3 * A PHP diff engine for phpwiki. (Taken from phpwiki-1.3.3)
5 * Copyright © 2000, 2001 Geoffrey T. Dairiki <dairiki@dairiki.org>
6 * You may copy this code freely under the conditions of the GPL.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License along
19 * with this program; if not, write to the Free Software Foundation, Inc.,
20 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
21 * http://www.gnu.org/copyleft/gpl.html
24 * @ingroup DifferenceEngine
25 * @defgroup DifferenceEngine DifferenceEngine
31 * @ingroup DifferenceEngine
33 abstract class DiffOp
{
39 public function getType() {
43 public function getOrig() {
47 public function getClosing( $i = null ) {
49 return $this->closing
;
51 if( array_key_exists( $i, $this->closing
) ) {
52 return $this->closing
[$i];
57 abstract public function reverse();
62 public function norig() {
63 return $this->orig ?
count( $this->orig
) : 0;
69 public function nclosing() {
70 return $this->closing ?
count( $this->closing
) : 0;
77 * @ingroup DifferenceEngine
79 class DiffOpCopy
extends DiffOp
{
80 public $type = 'copy';
82 public function __construct( $orig, $closing = false ) {
83 if ( !is_array( $closing ) ) {
87 $this->closing
= $closing;
93 public function reverse() {
94 return new DiffOpCopy( $this->closing
, $this->orig
);
101 * @ingroup DifferenceEngine
103 class DiffOpDelete
extends DiffOp
{
104 public $type = 'delete';
106 public function __construct( $lines ) {
107 $this->orig
= $lines;
108 $this->closing
= false;
114 public function reverse() {
115 return new DiffOpAdd( $this->orig
);
122 * @ingroup DifferenceEngine
124 class DiffOpAdd
extends DiffOp
{
125 public $type = 'add';
127 public function __construct( $lines ) {
128 $this->closing
= $lines;
133 * @return DiffOpDelete
135 public function reverse() {
136 return new DiffOpDelete( $this->closing
);
143 * @ingroup DifferenceEngine
145 class DiffOpChange
extends DiffOp
{
146 public $type = 'change';
148 public function __construct( $orig, $closing ) {
150 $this->closing
= $closing;
154 * @return DiffOpChange
156 public function reverse() {
157 return new DiffOpChange( $this->closing
, $this->orig
);
162 * Class used internally by Diff to actually compute the diffs.
164 * The algorithm used here is mostly lifted from the perl module
165 * Algorithm::Diff (version 1.06) by Ned Konz, which is available at:
166 * http://www.perl.com/CPAN/authors/id/N/NE/NEDKONZ/Algorithm-Diff-1.06.zip
168 * More ideas are taken from:
169 * http://www.ics.uci.edu/~eppstein/161/960229.html
171 * Some ideas are (and a bit of code) are from from analyze.c, from GNU
172 * diffutils-2.7, which can be found at:
173 * ftp://gnudist.gnu.org/pub/gnu/diffutils/diffutils-2.7.tar.gz
175 * closingly, some ideas (subdivision by NCHUNKS > 2, and some optimizations)
178 * Line length limits for robustness added by Tim Starling, 2005-08-31
179 * Alternative implementation added by Guy Van den Broeck, 2008-07-30
181 * @author Geoffrey T. Dairiki, Tim Starling, Guy Van den Broeck
183 * @ingroup DifferenceEngine
186 const MAX_XREF_LENGTH
= 10000;
188 protected $xchanged, $ychanged;
190 protected $xv = array(), $yv = array();
191 protected $xind = array(), $yind = array();
193 protected $seq = array(), $in_seq = array();
202 public function diff( $from_lines, $to_lines ) {
203 wfProfileIn( __METHOD__
);
205 // Diff and store locally
206 $this->diffLocal( $from_lines, $to_lines );
208 // Merge edits when possible
209 $this->shiftBoundaries( $from_lines, $this->xchanged
, $this->ychanged
);
210 $this->shiftBoundaries( $to_lines, $this->ychanged
, $this->xchanged
);
212 // Compute the edit operations.
213 $n_from = count( $from_lines );
214 $n_to = count( $to_lines );
218 while ( $xi < $n_from ||
$yi < $n_to ) {
219 assert( '$yi < $n_to || $this->xchanged[$xi]' );
220 assert( '$xi < $n_from || $this->ychanged[$yi]' );
222 // Skip matching "snake".
224 while ( $xi < $n_from && $yi < $n_to
225 && !$this->xchanged
[$xi] && !$this->ychanged
[$yi]
227 $copy[] = $from_lines[$xi++
];
231 $edits[] = new DiffOpCopy( $copy );
234 // Find deletes & adds.
236 while ( $xi < $n_from && $this->xchanged
[$xi] ) {
237 $delete[] = $from_lines[$xi++
];
241 while ( $yi < $n_to && $this->ychanged
[$yi] ) {
242 $add[] = $to_lines[$yi++
];
245 if ( $delete && $add ) {
246 $edits[] = new DiffOpChange( $delete, $add );
247 } elseif ( $delete ) {
248 $edits[] = new DiffOpDelete( $delete );
250 $edits[] = new DiffOpAdd( $add );
253 wfProfileOut( __METHOD__
);
262 private function diffLocal( $from_lines, $to_lines ) {
263 global $wgExternalDiffEngine;
264 wfProfileIn( __METHOD__
);
266 if ( $wgExternalDiffEngine == 'wikidiff3' ) {
268 $wikidiff3 = new WikiDiff3();
269 $wikidiff3->diff( $from_lines, $to_lines );
270 $this->xchanged
= $wikidiff3->removed
;
271 $this->ychanged
= $wikidiff3->added
;
275 $n_from = count( $from_lines );
276 $n_to = count( $to_lines );
277 $this->xchanged
= $this->ychanged
= array();
278 $this->xv
= $this->yv
= array();
279 $this->xind
= $this->yind
= array();
281 unset( $this->in_seq
);
284 // Skip leading common lines.
285 for ( $skip = 0; $skip < $n_from && $skip < $n_to; $skip++
) {
286 if ( $from_lines[$skip] !== $to_lines[$skip] ) {
289 $this->xchanged
[$skip] = $this->ychanged
[$skip] = false;
291 // Skip trailing common lines.
294 for ( $endskip = 0; --$xi > $skip && --$yi > $skip; $endskip++
) {
295 if ( $from_lines[$xi] !== $to_lines[$yi] ) {
298 $this->xchanged
[$xi] = $this->ychanged
[$yi] = false;
301 // Ignore lines which do not exist in both files.
302 for ( $xi = $skip; $xi < $n_from - $endskip; $xi++
) {
303 $xhash[$this->lineHash( $from_lines[$xi] )] = 1;
306 for ( $yi = $skip; $yi < $n_to - $endskip; $yi++
) {
307 $line = $to_lines[$yi];
308 if ( ( $this->ychanged
[$yi] = empty( $xhash[$this->lineHash( $line )] ) ) ) {
311 $yhash[$this->lineHash( $line )] = 1;
315 for ( $xi = $skip; $xi < $n_from - $endskip; $xi++
) {
316 $line = $from_lines[$xi];
317 if ( ( $this->xchanged
[$xi] = empty( $yhash[$this->lineHash( $line )] ) ) ) {
325 $this->compareSeq( 0, count( $this->xv
), 0, count( $this->yv
) );
327 wfProfileOut( __METHOD__
);
331 * Returns the whole line if it's small enough, or the MD5 hash otherwise
332 * @param $line string
335 private function lineHash( $line ) {
336 if ( strlen( $line ) > self
::MAX_XREF_LENGTH
) {
344 * Divide the Largest Common Subsequence (LCS) of the sequences
345 * [XOFF, XLIM) and [YOFF, YLIM) into NCHUNKS approximately equally
348 * Returns (LCS, PTS). LCS is the length of the LCS. PTS is an
349 * array of NCHUNKS+1 (X, Y) indexes giving the diving points between
350 * sub sequences. The first sub-sequence is contained in [X0, X1),
351 * [Y0, Y1), the second in [X1, X2), [Y1, Y2) and so on. Note
352 * that (X0, Y0) == (XOFF, YOFF) and
353 * (X[NCHUNKS], Y[NCHUNKS]) == (XLIM, YLIM).
355 * This function assumes that the first lines of the specified portions
356 * of the two files do not match, and likewise that the last lines do not
357 * match. The caller must trim matching lines from the beginning and end
358 * of the portions it is going to specify.
366 private function diag( $xoff, $xlim, $yoff, $ylim, $nchunks ) {
369 if ( $xlim - $xoff > $ylim - $yoff ) {
370 // Things seems faster (I'm not sure I understand why)
371 // when the shortest sequence in X.
373 list( $xoff, $xlim, $yoff, $ylim ) = array( $yoff, $ylim, $xoff, $xlim );
377 for ( $i = $ylim - 1; $i >= $yoff; $i-- ) {
378 $ymatches[$this->xv
[$i]][] = $i;
381 for ( $i = $ylim - 1; $i >= $yoff; $i-- ) {
382 $ymatches[$this->yv
[$i]][] = $i;
387 $this->seq
[0] = $yoff - 1;
388 $this->in_seq
= array();
391 $numer = $xlim - $xoff +
$nchunks - 1;
393 for ( $chunk = 0; $chunk < $nchunks; $chunk++
) {
395 for ( $i = 0; $i <= $this->lcs
; $i++
) {
396 $ymids[$i][$chunk - 1] = $this->seq
[$i];
400 $x1 = $xoff +
(int)( ( $numer +
( $xlim - $xoff ) * $chunk ) / $nchunks );
401 for ( ; $x < $x1; $x++
) {
402 $line = $flip ?
$this->yv
[$x] : $this->xv
[$x];
403 if ( empty( $ymatches[$line] ) ) {
408 $matches = $ymatches[$line];
410 while ( list( , $y ) = each( $matches ) ) {
411 if ( empty( $this->in_seq
[$y] ) ) {
412 $k = $this->lcsPos( $y );
414 $ymids[$k] = $ymids[$k - 1];
419 while ( list( , $y ) = each( $matches ) ) {
420 if ( $y > $this->seq
[$k - 1] ) {
421 assert( '$y < $this->seq[$k]' );
422 // Optimization: this is a common case:
423 // next match is just replacing previous match.
424 $this->in_seq
[$this->seq
[$k]] = false;
426 $this->in_seq
[$y] = 1;
427 } elseif ( empty( $this->in_seq
[$y] ) ) {
428 $k = $this->lcsPos( $y );
430 $ymids[$k] = $ymids[$k - 1];
436 $seps[] = $flip ?
array( $yoff, $xoff ) : array( $xoff, $yoff );
437 $ymid = $ymids[$this->lcs
];
438 for ( $n = 0; $n < $nchunks - 1; $n++
) {
439 $x1 = $xoff +
(int)( ( $numer +
( $xlim - $xoff ) * $n ) / $nchunks );
441 $seps[] = $flip ?
array( $y1, $x1 ) : array( $x1, $y1 );
443 $seps[] = $flip ?
array( $ylim, $xlim ) : array( $xlim, $ylim );
445 return array( $this->lcs
, $seps );
452 private function lcsPos( $ypos ) {
454 if ( $end == 0 ||
$ypos > $this->seq
[$end] ) {
455 $this->seq
[++
$this->lcs
] = $ypos;
456 $this->in_seq
[$ypos] = 1;
462 while ( $beg < $end ) {
463 $mid = (int)( ( $beg +
$end ) / 2 );
464 if ( $ypos > $this->seq
[$mid] ) {
471 assert( '$ypos != $this->seq[$end]' );
473 $this->in_seq
[$this->seq
[$end]] = false;
474 $this->seq
[$end] = $ypos;
475 $this->in_seq
[$ypos] = 1;
481 * Find LCS of two sequences.
483 * The results are recorded in the vectors $this->{x,y}changed[], by
484 * storing a 1 in the element for each line that is an insertion
485 * or deletion (ie. is not in the LCS).
487 * The subsequence of file 0 is [XOFF, XLIM) and likewise for file 1.
489 * Note that XLIM, YLIM are exclusive bounds.
490 * All line numbers are origin-0 and discarded lines are not counted.
496 private function compareSeq( $xoff, $xlim, $yoff, $ylim ) {
497 // Slide down the bottom initial diagonal.
498 while ( $xoff < $xlim && $yoff < $ylim && $this->xv
[$xoff] == $this->yv
[$yoff] ) {
503 // Slide up the top initial diagonal.
504 while ( $xlim > $xoff && $ylim > $yoff
505 && $this->xv
[$xlim - 1] == $this->yv
[$ylim - 1]
511 if ( $xoff == $xlim ||
$yoff == $ylim ) {
514 // This is ad hoc but seems to work well.
515 // $nchunks = sqrt(min($xlim - $xoff, $ylim - $yoff) / 2.5);
516 // $nchunks = max(2,min(8,(int)$nchunks));
517 $nchunks = min( 7, $xlim - $xoff, $ylim - $yoff ) +
1;
518 list( $lcs, $seps ) = $this->diag( $xoff, $xlim, $yoff, $ylim, $nchunks );
522 // X and Y sequences have no common subsequence:
524 while ( $yoff < $ylim ) {
525 $this->ychanged
[$this->yind
[$yoff++
]] = 1;
527 while ( $xoff < $xlim ) {
528 $this->xchanged
[$this->xind
[$xoff++
]] = 1;
531 // Use the partitions to split this problem into subproblems.
534 while ( $pt2 = next( $seps ) ) {
535 $this->compareSeq( $pt1[0], $pt2[0], $pt1[1], $pt2[1] );
542 * Adjust inserts/deletes of identical lines to join changes
543 * as much as possible.
545 * We do something when a run of changed lines include a
546 * line at one end and has an excluded, identical line at the other.
547 * We are free to choose which identical line is included.
548 * `compareseq' usually chooses the one at the beginning,
549 * but usually it is cleaner to consider the following identical line
550 * to be the "change".
552 * This is extracted verbatim from analyze.c (GNU diffutils-2.7).
554 private function shiftBoundaries( $lines, &$changed, $other_changed ) {
555 wfProfileIn( __METHOD__
);
559 assert( 'count($lines) == count($changed)' );
560 $len = count( $lines );
561 $other_len = count( $other_changed );
565 * Scan forwards to find beginning of another run of changes.
566 * Also keep track of the corresponding point in the other file.
568 * Throughout this code, $i and $j are adjusted together so that
569 * the first $i elements of $changed and the first $j elements
570 * of $other_changed both contain the same number of zeros
572 * Furthermore, $j is always kept so that $j == $other_len or
573 * $other_changed[$j] == false.
575 while ( $j < $other_len && $other_changed[$j] ) {
579 while ( $i < $len && !$changed[$i] ) {
580 assert( '$j < $other_len && ! $other_changed[$j]' );
583 while ( $j < $other_len && $other_changed[$j] ) {
594 // Find the end of this run of changes.
595 while ( ++
$i < $len && $changed[$i] ) {
601 * Record the length of this run of changes, so that
602 * we can later determine whether the run has grown.
604 $runlength = $i - $start;
607 * Move the changed region back, so long as the
608 * previous unchanged line matches the last changed one.
609 * This merges with previous changed regions.
611 while ( $start > 0 && $lines[$start - 1] == $lines[$i - 1] ) {
612 $changed[--$start] = 1;
613 $changed[--$i] = false;
614 while ( $start > 0 && $changed[$start - 1] ) {
618 while ( $other_changed[--$j] ) {
621 assert( '$j >= 0 && !$other_changed[$j]' );
625 * Set CORRESPONDING to the end of the changed run, at the last
626 * point where it corresponds to a changed run in the other file.
627 * CORRESPONDING == LEN means no such point has been found.
629 $corresponding = $j < $other_len ?
$i : $len;
632 * Move the changed region forward, so long as the
633 * first changed line matches the following unchanged one.
634 * This merges with following changed regions.
635 * Do this second, so that if there are no merges,
636 * the changed region is moved forward as far as possible.
638 while ( $i < $len && $lines[$start] == $lines[$i] ) {
639 $changed[$start++
] = false;
641 while ( $i < $len && $changed[$i] ) {
645 assert( '$j < $other_len && ! $other_changed[$j]' );
647 if ( $j < $other_len && $other_changed[$j] ) {
649 while ( $j < $other_len && $other_changed[$j] ) {
654 } while ( $runlength != $i - $start );
657 * If possible, move the fully-merged run of changes
658 * back to a corresponding run in the other file.
660 while ( $corresponding < $i ) {
661 $changed[--$start] = 1;
664 while ( $other_changed[--$j] ) {
667 assert( '$j >= 0 && !$other_changed[$j]' );
670 wfProfileOut( __METHOD__
);
675 * Class representing a 'diff' between two sequences of strings.
678 * @ingroup DifferenceEngine
689 * Computes diff between sequences of strings.
691 * @param $from_lines array An array of strings.
692 * Typically these are lines from a file.
693 * @param $to_lines array An array of strings.
695 public function __construct( $from_lines, $to_lines ) {
696 $eng = new DiffEngine
;
697 $this->edits
= $eng->diff( $from_lines, $to_lines );
701 * @return array|DiffOp[]
703 public function getEdits() {
708 * Compute reversed Diff.
712 * $diff = new Diff($lines1, $lines2);
713 * $rev = $diff->reverse();
714 * @return Object A Diff object representing the inverse of the
717 public function reverse() {
719 $rev->edits
= array();
720 /** @var DiffOp $edit */
721 foreach ( $this->edits
as $edit ) {
722 $rev->edits
[] = $edit->reverse();
729 * Check for empty diff.
731 * @return bool True if two sequences were identical.
733 public function isEmpty() {
734 foreach ( $this->edits
as $edit ) {
735 if ( $edit->type
!= 'copy' ) {
744 * Compute the length of the Longest Common Subsequence (LCS).
746 * This is mostly for diagnostic purposed.
748 * @return int The length of the LCS.
750 public function lcs() {
752 foreach ( $this->edits
as $edit ) {
753 if ( $edit->type
== 'copy' ) {
754 $lcs +
= count( $edit->orig
);
762 * Get the original set of lines.
764 * This reconstructs the $from_lines parameter passed to the
767 * @return array The original sequence of strings.
769 public function orig() {
772 foreach ( $this->edits
as $edit ) {
774 array_splice( $lines, count( $lines ), 0, $edit->orig
);
782 * Get the closing set of lines.
784 * This reconstructs the $to_lines parameter passed to the
787 * @return array The sequence of strings.
789 public function closing() {
792 foreach ( $this->edits
as $edit ) {
793 if ( $edit->closing
) {
794 array_splice( $lines, count( $lines ), 0, $edit->closing
);
803 * @todo document, bad name.
805 * @ingroup DifferenceEngine
807 class MappedDiff
extends Diff
{
811 * Computes diff between sequences of strings.
813 * This can be used to compute things like
814 * case-insensitve diffs, or diffs which ignore
815 * changes in white-space.
817 * @param $from_lines array An array of strings.
818 * Typically these are lines from a file.
820 * @param $to_lines array An array of strings.
822 * @param $mapped_from_lines array This array should
823 * have the same size number of elements as $from_lines.
824 * The elements in $mapped_from_lines and
825 * $mapped_to_lines are what is actually compared
826 * when computing the diff.
828 * @param $mapped_to_lines array This array should
829 * have the same number of elements as $to_lines.
831 public function __construct( $from_lines, $to_lines,
832 $mapped_from_lines, $mapped_to_lines ) {
833 wfProfileIn( __METHOD__
);
835 assert( 'count( $from_lines ) == count( $mapped_from_lines )' );
836 assert( 'count( $to_lines ) == count( $mapped_to_lines )' );
838 parent
::__construct( $mapped_from_lines, $mapped_to_lines );
841 $editCount = count( $this->edits
);
842 for ( $i = 0; $i < $editCount; $i++
) {
843 $orig = &$this->edits
[$i]->orig
;
844 if ( is_array( $orig ) ) {
845 $orig = array_slice( $from_lines, $xi, count( $orig ) );
846 $xi +
= count( $orig );
849 $closing = &$this->edits
[$i]->closing
;
850 if ( is_array( $closing ) ) {
851 $closing = array_slice( $to_lines, $yi, count( $closing ) );
852 $yi +
= count( $closing );
855 wfProfileOut( __METHOD__
);
860 * Additions by Axel Boldt follow, partly taken from diff.php, phpwiki-1.3.3
866 * @ingroup DifferenceEngine
868 class HWLDFWordAccumulator
{
869 public $insClass = ' class="diffchange diffchange-inline"';
870 public $delClass = ' class="diffchange diffchange-inline"';
872 private $lines = array();
880 private function flushGroup( $new_tag ) {
881 if ( $this->group
!== '' ) {
882 if ( $this->tag
== 'ins' ) {
883 $this->line
.= "<ins{$this->insClass}>" .
884 htmlspecialchars( $this->group
) . '</ins>';
885 } elseif ( $this->tag
== 'del' ) {
886 $this->line
.= "<del{$this->delClass}>" .
887 htmlspecialchars( $this->group
) . '</del>';
889 $this->line
.= htmlspecialchars( $this->group
);
893 $this->tag
= $new_tag;
899 private function flushLine( $new_tag ) {
900 $this->flushGroup( $new_tag );
901 if ( $this->line
!= '' ) {
902 array_push( $this->lines
, $this->line
);
904 # make empty lines visible by inserting an NBSP
905 array_push( $this->lines
, ' ' );
914 public function addWords( $words, $tag = '' ) {
915 if ( $tag != $this->tag
) {
916 $this->flushGroup( $tag );
919 foreach ( $words as $word ) {
920 // new-line should only come as first char of word.
924 if ( $word[0] == "\n" ) {
925 $this->flushLine( $tag );
926 $word = substr( $word, 1 );
928 assert( '!strstr( $word, "\n" )' );
929 $this->group
.= $word;
936 public function getLines() {
937 $this->flushLine( '~done' );
946 * @ingroup DifferenceEngine
948 class WordLevelDiff
extends MappedDiff
{
949 const MAX_LINE_LENGTH
= 10000;
953 * @param $closing_lines
955 public function __construct( $orig_lines, $closing_lines ) {
956 wfProfileIn( __METHOD__
);
958 list( $orig_words, $orig_stripped ) = $this->split( $orig_lines );
959 list( $closing_words, $closing_stripped ) = $this->split( $closing_lines );
961 parent
::__construct( $orig_words, $closing_words,
962 $orig_stripped, $closing_stripped );
963 wfProfileOut( __METHOD__
);
970 private function split( $lines ) {
971 wfProfileIn( __METHOD__
);
976 foreach ( $lines as $line ) {
977 # If the line is too long, just pretend the entire line is one big word
978 # This prevents resource exhaustion problems
985 if ( strlen( $line ) > self
::MAX_LINE_LENGTH
) {
990 if ( preg_match_all( '/ ( [^\S\n]+ | [0-9_A-Za-z\x80-\xff]+ | . ) (?: (?!< \n) [^\S\n])? /xs',
993 foreach ( $m[0] as $word ) {
996 foreach ( $m[1] as $stripped_word ) {
997 $stripped[] = $stripped_word;
1002 wfProfileOut( __METHOD__
);
1004 return array( $words, $stripped );
1010 public function orig() {
1011 wfProfileIn( __METHOD__
);
1012 $orig = new HWLDFWordAccumulator
;
1014 foreach ( $this->edits
as $edit ) {
1015 if ( $edit->type
== 'copy' ) {
1016 $orig->addWords( $edit->orig
);
1017 } elseif ( $edit->orig
) {
1018 $orig->addWords( $edit->orig
, 'del' );
1021 $lines = $orig->getLines();
1022 wfProfileOut( __METHOD__
);
1030 public function closing() {
1031 wfProfileIn( __METHOD__
);
1032 $closing = new HWLDFWordAccumulator
;
1034 foreach ( $this->edits
as $edit ) {
1035 if ( $edit->type
== 'copy' ) {
1036 $closing->addWords( $edit->closing
);
1037 } elseif ( $edit->closing
) {
1038 $closing->addWords( $edit->closing
, 'ins' );
1041 $lines = $closing->getLines();
1042 wfProfileOut( __METHOD__
);