPrChange` with previous run properties.
+- Added DOCX paragraph-format Track Changes using native `w:pPrChange` with previous paragraph properties.
+- Added format-aware planning that ignores harmless run fragmentation and splits revisions at both old/new run boundaries.
+- Added mixed text + formatting revision coverage and explicit section-format fail-closed behavior.
+- Kept HWPX format-only changes fail-closed until a verified native CharShape/ParaShape anchoring contract is available.
+- Preserved the v0.32.0 baseline; full suite: 425/425 passed.
+
+## 0.32.0 - Tracked body-paragraph comparison
+
+- Added top-level body block alignment so paragraph insert/delete preserves position around table anchors.
+- Added DOCX whole-paragraph Track Changes with revision-marked paragraph content and paragraph marks.
+- Added HWPX whole-paragraph insert/delete ranges with `paraend="1"`.
+- Added adjacent multi-paragraph, empty-paragraph, table-adjacent, and mixed structural/inline revision coverage.
+- Added fail-closed boundaries for unsafe paragraph imports and multi-section structural edits.
+- Preserved the v0.31.0 baseline; full suite: 418/418 passed.
+
+## 0.31.0 - Tracked DOCX table-row comparison
+
+- Added backend-neutral table-row revision planning.
+- Added native DOCX row insert/delete revisions with `w:trPr/w:ins` and `w:trPr/w:del`.
+- Kept HWPX row structural Track Changes fail-closed because no verified row-level OWPML grammar was established.
+- Added Accept-All => new / Reject-All => old structural projection tests.
+- Preserved the v0.30.0 baseline; full suite: 399/399 passed.
+
+## 0.30.0 - Structural table-row capability and repetition
+
+- Added backend-neutral structural table traversal and minimal row
+  clone/insert/remove contracts.
+- Added DOCX row cloning with fresh drawing identities while preserving native
+  row/cell structure, horizontal merges, nested tables, formatting, and media
+  relationships.
+- Added HWPX row cloning with refreshed paragraph/object identities, stale
+  layout-cache removal, preserved nested structure/media references, and native
+  merge-aware removal.
+- Added `render_template(..., sections=True, table_rows=True)` for an outer
+  Mustache section that wraps an entire table row.
+- Added sequence repetition, inverted row conditions, item context with parent
+  fallback, fragmented markers, nested-table content rendering, and single-pass
+  protection for template-like data values.
+- Vertical merge ambiguity and emptying a table fail closed.
+- `everywhere=False` retains the historical no-table template scope.
+- Preserved all v0.29.0 baselines; full suite: 388/388 passed.
+
+## 0.29.0 - Tracked text comparison
+
+- Added `Document.compare(new, author=..., date=..., everywhere=True)` as an
+  isolated high-level comparison feature.
+- Added backend-neutral `RevisionPlan`, `ParagraphRevision`, and `TextEdit`
+  models; alignment/diff policy remains outside DOCX/HWPX XML writers.
+- Added DOCX native revision output using `w:ins`, `w:del`, `w:delText`,
+  reviewer/date metadata, and `w:trackRevisions`.
+- Added HWPX native revision output using insert/delete begin/end markers,
+  `trackChanges`, `trackChangeAuthors`, and track-change configuration metadata.
+- Comparison is non-destructive and writes revisions into a cloned copy of the
+  newer document.
+- Structural paragraph/table-row insert/delete and unsafe complex inline
+  structures fail closed in this increment.
+- Added Accept->new / Reject->old projection tests, table-cell comparison,
+  metadata, no-op-copy, mismatch, and fail-closed regressions for both formats.
+- Preserved all v0.28.2 baselines; full suite: 355/355 passed.
+- Microsoft Word/Hancom GUI review of generated tracked-change files remains an
+  explicit external release QA item.
+
+## 0.25.0 - Logicless template v1
+
+- Added `pywpdoc.Document.render_template()` as a third isolated high-level
+  feature family.
+- Added independent template `parser`, `context`, `renderer`, and `api` modules.
+- Added `{{name}}`, dotted `{{customer.name}}`, current-context resolution,
+  run-split tags, body/table/header/footer traversal, and explicit missing-value
+  policies (`keep`, `empty`, `error`).
+- Template does not import `find` or `replace`; it reuses only lower-level
+  `pywpdoc.core.text` mapping/traversal/mutation primitives.
+- Mustache control tags are explicitly rejected in v1 rather than silently
+  mis-rendered; sections/loops/table-row repetition are reserved for later
+  independent submodules.
+- Preserved immutable v0.21.5 baseline; full suite is 284/284.
+
+## 0.24.0 - Isolated high-level replace feature
+
+- Added `pywpdoc.Document.replace_text()` as a thin facade over an independent
+  `pywpdoc.features.replace` module.
+- Extracted literal-span matching into `pywpdoc.core.text.match`, shared by
+  `find` and `replace` without sibling-feature imports.
+- Added `pywpdoc.core.text.mutation` for backend-neutral logical range mutation.
+- Added architecture tests prohibiting sibling feature dependencies.
+- Added real-document replacement isolation regression for HWPX and DOCX.
+- Preserved immutable v0.21.5 baseline; full suite is 268/268.
+
+# v0.21.4
+
+- Fixed DOCX snapshot cell-border serialization. HWPX native cell borders now replace the provisional Word borders instead of appending duplicate `w:top`/`w:left`/`w:bottom`/`w:right` nodes.
+- This fixes renderer-dependent missing table edges, including the lower outline of the application table in the school-notice Common rebuild sample.
+- Added a regression asserting exactly one OOXML node per cell edge and preserving the HWPX 0.5 mm bottom border as `w:bottom w:sz="11"`.
+- Full regression: 223/223 passed.
+
+# 0.21.2
+- Roll back the v0.21.0/v0.21.1 rich header/footer auto-reservation heuristic. HWPX save no longer estimates story height, rewrites header/footer areas, or injects a 0.5 mm clearance.
+- Keep the verified section/page geometry fix that invalidates stale `hp:linesegarray` caches after page size, orientation, or margin changes.
+- Keep the earlier rich-story `hp:subList/@textHeight="0"` policy and logical/control story synchronization.
+- Add a regression proving that a rich header preserves the caller-authored native header area instead of silently auto-fitting it.
+- Full regression: 219/219 passed.
+- Add a reverse-rebuild diagnostic for a real Hancom school-notice HWPX: page geometry, top-level block order, 3 tables, 3 rectangles, and 1 picture reconstruct semantically; the rebuilt document intentionally carries 0 persisted `linesegarray` caches versus 52 in the Hancom source, forcing Hancom to re-layout on open.
+
+# 0.21.1
+- Fix v0.21.0 rich-header over-reservation: HWPX now keeps the caller-authored header/footer area as the baseline and adds only the rich-story overflow beyond the corresponding page margin, plus a minimal 0.5 mm clearance.
+- Preserve v0.21 stale `hp:linesegarray` invalidation.
+- Add regressions for DOCX-like compact header/body spacing and repeated-save idempotency.
+
+# 0.21.0
+
+- Fix native HWPX section-geometry mutation to invalidate stale `hp:linesegarray` paragraph layout caches instead of carrying coordinates from the bundled skeleton into a changed page box.
+- Add native rich header/footer extent calculation from paragraph line metrics, table row/cell content, and cell margins.
+- Rich HWPX stories now reserve enough native header/footer area before save; final table-cell edits are included in the calculation.
+- Save-time normalization also upgrades older rich stories with a fixed `textHeight` to the content-sized form while preserving simple one-line stories.
+- Keep the fix entirely below the Common facade; DOCX behavior is unchanged.
+- Add 3 engine regressions; full regression is 220/220 passed.
+- Real Hancom GUI verification remains required before declaring visual parity with DOCX.
+
+# 0.20.0
+
+- Split document core-properties ownership into backend-local `hwpx/metadata.py` and `docx/metadata.py` modules.
+- Standalone HWPX and DOCX backends now return backend-owned `CoreProperties` objects while preserving `.native` escape hatches.
+- Verify shared string metadata fields: title, author, subject, comments, keywords, language, and last-modified-by.
+- Normalize shared created/modified datetime semantics to UTC-naive second precision inside each backend, not Common.
+- Add 10 metadata isolation/roundtrip regressions; full regression is 217/217 passed.
+
+# 0.19.0
+
+- Split picture/media ownership into backend-local `hwpx/media.py` and `docx/media.py` modules.
+- Standalone HWPX and DOCX backends now return backend-owned `Picture` objects from `add_picture()` and `pictures`.
+- Add a verified shared picture subset: physical width/height/size, resize, embedded bytes, save/reopen persistence.
+- Keep HWPX BinData/manifest/package behavior in the native HWPX engine; the standalone media module is an adapter rather than a duplicate implementation.
+- Keep DOCX HWPUNIT/core-length to EMU translation inside `docx/media.py`; Common remains delegation-only.
+- Fail closed for DOCX floating-picture requests (`treat_as_char=False`) until a verified anchor mapping exists.
+- Add 11 media isolation/contract regressions; full regression is 207/207 passed.
+
+# 0.18.0
+
+- Lower HWPX Section/Header/Footer behavior beneath the standalone wrappers.
+- Make `hwpx/section.py` and `hwpx/story.py` thin adapters only.
+- Add native HWPX policy modules `_hwpx_section.py`, `_hwpx_story.py`, `_hwpx_table.py`.
+- New HWPX stories use the skeleton's native `Header`/`머리말` style for the initial paragraph.
+- `HeaderFooter.text` no longer leaks nested-table cell text; `all_text` preserves recursive expert access.
+- Keep story-table width, cell margin, and TOP alignment behavior in the native HWPX layer.
+- 194 regression tests pass; original notice remains 12/12 exact rebuild.
+
+# v0.17.0
+
+- Split Section/page setup routing into backend-owned `section.py` modules.
+- Split Header/Footer block-story wrappers into backend-owned `story.py` modules.
+- Common remains delegation-only and receives no page geometry, story mutation, or header-table sizing logic.
+- DOCX even-page story access enables the required Word even/odd header setting inside the DOCX backend.
+- Header tables default to section content width independently in both backends.
+- Added section/story module isolation, persisted geometry, story block, add-section, support-gap, and HWPX mirror regressions.
+- Final regression: 186/186 passed; original notice 12/12 exact rebuild.
+
+# v0.16.0
+
+- Split standalone style ownership into `docx/styles.py` and `hwpx/styles.py`.
+- `Styles`, `Style`, `StyleFont`, and `StyleParagraphFormat` are backend-owned public types.
+- Added backend-local DOCX `clone_from` direct-format copying; Common remains delegation-only.
+- Paragraph/Run accept backend Style wrappers with one-way `paragraph -> styles` dependency.
+- Added style isolation, clone independence, apply/next/delete, common-delegation, and minimal protocol regressions.
+- Final regression: 167/167 passed; original notice 12/12 exact rebuild.
+
+# Changelog
+
+
+## 0.21.3
+
+- Common에 `Document.rebuild()` / `export_snapshot()` / `rebuild_from()` 구조 보존 경로를 추가했다. Common은 `DocumentSnapshot`만 전달하고 포맷별 export/import는 각 backend가 소유한다.
+- HWPX→HWPX `preserve_layout=True`는 native package payload를 보존해 변경 없는 rebuild에서 package part hash가 모두 동일하도록 했다. `hp:linesegarray`를 임의 삭제하지 않는다.
+- HWPX→DOCX portable importer를 추가했다. 페이지 geometry, HWPX layout slot, run/paragraph 서식, 표/병합/셀 테두리·채움, 그림, text box를 Word 구조로 변환한다.
+- DOCX 변환에서 `lineseg/@spacing`을 paragraph `space-after`로 중복 계산하던 문제를 제거했다. 한 줄 문단은 다음 cached line top을 이용해 measured advance를 유지한다.
+- HWPX text box 내부의 borderless positioning grid를 Word에서 별도 nested table로 중복 배치하지 않고 host paragraph/tab 구조로 변환해 fixed slot clipping을 방지한다.
+- Common HWPX same-format rebuild package fidelity, HWPX→DOCX reopen, inline positioning-grid flattening 회귀를 추가했다.
+
+## 0.15.0
+
+- Split table ownership inside each independent backend into `table.py` and `table_geometry.py`.
+- `backend.py` now composes/delegates table creation instead of owning Table/Row/Column/Cell implementations.
+- Keep Common API unchanged and thin; no table geometry or option normalization moved into Common.
+- Normalize standalone backend `Row.height`, `Column.width`, and `Cell.width` around shared physical Length semantics.
+- DOCX backend column-width setter now synchronizes rendered cell widths (`tcW`) as part of the backend contract, without changing python-docx native behavior.
+- Add architecture, geometry, merge, nested-table, vertical-alignment, width, and height regression gates.
+- Regression result: 153/153 passed; original notice exact rebuild remains 12/12.
+
+## 0.14.0
+
+- First feature-module isolation pass: split paragraph/run wrappers and text operations into backend-owned `paragraph.py` / `text.py` modules for both DOCX and HWPX.
+- Common remains delegation-only; it owns no text replacement or paragraph mutation behavior.
+- DOCX document-wide replace traversal moved out of the backend facade into `docx.text`.
+- HWPX text operations are routed through a narrow backend boundary while the mature native HWPX engine remains the implementation owner.
+- HWPX standalone backend `Paragraph.add_run()` now starts a new unstyled run from the paragraph style instead of leaking direct formatting from the previous sibling run; native expert API behavior remains unchanged.
+- Added module dependency gates plus backend-direct replace count/scope/direct-format isolation tests.
+- Regression result: 139/139 passed; original notice exact rebuild remains 12/12.
+
+## 0.13.0
+
+- Architecture-only refactor: independent HWPX backend, independent DOCX backend, thin common facade.
+- Added backend-neutral `core/` units, enums, errors, defaults, and minimal Protocol contracts.
+- Moved cross-run DOCX replace, page/table geometry, cell padding, vertical alignment, and nested-table sizing out of common and into the owning backend.
+- Preserved the legacy top-level HWPX expert API and the `common_api.UnifiedDocument` compatibility import.
+- Added AST dependency-boundary/geometry-delegation tests and direct standalone backend contract tests.
+- Regression result: 128/128 passed; original notice exact rebuild remains 12/12.
+
+## 0.12.3
+
+- Common-created DOCX/HWPX table cells now persist the same default vertical alignment: TOP.
+- Fixed the visual mismatch where HWPX native cells defaulted to CENTER while Word rendered missing `w:vAlign` as TOP.
+- `UnifiedCell.vertical_alignment` now maps TOP/CENTER/BOTTOM through one python-docx-shaped enum for both backends.
+- Common proxy setters now honor wrapper properties before falling through to native objects.
+- Persisted-XML cross-format tests now verify outer and nested cell vertical alignment in addition to physical geometry.
+- Regression suite expanded to 104 tests.
+
+## 0.12.2
+
+- Fixed common-API nested table geometry by sizing nested tables to the parent cell content box rather than the raw cell width.
+- Normalized common-created table cell padding to 1.8 mm horizontal and 0.5 mm vertical for both HWPX and DOCX.
+- Strengthened cross-format geometry tests to inspect persisted parent cell width, padding, content-box width, and nested table width.
+
+## 0.12.1
+
+- Common API preview geometry parity hotfix.
+- `UnifiedDocument.new()` now normalizes new DOCX/HWPX documents to the same A4 page and margin geometry; opened documents are never rewritten.
+- Common top-level `add_table()` now uses the full current body width when width is omitted, matching python-docx semantics instead of leaking the HWPX native 1-inch-per-column default.
+- Added persisted-XML cross-format geometry contract tests for page, margins, body width, table width, column widths, and nested-table width with 0.05 mm tolerance.
+- Regression suite expanded to 102 tests.
+
+
+## 0.12.0
+
+- HWPX v1.0 release-candidate stabilization.
+- Targeted edits preserve unrelated package-part bytes rather than reserializing unchanged XML.
+- Paragraph-style assignment uses generic charPr 3-way rebase so direct run formatting survives style changes.
+- `Paragraph.style = None` returns to verified default paragraph style.
+- `Run.clear()` invalidates stale line-layout cache.
+- Structured public error families: `HwpxValueError`, `HwpxStateError`, `HwpxUnsupportedError`; validation reports are serializable.
+- Experimental `python_hwpx_next.common_api.UnifiedDocument` backend-boundary preview for DOCX/HWPX.
+- Automatic regression expanded to 99 tests: 71 HWPX, 20 native cross-format, 8 common-preview.
+
+
+## 0.11.0
+
+- v1.0 전 core-semantics 안정화 릴리스.
+- `Paragraph.replace()` 추가: 여러 run/hyperlink 경계를 넘는 structure-preserving replace.
+- `Document.replace(..., everywhere=True)` 추가: body/table/shape/header/footer/note semantic stories를 대상으로 하고 mirrored header/footer 중복 편집을 방지하며 memo body는 보존.
+- visible Run/Hyperlink/tab/line-break 변경 시 stale `hp:linesegarray` 제거.
+- `Document.paragraphs`를 python-docx 의미로 정정: HWPX 내부 secPr carrier 및 table-only wrapper를 숨김.
+- `Row.cells`를 python-docx처럼 logical grid cells로 정정. HWPX physical XML cells는 `Row.physical_cells` escape hatch로 분리.
+- `HwpxError` / `HwpxValidationError` 추가. validation failure는 기존 `ValueError` 호환을 유지하면서 code/report를 노출.
+- `python-docx>=1.2,<2`를 test extra에 추가하고 `tests/test_cross_format_contract.py` 도입.
+- HWPX/DOCX 양쪽에서 mixed text, paragraph insertion, nested table visibility, merged-row logical cells, core properties, section margin/orientation, next paragraph style, cross-run replace 의미론을 병렬 검증.
+- 자동 회귀 80개로 확대.
+
+## 0.10.0
+
+- `WD_STYLE_TYPE` 추가. PARAGRAPH/CHARACTER는 지원하고 TABLE/LIST는 검증될 때까지 fail-closed.
+- `Styles` iteration을 python-docx 의미로 수정: ID가 아니라 `Style` 객체를 반환.
+- `Styles.default()`의 verified paragraph subset 추가.
+- `Style.next_paragraph_style`을 HWPX `nextStyleIDRef`에 매핑.
+- `Run.style` character-style facade 추가.
+- `Paragraph.contains_page_break` 추가.
+- section carrier 앞 `insert_paragraph_before()`가 `secPr`와 section-start controls를 새 첫 문단으로 이동하도록 안전화.
+- `Row.height` setter 및 vertical-merge-aware logical row height 계산 추가.
+- `Cell.merge(other_cell)`, `Cell.add_table()`, `Cell.tables`, `Cell.iter_inner_content()` 추가.
+- `Cell.text` setter가 nested table을 포함한 기존 cell story를 제거하고 single paragraph/run으로 교체하도록 python-docx 의미로 수정.
+- Header/Footer block container에 `tables`, `add_table()`, `iter_inner_content()` 추가. logical/control mirror equality를 테스트로 고정.
+- Header/Footer `add_paragraph(..., style=...)`가 style 인자를 실제 적용하도록 수정.
+- header/footer field의 charPr 선택식 우선순위 버그 정리.
+- `Document.tables`를 python-docx처럼 body top-level tables로 의미 정정하고 `Document.all_tables` recursive escape hatch 추가.
+- 자동 회귀 57개로 확대.
+
+## 0.9.0
+
+- `Table.columns` / `Column.width` 추가.
+- 병합 셀을 고려해 열 폭 변경 시 physical cell width와 table width를 함께 동기화.
+- 기존 logical column width 추정이 병합 셀을 우선해 잘못된 폭을 읽던 문제 수정; 단일-span 셀을 우선.
+- `Cell.vertical_alignment` 및 `WD_CELL_VERTICAL_ALIGNMENT` / `WD_ALIGN_VERTICAL` 추가.
+- `Section`에 python-docx-shaped page facade 추가: page width/height, margins, header/footer distance, gutter.
+- `Section.orientation` 및 `WD_ORIENT` 추가. HWPX `WIDELY`(portrait) / `NARROWLY`(landscape) mapping 사용.
+- `Document.core_properties` 추가: HWPX `Contents/content.hpf`의 검증된 metadata subset 편집.
+- `Document.settings` typed foundation 및 `CaretPosition` 추가.
+- `Style.delete()`가 dangling style ref를 만들지 않도록 참조 repoint + dense style-id compaction.
+- 자동 회귀 49개로 확대.
+
+## 0.8.0
+
+- `Paragraph.text` / `Run.text`가 `\t`, `\n`을 한컴 native mixed-content `hp:tab` / `hp:lineBreak`로 왕복.
+- `Paragraph.insert_paragraph_before()` 추가.
+- `Paragraph.clear()` / `Run.clear()` section setup 보존 강화.
+- `ParagraphFormat.tab_stops`, `TabStops`, `TabStop`, `WD_TAB_ALIGNMENT`, `WD_TAB_LEADER` 추가.
+- `Document.iter_inner_content()`와 `Paragraph.iter_inner_content()` 추가.
+- `Section.even_page_header/footer`, `odd_page_header/footer` 추가.
+- 기본 `Document.save()`를 atomic validated save로 강화.
+
+## 0.22.0 - PyWPDoc namespace architecture milestone
+
+- Added new `pywpdoc` public namespace and distribution name.
+- `pywpdoc.Document` is now the high-level format-neutral entry point.
+- Added direct backend namespaces `pywpdoc.hwpx` and `pywpdoc.docx`.
+- Preserved the verified `python_hwpx_next` v0.21.5 implementation/import path
+  during staged migration instead of performing a big-bang internal rename.
+- Added explicit common `Document.unwrap()` backend escape hatch.
+- Added inert `pywpdoc.features` boundary and dependency architecture tests.
+- Added immutable v0.21.5 baseline manifest and real-document regression gates.
+- Preserved v0.21.5 DOCX layout-slot `AT_LEAST` fix.
+- Legacy baseline: 224/224 passed; full suite: 236/236 passed.
+- HWPX->DOCX font visual parity remains a deferred known issue.
+
+
+## 0.23.0 - First isolated high-level feature
+
+- Added `pywpdoc.Document.find()` as the first high-level feature module.
+- Added backend-neutral `TextMap` and text traversal primitives under `pywpdoc.core.text`.
+- Added minimal `iter_text_paragraphs()` backend capability for HWPX/DOCX.
+- Added a separate read-only DOCX traversal module so search does not create missing header/footer parts.
+- Added cross-backend run-boundary, table/header, case/limit, direct-backend, read-only isolation, and architecture tests.
+- Preserved the immutable v0.21.5 baseline and all v0.22.0 migration tests.
+- Full suite: 249/249 passed.
+- HWPX->DOCX font visual parity remains deferred.
+
+## 0.26.0
+
+- added an experimental backend-neutral Drawing placement/transform model;
+- added pure Mustache section parsing and structural repeat plan primitives,
+  without enabling them in the public v1 renderer;
+- added repository-only DOCX OPC/package validation and an external Open XML
+  SDK validator adapter for release CI;
+- added architecture gates preventing Drawing/template structure from depending
+  on concrete backends;
+- preserved all existing public behavior and v0.21.5 baseline outputs.
+
+## 0.27.0
+
+- added backend-neutral `TextFragment` layer below `TextMap`;
+- DOCX high-level text features now mutate native `w:t` fragments rather than
+  replacing whole run text;
+- added safe transparent fragmentation across multiple `w:t`, hyperlinks,
+  bookmarks, proofing markers, and significant whitespace;
+- added hard/opaque boundary tracking for fields, revisions, SDT/content
+  controls, tabs/breaks, drawings/objects, math, and unknown semantic nodes;
+- `find`/`replace_text` skip matches crossing opaque boundaries;
+- `render_template` fails closed with `template-unsafe-text-boundary` for a tag
+  crossing such a boundary;
+- added DOCX fragmentation torture and save/reopen regression coverage;
+- preserved all v0.26 baselines and existing public template syntax.
+
+## 0.28.0
+
+- connected the isolated Mustache section parser to `render_template()` behind
+  the backward-compatible `sections=True` opt-in;
+- added normal (`#`) and inverted (`^`) inline sections, list repetition,
+  nested sections, current-context `{{.}}`, and parent-context fallback;
+- reused v0.27 TextFragment/TextMap handling so section markers may span safe
+  run/native-text fragmentation;
+- kept field/revision/SDT/tab/object boundaries fail-safe;
+- kept table-row/block repetition separate from inline text expansion;
+- preserved default control-tag fail-closed behavior and all earlier baseline
+  tests;
+- expanded the full suite to 329 tests.
+
+## 0.28.1
+
+- corrected the v0.28 cross-format section sample to request `Table Grid`
+  explicitly on both HWPX and DOCX instead of relying on backend defaults;
+- preserved `add_table(style=None)` behavior (DOCX remains `Normal Table`; HWPX
+  remains backend-owned);
+- added structural regressions proving DOCX `TableGrid` survives template
+  rendering/reopen and HWPX uses a solid black borderFill for the explicit grid;
+- added repository-only LibreOffice raster QA that verifies the rendered DOCX
+  contains visible horizontal and vertical table borders;
+- no runtime/document-generation behavior was changed outside the corrected
+  comparison fixture; full suite: 333/333 passed.
+
+## 0.28.2
+
+- fixed HWPX all-story paragraph traversal for real Hancom-authored files that
+  legally reuse paragraph `id` values (commonly `id="0"`) across unrelated
+  table-cell paragraphs;
+- replaced broad HWPX paragraph-id deduplication with semantic header/footer
+  mirror suppression and explicit memo-field exclusion;
+- fixed DOCX physical-cell traversal to use the underlying XML cell node as the
+  dedupe identity instead of Python `id(cell._tc)`, which can be reused for
+  short-lived lxml/python-docx proxy objects;
+- added user-authored Word/Hancom template documents as permanent real-authoring
+  regression fixtures;
+- added cross-format `find`, `replace_text`, `render_template(sections=True)`,
+  and save/reopen coverage for the manually authored fixtures;
+- preserved all v0.28.1 baselines; full suite: 340/340 passed.
+
+## 0.31.0
+
+- extended the backend-neutral comparison plan with structural `TableRowRevision`
+  entries while keeping v0.29 text revision objects backward-compatible;
+- added conservative minimum-cost alignment for top-level body table rows;
+- DOCX comparison now emits native row insertion/deletion revisions under
+  `w:trPr` using `w:ins` / `w:del`, alongside existing inline text revisions;
+- added old/new paragraph index mapping so row insertion/deletion may coexist
+  with later cell-text edits in the same table;
+- tracked deletion materializes the old DOCX row into the review copy and fails
+  closed for relationship-bearing/control-heavy rows or vertical merges until a
+  safe cross-package import layer exists;
+- HWPX structural row changes fail closed with
+  `compare-hwpx-table-row-track-unsupported`; the public OWPML model verified
+  for this increment exposes tracked insert/delete ranges inside paragraph text
+  but no row-level revision grammar equivalent to DOCX `w:trPr` revisions;
+- unchanged-row tables retain the v0.29 paragraph alignment path, preserving
+  text comparison behavior for same-structure complex/vertically merged tables;
+- added Accept-All/Reject-All row projection regressions for insertions,
+  deletions, mixed row/text changes, adjacent multi-row changes, metadata, and
+  non-destructive input handling;
+- preserved all v0.30.0 regression tests; full suite: **399/399 passed**.

pywpdoc-0.44.0rc1/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Rydolabs
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

pywpdoc-0.44.0rc1/MANIFEST.in

@@ -0,0 +1,9 @@
+include README.md README.ko.md LICENSE pyproject.toml CHANGELOG.md
+recursive-include src/pywpdoc/_engine/data *.hwpx
+include src/pywpdoc/_engine/py.typed
+prune tests
+prune docs
+prune examples
+prune scripts
+recursive-exclude .github *
+global-exclude __pycache__ *.py[cod] *.log