* Parse the XRef chain, following /Prev links for incremental updates. * * Handles hybrid-reference files (PDF 1.7 §7.5.8.4): when a traditional * xref table's trailer carries a /XRefStm entry, it points to a * supplementary xref stream containing entries for compressed objects * (whi
(
xrefParser: XRefParser,
startOffset: number,
)
| 330 | * walks every revision. |
| 331 | */ |
| 332 | private parseXRefChain( |
| 333 | xrefParser: XRefParser, |
| 334 | startOffset: number, |
| 335 | ): { xref: Map<number, XRefEntry>; trailer: PdfDict } { |
| 336 | const combinedXRef = new Map<number, XRefEntry>(); |
| 337 | let firstTrailer: PdfDict | null = null; |
| 338 | |
| 339 | // Track visited offsets to prevent infinite loops |
| 340 | const visited = new Set<number>(); |
| 341 | const queue: number[] = [startOffset]; |
| 342 | |
| 343 | while (queue.length > 0) { |
| 344 | const offset = queue.shift(); |
| 345 | |
| 346 | if (offset === undefined) { |
| 347 | break; |
| 348 | } |
| 349 | |
| 350 | // Circular reference check |
| 351 | if (visited.has(offset)) { |
| 352 | this.warnings.push(`Circular xref reference at offset ${offset}`); |
| 353 | continue; |
| 354 | } |
| 355 | |
| 356 | visited.add(offset); |
| 357 | |
| 358 | try { |
| 359 | const xrefData = xrefParser.parseAt(offset); |
| 360 | |
| 361 | // Hybrid xref: if this revision has a /XRefStm entry, it points to a |
| 362 | // supplementary xref stream whose entries override the table for the |
| 363 | // object numbers it covers. Apply the stream BEFORE the table so its |
| 364 | // entries win the "first-seen" merge below. |
| 365 | const xrefStmOffset = xrefData.trailer.getNumber("XRefStm")?.value; |
| 366 | |
| 367 | if (xrefStmOffset !== undefined && !visited.has(xrefStmOffset)) { |
| 368 | visited.add(xrefStmOffset); |
| 369 | |
| 370 | try { |
| 371 | const stmData = xrefParser.parseAt(xrefStmOffset); |
| 372 | |
| 373 | for (const [objNum, entry] of stmData.entries) { |
| 374 | if (!combinedXRef.has(objNum)) { |
| 375 | combinedXRef.set(objNum, entry); |
| 376 | } |
| 377 | } |
| 378 | } catch (error) { |
| 379 | // Hybrid stream is optional from the legacy table's perspective; |
| 380 | // if it's malformed, fall back to the table alone with a warning. |
| 381 | const message = error instanceof Error ? error.message : String(error); |
| 382 | |
| 383 | this.warnings.push(`Error parsing /XRefStm at ${xrefStmOffset}: ${message}`); |
| 384 | } |
| 385 | } |
| 386 | |
| 387 | // Merge table entries (first definition wins — covers anything the |
| 388 | // stream above didn't override, and any object not in the stream). |
| 389 | for (const [objNum, entry] of xrefData.entries) { |