| 1 |
<?php |
| 2 |
|
| 3 |
namespace ImportWP\Common\Importer\File; |
| 4 |
|
| 5 |
use ImportWP\Common\Importer\FileInterface; |
| 6 |
|
| 7 |
class XMLFile extends AbstractIndexedFile implements FileInterface |
| 8 |
{ |
| 9 |
|
| 10 |
private $base_path; |
| 11 |
private $base_path_segments = array(); |
| 12 |
private $record_offset = 0; |
| 13 |
private $chunk_size = 8192; |
| 14 |
private $depth = 0; |
| 15 |
private $open_nodes = array(); |
| 16 |
private $node_list = array(); |
| 17 |
private $chunk; |
| 18 |
private $chunk_offset = 0; |
| 19 |
|
| 20 |
private $record_counter = 0; |
| 21 |
private $record_start_index = 0; |
| 22 |
private $record_opened = false; |
| 23 |
private $last_record_incomplete = false; |
| 24 |
private $record_patch_cache_key = ''; |
| 25 |
|
| 26 |
/** |
| 27 |
* Set base path for records |
| 28 |
* |
| 29 |
* @param string $base_path |
| 30 |
*/ |
| 31 |
public function setRecordPath($base_path = '/') |
| 32 |
{ |
| 33 |
// strip slash from start |
| 34 |
if (strpos($base_path, '/') === 0) { |
| 35 |
$base_path = substr($base_path, 1); |
| 36 |
} |
| 37 |
|
| 38 |
// split basepath in segments |
| 39 |
$this->base_path_segments = explode('/', $base_path); |
| 40 |
array_filter($this->base_path_segments, function ($value) { |
| 41 |
return $value !== ""; |
| 42 |
}); |
| 43 |
|
| 44 |
if (!empty($this->base_path_segments)) { |
| 45 |
$this->record_patch_cache_key = implode('_', $this->base_path_segments); |
| 46 |
$base_path = array_pop($this->base_path_segments); |
| 47 |
} |
| 48 |
|
| 49 |
$this->base_path = $base_path; |
| 50 |
} |
| 51 |
|
| 52 |
/** |
| 53 |
* Get record |
| 54 |
* |
| 55 |
* @param int $index |
| 56 |
* |
| 57 |
* @return string |
| 58 |
*/ |
| 59 |
public function getRecord($index = 0) |
| 60 |
{ |
| 61 |
$record = parent::getRecord($index); |
| 62 |
|
| 63 |
if ($index === $this->getRecordCount() - 1 && true === $this->last_record_incomplete) { |
| 64 |
$record = $this->fixRecord($record); |
| 65 |
} |
| 66 |
|
| 67 |
return $this->wrapWithRecordTag($record); |
| 68 |
} |
| 69 |
|
| 70 |
/** |
| 71 |
* Fix a broken xml segment |
| 72 |
* |
| 73 |
* @param $record |
| 74 |
* |
| 75 |
* @return bool|string |
| 76 |
*/ |
| 77 |
private function fixRecord($record) |
| 78 |
{ |
| 79 |
$open = "/<[^\/][^>]+[^\/]>/"; |
| 80 |
$close = "/<\/[^>]+[^\/]>/"; |
| 81 |
$depth = 0; |
| 82 |
$started = false; |
| 83 |
$open_nodes = array(); |
| 84 |
|
| 85 |
$chunk = $record; |
| 86 |
|
| 87 |
while (preg_match( |
| 88 |
"/<[^?|!\-\-][^>]*>/", |
| 89 |
$chunk, |
| 90 |
$matches, |
| 91 |
PREG_OFFSET_CAPTURE |
| 92 |
) !== 0) { |
| 93 |
if (isset($matches[0]) && isset($matches[0][0])) { |
| 94 |
list($captured, $offset) = $matches[0]; |
| 95 |
|
| 96 |
if (preg_match($open, $captured) === 1) { |
| 97 |
$started = true; |
| 98 |
$depth++; |
| 99 |
$node_name = $this->getNodeName($captured); |
| 100 |
$open_nodes[] = $node_name; |
| 101 |
} else if (preg_match($close, $captured) === 1) { |
| 102 |
$depth--; |
| 103 |
array_pop($open_nodes); |
| 104 |
} |
| 105 |
|
| 106 |
$string_offset = $offset + strlen($captured); |
| 107 |
$chunk = substr($chunk, $string_offset); |
| 108 |
} |
| 109 |
|
| 110 |
if ($started && 0 === $depth) { |
| 111 |
// We should be at the end of the xml |
| 112 |
return substr($record, 0, strlen($chunk)); |
| 113 |
} |
| 114 |
} |
| 115 |
|
| 116 |
if (!empty($open_nodes)) { |
| 117 |
// Add closing tags for open nodes |
| 118 |
foreach ($open_nodes as $node) { |
| 119 |
$record .= "</{$node}>"; |
| 120 |
} |
| 121 |
} |
| 122 |
|
| 123 |
return $record; |
| 124 |
} |
| 125 |
|
| 126 |
/** |
| 127 |
* Get node name from xml tag |
| 128 |
* |
| 129 |
* @param $node |
| 130 |
* |
| 131 |
* @return bool|string |
| 132 |
*/ |
| 133 |
private function getNodeName($node) |
| 134 |
{ |
| 135 |
|
| 136 |
preg_match("/<\/?([a-zA-Z0-9\-_\.:]+)[ ]*[^>]*>/", $node, $matches); |
| 137 |
if (isset($matches[0]) && isset($matches[1])) { |
| 138 |
$name = $matches[1]; |
| 139 |
} else { |
| 140 |
$name = substr($node, 1, strlen($node) - 2); |
| 141 |
} |
| 142 |
|
| 143 |
return $name; |
| 144 |
} |
| 145 |
|
| 146 |
/** |
| 147 |
* Generate file position index. |
| 148 |
* |
| 149 |
* Read through entire xml file generating record start and end positions |
| 150 |
* |
| 151 |
* @return mixed |
| 152 |
*/ |
| 153 |
protected function generateIndex() |
| 154 |
{ |
| 155 |
$this->record_counter = 0; |
| 156 |
$this->record_start_index = 0; |
| 157 |
|
| 158 |
rewind($this->getFileHandle()); |
| 159 |
while (!feof($this->getFileHandle())) { |
| 160 |
|
| 161 |
$this->chunk .= $this->getChunk(); |
| 162 |
$this->readChunkXmlNodes(); |
| 163 |
|
| 164 |
// only read the first 1mb of file |
| 165 |
if ($this->is_processing && ($this->record_counter > 0 || ftell($this->getFileHandle()) > $this->process_max_size)) { |
| 166 |
break; |
| 167 |
} |
| 168 |
} |
| 169 |
|
| 170 |
// if file is incomplete do we try and fix the record, or skip it? |
| 171 |
if (!$this->is_processing && true === $this->record_opened) { |
| 172 |
$this->record_opened = false; |
| 173 |
$this->last_record_incomplete = true; |
| 174 |
$this->setIndex( |
| 175 |
$this->record_counter, |
| 176 |
$this->record_start_index, |
| 177 |
$this->chunk_offset |
| 178 |
); |
| 179 |
} |
| 180 |
} |
| 181 |
|
| 182 |
/** |
| 183 |
* Read chunk from file |
| 184 |
* |
| 185 |
* @return bool|string |
| 186 |
*/ |
| 187 |
private function getChunk() |
| 188 |
{ |
| 189 |
return fread($this->getFileHandle(), $this->chunk_size); |
| 190 |
} |
| 191 |
|
| 192 |
/** |
| 193 |
* Read all xml nodes in chunk |
| 194 |
*/ |
| 195 |
private function readChunkXmlNodes() |
| 196 |
{ |
| 197 |
$tags = [ |
| 198 |
["<?", "?>", true, 0], |
| 199 |
["<!--", "-->", true, 0], |
| 200 |
["<![CDATA[", "]]>", true, 0], |
| 201 |
["<!", ">", true, 0], |
| 202 |
["</", ">", false, -1], |
| 203 |
["<", "/>", false, 0], |
| 204 |
["<", ">", false, 1], |
| 205 |
]; |
| 206 |
|
| 207 |
while (preg_match("/<[^>]+>/", $this->chunk, $matches, PREG_OFFSET_CAPTURE) !== 0) { |
| 208 |
list($captured, $offset) = $matches[0]; |
| 209 |
|
| 210 |
$this->record_offset = $offset; |
| 211 |
|
| 212 |
foreach ($tags as $tag) { |
| 213 |
|
| 214 |
list($open_tag, $close_tag, $search_for_close, $depth_delta) = $tag; |
| 215 |
|
| 216 |
if (strpos($captured, $open_tag) === 0) { |
| 217 |
|
| 218 |
if (strlen($captured) - strlen($close_tag) === strpos($captured, $close_tag)) { |
| 219 |
// Tag Found |
| 220 |
if ($search_for_close === false) { |
| 221 |
$this->record_node($captured, $depth_delta); |
| 222 |
} |
| 223 |
break; |
| 224 |
} elseif ($search_for_close === true) { |
| 225 |
$close_tag_position = strpos($this->chunk, $close_tag, $offset); |
| 226 |
if ($close_tag_position !== false) { |
| 227 |
// Close tag is found in current chunk |
| 228 |
$captured = substr($this->chunk, $offset, $close_tag_position + strlen($close_tag) - $offset); |
| 229 |
if ($search_for_close === false) { |
| 230 |
$this->record_node($captured, $depth_delta); |
| 231 |
} |
| 232 |
} else { |
| 233 |
// We need to load next chunk to find |
| 234 |
return; |
| 235 |
} |
| 236 |
break; |
| 237 |
} |
| 238 |
} |
| 239 |
} |
| 240 |
|
| 241 |
$string_offset = $offset + strlen($captured); |
| 242 |
$this->chunk_offset += $string_offset; |
| 243 |
$this->chunk = substr($this->chunk, $string_offset); |
| 244 |
} |
| 245 |
return; |
| 246 |
|
| 247 |
// TODO: Match xmlns: <[^>]+xmlns:(.*?)="(.*?)"[^>]+> |
| 248 |
/*$open = "/<[^\/][^>]*[^\/]>/s"; |
| 249 |
$close = "/<\/[^>]*[^\/]>/"; |
| 250 |
$openclose = "/<[^\/][^>]+\/>/s"; |
| 251 |
|
| 252 |
while (preg_match("/<[^?|!\-\-][^>]*>/s", $this->chunk, $matches, |
| 253 |
PREG_OFFSET_CAPTURE) !== 0) { |
| 254 |
if (isset($matches[0]) && isset($matches[0][0])) { |
| 255 |
list($captured, $offset) = $matches[0]; |
| 256 |
|
| 257 |
$this->record_offset = $offset; |
| 258 |
|
| 259 |
if (preg_match($close, $captured) === 1) { |
| 260 |
$this->depth--; |
| 261 |
|
| 262 |
$node = array_pop($this->open_nodes); |
| 263 |
|
| 264 |
if ($this->isRecordEnd($node)) { |
| 265 |
$this->record_opened = false; |
| 266 |
$temp = $this->chunk_offset + $this->record_offset + strlen($captured); |
| 267 |
$this->setIndex($this->record_counter, |
| 268 |
$this->record_start_index, $temp); |
| 269 |
$this->record_counter++; |
| 270 |
} |
| 271 |
}elseif (preg_match($open, $captured) === 1) { |
| 272 |
$this->depth++; |
| 273 |
|
| 274 |
$node_name = $this->getNodeName($captured); |
| 275 |
|
| 276 |
if ($this->isRecordStart($node_name)) { |
| 277 |
$this->record_opened = true; |
| 278 |
$temp = $this->chunk_offset + $this->record_offset; |
| 279 |
$this->record_start_index = $temp; |
| 280 |
} |
| 281 |
|
| 282 |
$this->open_nodes[] = $node_name; |
| 283 |
$this->record_open_nodes(); |
| 284 |
|
| 285 |
} |
| 286 |
$string_offset = $offset + strlen($captured); |
| 287 |
$this->chunk_offset += $string_offset; |
| 288 |
$this->chunk = substr($this->chunk, $string_offset); |
| 289 |
} |
| 290 |
}*/ |
| 291 |
} |
| 292 |
|
| 293 |
private function record_node($node, $depth_delta) |
| 294 |
{ |
| 295 |
|
| 296 |
$this->depth += $depth_delta; |
| 297 |
$node_name = $this->getNodeName($node); |
| 298 |
|
| 299 |
if ($depth_delta > 0) { |
| 300 |
|
| 301 |
if ($this->isRecordStart($node_name)) { |
| 302 |
$this->record_opened = true; |
| 303 |
$this->record_start_index = $this->chunk_offset + $this->record_offset; |
| 304 |
} |
| 305 |
|
| 306 |
$this->open_nodes[] = $node_name; |
| 307 |
$this->record_open_nodes(); |
| 308 |
} elseif ($depth_delta < 0) { |
| 309 |
|
| 310 |
array_pop($this->open_nodes); |
| 311 |
if ($this->isRecordEnd($node_name)) { |
| 312 |
$this->record_opened = false; |
| 313 |
|
| 314 |
$this->setIndex( |
| 315 |
$this->record_counter, |
| 316 |
$this->record_start_index, |
| 317 |
$this->chunk_offset + $this->record_offset + strlen($node) |
| 318 |
); |
| 319 |
$this->record_counter++; |
| 320 |
} |
| 321 |
} |
| 322 |
|
| 323 |
// echo sprintf("%s %s - %d\n", str_repeat("\t", $this->depth), $node_name, $depth_delta); |
| 324 |
} |
| 325 |
|
| 326 |
/** |
| 327 |
* Check for record start node |
| 328 |
* |
| 329 |
* @param $node_name |
| 330 |
* |
| 331 |
* @return bool |
| 332 |
*/ |
| 333 |
private function isRecordStart($node_name) |
| 334 |
{ |
| 335 |
|
| 336 |
if ($this->base_path_segments === $this->open_nodes && $node_name === $this->base_path) { |
| 337 |
return true; |
| 338 |
} |
| 339 |
|
| 340 |
return false; |
| 341 |
} |
| 342 |
|
| 343 |
/** |
| 344 |
* Check for record end node |
| 345 |
* |
| 346 |
* @param $node_name |
| 347 |
* |
| 348 |
* @return bool |
| 349 |
*/ |
| 350 |
private function isRecordEnd($node_name) |
| 351 |
{ |
| 352 |
if ($this->record_opened === true && $node_name === $this->base_path) { |
| 353 |
|
| 354 |
// check to see if we are a nested duplicate node name /Events/Event/Instance and we are closing Event |
| 355 |
if (in_array($node_name, $this->open_nodes)) { |
| 356 |
$array_counts = array_count_values($this->open_nodes); |
| 357 |
if (isset($array_counts[$node_name]) && $array_counts[$node_name] > 0) { |
| 358 |
return false; |
| 359 |
} |
| 360 |
} |
| 361 |
|
| 362 |
return true; |
| 363 |
} |
| 364 |
|
| 365 |
return false; |
| 366 |
} |
| 367 |
|
| 368 |
public function wrapWithRecordTag($record) |
| 369 |
{ |
| 370 |
$namespace_attrs = ''; |
| 371 |
$unique_matches = array(); |
| 372 |
|
| 373 |
// clear sub namespaces |
| 374 |
$record = preg_replace('/xmlns[^=]*="[^"]*"/i', '', $record); |
| 375 |
|
| 376 |
// No need to check within cdata tags |
| 377 |
$temp = preg_replace('/<!\[CDATA\[.*?\]\]>/', '', $record); |
| 378 |
|
| 379 |
// $new_regex = '/<([^:!\/]+):[^ ]+[^>]*>/'; |
| 380 |
// $new_regex = '/<([^!\/>]+):[^>]+>/'; // breaking due to attr="http://", would match the : |
| 381 |
$new_regex = '/<([^!\/">]+):[^>]+>/'; |
| 382 |
|
| 383 |
while (preg_match($new_regex, $temp, $matches, PREG_OFFSET_CAPTURE) !== 0) { |
| 384 |
if (isset($matches[0]) && isset($matches[0][0]) && isset($matches[1]) && isset($matches[1][0])) { |
| 385 |
list($match, $offset) = $matches[0]; |
| 386 |
$namespace = trim($matches[1][0]); |
| 387 |
|
| 388 |
// TODO: update regex to not need this |
| 389 |
$namespace_parts = explode(' ', $namespace); |
| 390 |
$namespace = $namespace_parts[count($namespace_parts) - 1]; |
| 391 |
|
| 392 |
if (!in_array($namespace, $unique_matches)) { |
| 393 |
$unique_matches[] = $namespace; |
| 394 |
$namespace_attrs .= ' xmlns:' . $namespace . '="false"'; |
| 395 |
} |
| 396 |
|
| 397 |
$temp = substr($temp, $offset + strlen($match)); |
| 398 |
} |
| 399 |
} |
| 400 |
|
| 401 |
$open_tag = '<record' . $namespace_attrs . '>'; |
| 402 |
$close_tag = '</record>'; |
| 403 |
return $open_tag . $record . $close_tag; |
| 404 |
} |
| 405 |
|
| 406 |
private function record_open_nodes() |
| 407 |
{ |
| 408 |
$key = '/' . implode('/', $this->open_nodes); |
| 409 |
if (!in_array($key, $this->node_list, true)) { |
| 410 |
array_push($this->node_list, $key); |
| 411 |
} |
| 412 |
} |
| 413 |
|
| 414 |
public function get_node_list() |
| 415 |
{ |
| 416 |
|
| 417 |
$this->node_list = $this->config->get('node_list'); |
| 418 |
if (!is_array($this->node_list)) { |
| 419 |
$this->node_list = array(); |
| 420 |
} |
| 421 |
|
| 422 |
$loaded = $this->loadIndex(); |
| 423 |
if (empty($this->node_list) || !$loaded) { |
| 424 |
$this->generateIndex(); |
| 425 |
$this->storeIndexes(); |
| 426 |
|
| 427 |
$this->config->set('node_list', $this->node_list); |
| 428 |
} |
| 429 |
|
| 430 |
return $this->node_list; |
| 431 |
} |
| 432 |
|
| 433 |
protected final function getFileIndexKey() |
| 434 |
{ |
| 435 |
$key = sprintf('file_index-%s', $this->record_patch_cache_key); |
| 436 |
|
| 437 |
return $key; |
| 438 |
} |
| 439 |
} |
| 440 |
|
