odf.js 1.10.0 → 1.10.2

package/README.md

@@ -8,7 +8,7 @@
 
 **This package does not depend on `ooxml.js`**, even though the two do near-identical jobs for their respective formats: `ooxml.js` is a package signed, SBOM-attested, and branded exclusively around ECMA-376/OOXML, so depending on it here would be a permanently wrong signal for an OASIS-standard codec, and would force a breaking `ooxml.js` release every time an ODF-only fix needed the shared primitive layer. Instead, `odf.js` duplicates the small (~400-line) generic ZIP/XML/`Package` layer as its own code — kept deliberately structurally identical (plain, unmarked shapes, no branding) so TypeScript's structural typing makes the two packages' `Package`/`XmlNode`/`XmlElement` values freely interchangeable wherever a shared consumer (like `documents.js`) needs to treat them uniformly, without either package formally depending on the other.
 
-Both packages **do** depend on [`document-content-model`](https://github.com/ExaDev/document-content-model), the genuinely shared canonical schema for `ContentDocument`/`LayoutDocument` — the semantic content model (paragraphs, runs, tables, shapes, slides) both an ODF and an OOXML reader ultimately produce. `odf.js`'s typed readers return the real, imported `ContentSection`/`ContentSlide`/etc. types from that package, not a structurally-similar lookalike, so a downstream consumer (`documents.js`) can run an `.odt` through the exact same layout/pagination engine it already uses for `.docx`, unmodified.
+Both packages **do** depend on [`document-content-model`](https://github.com/ExaDev/document-schema.js), the genuinely shared canonical schema for `ContentDocument`/`LayoutDocument` — the semantic content model (paragraphs, runs, tables, shapes, slides) both an ODF and an OOXML reader ultimately produce. `odf.js`'s typed readers return the real, imported `ContentSection`/`ContentSlide`/etc. types from that package, not a structurally-similar lookalike, so a downstream consumer (`documents.js`) can run an `.odt` through the exact same layout/pagination engine it already uses for `.docx`, unmodified.
 
 ## Status
 
@@ -17,7 +17,7 @@ This package is under active development. What's built and shipped:
 - **Lossless core** — generic ZIP-of-XML primitives (`Package`/`XmlNode`/`XmlElement`, XML parse/build, zip/unzip, base64, the `packageCodec`/`xmlCodec` `z.codec()` pairs) with zero ODF-specific knowledge.
 - **Namespaces, media types, mimetype, manifest** (`src/ns.ts`, `src/media-type.ts`, `src/mimetype.ts`, `src/manifest.ts`) — full read *and* write, including `META-INF/manifest.xml`'s exhaustive per-part enumeration and the mimetype part's mandatory first-entry/stored/uncompressed byte layout, verified against real LibreOffice-produced output.
 - **Style interning** (`src/styles/`) — `StyleRegistry`: adopts a part's existing automatic styles on construction, finds-or-mints on `intern()`, fingerprints on canonical serialized properties plus parent style name (never `JSON.stringify`), and is collision-checked across all four style containers a document can have.
-- **Shared typed primitives** (`src/typed/shared/`) — ODF length-unit parsing (`cm`/`mm`/`in`/`pt`/`pc`/`px`), A1-style spreadsheet cell-reference computation with repeat-count cursor advancement, colour/geometry/master-page-size parsing into `document-content-model`'s own types, ODF's `text:s`/`text:tab`/`text:line-break` whitespace-run decoding, the read-side style cascade (`style:default-style` → parent chain → the referenced style — one layer shorter than OOXML's, since ODF has no separate direct-formatting layer on top), the `draw:transform`/`draw:g` group-flattening geometry resolver, an `svg:d`/`draw:points` vector-path grammar parser, and `meta.xml` reading.
+- **Shared typed primitives** (`src/typed/shared/`) — ODF length-unit parsing (`cm`/`mm`/`in`/`pt`/`pc`/`px`), A1-style spreadsheet cell-reference computation with repeat-count cursor advancement, colour/geometry/master-page-size parsing into `document-content-model`'s own types, ODF's `text:s`/`text:tab`/`text:line-break` whitespace-run decoding, the read-side style cascade (`style:default-style` → parent chain → the referenced style — one layer shorter than OOXML's, since ODF has no separate direct-formatting layer on top), the shared `text:p` → `ContentParagraph`/`ContentRun` and `table:table` → `ContentTable` readers (`readOdfParagraph`/`readOdfTable`) every typed reader below builds on, the `draw:transform`/`draw:g` group-flattening geometry resolver, an `svg:d`/`draw:points` vector-path grammar parser, and `meta.xml` reading.
 - **Typed readers** — `readOdt` (wordprocessing), `readOdp` (presentation slides: `draw:frame`/`draw:g` text/image/table content), `readOdg` (drawing pages: every vector primitive — `draw:rect`/`draw:ellipse`/`draw:circle`/`draw:line`/`draw:path`/`draw:polygon`/`draw:polyline`, plus a recognised `draw:custom-shape` preset subset — in real `draw:z-index`-aware paint order), and `readOds` (spreadsheets: geometry- and print-settings-rich, every `office:value-type` variant with its own OpenFormula string) each resolve a `Package` into `document-content-model`'s own `ContentSection`/`ContentSlide`/`ContentDrawPage`/`ContentSheet` shapes.
 - **`readOdfFormula`** resolves a standalone or embedded `.odf` formula's `content.xml` — bare MathML with no `office:document-content` wrapper, confirmed against real LibreOffice output — into its raw MathML nodes plus, when present, the formula's own native StarMath annotation string. There is no `document-content-model` pivot type for MathML, so this reader's own output type is its own.
 - **`readOdm`** resolves a `.odm` master document's own `content.xml` into an ordered list of chapter references (`{ name, href, filterName? }`, one per top-level linked `text:section`) without opening the external `.odt` files those references point at. A master document's chapters are genuinely external files by ODF design, not embedded package sub-documents — confirmed against real LibreOffice output, which never caches a chapter's own content inside the master document itself (see `src/typed/odm/read.ts`'s own top-of-file note).
@@ -74,11 +74,12 @@ Layered from a lossless core outward, mirroring `ooxml.js`'s own structure:
 
 - **`src/model/`** — `Package`/`XmlNode`/`XmlElement` and friends: a duplicate-by-design copy of `ooxml.js`'s equivalent, kept structurally identical (see [Why no `ooxml.js` dependency](#why-no-ooxmljs-dependency) above).
 - **`src/xml/`** — `parse.ts`/`build.ts` (XML string ⇄ `XmlNode[]` forest via `fast-xml-parser`), `fragment.ts`/`entities.ts` (production element/text-node construction and entity encoding — `odf.js` writes `manifest.xml` itself, unlike `ooxml.js`'s read-only stance on OPC relationships, so this needs to be real writing code, not test-only scaffolding), `query.ts` (shared tree-query helpers).
+- **`src/image/`** — `sniffImageFormat` (`ImageFormat`): a minimal PNG/JPEG magic-byte format sniffer with no ODF knowledge of its own, consumed by `src/manifest.ts` (choosing a binary part's manifest media type) and `src/typed/draw/shapes.ts` (resolving a `draw:image`'s own `ContentImageBlock.format`).
 - **`src/zip.ts`** — takes *ordered* `[path, entry]` tuples, not a `Record`, specifically so ODF's mimetype-first/stored/uncompressed requirement doesn't depend on `Record`/`Object.keys` insertion order surviving a Zod round trip.
 - **`src/package-io/`** — `write.ts` hoists a `mimetype` part first (stored) and `META-INF/manifest.xml` second, if present, before everything else in existing order — the one deliberate behavioural difference from `ooxml.js`'s own writer, and never fabricates either part as a side effect.
 - **`src/manifest.ts`** — unlike `ooxml.js` (which only ever *reads* OPC relationships, leaving writing to `documents.js`), `odf.js` owns manifest read **and** write, since the manifest is ODF's one mandatory part and its correctness is exhaustive.
 - **`src/styles/`** — `properties.ts` (the property-bag shape + real ODF attribute parsing), `serialize.ts` (canonical, deterministic property-bag → XML attributes), `registry.ts` (`StyleRegistry`, ODF's mandatory style-interning layer, no OOXML equivalent), `span.ts` (character-range wrapping into a formattable `text:span`, correctly splitting `text:s`/`text:tab` elements that straddle a boundary).
-- **`src/typed/shared/`** — the ODF-specific typed primitives every future format reader builds on: `units.ts`, `a1.ts`, `color.ts`/`geometry.ts` (parsing into `document-content-model`'s own types, never redefining them), `style.ts` (a thin re-export — ODF's style-properties concern is fully covered by `styles/properties.ts` and the cascade below), `text.ts` (whitespace-run decoding), `cascade.ts` (the read-side style-resolution walk), `metadata.ts` (`meta.xml` reading).
+- **`src/typed/shared/`** — the ODF-specific typed primitives every future format reader builds on: `units.ts`, `a1.ts`, `color.ts`/`geometry.ts` (parsing into `document-content-model`'s own types, never redefining them), `style.ts` (a thin re-export — ODF's style-properties concern is fully covered by `styles/properties.ts` and the cascade below), `text.ts` (whitespace-run decoding), `cascade.ts` (the read-side style-resolution walk), `paragraph.ts`/`table.ts` (`readOdfParagraph`/`readOdfTable`, the shared `text:p` → `ContentParagraph`/`ContentRun` and `table:table` → `ContentTable` readers `readOdt`, `readOds`, and `typed/draw/shapes.ts` all call), `transform.ts` (the `draw:transform`/`draw:g` group-flattening geometry resolver), `masterpage.ts` (master-page → page-layout page-size/print-settings resolution, shared by a `draw:page`'s own size and a spreadsheet's print settings), `path.ts` (an `svg:d`/`draw:points` vector-path grammar parser), `metadata.ts` (`meta.xml` reading).
 - **`src/typed/odt/`, `src/typed/odp/`, `src/typed/odg/`, `src/typed/ods/`** — the built `readOdt`/`readOdp`/`readOdg`/`readOds` readers; **`src/typed/draw/`** — the `draw:frame`/`draw:g`/vector-primitive shape vocabulary `readOdp` and `readOdg` both share; **`src/typed/formula/`, `src/typed/odm/`** — `readOdfFormula` (raw MathML, no `document-content-model` pivot) and `readOdm` (a `.odm` master document's own external chapter references — name/href/filter-name per linked `text:section`, never the linked content itself); **`src/typed/odb/`** — `readOdbInventory` (a `.odb` package's own connection info plus form/query/report/table *names*, never their content or the database engine's own storage).
 
 ## Why no `ooxml.js` dependency
@@ -112,8 +113,8 @@ Commits follow Conventional Commits (`feat:`, `fix:`, `test:`, `chore:`, …), e
 ## References
 
 - [ooxml.js](https://github.com/ExaDev/ooxml.js) — the sibling package doing the equivalent lossless-codec job for OOXML (docx/pptx/xlsx). Architecturally mirrored, deliberately not depended on — see [Why no `ooxml.js` dependency](#why-no-ooxmljs-dependency).
-- [document-content-model](https://github.com/ExaDev/document-content-model) — the canonical `ContentDocument`/`LayoutDocument` schema pivot both this package and `ooxml.js` depend on.
-- [documents.js](https://github.com/ExaDev/documents.js) — the intended downstream consumer, adding ODF ⇄ PDF conversion and a live-view ODF editor once this package's typed readers land.
+- [document-content-model](https://github.com/ExaDev/document-schema.js) — the canonical `ContentDocument`/`LayoutDocument` schema pivot both this package and `ooxml.js` depend on.
+- [documents.js](https://github.com/ExaDev/documents.js) — the downstream consumer, already built on this package's typed readers: its own `readOdtContent`/`readOdpContent`/`readOdsContent`/`readOdgContent` are thin adapters over `readOdt`/`readOdp`/`readOds`/`readOdg`, feeding both ODF ⇄ PDF conversion (`odtToPdf`/`odpToPdf`/`odsToPdf`/`odgToPdf` and their inverses) and independently-built live-view ODF editors (`OdtEditor`/`OdpEditor`/`OdsEditor`/`OdgEditor`).
 
 ## License
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "odf.js",
-  "version": "1.10.0",
+  "version": "1.10.2",
   "description": "Type-safe, lossless round-trip conversion between OpenDocument Format packages (odt, ods, odp) and JSON, hand-written and dependency-minimal, built on Zod 4 codecs.",
   "type": "module",
   "repository": {