Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
94.06% |
649 / 690 |
|
65.38% |
17 / 26 |
CRAP | |
0.00% |
0 / 1 |
| CommentParser | |
94.06% |
649 / 690 |
|
65.38% |
17 / 26 |
260.90 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
7 / 7 |
|
100.00% |
1 / 1 |
1 | |||
| parse | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
1 | |||
| nextInterestingLeafNode | |
95.45% |
21 / 22 |
|
0.00% |
0 / 1 |
8 | |||
| regexpAlternateGroup | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| getMessages | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| getTimestampRegexp | |
88.54% |
85 / 96 |
|
0.00% |
0 / 1 |
32.45 | |||
| getTimestampParser | |
93.85% |
122 / 130 |
|
0.00% |
0 / 1 |
52.63 | |||
| getLocalTimestampRegexps | |
100.00% |
9 / 9 |
|
100.00% |
1 / 1 |
1 | |||
| getLocalTimestampParsers | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
1 | |||
| getUsernameFromLink | |
97.06% |
33 / 34 |
|
0.00% |
0 / 1 |
17 | |||
| findSignature | |
100.00% |
43 / 43 |
|
100.00% |
1 / 1 |
15 | |||
| acceptOnlyNodesAllowingComments | |
100.00% |
17 / 17 |
|
100.00% |
1 / 1 |
11 | |||
| getCodepointOffset | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| findTimestamp | |
100.00% |
45 / 45 |
|
100.00% |
1 / 1 |
11 | |||
| adjustSigRange | |
100.00% |
14 / 14 |
|
100.00% |
1 / 1 |
3 | |||
| buildThreadItems | |
98.96% |
95 / 96 |
|
0.00% |
0 / 1 |
22 | |||
| computeTranscludedFrom | |
69.23% |
36 / 52 |
|
0.00% |
0 / 1 |
45.69 | |||
| titleCanExist | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
3 | |||
| parseTitle | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
2 | |||
| getTransclusionTitles | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
4 | |||
| getTransclusionRange | |
100.00% |
13 / 13 |
|
100.00% |
1 / 1 |
5 | |||
| truncateForId | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
2 | |||
| computeId | |
96.30% |
26 / 27 |
|
0.00% |
0 / 1 |
13 | |||
| computeName | |
91.67% |
11 / 12 |
|
0.00% |
0 / 1 |
4.01 | |||
| buildThreads | |
95.24% |
20 / 21 |
|
0.00% |
0 / 1 |
9 | |||
| computeIdsAndNames | |
100.00% |
9 / 9 |
|
100.00% |
1 / 1 |
3 | |||
| 1 | <?php |
| 2 | |
| 3 | namespace MediaWiki\Extension\DiscussionTools; |
| 4 | |
| 5 | use DateInterval; |
| 6 | use DateTime; |
| 7 | use DateTimeImmutable; |
| 8 | use DateTimeZone; |
| 9 | use InvalidArgumentException; |
| 10 | use LogicException; |
| 11 | use MediaWiki\Config\Config; |
| 12 | use MediaWiki\Extension\DiscussionTools\ThreadItem\ContentCommentItem; |
| 13 | use MediaWiki\Extension\DiscussionTools\ThreadItem\ContentHeadingItem; |
| 14 | use MediaWiki\Extension\DiscussionTools\ThreadItem\ContentThreadItem; |
| 15 | use MediaWiki\Language\Language; |
| 16 | use MediaWiki\Languages\LanguageConverterFactory; |
| 17 | use MediaWiki\Title\MalformedTitleException; |
| 18 | use MediaWiki\Title\TitleParser; |
| 19 | use MediaWiki\Title\TitleValue; |
| 20 | use MediaWiki\Utils\MWTimestamp; |
| 21 | use RuntimeException; |
| 22 | use Wikimedia\Assert\Assert; |
| 23 | use Wikimedia\IPUtils; |
| 24 | use Wikimedia\Parsoid\DOM\Element; |
| 25 | use Wikimedia\Parsoid\DOM\Node; |
| 26 | use Wikimedia\Parsoid\DOM\Text; |
| 27 | use Wikimedia\Parsoid\Utils\DOMCompat; |
| 28 | use Wikimedia\Parsoid\Utils\DOMUtils; |
| 29 | use Wikimedia\Timestamp\TimestampException; |
| 30 | |
| 31 | // TODO consider making timestamp parsing not a returned function |
| 32 | |
| 33 | class CommentParser { |
| 34 | |
| 35 | /** |
| 36 | * How far backwards we look for a signature associated with a timestamp before giving up. |
| 37 | * Note that this is not a hard limit on the length of signatures we detect. |
| 38 | */ |
| 39 | private const SIGNATURE_SCAN_LIMIT = 100; |
| 40 | |
| 41 | /** @var string[] */ |
| 42 | private array $dateFormat; |
| 43 | /** @var string[][] */ |
| 44 | private array $digits; |
| 45 | /** @var string[][] */ |
| 46 | private $contLangMessages; |
| 47 | private string $localTimezone; |
| 48 | /** @var string[][] */ |
| 49 | private array $timezones; |
| 50 | private string $specialContributionsName; |
| 51 | |
| 52 | private Element $rootNode; |
| 53 | private TitleValue $title; |
| 54 | |
| 55 | public function __construct( |
| 56 | private readonly Config $config, |
| 57 | private readonly Language $language, |
| 58 | private readonly LanguageConverterFactory $languageConverterFactory, |
| 59 | LanguageData $languageData, |
| 60 | private readonly TitleParser $titleParser, |
| 61 | ) { |
| 62 | $data = $languageData->getLocalData(); |
| 63 | $this->dateFormat = $data['dateFormat']; |
| 64 | $this->digits = $data['digits']; |
| 65 | $this->contLangMessages = $data['contLangMessages']; |
| 66 | $this->localTimezone = $data['localTimezone']; |
| 67 | $this->timezones = $data['timezones']; |
| 68 | $this->specialContributionsName = $data['specialContributionsName']; |
| 69 | } |
| 70 | |
| 71 | /** |
| 72 | * Parse a discussion page. |
| 73 | * |
| 74 | * @param Element $rootNode Root node of content to parse |
| 75 | * @param TitleValue $title Title of the page being parsed |
| 76 | */ |
| 77 | public function parse( Element $rootNode, TitleValue $title ): ContentThreadItemSet { |
| 78 | $this->rootNode = $rootNode; |
| 79 | $this->title = $title; |
| 80 | |
| 81 | $result = $this->buildThreadItems(); |
| 82 | $this->buildThreads( $result ); |
| 83 | $this->computeIdsAndNames( $result ); |
| 84 | |
| 85 | return $result; |
| 86 | } |
| 87 | |
| 88 | /** |
| 89 | * Return the next leaf node in the tree order that is likely a part of a discussion comment, |
| 90 | * rather than some boring "separator" element. |
| 91 | * |
| 92 | * Currently, this can return a Text node with content other than whitespace, or an Element node |
| 93 | * that is a "void element" or "text element", except some special cases that we treat as comment |
| 94 | * separators (isCommentSeparator()). |
| 95 | * |
| 96 | * @param ?Node $node Node after which to start searching |
| 97 | * (if null, start at the beginning of the document). |
| 98 | */ |
| 99 | private function nextInterestingLeafNode( ?Node $node ): Node { |
| 100 | $rootNode = $this->rootNode; |
| 101 | $treeWalker = new TreeWalker( |
| 102 | $rootNode, |
| 103 | NodeFilter::SHOW_ELEMENT | NodeFilter::SHOW_TEXT, |
| 104 | static function ( $n ) use ( $node ) { |
| 105 | // Skip past the starting node and its descendants |
| 106 | if ( $n === $node || $n->parentNode === $node ) { |
| 107 | return NodeFilter::FILTER_REJECT; |
| 108 | } |
| 109 | // Ignore some elements usually used as separators or headers (and their descendants) |
| 110 | if ( CommentUtils::isCommentSeparator( $n ) ) { |
| 111 | return NodeFilter::FILTER_REJECT; |
| 112 | } |
| 113 | // Ignore nodes with no rendering that mess up our indentation detection |
| 114 | if ( CommentUtils::isRenderingTransparentNode( $n ) ) { |
| 115 | return NodeFilter::FILTER_REJECT; |
| 116 | } |
| 117 | if ( CommentUtils::isCommentContent( $n ) ) { |
| 118 | return NodeFilter::FILTER_ACCEPT; |
| 119 | } |
| 120 | return NodeFilter::FILTER_SKIP; |
| 121 | } |
| 122 | ); |
| 123 | if ( $node ) { |
| 124 | $treeWalker->currentNode = $node; |
| 125 | } |
| 126 | $treeWalker->nextNode(); |
| 127 | if ( !$treeWalker->currentNode ) { |
| 128 | throw new RuntimeException( 'nextInterestingLeafNode not found' ); |
| 129 | } |
| 130 | return $treeWalker->currentNode; |
| 131 | } |
| 132 | |
| 133 | /** |
| 134 | * @param string[] $values Values to match |
| 135 | * @return string Regular expression |
| 136 | */ |
| 137 | private static function regexpAlternateGroup( array $values ): string { |
| 138 | return '(' . implode( '|', array_map( static function ( string $x ) { |
| 139 | return preg_quote( $x, '/' ); |
| 140 | }, $values ) ) . ')'; |
| 141 | } |
| 142 | |
| 143 | /** |
| 144 | * Get text of localisation messages in content language. |
| 145 | * |
| 146 | * @param string $contLangVariant Content language variant |
| 147 | * @param string[] $messages Message keys |
| 148 | * @return string[] Message values |
| 149 | */ |
| 150 | private function getMessages( string $contLangVariant, array $messages ): array { |
| 151 | return array_map( function ( string $key ) use ( $contLangVariant ) { |
| 152 | return $this->contLangMessages[$contLangVariant][$key]; |
| 153 | }, $messages ); |
| 154 | } |
| 155 | |
| 156 | /** |
| 157 | * Get a regexp that matches timestamps generated using the given date format. |
| 158 | * |
| 159 | * This only supports format characters that are used by the default date format in any of |
| 160 | * MediaWiki's languages, namely: D, d, F, G, H, i, j, l, M, n, Y, xg, xkY (and escape characters), |
| 161 | * and only dates when MediaWiki existed, let's say 2000 onwards (Thai dates before 1941 are |
| 162 | * complicated). |
| 163 | * |
| 164 | * @param string $contLangVariant Content language variant |
| 165 | * @param string $format Date format |
| 166 | * @param string $digitsRegexp Regular expression matching a single localised digit, e.g. '[0-9]' |
| 167 | * @param array $tzAbbrs Associative array mapping localised timezone abbreviations to |
| 168 | * IANA abbreviations, for the local timezone, e.g. [ 'EDT' => 'EDT', 'EST' => 'EST' ] |
| 169 | * @return string Regular expression |
| 170 | */ |
| 171 | private function getTimestampRegexp( |
| 172 | string $contLangVariant, string $format, string $digitsRegexp, array $tzAbbrs |
| 173 | ): string { |
| 174 | $formatLength = strlen( $format ); |
| 175 | $s = ''; |
| 176 | $raw = false; |
| 177 | // Adapted from Language::sprintfDate() |
| 178 | for ( $p = 0; $p < $formatLength; $p++ ) { |
| 179 | $num = false; |
| 180 | $code = $format[ $p ]; |
| 181 | if ( $code === 'x' && $p < $formatLength - 1 ) { |
| 182 | $code .= $format[++$p]; |
| 183 | } |
| 184 | if ( $code === 'xk' && $p < $formatLength - 1 ) { |
| 185 | $code .= $format[++$p]; |
| 186 | } |
| 187 | |
| 188 | switch ( $code ) { |
| 189 | case 'xx': |
| 190 | $s .= 'x'; |
| 191 | break; |
| 192 | case 'xg': |
| 193 | $s .= static::regexpAlternateGroup( |
| 194 | $this->getMessages( $contLangVariant, Language::MONTH_GENITIVE_MESSAGES ) |
| 195 | ); |
| 196 | break; |
| 197 | case 'xn': |
| 198 | $raw = true; |
| 199 | break; |
| 200 | case 'd': |
| 201 | $num = '2'; |
| 202 | break; |
| 203 | case 'D': |
| 204 | $s .= static::regexpAlternateGroup( |
| 205 | $this->getMessages( $contLangVariant, Language::WEEKDAY_ABBREVIATED_MESSAGES ) |
| 206 | ); |
| 207 | break; |
| 208 | case 'j': |
| 209 | $num = '1,2'; |
| 210 | break; |
| 211 | case 'l': |
| 212 | $s .= static::regexpAlternateGroup( |
| 213 | $this->getMessages( $contLangVariant, Language::WEEKDAY_MESSAGES ) |
| 214 | ); |
| 215 | break; |
| 216 | case 'F': |
| 217 | $s .= static::regexpAlternateGroup( |
| 218 | $this->getMessages( $contLangVariant, Language::MONTH_MESSAGES ) |
| 219 | ); |
| 220 | break; |
| 221 | case 'M': |
| 222 | $s .= static::regexpAlternateGroup( |
| 223 | $this->getMessages( $contLangVariant, Language::MONTH_ABBREVIATED_MESSAGES ) |
| 224 | ); |
| 225 | break; |
| 226 | case 'm': |
| 227 | $num = '2'; |
| 228 | break; |
| 229 | case 'n': |
| 230 | $num = '1,2'; |
| 231 | break; |
| 232 | case 'Y': |
| 233 | $num = '4'; |
| 234 | break; |
| 235 | case 'xkY': |
| 236 | $num = '4'; |
| 237 | break; |
| 238 | case 'G': |
| 239 | $num = '1,2'; |
| 240 | break; |
| 241 | case 'H': |
| 242 | $num = '2'; |
| 243 | break; |
| 244 | case 'i': |
| 245 | $num = '2'; |
| 246 | break; |
| 247 | case 's': |
| 248 | $num = '2'; |
| 249 | break; |
| 250 | case '\\': |
| 251 | // Backslash escaping |
| 252 | if ( $p < $formatLength - 1 ) { |
| 253 | $s .= preg_quote( $format[++$p], '/' ); |
| 254 | } else { |
| 255 | $s .= preg_quote( '\\', '/' ); |
| 256 | } |
| 257 | break; |
| 258 | case '"': |
| 259 | // Quoted literal |
| 260 | if ( $p < $formatLength - 1 ) { |
| 261 | $endQuote = strpos( $format, '"', $p + 1 ); |
| 262 | if ( $endQuote === false ) { |
| 263 | // No terminating quote, assume literal " |
| 264 | $s .= '"'; |
| 265 | } else { |
| 266 | $s .= preg_quote( substr( $format, $p + 1, $endQuote - $p - 1 ), '/' ); |
| 267 | $p = $endQuote; |
| 268 | } |
| 269 | } else { |
| 270 | // Quote at end of string, assume literal " |
| 271 | $s .= '"'; |
| 272 | } |
| 273 | break; |
| 274 | default: |
| 275 | // Copy whole characters together, instead of single bytes |
| 276 | $char = mb_substr( mb_strcut( $format, $p, 4 ), 0, 1 ); |
| 277 | $s .= preg_quote( $char, '/' ); |
| 278 | $p += strlen( $char ) - 1; |
| 279 | } |
| 280 | if ( $num !== false ) { |
| 281 | if ( $raw ) { |
| 282 | $s .= '([0-9]{' . $num . '})'; |
| 283 | $raw = false; |
| 284 | } else { |
| 285 | $s .= '(' . $digitsRegexp . '{' . $num . '})'; |
| 286 | } |
| 287 | } |
| 288 | // Ignore some invisible Unicode characters that often sneak into copy-pasted timestamps (T308448) |
| 289 | $s .= '[\\x{200E}\\x{200F}]?'; |
| 290 | } |
| 291 | |
| 292 | $tzRegexp = static::regexpAlternateGroup( array_keys( $tzAbbrs ) ); |
| 293 | |
| 294 | // Hard-coded parentheses and space like in Parser::pstPass2 |
| 295 | // Ignore some invisible Unicode characters that often sneak into copy-pasted timestamps (T245784) |
| 296 | // \uNNNN syntax can only be used from PHP 7.3 |
| 297 | return '/' . $s . ' [\\x{200E}\\x{200F}]?\\(' . $tzRegexp . '\\)/u'; |
| 298 | } |
| 299 | |
| 300 | /** |
| 301 | * Get a function that parses timestamps generated using the given date format, based on the result |
| 302 | * of matching the regexp returned by getTimestampRegexp() |
| 303 | * |
| 304 | * @param string $contLangVariant Content language variant |
| 305 | * @param string $format Date format, as used by MediaWiki |
| 306 | * @param array<int,string>|null $digits Localised digits from 0 to 9, e.g. `[ '0', '1', ..., '9' ]` |
| 307 | * @param string $localTimezone Local timezone IANA name, e.g. `America/New_York` |
| 308 | * @param array $tzAbbrs Map of localised timezone abbreviations to IANA abbreviations |
| 309 | * for the local timezone, e.g. [ 'EDT' => 'EDT', 'EST' => 'EST' ] |
| 310 | * @return callable Parser function |
| 311 | */ |
| 312 | private function getTimestampParser( |
| 313 | string $contLangVariant, string $format, ?array $digits, string $localTimezone, array $tzAbbrs |
| 314 | ): callable { |
| 315 | $untransformDigits = static function ( string $text ) use ( $digits ): int { |
| 316 | return (int)( $digits ? strtr( $text, array_flip( $digits ) ) : $text ); |
| 317 | }; |
| 318 | |
| 319 | $formatLength = strlen( $format ); |
| 320 | $matchingGroups = []; |
| 321 | for ( $p = 0; $p < $formatLength; $p++ ) { |
| 322 | $code = $format[$p]; |
| 323 | if ( $code === 'x' && $p < $formatLength - 1 ) { |
| 324 | $code .= $format[++$p]; |
| 325 | } |
| 326 | if ( $code === 'xk' && $p < $formatLength - 1 ) { |
| 327 | $code .= $format[++$p]; |
| 328 | } |
| 329 | |
| 330 | switch ( $code ) { |
| 331 | case 'xx': |
| 332 | case 'xn': |
| 333 | break; |
| 334 | case 'xg': |
| 335 | case 'd': |
| 336 | case 'j': |
| 337 | case 'D': |
| 338 | case 'l': |
| 339 | case 'F': |
| 340 | case 'M': |
| 341 | case 'm': |
| 342 | case 'n': |
| 343 | case 'Y': |
| 344 | case 'xkY': |
| 345 | case 'G': |
| 346 | case 'H': |
| 347 | case 'i': |
| 348 | case 's': |
| 349 | $matchingGroups[] = $code; |
| 350 | break; |
| 351 | case '\\': |
| 352 | // Backslash escaping |
| 353 | if ( $p < $formatLength - 1 ) { |
| 354 | $p++; |
| 355 | } |
| 356 | break; |
| 357 | case '"': |
| 358 | // Quoted literal |
| 359 | if ( $p < $formatLength - 1 ) { |
| 360 | $endQuote = strpos( $format, '"', $p + 1 ); |
| 361 | if ( $endQuote !== false ) { |
| 362 | $p = $endQuote; |
| 363 | } |
| 364 | } |
| 365 | break; |
| 366 | default: |
| 367 | break; |
| 368 | } |
| 369 | } |
| 370 | |
| 371 | return function ( array $match ) use ( |
| 372 | $matchingGroups, $untransformDigits, $localTimezone, $tzAbbrs, $contLangVariant |
| 373 | ) { |
| 374 | if ( is_array( $match[0] ) ) { |
| 375 | // Strip PREG_OFFSET_CAPTURE data |
| 376 | unset( $match['offset'] ); |
| 377 | $match = array_map( static function ( array $tuple ) { |
| 378 | return $tuple[0]; |
| 379 | }, $match ); |
| 380 | } |
| 381 | $year = 0; |
| 382 | $monthIdx = 0; |
| 383 | $day = 0; |
| 384 | $hour = 0; |
| 385 | $minute = 0; |
| 386 | foreach ( $matchingGroups as $i => $code ) { |
| 387 | $text = $match[$i + 1]; |
| 388 | switch ( $code ) { |
| 389 | case 'xg': |
| 390 | $monthIdx = array_search( |
| 391 | $text, |
| 392 | $this->getMessages( $contLangVariant, Language::MONTH_GENITIVE_MESSAGES ), |
| 393 | true |
| 394 | ); |
| 395 | break; |
| 396 | case 'd': |
| 397 | case 'j': |
| 398 | $day = $untransformDigits( $text ); |
| 399 | break; |
| 400 | case 'D': |
| 401 | case 'l': |
| 402 | // Day of the week - unused |
| 403 | break; |
| 404 | case 'F': |
| 405 | $monthIdx = array_search( |
| 406 | $text, |
| 407 | $this->getMessages( $contLangVariant, Language::MONTH_MESSAGES ), |
| 408 | true |
| 409 | ); |
| 410 | break; |
| 411 | case 'M': |
| 412 | $monthIdx = array_search( |
| 413 | $text, |
| 414 | $this->getMessages( $contLangVariant, Language::MONTH_ABBREVIATED_MESSAGES ), |
| 415 | true |
| 416 | ); |
| 417 | break; |
| 418 | case 'm': |
| 419 | case 'n': |
| 420 | $monthIdx = $untransformDigits( $text ) - 1; |
| 421 | break; |
| 422 | case 'Y': |
| 423 | $year = $untransformDigits( $text ); |
| 424 | break; |
| 425 | case 'xkY': |
| 426 | // Thai year |
| 427 | $year = $untransformDigits( $text ) - 543; |
| 428 | break; |
| 429 | case 'G': |
| 430 | case 'H': |
| 431 | $hour = $untransformDigits( $text ); |
| 432 | break; |
| 433 | case 'i': |
| 434 | $minute = $untransformDigits( $text ); |
| 435 | break; |
| 436 | case 's': |
| 437 | // Seconds - unused, because most timestamp formats omit them |
| 438 | break; |
| 439 | default: |
| 440 | throw new LogicException( 'Not implemented' ); |
| 441 | } |
| 442 | } |
| 443 | |
| 444 | // The last matching group is the timezone abbreviation |
| 445 | $tzAbbr = $tzAbbrs[ end( $match ) ]; |
| 446 | |
| 447 | // Most of the time, the timezone abbreviation is not necessary to parse the date, since we |
| 448 | // can assume all times are in the wiki's local timezone. |
| 449 | $date = new DateTime(); |
| 450 | // setTimezone must be called before setDate/setTime |
| 451 | $date->setTimezone( new DateTimeZone( $localTimezone ) ); |
| 452 | $date->setDate( $year, $monthIdx + 1, $day ); |
| 453 | $date->setTime( $hour, $minute, 0 ); |
| 454 | |
| 455 | // But during the "fall back" at the end of DST, some times will happen twice. |
| 456 | // Since the timezone abbreviation disambiguates the DST/non-DST times, we can detect |
| 457 | // when PHP chose the wrong one, and then try the other one. It appears that PHP always |
| 458 | // uses the later (non-DST) hour, but that behavior isn't documented, so we account for both. |
| 459 | $dateWarning = null; |
| 460 | if ( $date->format( 'T' ) !== $tzAbbr ) { |
| 461 | $altDate = clone $date; |
| 462 | if ( $date->format( 'I' ) ) { |
| 463 | // Parsed time is DST, try non-DST by advancing one hour |
| 464 | $altDate->add( new DateInterval( 'PT1H' ) ); |
| 465 | } else { |
| 466 | // Parsed time is non-DST, try DST by going back one hour |
| 467 | $altDate->sub( new DateInterval( 'PT1H' ) ); |
| 468 | } |
| 469 | if ( $altDate->format( 'T' ) === $tzAbbr ) { |
| 470 | $date = $altDate; |
| 471 | $dateWarning = 'Timestamp has timezone abbreviation for the wrong time'; |
| 472 | } else { |
| 473 | $dateWarning = 'Ambiguous time at DST switchover was parsed'; |
| 474 | } |
| 475 | } |
| 476 | |
| 477 | // Now set the timezone back to UTC for formatting |
| 478 | $date->setTimezone( new DateTimeZone( 'UTC' ) ); |
| 479 | $date = DateTimeImmutable::createFromMutable( $date ); |
| 480 | |
| 481 | // We require the date to be compatible with our libraries, for example zero or negative years (T352455) |
| 482 | // In PHP we need to check with MWTimestamp. |
| 483 | // In JS we need to check with Moment. |
| 484 | try { |
| 485 | // @phan-suppress-next-line PhanNoopNew |
| 486 | new MWTimestamp( $date->format( 'c' ) ); |
| 487 | } catch ( TimestampException ) { |
| 488 | return null; |
| 489 | } |
| 490 | |
| 491 | return [ |
| 492 | 'date' => $date, |
| 493 | 'warning' => $dateWarning, |
| 494 | ]; |
| 495 | }; |
| 496 | } |
| 497 | |
| 498 | /** |
| 499 | * Get a regexp that matches timestamps in the local date format, for each language variant. |
| 500 | * |
| 501 | * This calls getTimestampRegexp() with predefined data for the current wiki. |
| 502 | * |
| 503 | * @return string[] Regular expressions |
| 504 | */ |
| 505 | public function getLocalTimestampRegexps(): array { |
| 506 | $langConv = $this->languageConverterFactory->getLanguageConverter( $this->language ); |
| 507 | return array_map( function ( $contLangVariant ) { |
| 508 | return $this->getTimestampRegexp( |
| 509 | $contLangVariant, |
| 510 | $this->dateFormat[$contLangVariant], |
| 511 | '[' . implode( '', $this->digits[$contLangVariant] ) . ']', |
| 512 | $this->timezones[$contLangVariant] |
| 513 | ); |
| 514 | }, $langConv->getVariants() ); |
| 515 | } |
| 516 | |
| 517 | /** |
| 518 | * Get a function that parses timestamps in the local date format, for each language variant, |
| 519 | * based on the result of matching the regexp returned by getLocalTimestampRegexp(). |
| 520 | * |
| 521 | * This calls getTimestampParser() with predefined data for the current wiki. |
| 522 | * |
| 523 | * @return callable[] Parser functions |
| 524 | */ |
| 525 | private function getLocalTimestampParsers(): array { |
| 526 | $langConv = $this->languageConverterFactory->getLanguageConverter( $this->language ); |
| 527 | return array_map( function ( $contLangVariant ) { |
| 528 | return $this->getTimestampParser( |
| 529 | $contLangVariant, |
| 530 | $this->dateFormat[$contLangVariant], |
| 531 | $this->digits[$contLangVariant], |
| 532 | $this->localTimezone, |
| 533 | $this->timezones[$contLangVariant] |
| 534 | ); |
| 535 | }, $langConv->getVariants() ); |
| 536 | } |
| 537 | |
| 538 | /** |
| 539 | * Given a link node (`<a>`), if it's a link to a user-related page, return their username. |
| 540 | * |
| 541 | * @return array|null Array, or null: |
| 542 | * - string 'username' Username |
| 543 | * - string|null 'displayName' Display name (link text if link target was in the user namespace) |
| 544 | */ |
| 545 | private function getUsernameFromLink( Element $link ): ?array { |
| 546 | // Selflink: use title of current page |
| 547 | if ( DOMUtils::hasClass( $link, 'mw-selflink' ) ) { |
| 548 | $title = $this->title; |
| 549 | } else { |
| 550 | $titleString = CommentUtils::getTitleFromUrl( $link->getAttribute( 'href' ) ?? '', $this->config ) ?? ''; |
| 551 | // Performance optimization, skip strings that obviously don't contain a namespace |
| 552 | if ( $titleString === '' || !str_contains( $titleString, ':' ) ) { |
| 553 | return null; |
| 554 | } |
| 555 | $title = $this->parseTitle( $titleString ); |
| 556 | if ( !$title ) { |
| 557 | return null; |
| 558 | } |
| 559 | } |
| 560 | |
| 561 | $username = null; |
| 562 | $displayName = null; |
| 563 | $mainText = $title->getText(); |
| 564 | |
| 565 | if ( $title->inNamespace( NS_USER ) || $title->inNamespace( NS_USER_TALK ) ) { |
| 566 | $username = $mainText; |
| 567 | if ( str_contains( $username, '/' ) ) { |
| 568 | return null; |
| 569 | } |
| 570 | if ( $title->inNamespace( NS_USER ) ) { |
| 571 | // Use regex trim for consistency with JS implementation |
| 572 | $text = preg_replace( [ '/^[\s]+/u', '/[\s]+$/u' ], '', $link->textContent ?? '' ); |
| 573 | // Record the display name if it has been customised beyond changing case |
| 574 | if ( $text && mb_strtolower( $text ) !== mb_strtolower( $username ) ) { |
| 575 | $displayName = $text; |
| 576 | } |
| 577 | } |
| 578 | } elseif ( $title->inNamespace( NS_SPECIAL ) ) { |
| 579 | $parts = explode( '/', $mainText ); |
| 580 | if ( count( $parts ) === 2 && $parts[0] === $this->specialContributionsName ) { |
| 581 | // Normalize the username: users may link to their contributions with an unnormalized name |
| 582 | $userpage = $this->titleParser->makeTitleValueSafe( NS_USER, $parts[1] ); |
| 583 | if ( !$userpage ) { |
| 584 | return null; |
| 585 | } |
| 586 | $username = $userpage->getText(); |
| 587 | } |
| 588 | } |
| 589 | if ( $username === null ) { |
| 590 | return null; |
| 591 | } |
| 592 | if ( IPUtils::isIPv6( $username ) ) { |
| 593 | // Bot-generated links "Preceding unsigned comment added by" have non-standard case |
| 594 | $username = strtoupper( $username ); |
| 595 | } |
| 596 | return [ |
| 597 | 'username' => $username, |
| 598 | 'displayName' => $displayName, |
| 599 | ]; |
| 600 | } |
| 601 | |
| 602 | /** |
| 603 | * Find a user signature preceding a timestamp. |
| 604 | * |
| 605 | * The signature includes the timestamp node. |
| 606 | * |
| 607 | * A signature must contain at least one link to the user's userpage, discussion page or |
| 608 | * contributions (and may contain other links). The link may be nested in other elements. |
| 609 | * |
| 610 | * @param Text $timestampNode |
| 611 | * @param Node|null $until Node to stop searching at |
| 612 | * @return array Result, an associative array with the following keys: |
| 613 | * - Node[] `nodes` Sibling nodes comprising the signature, in reverse order (with |
| 614 | * $timestampNode or its parent node as the first element) |
| 615 | * - string|null `username` Username, null for unsigned comments |
| 616 | */ |
| 617 | private function findSignature( Text $timestampNode, ?Node $until = null ): array { |
| 618 | $sigUsername = null; |
| 619 | $sigDisplayName = null; |
| 620 | $length = 0; |
| 621 | $lastLinkNode = $timestampNode; |
| 622 | |
| 623 | CommentUtils::linearWalkBackwards( |
| 624 | $timestampNode, |
| 625 | function ( string $event, Node $node ) use ( |
| 626 | &$sigUsername, &$sigDisplayName, &$lastLinkNode, &$length, |
| 627 | $until, $timestampNode |
| 628 | ) { |
| 629 | if ( $event === 'enter' && $node === $until ) { |
| 630 | return true; |
| 631 | } |
| 632 | if ( $length >= static::SIGNATURE_SCAN_LIMIT ) { |
| 633 | return true; |
| 634 | } |
| 635 | if ( CommentUtils::isBlockElement( $node ) ) { |
| 636 | // Don't allow reaching into preceding paragraphs |
| 637 | return true; |
| 638 | } |
| 639 | |
| 640 | if ( $event === 'leave' && $node !== $timestampNode ) { |
| 641 | $length += $node instanceof Text ? |
| 642 | mb_strlen( CommentUtils::htmlTrim( $node->textContent ?? '' ) ) : 0; |
| 643 | } |
| 644 | |
| 645 | // Find the closest link before timestamp that links to the user's user page. |
| 646 | // |
| 647 | // Support timestamps being linked to the diff introducing the comment: |
| 648 | // if the timestamp node is the only child of a link node, use the link node instead |
| 649 | // |
| 650 | // Handle links nested in formatting elements. |
| 651 | if ( $event === 'leave' && $node instanceof Element && strtolower( $node->tagName ) === 'a' ) { |
| 652 | $classList = DOMCompat::getClassList( $node ); |
| 653 | // Generated timestamp links sometimes look like username links (e.g. on user talk pages) |
| 654 | // so ignore these. |
| 655 | if ( !$classList->contains( 'ext-discussiontools-init-timestamplink' ) ) { |
| 656 | $user = $this->getUsernameFromLink( $node ); |
| 657 | if ( $user ) { |
| 658 | // Accept the first link to the user namespace, then only accept links to that user |
| 659 | $sigUsername ??= $user['username']; |
| 660 | if ( $user['username'] === $sigUsername ) { |
| 661 | $lastLinkNode = $node; |
| 662 | if ( $user['displayName'] ) { |
| 663 | $sigDisplayName = $user['displayName']; |
| 664 | } |
| 665 | } |
| 666 | } |
| 667 | // Keep looking if a node with links wasn't a link to a user page |
| 668 | // "Doc James (talk · contribs · email)" |
| 669 | } |
| 670 | } |
| 671 | } |
| 672 | ); |
| 673 | |
| 674 | $range = new ImmutableRange( |
| 675 | $lastLinkNode->parentNode, |
| 676 | CommentUtils::childIndexOf( $lastLinkNode ), |
| 677 | $timestampNode->parentNode, |
| 678 | CommentUtils::childIndexOf( $timestampNode ) + 1 |
| 679 | ); |
| 680 | |
| 681 | // Expand the range so that it covers sibling nodes. |
| 682 | // This will include any wrapping formatting elements as part of the signature. |
| 683 | // |
| 684 | // Helpful accidental feature: users whose signature is not detected in full (due to |
| 685 | // text formatting) can just wrap it in a <span> to fix that. |
| 686 | // "Ten Pound Hammer • (What did I screw up now?)" |
| 687 | // "« Saper // dyskusja »" |
| 688 | // |
| 689 | // TODO Not sure if this is actually good, might be better to just use the range... |
| 690 | $sigNodes = array_reverse( CommentUtils::getCoveredSiblings( $range ) ); |
| 691 | |
| 692 | return [ |
| 693 | 'nodes' => $sigNodes, |
| 694 | 'username' => $sigUsername, |
| 695 | 'displayName' => $sigDisplayName, |
| 696 | ]; |
| 697 | } |
| 698 | |
| 699 | /** |
| 700 | * Callback for TreeWalker that will skip over nodes where we don't want to detect |
| 701 | * comments (or section headings). |
| 702 | * |
| 703 | * @return int Appropriate NodeFilter constant |
| 704 | */ |
| 705 | public static function acceptOnlyNodesAllowingComments( Node $node ): int { |
| 706 | if ( $node instanceof Element ) { |
| 707 | $tagName = strtolower( $node->tagName ); |
| 708 | // The table of contents has a heading that gets erroneously detected as a section |
| 709 | if ( $node->getAttribute( 'id' ) === 'toc' ) { |
| 710 | return NodeFilter::FILTER_REJECT; |
| 711 | } |
| 712 | // Don't detect comments within quotes (T275881) |
| 713 | if ( |
| 714 | $tagName === 'blockquote' || |
| 715 | $tagName === 'cite' || |
| 716 | $tagName === 'q' |
| 717 | ) { |
| 718 | return NodeFilter::FILTER_REJECT; |
| 719 | } |
| 720 | // Don't attempt to parse blocks marked 'mw-notalk' |
| 721 | if ( DOMUtils::hasClass( $node, 'mw-notalk' ) ) { |
| 722 | return NodeFilter::FILTER_REJECT; |
| 723 | } |
| 724 | // Don't detect comments within references. We can't add replies to them without bungling up |
| 725 | // the structure in some cases (T301213), and you're not supposed to do that anyway… |
| 726 | if ( |
| 727 | // <ol class="references"> is the only reliably consistent thing between the two parsers |
| 728 | $tagName === 'ol' && |
| 729 | DOMUtils::hasClass( $node, 'references' ) |
| 730 | ) { |
| 731 | return NodeFilter::FILTER_REJECT; |
| 732 | } |
| 733 | } |
| 734 | $parentNode = $node->parentNode; |
| 735 | // Don't detect comments within headings (but don't reject the headings themselves) |
| 736 | if ( $parentNode instanceof Element && preg_match( '/^h([1-6])$/i', $parentNode->tagName ) ) { |
| 737 | return NodeFilter::FILTER_REJECT; |
| 738 | } |
| 739 | return NodeFilter::FILTER_ACCEPT; |
| 740 | } |
| 741 | |
| 742 | /** |
| 743 | * Convert a byte offset within a text node to a unicode codepoint offset |
| 744 | * |
| 745 | * @param Text $node Text node |
| 746 | * @param int $byteOffset Byte offset |
| 747 | * @return int Codepoint offset |
| 748 | */ |
| 749 | private static function getCodepointOffset( Text $node, int $byteOffset ): int { |
| 750 | return mb_strlen( substr( $node->nodeValue ?? '', 0, $byteOffset ) ); |
| 751 | } |
| 752 | |
| 753 | /** |
| 754 | * Find a timestamps in a given text node |
| 755 | * |
| 756 | * @param Text $node |
| 757 | * @param string[] $timestampRegexps |
| 758 | * @return array|null Array with the following keys: |
| 759 | * - int 'offset' Length of extra text preceding the node that was used for matching (in bytes) |
| 760 | * - int 'parserIndex' Which of the regexps matched |
| 761 | * - array 'matchData' Regexp match data, which specifies the location of the match, |
| 762 | * and which can be parsed using getLocalTimestampParsers() (offsets are in bytes) |
| 763 | * - ImmutableRange 'range' Range covering the timestamp |
| 764 | */ |
| 765 | public function findTimestamp( Text $node, array $timestampRegexps ): ?array { |
| 766 | $nodeText = ''; |
| 767 | $offset = 0; |
| 768 | // Searched nodes (reverse order) |
| 769 | $nodes = []; |
| 770 | |
| 771 | while ( $node ) { |
| 772 | $nodeText = $node->nodeValue . $nodeText; |
| 773 | $nodes[] = $node; |
| 774 | |
| 775 | // In Parsoid HTML, entities are represented as a 'mw:Entity' node, rather than normal HTML |
| 776 | // entities. On Arabic Wikipedia, the "UTC" timezone name contains some non-breaking spaces, |
| 777 | // which apparently are often turned into entities by buggy editing tools. To handle |
| 778 | // this, we must piece together the text, so that our regexp can match those timestamps. |
| 779 | if ( |
| 780 | ( $previousSibling = $node->previousSibling ) && |
| 781 | $previousSibling instanceof Element && |
| 782 | $previousSibling->getAttribute( 'typeof' ) === 'mw:Entity' |
| 783 | ) { |
| 784 | $nodeText = $previousSibling->firstChild->nodeValue . $nodeText; |
| 785 | $offset += strlen( $previousSibling->firstChild->nodeValue ?? '' ); |
| 786 | $nodes[] = $previousSibling->firstChild; |
| 787 | |
| 788 | // If the entity is preceded by more text, do this again |
| 789 | if ( |
| 790 | $previousSibling->previousSibling && |
| 791 | $previousSibling->previousSibling instanceof Text |
| 792 | ) { |
| 793 | $offset += strlen( $previousSibling->previousSibling->nodeValue ?? '' ); |
| 794 | $node = $previousSibling->previousSibling; |
| 795 | } else { |
| 796 | $node = null; |
| 797 | } |
| 798 | } else { |
| 799 | $node = null; |
| 800 | } |
| 801 | } |
| 802 | |
| 803 | foreach ( $timestampRegexps as $i => $timestampRegexp ) { |
| 804 | $matchData = null; |
| 805 | // Allows us to mimic match.index in #getComments |
| 806 | if ( preg_match( $timestampRegexp, $nodeText, $matchData, PREG_OFFSET_CAPTURE ) ) { |
| 807 | $timestampLength = strlen( $matchData[0][0] ); |
| 808 | // Bytes at the end of the last node which aren't part of the match |
| 809 | $tailLength = strlen( $nodeText ) - $timestampLength - $matchData[0][1]; |
| 810 | // We are moving right to left, but we start to the right of the end of |
| 811 | // the timestamp if there is trailing garbage, so that is a negative offset. |
| 812 | $count = -$tailLength; |
| 813 | $endNode = $nodes[0]; |
| 814 | $endOffset = strlen( $endNode->nodeValue ?? '' ) - $tailLength; |
| 815 | |
| 816 | foreach ( $nodes as $n ) { |
| 817 | $count += strlen( $n->nodeValue ?? '' ); |
| 818 | // If we have counted to beyond the start of the timestamp, we are in the |
| 819 | // start node of the timestamp |
| 820 | if ( $count >= $timestampLength ) { |
| 821 | $startNode = $n; |
| 822 | // Offset is how much we overshot the start by |
| 823 | $startOffset = $count - $timestampLength; |
| 824 | break; |
| 825 | } |
| 826 | } |
| 827 | Assert::precondition( $endNode instanceof Node, 'endNode of timestamp is a Node' ); |
| 828 | Assert::precondition( $startNode instanceof Node, 'startNode of timestamp range found' ); |
| 829 | Assert::precondition( is_int( $startOffset ), 'startOffset of timestamp range found' ); |
| 830 | |
| 831 | $startOffset = static::getCodepointOffset( $startNode, $startOffset ); |
| 832 | $endOffset = static::getCodepointOffset( $endNode, $endOffset ); |
| 833 | |
| 834 | $range = new ImmutableRange( $startNode, $startOffset, $endNode, $endOffset ); |
| 835 | |
| 836 | return [ |
| 837 | 'matchData' => $matchData, |
| 838 | // Bytes at the start of the first node which aren't part of the match |
| 839 | // TODO: Remove this and use 'range' instead |
| 840 | 'offset' => $offset, |
| 841 | 'range' => $range, |
| 842 | 'parserIndex' => $i, |
| 843 | ]; |
| 844 | } |
| 845 | } |
| 846 | return null; |
| 847 | } |
| 848 | |
| 849 | /** |
| 850 | * @param Node[] $sigNodes |
| 851 | * @param array $match |
| 852 | * @param Text $node |
| 853 | */ |
| 854 | private function adjustSigRange( array $sigNodes, array $match, Text $node ): ImmutableRange { |
| 855 | $firstSigNode = end( $sigNodes ); |
| 856 | $lastSigNode = $sigNodes[0]; |
| 857 | |
| 858 | // TODO Document why this needs to be so complicated |
| 859 | $lastSigNodeOffsetByteOffset = |
| 860 | $match['matchData'][0][1] + strlen( $match['matchData'][0][0] ) - $match['offset']; |
| 861 | $lastSigNodeOffset = $lastSigNode === $node ? |
| 862 | static::getCodepointOffset( $node, $lastSigNodeOffsetByteOffset ) : |
| 863 | CommentUtils::childIndexOf( $lastSigNode ) + 1; |
| 864 | $sigRange = new ImmutableRange( |
| 865 | $firstSigNode->parentNode, |
| 866 | CommentUtils::childIndexOf( $firstSigNode ), |
| 867 | $lastSigNode === $node ? $node : $lastSigNode->parentNode, |
| 868 | $lastSigNodeOffset |
| 869 | ); |
| 870 | |
| 871 | return $sigRange; |
| 872 | } |
| 873 | |
| 874 | private function buildThreadItems(): ContentThreadItemSet { |
| 875 | $result = new ContentThreadItemSet(); |
| 876 | |
| 877 | $timestampRegexps = $this->getLocalTimestampRegexps(); |
| 878 | $dfParsers = $this->getLocalTimestampParsers(); |
| 879 | |
| 880 | $curCommentEnd = null; |
| 881 | |
| 882 | $treeWalker = new TreeWalker( |
| 883 | $this->rootNode, |
| 884 | NodeFilter::SHOW_ELEMENT | NodeFilter::SHOW_TEXT, |
| 885 | [ static::class, 'acceptOnlyNodesAllowingComments' ] |
| 886 | ); |
| 887 | while ( $node = $treeWalker->nextNode() ) { |
| 888 | if ( $node instanceof Element && preg_match( '/^h([1-6])$/i', $node->tagName, $match ) ) { |
| 889 | $headingNode = CommentUtils::getHeadlineNode( $node ); |
| 890 | $range = new ImmutableRange( |
| 891 | $headingNode, 0, $headingNode, $headingNode->childNodes->length |
| 892 | ); |
| 893 | $transcludedFrom = $this->computeTranscludedFrom( $range ); |
| 894 | $curComment = new ContentHeadingItem( $range, $transcludedFrom, (int)( $match[ 1 ] ) ); |
| 895 | $curComment->setRootNode( $this->rootNode ); |
| 896 | $result->addThreadItem( $curComment ); |
| 897 | $curCommentEnd = $node; |
| 898 | } elseif ( $node instanceof Text && ( $match = $this->findTimestamp( $node, $timestampRegexps ) ) ) { |
| 899 | $warnings = []; |
| 900 | $foundSignature = $this->findSignature( $node, $curCommentEnd ); |
| 901 | $author = $foundSignature['username']; |
| 902 | |
| 903 | if ( $author === null ) { |
| 904 | // Ignore timestamps for which we couldn't find a signature. It's probably not a real |
| 905 | // comment, but just a false match due to a copypasted timestamp. |
| 906 | continue; |
| 907 | } |
| 908 | |
| 909 | $sigRanges = []; |
| 910 | $timestampRanges = []; |
| 911 | |
| 912 | $sigRanges[] = $this->adjustSigRange( $foundSignature['nodes'], $match, $node ); |
| 913 | $timestampRanges[] = $match['range']; |
| 914 | |
| 915 | // Everything from the last comment up to here is the next comment |
| 916 | $startNode = $this->nextInterestingLeafNode( $curCommentEnd ); |
| 917 | $endNode = $foundSignature['nodes'][0]; |
| 918 | |
| 919 | // Skip to the end of the "paragraph". This only looks at tag names and can be fooled by CSS, but |
| 920 | // avoiding that would be more difficult and slower. |
| 921 | // |
| 922 | // If this skips over another potential signature, also skip it in the main TreeWalker loop, to |
| 923 | // avoid generating multiple comments when there is more than one signature on a single "line". |
| 924 | // Often this is done when someone edits their comment later and wants to add a note about that. |
| 925 | // (Or when another person corrects a typo, or strikes out a comment, etc.) Multiple comments |
| 926 | // within one paragraph/list-item result in a confusing double "Reply" button, and we also have |
| 927 | // no way to indicate which one you're replying to (this might matter in the future for |
| 928 | // notifications or something). |
| 929 | CommentUtils::linearWalk( |
| 930 | $endNode, |
| 931 | function ( string $event, Node $n ) use ( |
| 932 | &$endNode, &$sigRanges, &$timestampRanges, |
| 933 | $treeWalker, $timestampRegexps, $node |
| 934 | ) { |
| 935 | if ( CommentUtils::isBlockElement( $n ) || CommentUtils::isCommentSeparator( $n ) ) { |
| 936 | // Stop when entering or leaving a block node |
| 937 | return true; |
| 938 | } |
| 939 | if ( |
| 940 | $event === 'leave' && |
| 941 | $n instanceof Text && $n !== $node && |
| 942 | ( $match2 = $this->findTimestamp( $n, $timestampRegexps ) ) |
| 943 | ) { |
| 944 | // If this skips over another potential signature, also skip it in the main TreeWalker loop |
| 945 | $treeWalker->currentNode = $n; |
| 946 | // …and add it as another signature to this comment (regardless of the author and timestamp) |
| 947 | $foundSignature2 = $this->findSignature( $n, $node ); |
| 948 | if ( $foundSignature2['username'] !== null ) { |
| 949 | $sigRanges[] = $this->adjustSigRange( $foundSignature2['nodes'], $match2, $n ); |
| 950 | $timestampRanges[] = $match2['range']; |
| 951 | } |
| 952 | } |
| 953 | if ( $event === 'leave' ) { |
| 954 | // Take the last complete node which we skipped past |
| 955 | $endNode = $n; |
| 956 | } |
| 957 | } |
| 958 | ); |
| 959 | |
| 960 | $length = ( $endNode instanceof Text ) ? |
| 961 | mb_strlen( rtrim( $endNode->nodeValue ?? '', "\t\n\f\r " ) ) : |
| 962 | // PHP bug: childNodes can be null for comment nodes |
| 963 | // (it should always be a NodeList, even if the node can't have children) |
| 964 | ( $endNode->childNodes ? $endNode->childNodes->length : 0 ); |
| 965 | $range = new ImmutableRange( |
| 966 | $startNode->parentNode, |
| 967 | CommentUtils::childIndexOf( $startNode ), |
| 968 | $endNode, |
| 969 | $length |
| 970 | ); |
| 971 | $transcludedFrom = $this->computeTranscludedFrom( $range ); |
| 972 | |
| 973 | $startLevel = CommentUtils::getIndentLevel( $startNode, $this->rootNode ) + 1; |
| 974 | $endLevel = CommentUtils::getIndentLevel( $node, $this->rootNode ) + 1; |
| 975 | if ( $startLevel !== $endLevel ) { |
| 976 | $warnings[] = 'Comment starts and ends with different indentation'; |
| 977 | } |
| 978 | // Should this use the indent level of $startNode or $node? |
| 979 | $level = min( $startLevel, $endLevel ); |
| 980 | |
| 981 | $parserResult = $dfParsers[ $match['parserIndex'] ]( $match['matchData'] ); |
| 982 | if ( !$parserResult ) { |
| 983 | continue; |
| 984 | } |
| 985 | [ 'date' => $dateTime, 'warning' => $dateWarning ] = $parserResult; |
| 986 | |
| 987 | if ( $dateWarning ) { |
| 988 | $warnings[] = $dateWarning; |
| 989 | } |
| 990 | |
| 991 | $curComment = new ContentCommentItem( |
| 992 | $level, |
| 993 | $range, |
| 994 | $transcludedFrom, |
| 995 | $sigRanges, |
| 996 | $timestampRanges, |
| 997 | $dateTime, |
| 998 | $author, |
| 999 | $foundSignature['displayName'] |
| 1000 | ); |
| 1001 | $curComment->setRootNode( $this->rootNode ); |
| 1002 | if ( $warnings ) { |
| 1003 | $curComment->addWarnings( $warnings ); |
| 1004 | } |
| 1005 | if ( $result->isEmpty() ) { |
| 1006 | // Add a fake placeholder heading if there are any comments in the 0th section |
| 1007 | // (before the first real heading) |
| 1008 | $range = new ImmutableRange( $this->rootNode, 0, $this->rootNode, 0 ); |
| 1009 | $fakeHeading = new ContentHeadingItem( $range, false, null ); |
| 1010 | $fakeHeading->setRootNode( $this->rootNode ); |
| 1011 | $result->addThreadItem( $fakeHeading ); |
| 1012 | } |
| 1013 | $result->addThreadItem( $curComment ); |
| 1014 | $curCommentEnd = $curComment->getRange()->endContainer; |
| 1015 | } |
| 1016 | } |
| 1017 | |
| 1018 | return $result; |
| 1019 | } |
| 1020 | |
| 1021 | /** |
| 1022 | * Get the name of the page from which this thread item is transcluded (if any). Replies to |
| 1023 | * transcluded items must be posted on that page, instead of the current one. |
| 1024 | * |
| 1025 | * This is tricky, because we don't want to mark items as trancluded when they're just using a |
| 1026 | * template (e.g. {{ping|…}} or a non-substituted signature template). Sometimes the whole comment |
| 1027 | * can be template-generated (e.g. when using some wrapper templates), but as long as a reply can |
| 1028 | * be added outside of that template, we should not treat it as transcluded. |
| 1029 | * |
| 1030 | * The start/end boundary points of comment ranges and Parsoid transclusion ranges don't line up |
| 1031 | * exactly, even when to a human it's obvious that they cover the same content, making this more |
| 1032 | * complicated. |
| 1033 | * |
| 1034 | * @return string|bool `false` if this item is not transcluded. A string if it's transcluded |
| 1035 | * from a single page (the page title, in text form with spaces). `true` if it's transcluded, but |
| 1036 | * we can't determine the source. |
| 1037 | */ |
| 1038 | public function computeTranscludedFrom( ImmutableRange $commentRange ) { |
| 1039 | // Collapsed ranges should otherwise be impossible, but they're not (T299583) |
| 1040 | // TODO: See if we can fix the root cause, and remove this? |
| 1041 | if ( $commentRange->collapsed ) { |
| 1042 | return false; |
| 1043 | } |
| 1044 | |
| 1045 | // General approach: |
| 1046 | // |
| 1047 | // Compare the comment range to each transclusion range on the page, and if it overlaps any of |
| 1048 | // them, examine the overlap. There are a few cases: |
| 1049 | // |
| 1050 | // * Comment and transclusion do not overlap: |
| 1051 | // → Not transcluded. |
| 1052 | // * Comment contains the transclusion: |
| 1053 | // → Not transcluded (just a template). |
| 1054 | // * Comment is contained within the transclusion: |
| 1055 | // → Transcluded, we can determine the source page (unless it's a complex transclusion). |
| 1056 | // * Comment and transclusion overlap partially: |
| 1057 | // → Transcluded, but we can't determine the source page. |
| 1058 | // * Comment (almost) exactly matches the transclusion: |
| 1059 | // → Maybe transcluded (it could be that the source page only contains that single comment), |
| 1060 | // maybe not transcluded (it could be a wrapper template that covers a single comment). |
| 1061 | // This is very sad, and we decide based on the namespace. |
| 1062 | // |
| 1063 | // Most transclusion ranges on the page trivially fall in the "do not overlap" or "contains" |
| 1064 | // cases, and we only have to carefully examine the two transclusion ranges that contain the |
| 1065 | // first and last node of the comment range. |
| 1066 | // |
| 1067 | // To check for almost exact matches, we walk between the relevant boundary points, and if we |
| 1068 | // only find uninteresting nodes (that would be ignored when detecting comments), we treat them |
| 1069 | // like exact matches. |
| 1070 | |
| 1071 | $startTransclNode = CommentUtils::getTranscludedFromElement( |
| 1072 | CommentUtils::getRangeFirstNode( $commentRange ) |
| 1073 | ); |
| 1074 | $endTransclNode = CommentUtils::getTranscludedFromElement( |
| 1075 | CommentUtils::getRangeLastNode( $commentRange ) |
| 1076 | ); |
| 1077 | |
| 1078 | // We only have to examine the two transclusion ranges that contain the first/last node of the |
| 1079 | // comment range (if they exist). Ignore ranges outside the comment or in the middle of it. |
| 1080 | $transclNodes = []; |
| 1081 | if ( $startTransclNode ) { |
| 1082 | $transclNodes[] = $startTransclNode; |
| 1083 | } |
| 1084 | if ( $endTransclNode && $endTransclNode !== $startTransclNode ) { |
| 1085 | $transclNodes[] = $endTransclNode; |
| 1086 | } |
| 1087 | |
| 1088 | foreach ( $transclNodes as $transclNode ) { |
| 1089 | $transclRange = static::getTransclusionRange( $transclNode ); |
| 1090 | $compared = CommentUtils::compareRanges( $commentRange, $transclRange ); |
| 1091 | $transclTitles = $this->getTransclusionTitles( $transclNode ); |
| 1092 | $simpleTransclTitle = count( $transclTitles ) === 1 && $transclTitles[0] !== null ? |
| 1093 | $this->parseTitle( $transclTitles[0] ) : null; |
| 1094 | |
| 1095 | switch ( $compared ) { |
| 1096 | case 'equal': |
| 1097 | // Comment (almost) exactly matches the transclusion |
| 1098 | if ( $simpleTransclTitle === null ) { |
| 1099 | // Allow replying to some accidental complex transclusions consisting of only templates |
| 1100 | // and wikitext (T313093) |
| 1101 | if ( count( $transclTitles ) > 1 ) { |
| 1102 | foreach ( $transclTitles as $transclTitleString ) { |
| 1103 | if ( $transclTitleString !== null ) { |
| 1104 | $transclTitle = $this->parseTitle( $transclTitleString ); |
| 1105 | if ( $transclTitle && !$transclTitle->inNamespace( NS_TEMPLATE ) ) { |
| 1106 | return true; |
| 1107 | } |
| 1108 | } |
| 1109 | } |
| 1110 | // Continue examining the other ranges. |
| 1111 | break; |
| 1112 | } |
| 1113 | // Multi-template transclusion, or a parser function call, or template-affected wikitext outside |
| 1114 | // of a template call, or a mix of the above |
| 1115 | return true; |
| 1116 | |
| 1117 | } elseif ( $simpleTransclTitle->inNamespace( NS_TEMPLATE ) ) { |
| 1118 | // Is that a subpage transclusion with a single comment, or a wrapper template |
| 1119 | // transclusion on this page? We don't know, but let's guess based on the namespace. |
| 1120 | // (T289873) |
| 1121 | // Continue examining the other ranges. |
| 1122 | break; |
| 1123 | } elseif ( !$this->titleCanExist( $simpleTransclTitle ) ) { |
| 1124 | // Special page transclusion (T344622) or something else weird. Don't return the title, |
| 1125 | // since it's useless for replying, and can't be stored in the permalink database. |
| 1126 | return true; |
| 1127 | } else { |
| 1128 | Assert::precondition( $transclTitles[0] !== null, "Simple transclusion found" ); |
| 1129 | return strtr( $transclTitles[0], '_', ' ' ); |
| 1130 | } |
| 1131 | |
| 1132 | case 'contains': |
| 1133 | // Comment contains the transclusion |
| 1134 | |
| 1135 | // If the entire transclusion is contained within the comment range, that's just a |
| 1136 | // template. This is the same as a transclusion in the middle of the comment, which we |
| 1137 | // ignored earlier, it just takes us longer to get here in this case. |
| 1138 | |
| 1139 | // Continue examining the other ranges. |
| 1140 | break; |
| 1141 | |
| 1142 | case 'contained': |
| 1143 | // Comment is contained within the transclusion |
| 1144 | if ( $simpleTransclTitle === null ) { |
| 1145 | return true; |
| 1146 | } elseif ( !$this->titleCanExist( $simpleTransclTitle ) ) { |
| 1147 | // Special page transclusion (T344622) or something else weird. Don't return the title, |
| 1148 | // since it's useless for replying, and can't be stored in the permalink database. |
| 1149 | return true; |
| 1150 | } else { |
| 1151 | Assert::precondition( $transclTitles[0] !== null, "Simple transclusion found" ); |
| 1152 | return strtr( $transclTitles[0], '_', ' ' ); |
| 1153 | } |
| 1154 | |
| 1155 | case 'after': |
| 1156 | case 'before': |
| 1157 | // Comment and transclusion do not overlap |
| 1158 | |
| 1159 | // This should be impossible, because we ignored these ranges earlier. |
| 1160 | throw new LogicException( 'Unexpected transclusion or comment range' ); |
| 1161 | |
| 1162 | case 'overlapstart': |
| 1163 | case 'overlapend': |
| 1164 | // Comment and transclusion overlap partially |
| 1165 | return true; |
| 1166 | |
| 1167 | default: |
| 1168 | throw new LogicException( 'Unexpected return value from compareRanges()' ); |
| 1169 | } |
| 1170 | } |
| 1171 | |
| 1172 | // If we got here, the comment range was not contained by or overlapping any of the transclusion |
| 1173 | // ranges. Comment is not transcluded. |
| 1174 | return false; |
| 1175 | } |
| 1176 | |
| 1177 | private function titleCanExist( TitleValue $title ): bool { |
| 1178 | return $title->getNamespace() >= NS_MAIN && |
| 1179 | !$title->isExternal() && |
| 1180 | $title->getText() !== ''; |
| 1181 | } |
| 1182 | |
| 1183 | private function parseTitle( string $titleString ): ?TitleValue { |
| 1184 | try { |
| 1185 | return $this->titleParser->parseTitle( $titleString ); |
| 1186 | } catch ( MalformedTitleException ) { |
| 1187 | return null; |
| 1188 | } |
| 1189 | } |
| 1190 | |
| 1191 | /** |
| 1192 | * Return the page titles for each part of the transclusion, or nulls for each part that isn't |
| 1193 | * transcluded from another page. |
| 1194 | * |
| 1195 | * If the node represents a single-page transclusion, this will return an array containing a |
| 1196 | * single string. |
| 1197 | * |
| 1198 | * @return array<string|null> |
| 1199 | */ |
| 1200 | private function getTransclusionTitles( Element $node ): array { |
| 1201 | $dataMw = json_decode( $node->getAttribute( 'data-mw' ) ?? '', true ); |
| 1202 | $out = []; |
| 1203 | |
| 1204 | foreach ( $dataMw['parts'] ?? [] as $part ) { |
| 1205 | if ( |
| 1206 | !is_string( $part ) && |
| 1207 | // 'href' will be unset if this is a parser function rather than a template |
| 1208 | isset( $part['template']['target']['href'] ) |
| 1209 | ) { |
| 1210 | $parsoidHref = $part['template']['target']['href']; |
| 1211 | Assert::precondition( str_starts_with( $parsoidHref, './' ), 'href has valid format' ); |
| 1212 | $out[] = rawurldecode( substr( $parsoidHref, 2 ) ); |
| 1213 | } else { |
| 1214 | $out[] = null; |
| 1215 | } |
| 1216 | } |
| 1217 | |
| 1218 | return $out; |
| 1219 | } |
| 1220 | |
| 1221 | /** |
| 1222 | * Given a transclusion's first node (e.g. returned by CommentUtils::getTranscludedFromElement()), |
| 1223 | * return a range starting before the node and ending after the transclusion's last node. |
| 1224 | */ |
| 1225 | private function getTransclusionRange( Element $startNode ): ImmutableRange { |
| 1226 | $endNode = $startNode; |
| 1227 | while ( |
| 1228 | // Phan doesn't realize that the conditions on $nextSibling can terminate the loop |
| 1229 | // @phan-suppress-next-line PhanInfiniteLoop |
| 1230 | $endNode && |
| 1231 | ( $nextSibling = $endNode->nextSibling ) && |
| 1232 | $nextSibling instanceof Element && |
| 1233 | $nextSibling->getAttribute( 'about' ) === $endNode->getAttribute( 'about' ) |
| 1234 | ) { |
| 1235 | $endNode = $nextSibling; |
| 1236 | } |
| 1237 | |
| 1238 | $range = new ImmutableRange( |
| 1239 | $startNode->parentNode, |
| 1240 | CommentUtils::childIndexOf( $startNode ), |
| 1241 | $endNode->parentNode, |
| 1242 | CommentUtils::childIndexOf( $endNode ) + 1 |
| 1243 | ); |
| 1244 | |
| 1245 | return $range; |
| 1246 | } |
| 1247 | |
| 1248 | /** |
| 1249 | * Truncate user generated parts of IDs so full ID always fits within a database field of length 255 |
| 1250 | * |
| 1251 | * nb: Text should already have had spaces replaced with underscores by this point. |
| 1252 | * |
| 1253 | * @param string $text Text |
| 1254 | * @param bool $legacy Generate legacy ID, not needed in JS implementation |
| 1255 | * @return string Truncated text |
| 1256 | */ |
| 1257 | private function truncateForId( string $text, bool $legacy = false ): string { |
| 1258 | $truncated = $this->language->truncateForDatabase( $text, 80, '' ); |
| 1259 | if ( !$legacy ) { |
| 1260 | $truncated = trim( $truncated, '_' ); |
| 1261 | } |
| 1262 | return $truncated; |
| 1263 | } |
| 1264 | |
| 1265 | /** |
| 1266 | * Given a thread item, return an identifier for it that is unique within the page. |
| 1267 | * |
| 1268 | * @param ContentThreadItem $threadItem |
| 1269 | * @param ContentThreadItemSet $previousItems |
| 1270 | * @param bool $legacy Generate legacy ID, not needed in JS implementation |
| 1271 | */ |
| 1272 | private function computeId( |
| 1273 | ContentThreadItem $threadItem, ContentThreadItemSet $previousItems, bool $legacy = false |
| 1274 | ): string { |
| 1275 | $id = null; |
| 1276 | |
| 1277 | if ( $threadItem instanceof ContentHeadingItem && $threadItem->isPlaceholderHeading() ) { |
| 1278 | // The range points to the root note, using it like below results in silly values |
| 1279 | $id = 'h-'; |
| 1280 | } elseif ( $threadItem instanceof ContentHeadingItem ) { |
| 1281 | $id = 'h-' . $this->truncateForId( $threadItem->getLinkableId(), $legacy ); |
| 1282 | } elseif ( $threadItem instanceof ContentCommentItem ) { |
| 1283 | $id = 'c-' . $this->truncateForId( str_replace( ' ', '_', $threadItem->getAuthor() ), $legacy ) . |
| 1284 | '-' . $threadItem->getTimestampString(); |
| 1285 | } else { |
| 1286 | throw new InvalidArgumentException( 'Unknown ThreadItem type' ); |
| 1287 | } |
| 1288 | |
| 1289 | // If there would be multiple comments with the same ID (i.e. the user left multiple comments |
| 1290 | // in one edit, or within a minute), add the parent ID to disambiguate them. |
| 1291 | $threadItemParent = $threadItem->getParent(); |
| 1292 | if ( $threadItemParent instanceof ContentHeadingItem && !$threadItemParent->isPlaceholderHeading() ) { |
| 1293 | $id .= '-' . $this->truncateForId( $threadItemParent->getLinkableId(), $legacy ); |
| 1294 | } elseif ( $threadItemParent instanceof ContentCommentItem ) { |
| 1295 | $id .= '-' . $this->truncateForId( str_replace( ' ', '_', $threadItemParent->getAuthor() ), $legacy ) . |
| 1296 | '-' . $threadItemParent->getTimestampString(); |
| 1297 | } |
| 1298 | |
| 1299 | if ( $threadItem instanceof ContentHeadingItem ) { |
| 1300 | // To avoid old threads re-appearing on popular pages when someone uses a vague title |
| 1301 | // (e.g. dozens of threads titled "question" on [[Wikipedia:Help desk]]: https://w.wiki/fbN), |
| 1302 | // include the oldest timestamp in the thread (i.e. date the thread was started) in the |
| 1303 | // heading ID. |
| 1304 | $oldestComment = $threadItem->getOldestReply(); |
| 1305 | if ( $oldestComment ) { |
| 1306 | $id .= '-' . $oldestComment->getTimestampString(); |
| 1307 | } |
| 1308 | } |
| 1309 | |
| 1310 | if ( $previousItems->findCommentById( $id ) ) { |
| 1311 | // Well, that's tough |
| 1312 | if ( !$legacy ) { |
| 1313 | $threadItem->addWarning( 'Duplicate comment ID' ); |
| 1314 | } |
| 1315 | // Finally, disambiguate by adding sequential numbers, to allow replying to both comments |
| 1316 | $number = 1; |
| 1317 | while ( $previousItems->findCommentById( "$id-$number" ) ) { |
| 1318 | $number++; |
| 1319 | } |
| 1320 | $id = "$id-$number"; |
| 1321 | } |
| 1322 | |
| 1323 | return $id; |
| 1324 | } |
| 1325 | |
| 1326 | /** |
| 1327 | * Given a thread item, return an identifier for it that is consistent across all pages and |
| 1328 | * revisions where this comment might appear. |
| 1329 | * |
| 1330 | * Multiple comments on a page can have the same name; use ID to distinguish them. |
| 1331 | */ |
| 1332 | private function computeName( ContentThreadItem $threadItem ): string { |
| 1333 | $name = null; |
| 1334 | |
| 1335 | if ( $threadItem instanceof ContentHeadingItem ) { |
| 1336 | $name = 'h-'; |
| 1337 | $mainComment = $threadItem->getOldestReply(); |
| 1338 | } elseif ( $threadItem instanceof ContentCommentItem ) { |
| 1339 | $name = 'c-'; |
| 1340 | $mainComment = $threadItem; |
| 1341 | } else { |
| 1342 | throw new InvalidArgumentException( 'Unknown ThreadItem type' ); |
| 1343 | } |
| 1344 | |
| 1345 | if ( $mainComment ) { |
| 1346 | $name .= $this->truncateForId( str_replace( ' ', '_', $mainComment->getAuthor() ) ) . |
| 1347 | '-' . $mainComment->getTimestampString(); |
| 1348 | } |
| 1349 | |
| 1350 | return $name; |
| 1351 | } |
| 1352 | |
| 1353 | private function buildThreads( ContentThreadItemSet $result ): void { |
| 1354 | $lastHeading = null; |
| 1355 | $replies = []; |
| 1356 | |
| 1357 | foreach ( $result->getThreadItems() as $threadItem ) { |
| 1358 | if ( count( $replies ) < $threadItem->getLevel() ) { |
| 1359 | // Someone skipped an indentation level (or several). Pretend that the previous reply |
| 1360 | // covers multiple indentation levels, so that following comments get connected to it. |
| 1361 | $threadItem->addWarning( 'Comment skips indentation level' ); |
| 1362 | while ( count( $replies ) < $threadItem->getLevel() ) { |
| 1363 | $replies[] = end( $replies ); |
| 1364 | } |
| 1365 | } |
| 1366 | |
| 1367 | if ( $threadItem instanceof ContentHeadingItem ) { |
| 1368 | // New root (thread) |
| 1369 | // Attach as a sub-thread to preceding higher-level heading. |
| 1370 | // Any replies will appear in the tree twice, under the main-thread and the sub-thread. |
| 1371 | $maybeParent = $lastHeading; |
| 1372 | while ( $maybeParent && $maybeParent->getHeadingLevel() >= $threadItem->getHeadingLevel() ) { |
| 1373 | $maybeParent = $maybeParent->getParent(); |
| 1374 | } |
| 1375 | if ( $maybeParent ) { |
| 1376 | $threadItem->setParent( $maybeParent ); |
| 1377 | $maybeParent->addReply( $threadItem ); |
| 1378 | } |
| 1379 | $lastHeading = $threadItem; |
| 1380 | } elseif ( isset( $replies[ $threadItem->getLevel() - 1 ] ) ) { |
| 1381 | // Add as a reply to the closest less-nested comment |
| 1382 | $threadItem->setParent( $replies[ $threadItem->getLevel() - 1 ] ); |
| 1383 | $threadItem->getParent()->addReply( $threadItem ); |
| 1384 | } else { |
| 1385 | $threadItem->addWarning( 'Comment could not be connected to a thread' ); |
| 1386 | } |
| 1387 | |
| 1388 | $replies[ $threadItem->getLevel() ] = $threadItem; |
| 1389 | // Cut off more deeply nested replies |
| 1390 | array_splice( $replies, $threadItem->getLevel() + 1 ); |
| 1391 | } |
| 1392 | } |
| 1393 | |
| 1394 | /** |
| 1395 | * Set the IDs and names used to refer to comments and headings. |
| 1396 | * This has to be a separate pass because we don't have the list of replies before |
| 1397 | * this point. |
| 1398 | */ |
| 1399 | private function computeIdsAndNames( ContentThreadItemSet $result ): void { |
| 1400 | foreach ( $result->getThreadItems() as $threadItem ) { |
| 1401 | $name = $this->computeName( $threadItem ); |
| 1402 | $threadItem->setName( $name ); |
| 1403 | |
| 1404 | $id = $this->computeId( $threadItem, $result ); |
| 1405 | $threadItem->setId( $id ); |
| 1406 | $legacyId = $this->computeId( $threadItem, $result, true ); |
| 1407 | if ( $legacyId !== $id ) { |
| 1408 | $threadItem->setLegacyId( $legacyId ); |
| 1409 | } |
| 1410 | |
| 1411 | $result->updateIdAndNameMaps( $threadItem ); |
| 1412 | } |
| 1413 | } |
| 1414 | } |