| 1 |
<?php |
| 2 |
|
| 3 |
namespace ImportWP\Common\Importer\File; |
| 4 |
|
| 5 |
use ImportWP\Common\Importer\FileInterface; |
| 6 |
|
| 7 |
class JSONFile extends AbstractIndexedFile implements FileInterface |
| 8 |
{ |
| 9 |
/** |
| 10 |
* Path segments to the array that holds records. |
| 11 |
* Empty means the root value is an array of records. |
| 12 |
* |
| 13 |
* @var string[] |
| 14 |
*/ |
| 15 |
private $base_path_segments = []; |
| 16 |
|
| 17 |
/** |
| 18 |
* Cache key fragment derived from the record path. |
| 19 |
* |
| 20 |
* @var string |
| 21 |
*/ |
| 22 |
private $record_path_cache_key = 'root'; |
| 23 |
|
| 24 |
/** |
| 25 |
* Discovered record paths (path => sample count). |
| 26 |
* |
| 27 |
* @var array |
| 28 |
*/ |
| 29 |
private $path_list = []; |
| 30 |
|
| 31 |
private $chunk = ''; |
| 32 |
private $chunk_size = 8192; |
| 33 |
private $chunk_offset = 0; |
| 34 |
|
| 35 |
private $record_counter = 0; |
| 36 |
private $record_start_index = 0; |
| 37 |
private $record_offset = 0; |
| 38 |
|
| 39 |
// Stream parser state — must persist across chunk reads |
| 40 |
private $parse_segment_index = 0; |
| 41 |
private $parse_awaiting_value = false; |
| 42 |
private $parse_depth = 0; |
| 43 |
private $parse_search_depth = 0; |
| 44 |
private $parse_path_matched = false; |
| 45 |
private $parse_tracking = false; |
| 46 |
private $parse_record_depth = 0; |
| 47 |
private $parse_skip_depth = 0; |
| 48 |
|
| 49 |
/** |
| 50 |
* Set base path for records. |
| 51 |
* |
| 52 |
* Examples: "data", "results/items", "/" or "" for a root array. |
| 53 |
* |
| 54 |
* @param string $path |
| 55 |
*/ |
| 56 |
public function setRecordPath($path = '/') |
| 57 |
{ |
| 58 |
if ($path === null) { |
| 59 |
$path = '/'; |
| 60 |
} |
| 61 |
|
| 62 |
$path = trim((string) $path); |
| 63 |
if (strpos($path, '/') === 0) { |
| 64 |
$path = substr($path, 1); |
| 65 |
} |
| 66 |
$path = rtrim($path, '/'); |
| 67 |
|
| 68 |
$segments = $path === '' ? [] : explode('/', $path); |
| 69 |
$segments = array_values(array_filter($segments, function ($value) { |
| 70 |
return $value !== ''; |
| 71 |
})); |
| 72 |
|
| 73 |
$new_key = empty($segments) ? 'root' : implode('_', $segments); |
| 74 |
if ($new_key !== $this->record_path_cache_key) { |
| 75 |
$this->resetIndexState(); |
| 76 |
} |
| 77 |
|
| 78 |
$this->base_path_segments = $segments; |
| 79 |
$this->record_path_cache_key = $new_key; |
| 80 |
} |
| 81 |
|
| 82 |
public function getFileIndexKey() |
| 83 |
{ |
| 84 |
return sprintf('file_index-%s', $this->record_path_cache_key); |
| 85 |
} |
| 86 |
|
| 87 |
/** |
| 88 |
* Discover candidate record paths (arrays of objects). |
| 89 |
* |
| 90 |
* @return array path => count |
| 91 |
*/ |
| 92 |
public function get_path_list() |
| 93 |
{ |
| 94 |
$this->path_list = $this->config ? $this->config->get('json_path_list') : null; |
| 95 |
if (!is_array($this->path_list)) { |
| 96 |
$this->path_list = []; |
| 97 |
} |
| 98 |
|
| 99 |
if (!empty($this->path_list)) { |
| 100 |
return $this->path_list; |
| 101 |
} |
| 102 |
|
| 103 |
$decoded = $this->decodeFileSample(); |
| 104 |
if (!is_array($decoded)) { |
| 105 |
return []; |
| 106 |
} |
| 107 |
|
| 108 |
$this->path_list = []; |
| 109 |
$this->collectPaths($decoded, '', $this->path_list); |
| 110 |
if ($this->config) { |
| 111 |
$this->config->set('json_path_list', $this->path_list); |
| 112 |
} |
| 113 |
|
| 114 |
return $this->path_list; |
| 115 |
} |
| 116 |
|
| 117 |
/** |
| 118 |
* Return records at the current record path by decoding JSON. |
| 119 |
* Used for preview so we are not dependent on a stream index. |
| 120 |
* |
| 121 |
* @param int $limit |
| 122 |
* @return array |
| 123 |
*/ |
| 124 |
public function getDecodedRecords($limit = 1) |
| 125 |
{ |
| 126 |
$decoded = $this->decodeFileSample(); |
| 127 |
if (!is_array($decoded)) { |
| 128 |
return []; |
| 129 |
} |
| 130 |
|
| 131 |
$records = $this->resolveRecordsAtPath($decoded); |
| 132 |
if (!is_array($records) || !$this->isList($records)) { |
| 133 |
return []; |
| 134 |
} |
| 135 |
|
| 136 |
if ($limit > 0) { |
| 137 |
return array_slice($records, 0, $limit); |
| 138 |
} |
| 139 |
|
| 140 |
return $records; |
| 141 |
} |
| 142 |
|
| 143 |
/** |
| 144 |
* Generate record file positions. |
| 145 |
*/ |
| 146 |
public function generateIndex() |
| 147 |
{ |
| 148 |
$this->record_counter = 0; |
| 149 |
$this->record_start_index = 0; |
| 150 |
$this->chunk = ''; |
| 151 |
$this->chunk_offset = 0; |
| 152 |
|
| 153 |
$this->parse_segment_index = 0; |
| 154 |
$this->parse_awaiting_value = false; |
| 155 |
$this->parse_depth = 0; |
| 156 |
$this->parse_search_depth = 0; |
| 157 |
$this->parse_path_matched = empty($this->base_path_segments); |
| 158 |
$this->parse_tracking = false; |
| 159 |
$this->parse_record_depth = 0; |
| 160 |
$this->parse_skip_depth = 0; |
| 161 |
|
| 162 |
rewind($this->getFileHandle()); |
| 163 |
while (!feof($this->getFileHandle())) { |
| 164 |
if ($this->is_processing && $this->chunk_offset > $this->process_max_size) { |
| 165 |
break; |
| 166 |
} |
| 167 |
|
| 168 |
$this->chunk .= $this->getChunk(); |
| 169 |
$this->processChunk(); |
| 170 |
} |
| 171 |
} |
| 172 |
|
| 173 |
/** |
| 174 |
* Read chunk from file. |
| 175 |
* |
| 176 |
* @return bool|string |
| 177 |
*/ |
| 178 |
public function getChunk() |
| 179 |
{ |
| 180 |
return fread($this->getFileHandle(), $this->chunk_size); |
| 181 |
} |
| 182 |
|
| 183 |
public function processChunk() |
| 184 |
{ |
| 185 |
$regex_parts = [ |
| 186 |
'"[^"\\\\]*(?:\\\\.[^"\\\\]*)*"', |
| 187 |
'([^,"\'{}\[\]:\s]+)', |
| 188 |
'({)', |
| 189 |
'(})', |
| 190 |
'(\[)', |
| 191 |
'(\])', |
| 192 |
'(:)', |
| 193 |
'(,)', |
| 194 |
]; |
| 195 |
|
| 196 |
$regex = '/' . implode('|', $regex_parts) . '/s'; |
| 197 |
|
| 198 |
while (preg_match($regex, $this->chunk, $matches, PREG_OFFSET_CAPTURE) !== 0) { |
| 199 |
list($captured, $offset) = $matches[0]; |
| 200 |
$this->record_offset = $offset; |
| 201 |
|
| 202 |
$is_string = strlen($captured) >= 2 && $captured[0] === '"'; |
| 203 |
$is_open_object = $captured === '{'; |
| 204 |
$is_close_object = $captured === '}'; |
| 205 |
$is_open_array = $captured === '['; |
| 206 |
$is_close_array = $captured === ']'; |
| 207 |
$is_colon = $captured === ':'; |
| 208 |
$is_structure = $is_open_object || $is_close_object || $is_open_array || $is_close_array; |
| 209 |
|
| 210 |
// Skipping an unmatched value subtree |
| 211 |
if ($this->parse_skip_depth > 0) { |
| 212 |
if ($is_open_object || $is_open_array) { |
| 213 |
$this->parse_skip_depth++; |
| 214 |
$this->parse_depth++; |
| 215 |
} elseif ($is_close_object || $is_close_array) { |
| 216 |
$this->parse_skip_depth--; |
| 217 |
$this->parse_depth--; |
| 218 |
} |
| 219 |
$this->consumeToken($offset, $captured); |
| 220 |
continue; |
| 221 |
} |
| 222 |
|
| 223 |
if ($this->parse_tracking) { |
| 224 |
if ($is_open_object || $is_open_array) { |
| 225 |
$this->parse_record_depth++; |
| 226 |
$this->parse_depth++; |
| 227 |
} elseif ($is_close_object || $is_close_array) { |
| 228 |
$this->parse_record_depth--; |
| 229 |
$this->parse_depth--; |
| 230 |
if ($this->parse_record_depth === 0 && $is_close_object) { |
| 231 |
$this->setIndex( |
| 232 |
$this->record_counter, |
| 233 |
$this->record_start_index, |
| 234 |
$this->chunk_offset + $this->record_offset + strlen($captured) |
| 235 |
); |
| 236 |
$this->record_counter++; |
| 237 |
$this->parse_tracking = false; |
| 238 |
} elseif ($this->parse_record_depth < 0) { |
| 239 |
$this->parse_tracking = false; |
| 240 |
$this->parse_path_matched = false; |
| 241 |
$this->parse_awaiting_value = false; |
| 242 |
$this->parse_segment_index = 0; |
| 243 |
$this->parse_search_depth = 0; |
| 244 |
} |
| 245 |
} |
| 246 |
$this->consumeToken($offset, $captured); |
| 247 |
continue; |
| 248 |
} |
| 249 |
|
| 250 |
if ($this->parse_path_matched) { |
| 251 |
if ($is_open_object) { |
| 252 |
$this->parse_tracking = true; |
| 253 |
$this->parse_record_depth = 1; |
| 254 |
$this->parse_depth++; |
| 255 |
$this->record_start_index = $this->chunk_offset + $this->record_offset; |
| 256 |
} elseif ($is_close_array || $is_close_object) { |
| 257 |
$this->parse_path_matched = false; |
| 258 |
$this->parse_awaiting_value = false; |
| 259 |
$this->parse_segment_index = 0; |
| 260 |
$this->parse_search_depth = 0; |
| 261 |
$this->parse_depth--; |
| 262 |
} elseif ($is_open_array) { |
| 263 |
$this->parse_depth++; |
| 264 |
} |
| 265 |
$this->consumeToken($offset, $captured); |
| 266 |
continue; |
| 267 |
} |
| 268 |
|
| 269 |
if ($this->parse_awaiting_value) { |
| 270 |
if ($is_colon) { |
| 271 |
$this->consumeToken($offset, $captured); |
| 272 |
continue; |
| 273 |
} |
| 274 |
|
| 275 |
$is_last_segment = $this->parse_segment_index >= count($this->base_path_segments); |
| 276 |
|
| 277 |
if ($is_last_segment && $is_open_array) { |
| 278 |
$this->parse_path_matched = true; |
| 279 |
$this->parse_awaiting_value = false; |
| 280 |
$this->parse_depth++; |
| 281 |
} elseif (!$is_last_segment && $is_open_object) { |
| 282 |
$this->parse_awaiting_value = false; |
| 283 |
$this->parse_depth++; |
| 284 |
$this->parse_search_depth = $this->parse_depth; |
| 285 |
} elseif ($is_open_object || $is_open_array) { |
| 286 |
$this->parse_awaiting_value = false; |
| 287 |
$this->parse_segment_index = 0; |
| 288 |
$this->parse_search_depth = 0; |
| 289 |
$this->parse_skip_depth = 1; |
| 290 |
$this->parse_depth++; |
| 291 |
} else { |
| 292 |
$this->parse_awaiting_value = false; |
| 293 |
$this->parse_segment_index = 0; |
| 294 |
$this->parse_search_depth = 0; |
| 295 |
} |
| 296 |
|
| 297 |
$this->consumeToken($offset, $captured); |
| 298 |
continue; |
| 299 |
} |
| 300 |
|
| 301 |
if (empty($this->base_path_segments)) { |
| 302 |
if ($is_open_array && $this->parse_depth === 0) { |
| 303 |
$this->parse_path_matched = true; |
| 304 |
$this->parse_depth++; |
| 305 |
} elseif ($is_structure) { |
| 306 |
if ($is_open_object || $is_open_array) { |
| 307 |
$this->parse_depth++; |
| 308 |
} else { |
| 309 |
$this->parse_depth--; |
| 310 |
} |
| 311 |
} |
| 312 |
$this->consumeToken($offset, $captured); |
| 313 |
continue; |
| 314 |
} |
| 315 |
|
| 316 |
// Enter root object so we can match top-level path keys |
| 317 |
if ($is_open_object && $this->parse_depth === 0 && $this->parse_segment_index === 0 && !$this->parse_awaiting_value) { |
| 318 |
$this->parse_depth++; |
| 319 |
$this->parse_search_depth = 1; |
| 320 |
$this->consumeToken($offset, $captured); |
| 321 |
continue; |
| 322 |
} |
| 323 |
|
| 324 |
// Looking for the next path key at search_depth |
| 325 |
if ($is_string && $this->parse_depth === $this->parse_search_depth) { |
| 326 |
$key = $this->unquote($captured); |
| 327 |
if (isset($this->base_path_segments[$this->parse_segment_index]) && $key === $this->base_path_segments[$this->parse_segment_index]) { |
| 328 |
$this->parse_segment_index++; |
| 329 |
$this->parse_awaiting_value = true; |
| 330 |
} |
| 331 |
} elseif ($is_structure) { |
| 332 |
if ($is_open_object || $is_open_array) { |
| 333 |
$this->parse_depth++; |
| 334 |
} else { |
| 335 |
$this->parse_depth--; |
| 336 |
if ($this->parse_depth < $this->parse_search_depth) { |
| 337 |
$this->parse_segment_index = min($this->parse_segment_index, $this->segmentsReachedAtDepth($this->parse_depth)); |
| 338 |
$this->parse_search_depth = max(0, $this->parse_depth); |
| 339 |
} |
| 340 |
} |
| 341 |
} |
| 342 |
|
| 343 |
$this->consumeToken($offset, $captured); |
| 344 |
} |
| 345 |
} |
| 346 |
|
| 347 |
/** |
| 348 |
* @param int $depth |
| 349 |
* @return int |
| 350 |
*/ |
| 351 |
private function segmentsReachedAtDepth($depth) |
| 352 |
{ |
| 353 |
return max(0, min($depth, count($this->base_path_segments))); |
| 354 |
} |
| 355 |
|
| 356 |
/** |
| 357 |
* @param int $offset |
| 358 |
* @param string $captured |
| 359 |
*/ |
| 360 |
private function consumeToken($offset, $captured) |
| 361 |
{ |
| 362 |
$string_offset = $offset + strlen($captured); |
| 363 |
$this->chunk_offset += $string_offset; |
| 364 |
$this->chunk = substr($this->chunk, $string_offset); |
| 365 |
} |
| 366 |
|
| 367 |
/** |
| 368 |
* @param string $token |
| 369 |
* @return string |
| 370 |
*/ |
| 371 |
private function unquote($token) |
| 372 |
{ |
| 373 |
if (strlen($token) >= 2 && $token[0] === '"' && substr($token, -1) === '"') { |
| 374 |
return stripcslashes(substr($token, 1, -1)); |
| 375 |
} |
| 376 |
|
| 377 |
return $token; |
| 378 |
} |
| 379 |
|
| 380 |
/** |
| 381 |
* @return array|null |
| 382 |
*/ |
| 383 |
private function decodeFileSample() |
| 384 |
{ |
| 385 |
$handle = $this->getFileHandle(); |
| 386 |
$current = ftell($handle); |
| 387 |
rewind($handle); |
| 388 |
|
| 389 |
$sample = ''; |
| 390 |
$max = max($this->process_max_size, 1000000); |
| 391 |
while (!feof($handle) && strlen($sample) < $max) { |
| 392 |
$sample .= fread($handle, $this->chunk_size); |
| 393 |
} |
| 394 |
|
| 395 |
$decoded = json_decode($sample, true); |
| 396 |
if (json_last_error() !== JSON_ERROR_NONE) { |
| 397 |
$decoded = $this->decodePartialSample($sample); |
| 398 |
} |
| 399 |
|
| 400 |
fseek($handle, $current); |
| 401 |
|
| 402 |
return is_array($decoded) ? $decoded : null; |
| 403 |
} |
| 404 |
|
| 405 |
/** |
| 406 |
* Walk to the record array for the current base path. |
| 407 |
* |
| 408 |
* @param array $decoded |
| 409 |
* @return array|null |
| 410 |
*/ |
| 411 |
private function resolveRecordsAtPath(array $decoded) |
| 412 |
{ |
| 413 |
if (empty($this->base_path_segments)) { |
| 414 |
return $decoded; |
| 415 |
} |
| 416 |
|
| 417 |
$current = $decoded; |
| 418 |
foreach ($this->base_path_segments as $segment) { |
| 419 |
if (!is_array($current) || !array_key_exists($segment, $current)) { |
| 420 |
return null; |
| 421 |
} |
| 422 |
$current = $current[$segment]; |
| 423 |
} |
| 424 |
|
| 425 |
return $current; |
| 426 |
} |
| 427 |
|
| 428 |
/** |
| 429 |
* Attempt to decode a truncated JSON sample by closing open braces/brackets. |
| 430 |
* |
| 431 |
* @param string $sample |
| 432 |
* @return array|null |
| 433 |
*/ |
| 434 |
private function decodePartialSample($sample) |
| 435 |
{ |
| 436 |
$in_string = false; |
| 437 |
$escape = false; |
| 438 |
$stack = []; |
| 439 |
|
| 440 |
$len = strlen($sample); |
| 441 |
for ($i = 0; $i < $len; $i++) { |
| 442 |
$ch = $sample[$i]; |
| 443 |
if ($in_string) { |
| 444 |
if ($escape) { |
| 445 |
$escape = false; |
| 446 |
} elseif ($ch === '\\') { |
| 447 |
$escape = true; |
| 448 |
} elseif ($ch === '"') { |
| 449 |
$in_string = false; |
| 450 |
} |
| 451 |
continue; |
| 452 |
} |
| 453 |
|
| 454 |
if ($ch === '"') { |
| 455 |
$in_string = true; |
| 456 |
} elseif ($ch === '{' || $ch === '[') { |
| 457 |
$stack[] = $ch; |
| 458 |
} elseif ($ch === '}' || $ch === ']') { |
| 459 |
array_pop($stack); |
| 460 |
} |
| 461 |
} |
| 462 |
|
| 463 |
if ($in_string) { |
| 464 |
$sample .= '"'; |
| 465 |
} |
| 466 |
|
| 467 |
$sample = rtrim($sample); |
| 468 |
$sample = preg_replace('/[,:]\s*$/', '', $sample); |
| 469 |
|
| 470 |
while (!empty($stack)) { |
| 471 |
$open = array_pop($stack); |
| 472 |
$sample .= $open === '{' ? '}' : ']'; |
| 473 |
} |
| 474 |
|
| 475 |
$decoded = json_decode($sample, true); |
| 476 |
return json_last_error() === JSON_ERROR_NONE ? $decoded : null; |
| 477 |
} |
| 478 |
|
| 479 |
/** |
| 480 |
* @param mixed $data |
| 481 |
* @param string $prefix |
| 482 |
* @param array $paths |
| 483 |
*/ |
| 484 |
private function collectPaths($data, $prefix, &$paths) |
| 485 |
{ |
| 486 |
if (!is_array($data)) { |
| 487 |
return; |
| 488 |
} |
| 489 |
|
| 490 |
if ($this->isList($data)) { |
| 491 |
if ($this->isListOfObjects($data)) { |
| 492 |
$key = $prefix === '' ? '/' : $prefix; |
| 493 |
$paths[$key] = count($data); |
| 494 |
} |
| 495 |
return; |
| 496 |
} |
| 497 |
|
| 498 |
foreach ($data as $key => $value) { |
| 499 |
if (!is_array($value)) { |
| 500 |
continue; |
| 501 |
} |
| 502 |
$path = $prefix === '' ? (string) $key : $prefix . '/' . $key; |
| 503 |
$this->collectPaths($value, $path, $paths); |
| 504 |
} |
| 505 |
} |
| 506 |
|
| 507 |
/** |
| 508 |
* @param array $data |
| 509 |
* @return bool |
| 510 |
*/ |
| 511 |
private function isList(array $data) |
| 512 |
{ |
| 513 |
if ($data === []) { |
| 514 |
return true; |
| 515 |
} |
| 516 |
|
| 517 |
return array_keys($data) === range(0, count($data) - 1); |
| 518 |
} |
| 519 |
|
| 520 |
/** |
| 521 |
* @param array $data |
| 522 |
* @return bool |
| 523 |
*/ |
| 524 |
private function isListOfObjects(array $data) |
| 525 |
{ |
| 526 |
if (empty($data) || !$this->isList($data)) { |
| 527 |
return false; |
| 528 |
} |
| 529 |
|
| 530 |
foreach ($data as $item) { |
| 531 |
if (!is_array($item) || $this->isList($item)) { |
| 532 |
return false; |
| 533 |
} |
| 534 |
} |
| 535 |
|
| 536 |
return true; |
| 537 |
} |
| 538 |
} |
| 539 |
|
