Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
100.00% |
165 / 165 |
|
100.00% |
20 / 20 |
CRAP | |
100.00% |
1 / 1 |
| Document | |
100.00% |
165 / 165 |
|
100.00% |
20 / 20 |
78 | |
100.00% |
1 / 1 |
| __construct | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| getDecodeBudget | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| fromFile | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
3 | |||
| getTrailer | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| getObjectNumbers | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| getRoot | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
2 | |||
| getObject | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
4 | |||
| resolve | |
100.00% |
8 / 8 |
|
100.00% |
1 / 1 |
3 | |||
| getOrResolveFontInfo | |
100.00% |
12 / 12 |
|
100.00% |
1 / 1 |
4 | |||
| parseAt | |
100.00% |
8 / 8 |
|
100.00% |
1 / 1 |
3 | |||
| getFromObjectStream | |
100.00% |
13 / 13 |
|
100.00% |
1 / 1 |
4 | |||
| load | |
100.00% |
11 / 11 |
|
100.00% |
1 / 1 |
4 | |||
| loadViaXref | |
100.00% |
25 / 25 |
|
100.00% |
1 / 1 |
9 | |||
| mergeXrefSection | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
3 | |||
| isClassicXref | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
2 | |||
| loadViaRepair | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
2 | |||
| expandObjectStreamsFromRepair | |
100.00% |
19 / 19 |
|
100.00% |
1 / 1 |
10 | |||
| findCatalogReference | |
100.00% |
16 / 16 |
|
100.00% |
1 / 1 |
10 | |||
| isUsable | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
8 | |||
| looksLikeObjectAt | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
3 | |||
| 1 | <?php |
| 2 | declare(strict_types=1); |
| 3 | /** |
| 4 | * Pop PHP Framework (https://www.popphp.org/) |
| 5 | * |
| 6 | * @link https://github.com/popphp/popphp-framework |
| 7 | * @author Nick Sagona, III <nick@popphp.org> |
| 8 | * @copyright Copyright (c) 2009-2026 Nick Sagona, III |
| 9 | * @license https://www.popphp.org/license New BSD License |
| 10 | */ |
| 11 | |
| 12 | /** |
| 13 | * @namespace |
| 14 | */ |
| 15 | namespace Pop\Pdf\Extract; |
| 16 | |
| 17 | use Pop\Pdf\Extract\Filter\Budget; |
| 18 | |
| 19 | /** |
| 20 | * Pdf extract document class |
| 21 | * |
| 22 | * @category Pop |
| 23 | * @package Pop\Pdf |
| 24 | * @author Nick Sagona, III <nick@popphp.org> |
| 25 | * @copyright Copyright (c) 2009-2026 Nick Sagona, III |
| 26 | * @license https://www.popphp.org/license New BSD License |
| 27 | * @version 6.0.0 |
| 28 | */ |
| 29 | class Document |
| 30 | { |
| 31 | |
| 32 | /** |
| 33 | * Maximum bytes retained by the font info cache |
| 34 | */ |
| 35 | protected const MAX_FONT_INFO_CACHE_BYTES = 67108864; // 64MB |
| 36 | |
| 37 | /** |
| 38 | * Maximum total bytes this document may decode across every stream over its lifetime. |
| 39 | * |
| 40 | * The Budget only throws once a charge pushes the running total past this ceiling - by |
| 41 | * then, the decode that tipped it over (up to Flate's own 64MB per-call cap) has already |
| 42 | * completed and its output is retained in memory alongside every prior charged chunk, so |
| 43 | * real peak usage runs measurably higher than this number, on top of the PHP process's own |
| 44 | * baseline footprint. This must stay comfortably below common PHP memory_limit floors (128M |
| 45 | * on conservative/shared hosting) or the process hits a hard, uncatchable OOM fatal before |
| 46 | * the Budget ever gets to throw its catchable Exception - which is exactly what this |
| 47 | * constant exists to prevent. |
| 48 | */ |
| 49 | protected const MAX_TOTAL_DECODED_BYTES = 67108864; // 64MB |
| 50 | |
| 51 | /** |
| 52 | * Raw PDF data |
| 53 | * @var string |
| 54 | */ |
| 55 | protected string $data; |
| 56 | |
| 57 | /** |
| 58 | * Object number to xref location map |
| 59 | * @var array |
| 60 | */ |
| 61 | protected array $offsets = []; |
| 62 | |
| 63 | /** |
| 64 | * Trailer dictionary |
| 65 | * @var array |
| 66 | */ |
| 67 | protected array $trailer = []; |
| 68 | |
| 69 | /** |
| 70 | * Resolved object cache |
| 71 | * @var array |
| 72 | */ |
| 73 | protected array $cache = []; |
| 74 | |
| 75 | /** |
| 76 | * Expanded object stream cache, keyed by stream object number |
| 77 | * @var array |
| 78 | */ |
| 79 | protected array $objectStreamCache = []; |
| 80 | |
| 81 | /** |
| 82 | * Positional view of each expanded object stream (index within the stream |
| 83 | * to object value), keyed by stream object number. Built once alongside |
| 84 | * objectStreamCache so index lookups don't re-run array_values() on every call. |
| 85 | * @var array |
| 86 | */ |
| 87 | protected array $objectStreamIndexCache = []; |
| 88 | |
| 89 | /** |
| 90 | * Object numbers currently being resolved, used to detect circular references |
| 91 | * @var array |
| 92 | */ |
| 93 | protected array $resolving = []; |
| 94 | |
| 95 | /** |
| 96 | * Object stream numbers currently being expanded, used to detect circular references |
| 97 | * @var array |
| 98 | */ |
| 99 | protected array $expandingStreams = []; |
| 100 | |
| 101 | /** |
| 102 | * Resolved FontInfo cache, keyed by a caller-supplied cache key |
| 103 | * @var array |
| 104 | */ |
| 105 | protected array $fontInfoCache = []; |
| 106 | |
| 107 | /** |
| 108 | * Running total of bytes retained in the font info cache |
| 109 | * @var int |
| 110 | */ |
| 111 | protected int $fontInfoCacheBytes = 0; |
| 112 | |
| 113 | /** |
| 114 | * Most-recently-used font info cache key, used once the byte budget is exhausted |
| 115 | * @var ?string |
| 116 | */ |
| 117 | protected ?string $fontInfoMruKey = null; |
| 118 | |
| 119 | /** |
| 120 | * Most-recently-used font info cache value, used once the byte budget is exhausted |
| 121 | * @var mixed |
| 122 | */ |
| 123 | protected mixed $fontInfoMruValue = null; |
| 124 | |
| 125 | /** |
| 126 | * Total decoded-byte budget shared by every stream this document decodes |
| 127 | * @var Budget |
| 128 | */ |
| 129 | protected Budget $decodeBudget; |
| 130 | |
| 131 | /** |
| 132 | * Constructor |
| 133 | * |
| 134 | * Instantiate a document from raw PDF data. |
| 135 | * |
| 136 | * @param string $data |
| 137 | */ |
| 138 | public function __construct(string $data) |
| 139 | { |
| 140 | $this->decodeBudget = new Budget(self::MAX_TOTAL_DECODED_BYTES); |
| 141 | $this->data = $data; |
| 142 | $this->load(); |
| 143 | } |
| 144 | |
| 145 | /** |
| 146 | * Get this document's shared decode budget |
| 147 | * |
| 148 | * @return Budget |
| 149 | */ |
| 150 | public function getDecodeBudget(): Budget |
| 151 | { |
| 152 | return $this->decodeBudget; |
| 153 | } |
| 154 | |
| 155 | /** |
| 156 | * Create a document from a PDF file |
| 157 | * |
| 158 | * @param string $file |
| 159 | * @throws Exception |
| 160 | * @return Document |
| 161 | */ |
| 162 | public static function fromFile(string $file): Document |
| 163 | { |
| 164 | if (!file_exists($file)) { |
| 165 | throw new Exception('Error: That PDF file does not exist.'); |
| 166 | } |
| 167 | |
| 168 | // Suppressed: a read failure here (e.g. permission denied) is |
| 169 | // already converted into a typed Exception below, so the native |
| 170 | // PHP warning would just be noise on an already-handled failure. |
| 171 | $data = @file_get_contents($file); |
| 172 | |
| 173 | if ($data === false) { |
| 174 | throw new Exception('Error: Could not read that PDF file.'); |
| 175 | } |
| 176 | |
| 177 | return new self($data); |
| 178 | } |
| 179 | |
| 180 | /** |
| 181 | * Get the trailer dictionary |
| 182 | * |
| 183 | * @return array |
| 184 | */ |
| 185 | public function getTrailer(): array |
| 186 | { |
| 187 | return $this->trailer; |
| 188 | } |
| 189 | |
| 190 | /** |
| 191 | * Get every object number this document's xref exposes |
| 192 | * |
| 193 | * @return array |
| 194 | */ |
| 195 | public function getObjectNumbers(): array |
| 196 | { |
| 197 | return array_keys($this->offsets); |
| 198 | } |
| 199 | |
| 200 | /** |
| 201 | * Get the resolved document catalog (Root) |
| 202 | * |
| 203 | * @throws Exception |
| 204 | * @return array |
| 205 | */ |
| 206 | public function getRoot(): array |
| 207 | { |
| 208 | $root = $this->resolve($this->trailer['Root'] ?? null); |
| 209 | |
| 210 | if (!is_array($root)) { |
| 211 | throw new Exception('Error: Could not resolve the PDF document catalog (Root).'); |
| 212 | } |
| 213 | |
| 214 | return $root; |
| 215 | } |
| 216 | |
| 217 | /** |
| 218 | * Get an object by object number, from the cache or by parsing/expanding it |
| 219 | * |
| 220 | * @param int $objNum |
| 221 | * @return mixed |
| 222 | */ |
| 223 | public function getObject(int $objNum): mixed |
| 224 | { |
| 225 | if (array_key_exists($objNum, $this->cache)) { |
| 226 | return $this->cache[$objNum]; |
| 227 | } |
| 228 | |
| 229 | if (!isset($this->offsets[$objNum])) { |
| 230 | return null; |
| 231 | } |
| 232 | |
| 233 | $location = $this->offsets[$objNum]; |
| 234 | $value = isset($location['inStream']) |
| 235 | ? $this->getFromObjectStream($location['inStream'], $location['index']) |
| 236 | : $this->parseAt($location['offset']); |
| 237 | |
| 238 | $this->cache[$objNum] = $value; |
| 239 | |
| 240 | return $value; |
| 241 | } |
| 242 | |
| 243 | /** |
| 244 | * Resolve a value, following indirect references until a direct value is reached |
| 245 | * |
| 246 | * @param mixed $value |
| 247 | * @throws Exception |
| 248 | * @return mixed |
| 249 | */ |
| 250 | public function resolve(mixed $value): mixed |
| 251 | { |
| 252 | if ($value instanceof Value\Reference) { |
| 253 | $objNum = $value->objNum; |
| 254 | |
| 255 | if (isset($this->resolving[$objNum])) { |
| 256 | throw new Exception("Error: Circular reference detected while resolving object {$objNum}."); |
| 257 | } |
| 258 | |
| 259 | $this->resolving[$objNum] = true; |
| 260 | |
| 261 | try { |
| 262 | return $this->resolve($this->getObject($objNum)); |
| 263 | } finally { |
| 264 | unset($this->resolving[$objNum]); |
| 265 | } |
| 266 | } |
| 267 | |
| 268 | return $value; |
| 269 | } |
| 270 | |
| 271 | /** |
| 272 | * Get a cached FontInfo result for a key, or compute and (budget permitting) cache it |
| 273 | * |
| 274 | * @param string $key |
| 275 | * @param callable $factory |
| 276 | * @return mixed |
| 277 | */ |
| 278 | public function getOrResolveFontInfo(string $key, callable $factory): mixed |
| 279 | { |
| 280 | if (array_key_exists($key, $this->fontInfoCache)) { |
| 281 | return $this->fontInfoCache[$key]; |
| 282 | } |
| 283 | |
| 284 | if ($this->fontInfoMruKey === $key) { |
| 285 | return $this->fontInfoMruValue; |
| 286 | } |
| 287 | |
| 288 | $result = $factory(); |
| 289 | |
| 290 | // Bound how much decoded font data (e.g. decompressed embedded |
| 291 | // TrueType programs) this cache may retain for the document's whole |
| 292 | // lifetime - a PDF with more/larger distinct fonts than the budget |
| 293 | // still works correctly, it just stops benefiting from caching once |
| 294 | // exhausted, rather than retaining every font's data forever (a |
| 295 | // 206KB PDF with 20 fonts each decompressing to 10MB was confirmed |
| 296 | // to otherwise inflate peak memory ~8x during Phase D's final |
| 297 | // review). |
| 298 | $size = strlen(serialize($result)); |
| 299 | |
| 300 | if (($this->fontInfoCacheBytes + $size) <= self::MAX_FONT_INFO_CACHE_BYTES) { |
| 301 | $this->fontInfoCache[$key] = $result; |
| 302 | $this->fontInfoCacheBytes += $size; |
| 303 | } else { |
| 304 | // Even once the budget is exhausted, always keep the single |
| 305 | // MOST RECENTLY resolved result cached - consecutive runs |
| 306 | // overwhelmingly share the SAME font (Interpreter only |
| 307 | // re-resolves on Tf), so this collapses what would otherwise be |
| 308 | // a per-run recompute back down to a per-font-activation one, |
| 309 | // without giving up the overall memory ceiling. Without this, a |
| 310 | // single font whose resolved size alone exceeds the budget |
| 311 | // (e.g. one 70MB embedded TrueType program) would be |
| 312 | // re-decompressed on every single run referencing it - a worse |
| 313 | // CPU DoS than the memory regression this cache was added to |
| 314 | // fix (confirmed during Phase D's final re-review). |
| 315 | $this->fontInfoMruKey = $key; |
| 316 | $this->fontInfoMruValue = $result; |
| 317 | } |
| 318 | |
| 319 | return $result; |
| 320 | } |
| 321 | |
| 322 | /** |
| 323 | * Parse an object directly at a byte offset |
| 324 | * |
| 325 | * @param int $offset |
| 326 | * @throws Exception |
| 327 | * @return mixed |
| 328 | */ |
| 329 | protected function parseAt(int $offset): mixed |
| 330 | { |
| 331 | $tokenizer = new Tokenizer($this->data, $offset); |
| 332 | $tokenizer->next(); // object number |
| 333 | $tokenizer->next(); // generation number |
| 334 | $objToken = $tokenizer->next(); |
| 335 | |
| 336 | if (($objToken['type'] !== 'keyword') || ($objToken['value'] !== 'obj')) { |
| 337 | throw new Exception('Error: Expected obj keyword while resolving a PDF object.'); |
| 338 | } |
| 339 | |
| 340 | $parser = new ObjectParser($tokenizer); |
| 341 | |
| 342 | return $parser->parseValue(); |
| 343 | } |
| 344 | |
| 345 | /** |
| 346 | * Get an object at an index within an object stream, expanding and caching the stream if needed |
| 347 | * |
| 348 | * @param int $streamObjNum |
| 349 | * @param int $index |
| 350 | * @throws Exception |
| 351 | * @return mixed |
| 352 | */ |
| 353 | protected function getFromObjectStream(int $streamObjNum, int $index): mixed |
| 354 | { |
| 355 | if (!isset($this->objectStreamCache[$streamObjNum])) { |
| 356 | if (isset($this->expandingStreams[$streamObjNum])) { |
| 357 | throw new Exception( |
| 358 | "Error: Circular object stream reference detected while expanding object {$streamObjNum}." |
| 359 | ); |
| 360 | } |
| 361 | |
| 362 | $this->expandingStreams[$streamObjNum] = true; |
| 363 | |
| 364 | try { |
| 365 | $streamObj = $this->getObject($streamObjNum); |
| 366 | |
| 367 | if (!($streamObj instanceof Value\Stream)) { |
| 368 | throw new Exception("Error: Object {$streamObjNum} is not a valid object stream."); |
| 369 | } |
| 370 | |
| 371 | $this->objectStreamCache[$streamObjNum] = ObjectStream::expand($streamObj, $this->decodeBudget); |
| 372 | $this->objectStreamIndexCache[$streamObjNum] = array_values($this->objectStreamCache[$streamObjNum]); |
| 373 | } finally { |
| 374 | unset($this->expandingStreams[$streamObjNum]); |
| 375 | } |
| 376 | } |
| 377 | |
| 378 | return $this->objectStreamIndexCache[$streamObjNum][$index] ?? null; |
| 379 | } |
| 380 | |
| 381 | /** |
| 382 | * Load offsets/trailer via xref, falling back to brute-force repair if unusable |
| 383 | * |
| 384 | * @throws Exception |
| 385 | * @return void |
| 386 | */ |
| 387 | protected function load(): void |
| 388 | { |
| 389 | try { |
| 390 | [$offsets, $trailer] = $this->loadViaXref(); |
| 391 | } catch (\Throwable $e) { |
| 392 | // Any lower-layer failure - not just this namespace's own |
| 393 | // Extract\Exception, but raw PHP errors from malformed data |
| 394 | // (e.g. a TypeError from a corrupt xref stream's /W array) - |
| 395 | // must trigger the repair fallback rather than leak out. |
| 396 | $offsets = []; |
| 397 | $trailer = []; |
| 398 | } |
| 399 | |
| 400 | if (!$this->isUsable($offsets, $trailer)) { |
| 401 | [$offsets, $trailer, $preResolved] = $this->loadViaRepair(); |
| 402 | $this->cache = $preResolved + $this->cache; |
| 403 | } |
| 404 | |
| 405 | if (isset($trailer['Encrypt'])) { |
| 406 | throw new Exception('Error: Encrypted PDFs are not currently supported for text extraction.'); |
| 407 | } |
| 408 | |
| 409 | $this->offsets = $offsets; |
| 410 | $this->trailer = $trailer; |
| 411 | } |
| 412 | |
| 413 | /** |
| 414 | * Load offsets/trailer by following the startxref chain (classic tables and/or xref streams) |
| 415 | * |
| 416 | * @throws Exception |
| 417 | * @return array |
| 418 | */ |
| 419 | protected function loadViaXref(): array |
| 420 | { |
| 421 | $startXrefPos = strrpos($this->data, 'startxref'); |
| 422 | if ($startXrefPos === false) { |
| 423 | throw new Exception('Error: No startxref marker found.'); |
| 424 | } |
| 425 | |
| 426 | $tokenizer = new Tokenizer($this->data, $startXrefPos + strlen('startxref')); |
| 427 | $posToken = $tokenizer->next(); |
| 428 | |
| 429 | if ($posToken['type'] !== 'number') { |
| 430 | throw new Exception('Error: Malformed startxref value.'); |
| 431 | } |
| 432 | |
| 433 | $offsets = []; |
| 434 | $trailer = []; |
| 435 | $visited = []; |
| 436 | $xrefPos = (int) $posToken['value']; |
| 437 | |
| 438 | while (($xrefPos !== null) && (!isset($visited[$xrefPos]))) { |
| 439 | $visited[$xrefPos] = true; |
| 440 | |
| 441 | $section = $this->isClassicXref($xrefPos) |
| 442 | ? Xref\Table::parse($this->data, $xrefPos) |
| 443 | : Xref\Stream::parse($this->data, $xrefPos, $this->decodeBudget); |
| 444 | |
| 445 | $this->mergeXrefSection($section, $offsets, $trailer); |
| 446 | |
| 447 | // A hybrid-reference file's classic xref table may point to a |
| 448 | // supplemental cross-reference stream (for compressed objects |
| 449 | // the classic table can't express) via /XRefStm, alongside a |
| 450 | // /Prev continuing the classic chain - both must be merged, |
| 451 | // per PDF spec 7.5.8.4, not treated as mutually exclusive. |
| 452 | if (isset($section['trailer']['XRefStm'])) { |
| 453 | $xrefStmPos = (int) $section['trailer']['XRefStm']; |
| 454 | if (!isset($visited[$xrefStmPos])) { |
| 455 | $visited[$xrefStmPos] = true; |
| 456 | $xrefStmSection = Xref\Stream::parse($this->data, $xrefStmPos, $this->decodeBudget); |
| 457 | $this->mergeXrefSection($xrefStmSection, $offsets, $trailer); |
| 458 | } |
| 459 | } |
| 460 | |
| 461 | $xrefPos = isset($section['trailer']['Prev']) ? (int) $section['trailer']['Prev'] : null; |
| 462 | } |
| 463 | |
| 464 | return [$offsets, $trailer]; |
| 465 | } |
| 466 | |
| 467 | /** |
| 468 | * Merge one xref section's offsets/trailer into the accumulated result |
| 469 | * |
| 470 | * @param array $section |
| 471 | * @param array $offsets |
| 472 | * @param array $trailer |
| 473 | * @return void |
| 474 | */ |
| 475 | protected function mergeXrefSection(array $section, array &$offsets, array &$trailer): void |
| 476 | { |
| 477 | foreach ($section['offsets'] as $objNum => $location) { |
| 478 | if (!isset($offsets[$objNum])) { |
| 479 | $offsets[$objNum] = $location; |
| 480 | } |
| 481 | } |
| 482 | |
| 483 | $trailer = $trailer + $section['trailer']; |
| 484 | } |
| 485 | |
| 486 | /** |
| 487 | * Determine if the xref section at a position is a classic table (vs. an xref stream) |
| 488 | * |
| 489 | * @param int $pos |
| 490 | * @return bool |
| 491 | */ |
| 492 | protected function isClassicXref(int $pos): bool |
| 493 | { |
| 494 | $tokenizer = new Tokenizer($this->data, $pos); |
| 495 | $token = $tokenizer->next(); |
| 496 | |
| 497 | return ($token['type'] === 'keyword') && ($token['value'] === 'xref'); |
| 498 | } |
| 499 | |
| 500 | /** |
| 501 | * Load offsets/trailer via brute-force repair scan |
| 502 | * |
| 503 | * @return array |
| 504 | */ |
| 505 | protected function loadViaRepair(): array |
| 506 | { |
| 507 | $result = Repair::scan($this->data); |
| 508 | $trailer = $result['trailer']; |
| 509 | |
| 510 | if (!isset($trailer['Root'])) { |
| 511 | $trailer['Root'] = $this->findCatalogReference($result['offsets']); |
| 512 | } |
| 513 | |
| 514 | $preResolved = $this->expandObjectStreamsFromRepair($result['offsets']); |
| 515 | |
| 516 | return [$result['offsets'], $trailer, $preResolved]; |
| 517 | } |
| 518 | |
| 519 | /** |
| 520 | * A brute-force repair scan only finds objects that appear as literal |
| 521 | * "N G obj ... endobj" text - objects packed inside a /Type /ObjStm |
| 522 | * container's stream body don't match that pattern at all, since |
| 523 | * they're serialized inline within the ObjStm's own stream data rather |
| 524 | * than as their own "obj" markers. Without this second pass, any object |
| 525 | * that only exists inside an object stream would be silently |
| 526 | * unrecoverable after repair fires. |
| 527 | * |
| 528 | * These recovered objects don't have a byte offset the way normal |
| 529 | * repair-scanned objects do - they're already fully parsed values, not |
| 530 | * "here's where to find it" locations - so instead of trying to fit |
| 531 | * them into the offsets/inStream shape, this returns a ready-to-use |
| 532 | * [objNum => value] map that the caller seeds directly into the object |
| 533 | * cache (getObject() already checks the cache before consulting |
| 534 | * offsets). |
| 535 | * |
| 536 | * @param array $offsets |
| 537 | * @return array |
| 538 | */ |
| 539 | protected function expandObjectStreamsFromRepair(array $offsets): array |
| 540 | { |
| 541 | $preResolved = []; |
| 542 | |
| 543 | foreach ($offsets as $location) { |
| 544 | if (!isset($location['offset'])) { |
| 545 | continue; |
| 546 | } |
| 547 | |
| 548 | try { |
| 549 | $value = $this->parseAt($location['offset']); |
| 550 | } catch (\Throwable $e) { |
| 551 | continue; |
| 552 | } |
| 553 | |
| 554 | if (!($value instanceof Value\Stream)) { |
| 555 | continue; |
| 556 | } |
| 557 | |
| 558 | $type = $value->dict['Type'] ?? null; |
| 559 | if (!($type instanceof Value\Name) || ($type->name !== 'ObjStm')) { |
| 560 | continue; |
| 561 | } |
| 562 | |
| 563 | try { |
| 564 | $expanded = ObjectStream::expand($value, $this->decodeBudget); |
| 565 | } catch (\Throwable $e) { |
| 566 | continue; |
| 567 | } |
| 568 | |
| 569 | foreach ($expanded as $objNum => $objValue) { |
| 570 | if (!isset($preResolved[$objNum])) { |
| 571 | $preResolved[$objNum] = $objValue; |
| 572 | } |
| 573 | } |
| 574 | } |
| 575 | |
| 576 | return $preResolved; |
| 577 | } |
| 578 | |
| 579 | /** |
| 580 | * Scan repair-recovered offsets for an object that looks like the document catalog |
| 581 | * |
| 582 | * @param array $offsets |
| 583 | * @return ?Value\Reference |
| 584 | */ |
| 585 | protected function findCatalogReference(array $offsets): ?Value\Reference |
| 586 | { |
| 587 | foreach ($offsets as $objNum => $location) { |
| 588 | if (!isset($location['offset'])) { |
| 589 | continue; |
| 590 | } |
| 591 | |
| 592 | try { |
| 593 | $tokenizer = new Tokenizer($this->data, $location['offset']); |
| 594 | $tokenizer->next(); |
| 595 | $genToken = $tokenizer->next(); |
| 596 | $objToken = $tokenizer->next(); |
| 597 | |
| 598 | if (($objToken['type'] === 'keyword') && ($objToken['value'] === 'obj')) { |
| 599 | $parser = new ObjectParser($tokenizer); |
| 600 | $value = $parser->parseValue(); |
| 601 | |
| 602 | if (is_array($value) && isset($value['Type']) && |
| 603 | ($value['Type'] instanceof Value\Name) && ($value['Type']->name === 'Catalog')) { |
| 604 | return new Value\Reference($objNum, (int) $genToken['value']); |
| 605 | } |
| 606 | } |
| 607 | } catch (Exception $e) { |
| 608 | continue; |
| 609 | } |
| 610 | } |
| 611 | |
| 612 | return null; |
| 613 | } |
| 614 | |
| 615 | /** |
| 616 | * Determine if xref-derived offsets/trailer look usable (vs. needing repair) |
| 617 | * |
| 618 | * @param array $offsets |
| 619 | * @param array $trailer |
| 620 | * @return bool |
| 621 | */ |
| 622 | protected function isUsable(array $offsets, array $trailer): bool |
| 623 | { |
| 624 | if (empty($offsets) || !isset($trailer['Root'])) { |
| 625 | return false; |
| 626 | } |
| 627 | |
| 628 | $root = $trailer['Root']; |
| 629 | if (!($root instanceof Value\Reference) || !isset($offsets[$root->objNum])) { |
| 630 | return false; |
| 631 | } |
| 632 | |
| 633 | $sample = array_slice($offsets, 0, 5, true); |
| 634 | foreach ($sample as $location) { |
| 635 | if (isset($location['offset']) && !$this->looksLikeObjectAt($location['offset'])) { |
| 636 | return false; |
| 637 | } |
| 638 | } |
| 639 | |
| 640 | return true; |
| 641 | } |
| 642 | |
| 643 | /** |
| 644 | * Determine if a byte offset looks like the start of an "N G obj" object |
| 645 | * |
| 646 | * @param int $offset |
| 647 | * @return bool |
| 648 | */ |
| 649 | protected function looksLikeObjectAt(int $offset): bool |
| 650 | { |
| 651 | if (($offset < 0) || ($offset >= strlen($this->data))) { |
| 652 | return false; |
| 653 | } |
| 654 | |
| 655 | $chunk = substr($this->data, $offset, 32); |
| 656 | |
| 657 | return (bool) preg_match('/^\s*\d+\s+\d+\s+obj\b/', $chunk); |
| 658 | } |
| 659 | |
| 660 | } |