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

package/dist/io/ooxml/zip-container.js

@@ -1,228 +1,228 @@
-// End of central directory -> central directory -> local headers, in one
-// pass. Pure and synchronous — no decompression happens here (that is
-// inflate.js's job, the one async file in this directory) and no XML is
-// read (that is a later parça's job). This file only answers "which zip
-// members exist, and where are their compressed bytes" — ported from
-// scripts/opc-bytes.py's read_package(), same field offsets, same
-// rejections (via zip-reject.js), same physical-order sort at the end.
-//
-// Every multi-byte field in a zip file is little-endian — DataView's
-// `littleEndian=true` argument on every read is not optional decoration,
-// it is the one thing that makes these offsets mean what the zip spec says
-// they mean.
-
-import {
-  rejectZip64Locator,
-  rejectZip64EocdSentinel,
-  rejectEntryZip64Sentinel,
-  rejectEntryEncrypted,
-  rejectEntryDataDescriptor,
-  rejectEntryUnknownMethod,
-} from "./zip-reject.js";
-
-const EOCD_SIG = [0x50, 0x4b, 0x05, 0x06]; // "PK\x05\x06"
-const CENTRAL_SIG = [0x50, 0x4b, 0x01, 0x02]; // "PK\x01\x02"
-const LOCAL_SIG = [0x50, 0x4b, 0x03, 0x04]; // "PK\x03\x04"
-
-const LOC_CRC = 14;
-const LOC_CSIZE = 18;
-const LOC_USIZE = 22;
-const LOC_NAMELEN = 26;
-const LOC_EXTRALEN = 28;
-const LOC_FIXED = 30;
-
-const CEN_FLAGS = 8;
-const CEN_METHOD = 10;
-const CEN_CRC = 16;
-const CEN_CSIZE = 20;
-const CEN_USIZE = 24;
-const CEN_NAMELEN = 28;
-const CEN_EXTRALEN = 30;
-const CEN_COMMENTLEN = 32;
-const CEN_LOCAL_OFFSET = 42;
-const CEN_FIXED = 46;
-
-const EOCD_TOTAL_ENTRIES = 10;
-const EOCD_CD_SIZE = 12;
-const EOCD_CD_OFFSET = 16;
-
-/**
- * @typedef {object} ZipEntry
- * @property {string} name
- * @property {number} method          0 = store, 8 = deflate (already checked, only these two survive)
- * @property {number} compressedSize
- * @property {number} uncompressedSize
- * @property {number} crc32           the RECORDED crc, not yet verified against decompressed content — inflate.js does that
- * @property {number} localHeaderOffset  the local header starts here; it ends at
- *   `compressedRange.start`, so the local-header bytes are that slice
- * @property {{start: number, end: number}} compressedRange  byte offsets into the ORIGINAL buffer passed to parseZipContainer
- * @property {{start: number, end: number}} centralEntryRange  this entry's own
- *   central-directory record's byte range — the repack writer (opc-write.js)
- *   re-emits it verbatim, patching only the numeric fields (opc-bytes.py parity)
- */
-
-/**
- * @param {Uint8Array} bytes
- * @param {number} offset
- * @param {number[]} sig
- * @returns {boolean}
- */
-function matchSig(bytes, offset, sig) {
-  for (let i = 0; i < sig.length; i++) {
-    if (bytes[offset + i] !== sig[i]) return false;
-  }
-  return true;
-}
-
-/**
- * @param {Uint8Array} bytes
- * @param {number[]} sig
- * @param {number} searchStart
- * @returns {number}
- */
-function findSigBackwards(bytes, sig, searchStart) {
-  outer: for (let i = searchStart; i >= 0; i--) {
-    for (let j = 0; j < sig.length; j++) {
-      if (bytes[i + j] !== sig[j]) continue outer;
-    }
-    return i;
-  }
-  return -1;
-}
-
-/**
- * The EOCD sits at the end, but a zip comment (up to 65535 bytes) may follow
- * it — scan backwards over the largest legal window, matching
- * opc-bytes.py's `_find_eocd`.
- * @param {Uint8Array} bytes
- * @returns {number}
- */
-function findEocd(bytes) {
-  const window = Math.min(bytes.length, 22 + 0xffff);
-  const searchStart = bytes.length - window;
-  const idx = findSigBackwards(bytes, EOCD_SIG, bytes.length - 4);
-  if (idx < 0 || idx < searchStart) {
-    throw new Error("zip: no end-of-central-directory record — not a zip file");
-  }
-  return idx;
-}
-
-/**
- * @param {Uint8Array} bytes  the full, undecompressed .docx/.zip file contents
- * @returns {ZipEntry[]}  in PHYSICAL order (by local header offset), matching opc-bytes.py's own final sort
- */
-export function parseZipContainer(bytes) {
-  const view = new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength);
-  const eocdOffset = findEocd(bytes);
-
-  rejectZip64Locator(bytes, eocdOffset);
-
-  const totalEntries = view.getUint16(eocdOffset + EOCD_TOTAL_ENTRIES, true);
-  const cdSize = view.getUint32(eocdOffset + EOCD_CD_SIZE, true);
-  const cdOffset = view.getUint32(eocdOffset + EOCD_CD_OFFSET, true);
-  const thisDisk = view.getUint16(eocdOffset + 4, true);
-  const diskWithCd = view.getUint16(eocdOffset + 6, true);
-  rejectZip64EocdSentinel({ diskWithCd, thisDisk, cdOffset, cdSize });
-
-  /** @type {ZipEntry[]} */
-  const entries = [];
-  let pos = cdOffset;
-  for (let i = 0; i < totalEntries; i++) {
-    if (!matchSig(bytes, pos, CENTRAL_SIG)) {
-      throw new Error(`zip: central directory entry ${i} has a bad signature at offset ${pos}`);
-    }
-    const nameLen = view.getUint16(pos + CEN_NAMELEN, true);
-    const extraLen = view.getUint16(pos + CEN_EXTRALEN, true);
-    const commentLen = view.getUint16(pos + CEN_COMMENTLEN, true);
-
-    const name = new TextDecoder("utf-8", { fatal: true }).decode(
-      bytes.subarray(pos + CEN_FIXED, pos + CEN_FIXED + nameLen)
-    );
-    const flags = view.getUint16(pos + CEN_FLAGS, true);
-    const method = view.getUint16(pos + CEN_METHOD, true);
-    const crc32 = view.getUint32(pos + CEN_CRC, true);
-    const csize = view.getUint32(pos + CEN_CSIZE, true);
-    const usize = view.getUint32(pos + CEN_USIZE, true);
-    const localOffset = view.getUint32(pos + CEN_LOCAL_OFFSET, true);
-
-    rejectEntryEncrypted(name, flags);
-    rejectEntryDataDescriptor(name, flags);
-    rejectEntryUnknownMethod(name, method);
-    rejectEntryZip64Sentinel(name, csize, usize, localOffset);
-
-    if (!matchSig(bytes, localOffset, LOCAL_SIG)) {
-      throw new Error(`zip: ${JSON.stringify(name)}: bad local header signature at offset ${localOffset}`);
-    }
-    const locNameLen = view.getUint16(localOffset + LOC_NAMELEN, true);
-    const locExtraLen = view.getUint16(localOffset + LOC_EXTRALEN, true);
-    const headerLen = LOC_FIXED + locNameLen + locExtraLen;
-    const dataStart = localOffset + headerLen;
-    const dataEnd = dataStart + csize;
-    if (dataEnd > bytes.length) {
-      throw new Error(`zip: ${JSON.stringify(name)}: truncated compressed data (need ${csize} bytes at ${dataStart}, file ends at ${bytes.length})`);
-    }
-
-    const centralEntryEnd = pos + CEN_FIXED + nameLen + extraLen + commentLen;
-    entries.push({
-      name,
-      method,
-      compressedSize: csize,
-      uncompressedSize: usize,
-      crc32,
-      localHeaderOffset: localOffset,
-      compressedRange: { start: dataStart, end: dataEnd },
-      centralEntryRange: { start: pos, end: centralEntryEnd },
-    });
-
-    pos = centralEntryEnd;
-  }
-
-  entries.sort((a, b) => a.localHeaderOffset - b.localHeaderOffset);
-  return entries;
-}
-
-/**
- * ADR-0020 does not design real OPC relationship resolution yet ([Content_
- * Types].xml / _rels/.rels parsing is XML parsing, out of scope for the zip
- * layer). Measured 2026-08-27: all 46 corpus documents' _rels/.rels resolve
- * their officeDocument relationship to exactly "word/document.xml" — so
- * that fixed path is used AS A NAMED, VISIBLE ASSUMPTION, not silently. A
- * document that does not use it is REJECTED with its own specific message
- * (ADR-0020 DOĞRULANMALI: real _rels resolution belongs in import.js, the
- * orchestrator that will have XML-reading capability once xml-scan.js
- * exists — this function is a placeholder for that, not a substitute).
- * @param {ZipEntry[]} entries
- * @returns {ZipEntry}
- */
-export function resolveMainDocumentPart(entries) {
-  const found = entries.find((e) => e.name === "word/document.xml");
-  if (found === undefined) {
-    throw new Error(
-      "zip: this package does not use the fixed path \"word/document.xml\" for its main document part — " +
-        "real _rels/.rels relationship resolution is not implemented yet (ADR-0020 DOĞRULANMALI, deferred " +
-        "to import.js). Measured 46/46 real corpus documents use this fixed path; this one does not."
-    );
-  }
-  return found;
-}
-
-/**
- * Same fixed-path assumption as `resolveMainDocumentPart`, for the styles
- * part. Measured 2026-08-27: all 46 real corpus documents use exactly
- * "word/styles.xml" — real `word/document.xml`'s own `_rels` would point to
- * it via the styles relationship type, but that resolution is the same
- * deferred XML-reading work (ADR-0020 DOĞRULANMALI #16, `import.js`).
- * @param {ZipEntry[]} entries
- * @returns {ZipEntry}
- */
-export function resolveStylesPart(entries) {
-  const found = entries.find((e) => e.name === "word/styles.xml");
-  if (found === undefined) {
-    throw new Error(
-      "zip: this package does not use the fixed path \"word/styles.xml\" for its styles part — " +
-        "real _rels/.rels relationship resolution is not implemented yet (ADR-0020 DOĞRULANMALI, deferred " +
-        "to import.js). Measured 46/46 real corpus documents use this fixed path; this one does not."
-    );
-  }
-  return found;
-}
+// End of central directory -> central directory -> local headers, in one
+// pass. Pure and synchronous — no decompression happens here (that is
+// inflate.js's job, the one async file in this directory) and no XML is
+// read (that is a later parça's job). This file only answers "which zip
+// members exist, and where are their compressed bytes" — ported from
+// scripts/opc-bytes.py's read_package(), same field offsets, same
+// rejections (via zip-reject.js), same physical-order sort at the end.
+//
+// Every multi-byte field in a zip file is little-endian — DataView's
+// `littleEndian=true` argument on every read is not optional decoration,
+// it is the one thing that makes these offsets mean what the zip spec says
+// they mean.
+
+import {
+  rejectZip64Locator,
+  rejectZip64EocdSentinel,
+  rejectEntryZip64Sentinel,
+  rejectEntryEncrypted,
+  rejectEntryDataDescriptor,
+  rejectEntryUnknownMethod,
+} from "./zip-reject.js";
+
+const EOCD_SIG = [0x50, 0x4b, 0x05, 0x06]; // "PK\x05\x06"
+const CENTRAL_SIG = [0x50, 0x4b, 0x01, 0x02]; // "PK\x01\x02"
+const LOCAL_SIG = [0x50, 0x4b, 0x03, 0x04]; // "PK\x03\x04"
+
+const LOC_CRC = 14;
+const LOC_CSIZE = 18;
+const LOC_USIZE = 22;
+const LOC_NAMELEN = 26;
+const LOC_EXTRALEN = 28;
+const LOC_FIXED = 30;
+
+const CEN_FLAGS = 8;
+const CEN_METHOD = 10;
+const CEN_CRC = 16;
+const CEN_CSIZE = 20;
+const CEN_USIZE = 24;
+const CEN_NAMELEN = 28;
+const CEN_EXTRALEN = 30;
+const CEN_COMMENTLEN = 32;
+const CEN_LOCAL_OFFSET = 42;
+const CEN_FIXED = 46;
+
+const EOCD_TOTAL_ENTRIES = 10;
+const EOCD_CD_SIZE = 12;
+const EOCD_CD_OFFSET = 16;
+
+/**
+ * @typedef {object} ZipEntry
+ * @property {string} name
+ * @property {number} method          0 = store, 8 = deflate (already checked, only these two survive)
+ * @property {number} compressedSize
+ * @property {number} uncompressedSize
+ * @property {number} crc32           the RECORDED crc, not yet verified against decompressed content — inflate.js does that
+ * @property {number} localHeaderOffset  the local header starts here; it ends at
+ *   `compressedRange.start`, so the local-header bytes are that slice
+ * @property {{start: number, end: number}} compressedRange  byte offsets into the ORIGINAL buffer passed to parseZipContainer
+ * @property {{start: number, end: number}} centralEntryRange  this entry's own
+ *   central-directory record's byte range — the repack writer (opc-write.js)
+ *   re-emits it verbatim, patching only the numeric fields (opc-bytes.py parity)
+ */
+
+/**
+ * @param {Uint8Array} bytes
+ * @param {number} offset
+ * @param {number[]} sig
+ * @returns {boolean}
+ */
+function matchSig(bytes, offset, sig) {
+  for (let i = 0; i < sig.length; i++) {
+    if (bytes[offset + i] !== sig[i]) return false;
+  }
+  return true;
+}
+
+/**
+ * @param {Uint8Array} bytes
+ * @param {number[]} sig
+ * @param {number} searchStart
+ * @returns {number}
+ */
+function findSigBackwards(bytes, sig, searchStart) {
+  outer: for (let i = searchStart; i >= 0; i--) {
+    for (let j = 0; j < sig.length; j++) {
+      if (bytes[i + j] !== sig[j]) continue outer;
+    }
+    return i;
+  }
+  return -1;
+}
+
+/**
+ * The EOCD sits at the end, but a zip comment (up to 65535 bytes) may follow
+ * it — scan backwards over the largest legal window, matching
+ * opc-bytes.py's `_find_eocd`.
+ * @param {Uint8Array} bytes
+ * @returns {number}
+ */
+function findEocd(bytes) {
+  const window = Math.min(bytes.length, 22 + 0xffff);
+  const searchStart = bytes.length - window;
+  const idx = findSigBackwards(bytes, EOCD_SIG, bytes.length - 4);
+  if (idx < 0 || idx < searchStart) {
+    throw new Error("zip: no end-of-central-directory record — not a zip file");
+  }
+  return idx;
+}
+
+/**
+ * @param {Uint8Array} bytes  the full, undecompressed .docx/.zip file contents
+ * @returns {ZipEntry[]}  in PHYSICAL order (by local header offset), matching opc-bytes.py's own final sort
+ */
+export function parseZipContainer(bytes) {
+  const view = new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength);
+  const eocdOffset = findEocd(bytes);
+
+  rejectZip64Locator(bytes, eocdOffset);
+
+  const totalEntries = view.getUint16(eocdOffset + EOCD_TOTAL_ENTRIES, true);
+  const cdSize = view.getUint32(eocdOffset + EOCD_CD_SIZE, true);
+  const cdOffset = view.getUint32(eocdOffset + EOCD_CD_OFFSET, true);
+  const thisDisk = view.getUint16(eocdOffset + 4, true);
+  const diskWithCd = view.getUint16(eocdOffset + 6, true);
+  rejectZip64EocdSentinel({ diskWithCd, thisDisk, cdOffset, cdSize });
+
+  /** @type {ZipEntry[]} */
+  const entries = [];
+  let pos = cdOffset;
+  for (let i = 0; i < totalEntries; i++) {
+    if (!matchSig(bytes, pos, CENTRAL_SIG)) {
+      throw new Error(`zip: central directory entry ${i} has a bad signature at offset ${pos}`);
+    }
+    const nameLen = view.getUint16(pos + CEN_NAMELEN, true);
+    const extraLen = view.getUint16(pos + CEN_EXTRALEN, true);
+    const commentLen = view.getUint16(pos + CEN_COMMENTLEN, true);
+
+    const name = new TextDecoder("utf-8", { fatal: true }).decode(
+      bytes.subarray(pos + CEN_FIXED, pos + CEN_FIXED + nameLen)
+    );
+    const flags = view.getUint16(pos + CEN_FLAGS, true);
+    const method = view.getUint16(pos + CEN_METHOD, true);
+    const crc32 = view.getUint32(pos + CEN_CRC, true);
+    const csize = view.getUint32(pos + CEN_CSIZE, true);
+    const usize = view.getUint32(pos + CEN_USIZE, true);
+    const localOffset = view.getUint32(pos + CEN_LOCAL_OFFSET, true);
+
+    rejectEntryEncrypted(name, flags);
+    rejectEntryDataDescriptor(name, flags);
+    rejectEntryUnknownMethod(name, method);
+    rejectEntryZip64Sentinel(name, csize, usize, localOffset);
+
+    if (!matchSig(bytes, localOffset, LOCAL_SIG)) {
+      throw new Error(`zip: ${JSON.stringify(name)}: bad local header signature at offset ${localOffset}`);
+    }
+    const locNameLen = view.getUint16(localOffset + LOC_NAMELEN, true);
+    const locExtraLen = view.getUint16(localOffset + LOC_EXTRALEN, true);
+    const headerLen = LOC_FIXED + locNameLen + locExtraLen;
+    const dataStart = localOffset + headerLen;
+    const dataEnd = dataStart + csize;
+    if (dataEnd > bytes.length) {
+      throw new Error(`zip: ${JSON.stringify(name)}: truncated compressed data (need ${csize} bytes at ${dataStart}, file ends at ${bytes.length})`);
+    }
+
+    const centralEntryEnd = pos + CEN_FIXED + nameLen + extraLen + commentLen;
+    entries.push({
+      name,
+      method,
+      compressedSize: csize,
+      uncompressedSize: usize,
+      crc32,
+      localHeaderOffset: localOffset,
+      compressedRange: { start: dataStart, end: dataEnd },
+      centralEntryRange: { start: pos, end: centralEntryEnd },
+    });
+
+    pos = centralEntryEnd;
+  }
+
+  entries.sort((a, b) => a.localHeaderOffset - b.localHeaderOffset);
+  return entries;
+}
+
+/**
+ * ADR-0020 does not design real OPC relationship resolution yet ([Content_
+ * Types].xml / _rels/.rels parsing is XML parsing, out of scope for the zip
+ * layer). Measured 2026-08-27: all 46 corpus documents' _rels/.rels resolve
+ * their officeDocument relationship to exactly "word/document.xml" — so
+ * that fixed path is used AS A NAMED, VISIBLE ASSUMPTION, not silently. A
+ * document that does not use it is REJECTED with its own specific message
+ * (ADR-0020 DOĞRULANMALI: real _rels resolution belongs in import.js, the
+ * orchestrator that will have XML-reading capability once xml-scan.js
+ * exists — this function is a placeholder for that, not a substitute).
+ * @param {ZipEntry[]} entries
+ * @returns {ZipEntry}
+ */
+export function resolveMainDocumentPart(entries) {
+  const found = entries.find((e) => e.name === "word/document.xml");
+  if (found === undefined) {
+    throw new Error(
+      "zip: this package does not use the fixed path \"word/document.xml\" for its main document part — " +
+        "real _rels/.rels relationship resolution is not implemented yet (ADR-0020 DOĞRULANMALI, deferred " +
+        "to import.js). Measured 46/46 real corpus documents use this fixed path; this one does not."
+    );
+  }
+  return found;
+}
+
+/**
+ * Same fixed-path assumption as `resolveMainDocumentPart`, for the styles
+ * part. Measured 2026-08-27: all 46 real corpus documents use exactly
+ * "word/styles.xml" — real `word/document.xml`'s own `_rels` would point to
+ * it via the styles relationship type, but that resolution is the same
+ * deferred XML-reading work (ADR-0020 DOĞRULANMALI #16, `import.js`).
+ * @param {ZipEntry[]} entries
+ * @returns {ZipEntry}
+ */
+export function resolveStylesPart(entries) {
+  const found = entries.find((e) => e.name === "word/styles.xml");
+  if (found === undefined) {
+    throw new Error(
+      "zip: this package does not use the fixed path \"word/styles.xml\" for its styles part — " +
+        "real _rels/.rels relationship resolution is not implemented yet (ADR-0020 DOĞRULANMALI, deferred " +
+        "to import.js). Measured 46/46 real corpus documents use this fixed path; this one does not."
+    );
+  }
+  return found;
+}