@sametalmdr/word-editor 0.2.0-alpha → 0.3.0-alpha

package/dist/io/ooxml/inflate.js

@@ -1,126 +1,126 @@
-// The ONE file in src/io/ooxml/ that touches a platform global (ADR-0020
-// Karar 8, §1/§7's own precedent in src/measure/probe.js). Decompression is
-// lazy and per-entry: zip-container.js never decompresses anything, so a
-// caller only pays this async cost for the members it actually needs —
-// e.g. it can list every part in a package and read sizes/names without
-// ever calling this function.
-//
-// `typeof DecompressionStream !== "undefined"` is checked INSIDE the
-// function, never at module scope (invariant #5, ssr-safety) — importing
-// this file must never touch a global, only calling inflateEntry() may.
-// Measured 2026-08-27: DecompressionStream is a Node v24.13.1 global,
-// matching ADR-0020's own claim, but WHICH Node 18.x point release first
-// made it global (vs. requiring an import) was not checked — the lazy
-// `typeof` guard means that question never has to be answered, the code is
-// correct either way, and a missing API fails at the first call, not at
-// import time.
-//
-// CRC-32 verification happens HERE, not in zip-container.js, because it
-// needs decompressed bytes — the pure/sync layer never sees them. Ported
-// check from scripts/opc-bytes.py's read_package(): "CRC-32 mismatch in the
-// SOURCE file — refusing to touch it" (never trust the central directory's
-// claim alone).
-
-const METHOD_STORE = 0;
-
-/** @type {Uint32Array | null} */
-let crcTable = null;
-
-function getCrcTable() {
-  if (crcTable !== null) return crcTable;
-  const table = new Uint32Array(256);
-  for (let n = 0; n < 256; n++) {
-    let c = n;
-    for (let k = 0; k < 8; k++) {
-      c = c & 1 ? 0xedb88320 ^ (c >>> 1) : c >>> 1;
-    }
-    table[n] = c >>> 0;
-  }
-  crcTable = table;
-  return table;
-}
-
-/**
- * Standard CRC-32 (ISO 3309 / PKZIP / zlib) — the same algorithm
- * `zlib.crc32` implements in Python, bit for bit. A fixed, well-known
- * specification, not a guess.
- * @param {Uint8Array} bytes
- * @returns {number}
- */
-export function crc32(bytes) {
-  const table = getCrcTable();
-  let crc = 0xffffffff;
-  for (let i = 0; i < bytes.length; i++) {
-    crc = table[(crc ^ bytes[i]) & 0xff] ^ (crc >>> 8);
-  }
-  return (crc ^ 0xffffffff) >>> 0;
-}
-
-/**
- * @param {ReadableStream<Uint8Array>} stream
- * @returns {Promise<Uint8Array>}
- */
-async function collect(stream) {
-  const reader = stream.getReader();
-  /** @type {Uint8Array[]} */
-  const chunks = [];
-  let total = 0;
-  for (;;) {
-    const { done, value } = await reader.read();
-    if (done) break;
-    chunks.push(value);
-    total += value.length;
-  }
-  const out = new Uint8Array(total);
-  let offset = 0;
-  for (const chunk of chunks) {
-    out.set(chunk, offset);
-    offset += chunk.length;
-  }
-  return out;
-}
-
-/**
- * Decompresses one zip entry's compressed bytes and verifies the result
- * against its RECORDED crc32 — never trusts the central directory's claim
- * alone (the exact discipline scripts/opc-bytes.py's read_package() already
- * applies). Throws if the platform has no DecompressionStream, or if the
- * decompressed content's own CRC-32 does not match the entry's recorded one.
- *
- * @param {Uint8Array} bytes  the full original buffer parseZipContainer() was given
- * @param {import("./zip-container.js").ZipEntry} entry
- * @returns {Promise<Uint8Array>}
- */
-export async function inflateEntry(bytes, entry) {
-  const compressed = bytes.subarray(entry.compressedRange.start, entry.compressedRange.end);
-
-  let content;
-  if (entry.method === METHOD_STORE) {
-    content = compressed;
-  } else {
-    if (typeof DecompressionStream === "undefined") {
-      throw new Error(
-        `inflate: ${JSON.stringify(entry.name)} needs DecompressionStream ("deflate-raw"), which this ` +
-          "runtime does not provide"
-      );
-    }
-    const source = new ReadableStream({
-      start(controller) {
-        controller.enqueue(compressed);
-        controller.close();
-      },
-    });
-    const decompressedStream = source.pipeThrough(new DecompressionStream("deflate-raw"));
-    content = await collect(decompressedStream);
-  }
-
-  const actual = crc32(content);
-  if (actual !== entry.crc32) {
-    throw new Error(
-      `inflate: ${JSON.stringify(entry.name)}: CRC-32 mismatch after decompression ` +
-        `(expected ${entry.crc32.toString(16)}, got ${actual.toString(16)}) — refusing to use this content`
-    );
-  }
-
-  return content;
-}
+// The ONE file in src/io/ooxml/ that touches a platform global (ADR-0020
+// Karar 8, §1/§7's own precedent in src/measure/probe.js). Decompression is
+// lazy and per-entry: zip-container.js never decompresses anything, so a
+// caller only pays this async cost for the members it actually needs —
+// e.g. it can list every part in a package and read sizes/names without
+// ever calling this function.
+//
+// `typeof DecompressionStream !== "undefined"` is checked INSIDE the
+// function, never at module scope (invariant #5, ssr-safety) — importing
+// this file must never touch a global, only calling inflateEntry() may.
+// Measured 2026-08-27: DecompressionStream is a Node v24.13.1 global,
+// matching ADR-0020's own claim, but WHICH Node 18.x point release first
+// made it global (vs. requiring an import) was not checked — the lazy
+// `typeof` guard means that question never has to be answered, the code is
+// correct either way, and a missing API fails at the first call, not at
+// import time.
+//
+// CRC-32 verification happens HERE, not in zip-container.js, because it
+// needs decompressed bytes — the pure/sync layer never sees them. Ported
+// check from scripts/opc-bytes.py's read_package(): "CRC-32 mismatch in the
+// SOURCE file — refusing to touch it" (never trust the central directory's
+// claim alone).
+
+const METHOD_STORE = 0;
+
+/** @type {Uint32Array | null} */
+let crcTable = null;
+
+function getCrcTable() {
+  if (crcTable !== null) return crcTable;
+  const table = new Uint32Array(256);
+  for (let n = 0; n < 256; n++) {
+    let c = n;
+    for (let k = 0; k < 8; k++) {
+      c = c & 1 ? 0xedb88320 ^ (c >>> 1) : c >>> 1;
+    }
+    table[n] = c >>> 0;
+  }
+  crcTable = table;
+  return table;
+}
+
+/**
+ * Standard CRC-32 (ISO 3309 / PKZIP / zlib) — the same algorithm
+ * `zlib.crc32` implements in Python, bit for bit. A fixed, well-known
+ * specification, not a guess.
+ * @param {Uint8Array} bytes
+ * @returns {number}
+ */
+export function crc32(bytes) {
+  const table = getCrcTable();
+  let crc = 0xffffffff;
+  for (let i = 0; i < bytes.length; i++) {
+    crc = table[(crc ^ bytes[i]) & 0xff] ^ (crc >>> 8);
+  }
+  return (crc ^ 0xffffffff) >>> 0;
+}
+
+/**
+ * @param {ReadableStream<Uint8Array>} stream
+ * @returns {Promise<Uint8Array>}
+ */
+async function collect(stream) {
+  const reader = stream.getReader();
+  /** @type {Uint8Array[]} */
+  const chunks = [];
+  let total = 0;
+  for (;;) {
+    const { done, value } = await reader.read();
+    if (done) break;
+    chunks.push(value);
+    total += value.length;
+  }
+  const out = new Uint8Array(total);
+  let offset = 0;
+  for (const chunk of chunks) {
+    out.set(chunk, offset);
+    offset += chunk.length;
+  }
+  return out;
+}
+
+/**
+ * Decompresses one zip entry's compressed bytes and verifies the result
+ * against its RECORDED crc32 — never trusts the central directory's claim
+ * alone (the exact discipline scripts/opc-bytes.py's read_package() already
+ * applies). Throws if the platform has no DecompressionStream, or if the
+ * decompressed content's own CRC-32 does not match the entry's recorded one.
+ *
+ * @param {Uint8Array} bytes  the full original buffer parseZipContainer() was given
+ * @param {import("./zip-container.js").ZipEntry} entry
+ * @returns {Promise<Uint8Array>}
+ */
+export async function inflateEntry(bytes, entry) {
+  const compressed = bytes.subarray(entry.compressedRange.start, entry.compressedRange.end);
+
+  let content;
+  if (entry.method === METHOD_STORE) {
+    content = compressed;
+  } else {
+    if (typeof DecompressionStream === "undefined") {
+      throw new Error(
+        `inflate: ${JSON.stringify(entry.name)} needs DecompressionStream ("deflate-raw"), which this ` +
+          "runtime does not provide"
+      );
+    }
+    const source = new ReadableStream({
+      start(controller) {
+        controller.enqueue(compressed);
+        controller.close();
+      },
+    });
+    const decompressedStream = source.pipeThrough(new DecompressionStream("deflate-raw"));
+    content = await collect(decompressedStream);
+  }
+
+  const actual = crc32(content);
+  if (actual !== entry.crc32) {
+    throw new Error(
+      `inflate: ${JSON.stringify(entry.name)}: CRC-32 mismatch after decompression ` +
+        `(expected ${entry.crc32.toString(16)}, got ${actual.toString(16)}) — refusing to use this content`
+    );
+  }
+
+  return content;
+}

package/dist/io/ooxml/media.js

@@ -1,87 +1,87 @@
-// Turns the retained package's media parts into `data:` URIs the renderer can
-// put on a Doc.image's `<img src>`. Read-only: it inflates `word/media/*` from
-// the ALREADY-retained package bytes (provenance.packageBytes/entries) with no
-// network access at all (invariant #3 — no request is ever issued), and the
-// export path is untouched (the media part stays a verbatim OPC part). base64
-// is a pure encoder rather than btoa/Buffer so nothing depends on a platform
-// global.
-
-import { inflateEntry } from "./inflate.js";
-
-/** @type {Record<string, string>} */
-const IMAGE_MIME = {
-  png: "image/png",
-  jpg: "image/jpeg",
-  jpeg: "image/jpeg",
-  gif: "image/gif",
-  bmp: "image/bmp",
-  tif: "image/tiff",
-  tiff: "image/tiff",
-  svg: "image/svg+xml",
-  webp: "image/webp",
-  emf: "image/emf",
-  wmf: "image/wmf",
-};
-
-/**
- * The image MIME type for a media part, from its extension. A wrong MIME breaks
- * the `<img>`, so this maps the extensions the corpus actually uses; an unknown
- * one falls back to the generic binary type (the browser sniffs common formats).
- * @param {string} partName
- * @returns {string}
- */
-function mimeForPart(partName) {
-  const dot = partName.lastIndexOf(".");
-  const ext = dot === -1 ? "" : partName.slice(dot + 1).toLowerCase();
-  return IMAGE_MIME[ext] ?? "application/octet-stream";
-}
-
-const B64 = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
-
-/**
- * Standard base64 of a byte array — pure, no `btoa`/`Buffer` (invariant #5:
- * nothing platform-specific at import or here).
- * @param {Uint8Array} bytes
- * @returns {string}
- */
-function toBase64(bytes) {
-  let out = "";
-  for (let i = 0; i < bytes.length; i += 3) {
-    const b0 = bytes[i];
-    const b1 = i + 1 < bytes.length ? bytes[i + 1] : -1;
-    const b2 = i + 2 < bytes.length ? bytes[i + 2] : -1;
-    out += B64[b0 >> 2];
-    out += B64[((b0 & 3) << 4) | (b1 < 0 ? 0 : b1 >> 4)];
-    out += b1 < 0 ? "=" : B64[((b1 & 15) << 2) | (b2 < 0 ? 0 : b2 >> 6)];
-    out += b2 < 0 ? "=" : B64[b2 & 63];
-  }
-  return out;
-}
-
-/**
- * `mediaPartName` -> `data:` URI for every image `doc` references, read from the
- * retained package. Only the parts actually referenced by `Doc.images` are
- * decoded. An empty/absent package (convertDocument's provenance, no images) or
- * a missing part yields no entry for that name — the renderer then draws the
- * `<img>` without a `src`.
- *
- * @param {import("../../model/node.js").Doc} doc
- * @param {import("./provenance.js").Provenance} provenance
- * @returns {Promise<Map<string, string>>}
- */
-export async function buildMediaMap(doc, provenance) {
-  /** @type {Map<string, string>} */
-  const map = new Map();
-  const { packageBytes, entries } = provenance;
-  if (packageBytes === undefined || entries === undefined) {
-    return map;
-  }
-  const wanted = new Set(doc.images.map((im) => im.mediaPartName));
-  for (const partName of wanted) {
-    const entry = entries.find((e) => e.name === partName);
-    if (entry === undefined) continue;
-    const bytes = await inflateEntry(packageBytes, entry);
-    map.set(partName, `data:${mimeForPart(partName)};base64,${toBase64(bytes)}`);
-  }
-  return map;
-}
+// Turns the retained package's media parts into `data:` URIs the renderer can
+// put on a Doc.image's `<img src>`. Read-only: it inflates `word/media/*` from
+// the ALREADY-retained package bytes (provenance.packageBytes/entries) with no
+// network access at all (invariant #3 — no request is ever issued), and the
+// export path is untouched (the media part stays a verbatim OPC part). base64
+// is a pure encoder rather than btoa/Buffer so nothing depends on a platform
+// global.
+
+import { inflateEntry } from "./inflate.js";
+
+/** @type {Record<string, string>} */
+const IMAGE_MIME = {
+  png: "image/png",
+  jpg: "image/jpeg",
+  jpeg: "image/jpeg",
+  gif: "image/gif",
+  bmp: "image/bmp",
+  tif: "image/tiff",
+  tiff: "image/tiff",
+  svg: "image/svg+xml",
+  webp: "image/webp",
+  emf: "image/emf",
+  wmf: "image/wmf",
+};
+
+/**
+ * The image MIME type for a media part, from its extension. A wrong MIME breaks
+ * the `<img>`, so this maps the extensions the corpus actually uses; an unknown
+ * one falls back to the generic binary type (the browser sniffs common formats).
+ * @param {string} partName
+ * @returns {string}
+ */
+function mimeForPart(partName) {
+  const dot = partName.lastIndexOf(".");
+  const ext = dot === -1 ? "" : partName.slice(dot + 1).toLowerCase();
+  return IMAGE_MIME[ext] ?? "application/octet-stream";
+}
+
+const B64 = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
+
+/**
+ * Standard base64 of a byte array — pure, no `btoa`/`Buffer` (invariant #5:
+ * nothing platform-specific at import or here).
+ * @param {Uint8Array} bytes
+ * @returns {string}
+ */
+function toBase64(bytes) {
+  let out = "";
+  for (let i = 0; i < bytes.length; i += 3) {
+    const b0 = bytes[i];
+    const b1 = i + 1 < bytes.length ? bytes[i + 1] : -1;
+    const b2 = i + 2 < bytes.length ? bytes[i + 2] : -1;
+    out += B64[b0 >> 2];
+    out += B64[((b0 & 3) << 4) | (b1 < 0 ? 0 : b1 >> 4)];
+    out += b1 < 0 ? "=" : B64[((b1 & 15) << 2) | (b2 < 0 ? 0 : b2 >> 6)];
+    out += b2 < 0 ? "=" : B64[b2 & 63];
+  }
+  return out;
+}
+
+/**
+ * `mediaPartName` -> `data:` URI for every image `doc` references, read from the
+ * retained package. Only the parts actually referenced by `Doc.images` are
+ * decoded. An empty/absent package (convertDocument's provenance, no images) or
+ * a missing part yields no entry for that name — the renderer then draws the
+ * `<img>` without a `src`.
+ *
+ * @param {import("../../model/node.js").Doc} doc
+ * @param {import("./provenance.js").Provenance} provenance
+ * @returns {Promise<Map<string, string>>}
+ */
+export async function buildMediaMap(doc, provenance) {
+  /** @type {Map<string, string>} */
+  const map = new Map();
+  const { packageBytes, entries } = provenance;
+  if (packageBytes === undefined || entries === undefined) {
+    return map;
+  }
+  const wanted = new Set(doc.images.map((im) => im.mediaPartName));
+  for (const partName of wanted) {
+    const entry = entries.find((e) => e.name === partName);
+    if (entry === undefined) continue;
+    const bytes = await inflateEntry(packageBytes, entry);
+    map.set(partName, `data:${mimeForPart(partName)};base64,${toBase64(bytes)}`);
+  }
+  return map;
+}

package/dist/io/ooxml/ns.d.ts

@@ -3,6 +3,8 @@ export const WP_NS: "http://schemas.openxmlformats.org/drawingml/2006/wordproces
 export const A_NS: "http://schemas.openxmlformats.org/drawingml/2006/main";
 export const WPS_NS: "http://schemas.microsoft.com/office/word/2010/wordprocessingShape";
 export const WPG_NS: "http://schemas.microsoft.com/office/word/2010/wordprocessingGroup";
+export const WP14_NS: "http://schemas.microsoft.com/office/word/2010/wordprocessingDrawing";
+export const WORD_COMPAT_URI: "http://schemas.microsoft.com/office/word";
 export const PIC_NS: "http://schemas.openxmlformats.org/drawingml/2006/picture";
 export const MC_NS: "http://schemas.openxmlformats.org/markup-compatibility/2006";
 export const R_NS: "http://schemas.openxmlformats.org/officeDocument/2006/relationships";

package/dist/io/ooxml/ns.js

@@ -1,21 +1,26 @@
-// Shared OOXML namespace URIs. One definition, several consumers
-// (drawingml.js, sections.js, import-gate.js, custgeom.js) — a namespace URI
-// written twice risks the exact drift this project's own
-// scripts/network-patterns.js already had once (a copy in verify.js, a copy
-// in the test, one updated when P2 added the SVG namespace, the other not).
-
-export const W_NS = "http://schemas.openxmlformats.org/wordprocessingml/2006/main";
-export const WP_NS = "http://schemas.openxmlformats.org/drawingml/2006/wordprocessingDrawing";
-export const A_NS = "http://schemas.openxmlformats.org/drawingml/2006/main";
-export const WPS_NS = "http://schemas.microsoft.com/office/word/2010/wordprocessingShape";
-export const WPG_NS = "http://schemas.microsoft.com/office/word/2010/wordprocessingGroup";
-export const PIC_NS = "http://schemas.openxmlformats.org/drawingml/2006/picture";
-export const MC_NS = "http://schemas.openxmlformats.org/markup-compatibility/2006";
-// The `r:` prefix on `<a:blip r:embed="rIdN">` — an officeDocument relationship id.
-export const R_NS = "http://schemas.openxmlformats.org/officeDocument/2006/relationships";
-// The namespace of `<Relationship>` inside a *_rels/*.rels part (a different URI
-// from R_NS — the package relationships schema, not the officeDocument one).
-export const PKG_RELS_NS = "http://schemas.openxmlformats.org/package/2006/relationships";
-// The `xml:` prefix's own namespace — predeclared, never written in a document
-// (`xml:space` on `w:t`). An identifier, already on verify's whitelist.
-export const XML_NS = "http://www.w3.org/XML/1998/namespace";
+// Shared OOXML namespace URIs. One definition, several consumers
+// (drawingml.js, sections.js, import-gate.js, custgeom.js) — a namespace URI
+// written twice risks the exact drift this project's own
+// scripts/network-patterns.js already had once (a copy in verify.js, a copy
+// in the test, one updated when P2 added the SVG namespace, the other not).
+
+export const W_NS = "http://schemas.openxmlformats.org/wordprocessingml/2006/main";
+export const WP_NS = "http://schemas.openxmlformats.org/drawingml/2006/wordprocessingDrawing";
+export const A_NS = "http://schemas.openxmlformats.org/drawingml/2006/main";
+export const WPS_NS = "http://schemas.microsoft.com/office/word/2010/wordprocessingShape";
+export const WPG_NS = "http://schemas.microsoft.com/office/word/2010/wordprocessingGroup";
+// Word 2010's drawing extensions — checkFit reads wp14:sizeRelH (a box sized as a share of the page).
+export const WP14_NS = "http://schemas.microsoft.com/office/word/2010/wordprocessingDrawing";
+// Not an element namespace: the value of w:compatSetting@w:uri that names Word's own compatibility
+// settings (compatibilityMode). An identifier, never fetched.
+export const WORD_COMPAT_URI = "http://schemas.microsoft.com/office/word";
+export const PIC_NS = "http://schemas.openxmlformats.org/drawingml/2006/picture";
+export const MC_NS = "http://schemas.openxmlformats.org/markup-compatibility/2006";
+// The `r:` prefix on `<a:blip r:embed="rIdN">` — an officeDocument relationship id.
+export const R_NS = "http://schemas.openxmlformats.org/officeDocument/2006/relationships";
+// The namespace of `<Relationship>` inside a *_rels/*.rels part (a different URI
+// from R_NS — the package relationships schema, not the officeDocument one).
+export const PKG_RELS_NS = "http://schemas.openxmlformats.org/package/2006/relationships";
+// The `xml:` prefix's own namespace — predeclared, never written in a document
+// (`xml:space` on `w:t`). An identifier, already on verify's whitelist.
+export const XML_NS = "http://www.w3.org/XML/1998/namespace";