Chris@0: <?php
Chris@0: 
Chris@0: namespace Drupal\Component\Diff\Engine;
Chris@0: 
Chris@0: /**
Chris@0:  * Class used internally by Diff to actually compute the diffs.
Chris@0:  *
Chris@0:  * The algorithm used here is mostly lifted from the perl module
Chris@0:  * Algorithm::Diff (version 1.06) by Ned Konz, which is available at:
Chris@0:  *   http://www.perl.com/CPAN/authors/id/N/NE/NEDKONZ/Algorithm-Diff-1.06.zip
Chris@0:  *
Chris@0:  * More ideas are taken from:
Chris@0:  *   http://www.ics.uci.edu/~eppstein/161/960229.html
Chris@0:  *
Chris@0:  * Some ideas (and a bit of code) are from analyze.c, from GNU
Chris@0:  * diffutils-2.7, which can be found at:
Chris@0:  *   ftp://gnudist.gnu.org/pub/gnu/diffutils/diffutils-2.7.tar.gz
Chris@0:  *
Chris@0:  * closingly, some ideas (subdivision by NCHUNKS > 2, and some optimizations)
Chris@0:  * are my own.
Chris@0:  *
Chris@0:  * Line length limits for robustness added by Tim Starling, 2005-08-31
Chris@0:  *
Chris@0:  * @author Geoffrey T. Dairiki, Tim Starling
Chris@0:  * @private
Chris@0:  * @subpackage DifferenceEngine
Chris@0:  */
Chris@0: class DiffEngine {
Chris@0: 
Chris@0:   const USE_ASSERTS = FALSE;
Chris@0: 
Chris@0:   const MAX_XREF_LENGTH = 10000;
Chris@0: 
Chris@0:   public function diff($from_lines, $to_lines) {
Chris@0: 
Chris@0:     $n_from = sizeof($from_lines);
Chris@0:     $n_to = sizeof($to_lines);
Chris@0: 
Chris@0:     $this->xchanged = $this->ychanged = [];
Chris@0:     $this->xv = $this->yv = [];
Chris@0:     $this->xind = $this->yind = [];
Chris@0:     unset($this->seq);
Chris@0:     unset($this->in_seq);
Chris@0:     unset($this->lcs);
Chris@0: 
Chris@0:     // Skip leading common lines.
Chris@0:     for ($skip = 0; $skip < $n_from && $skip < $n_to; $skip++) {
Chris@0:       if ($from_lines[$skip] !== $to_lines[$skip]) {
Chris@0:         break;
Chris@0:       }
Chris@0:       $this->xchanged[$skip] = $this->ychanged[$skip] = FALSE;
Chris@0:     }
Chris@0:     // Skip trailing common lines.
Chris@0:     $xi = $n_from;
Chris@0:     $yi = $n_to;
Chris@0:     for ($endskip = 0; --$xi > $skip && --$yi > $skip; $endskip++) {
Chris@0:       if ($from_lines[$xi] !== $to_lines[$yi]) {
Chris@0:         break;
Chris@0:       }
Chris@0:       $this->xchanged[$xi] = $this->ychanged[$yi] = FALSE;
Chris@0:     }
Chris@0: 
Chris@0:     // Ignore lines which do not exist in both files.
Chris@0:     for ($xi = $skip; $xi < $n_from - $endskip; $xi++) {
Chris@0:       $xhash[$this->_line_hash($from_lines[$xi])] = 1;
Chris@0:     }
Chris@0: 
Chris@0:     for ($yi = $skip; $yi < $n_to - $endskip; $yi++) {
Chris@0:       $line = $to_lines[$yi];
Chris@0:       if ($this->ychanged[$yi] = empty($xhash[$this->_line_hash($line)])) {
Chris@0:         continue;
Chris@0:       }
Chris@0:       $yhash[$this->_line_hash($line)] = 1;
Chris@0:       $this->yv[] = $line;
Chris@0:       $this->yind[] = $yi;
Chris@0:     }
Chris@0:     for ($xi = $skip; $xi < $n_from - $endskip; $xi++) {
Chris@0:       $line = $from_lines[$xi];
Chris@0:       if ($this->xchanged[$xi] = empty($yhash[$this->_line_hash($line)])) {
Chris@0:         continue;
Chris@0:       }
Chris@0:       $this->xv[] = $line;
Chris@0:       $this->xind[] = $xi;
Chris@0:     }
Chris@0: 
Chris@0:     // Find the LCS.
Chris@0:     $this->_compareseq(0, sizeof($this->xv), 0, sizeof($this->yv));
Chris@0: 
Chris@0:     // Merge edits when possible
Chris@0:     $this->_shift_boundaries($from_lines, $this->xchanged, $this->ychanged);
Chris@0:     $this->_shift_boundaries($to_lines, $this->ychanged, $this->xchanged);
Chris@0: 
Chris@0:     // Compute the edit operations.
Chris@0:     $edits = [];
Chris@0:     $xi = $yi = 0;
Chris@0:     while ($xi < $n_from || $yi < $n_to) {
Chris@0:       $this::USE_ASSERTS && assert($yi < $n_to || $this->xchanged[$xi]);
Chris@0:       $this::USE_ASSERTS && assert($xi < $n_from || $this->ychanged[$yi]);
Chris@0: 
Chris@0:       // Skip matching "snake".
Chris@0:       $copy = [];
Chris@0:       while ( $xi < $n_from && $yi < $n_to && !$this->xchanged[$xi] && !$this->ychanged[$yi]) {
Chris@0:         $copy[] = $from_lines[$xi++];
Chris@0:         ++$yi;
Chris@0:       }
Chris@0:       if ($copy) {
Chris@0:         $edits[] = new DiffOpCopy($copy);
Chris@0:       }
Chris@0:       // Find deletes & adds.
Chris@0:       $delete = [];
Chris@0:       while ($xi < $n_from && $this->xchanged[$xi]) {
Chris@0:         $delete[] = $from_lines[$xi++];
Chris@0:       }
Chris@0:       $add = [];
Chris@0:       while ($yi < $n_to && $this->ychanged[$yi]) {
Chris@0:         $add[] = $to_lines[$yi++];
Chris@0:       }
Chris@0:       if ($delete && $add) {
Chris@0:         $edits[] = new DiffOpChange($delete, $add);
Chris@0:       }
Chris@0:       elseif ($delete) {
Chris@0:         $edits[] = new DiffOpDelete($delete);
Chris@0:       }
Chris@0:       elseif ($add) {
Chris@0:         $edits[] = new DiffOpAdd($add);
Chris@0:       }
Chris@0:     }
Chris@0:     return $edits;
Chris@0:   }
Chris@0: 
Chris@0:   /**
Chris@0:    * Returns the whole line if it's small enough, or the MD5 hash otherwise.
Chris@0:    */
Chris@0:   protected function _line_hash($line) {
Chris@17:     if (mb_strlen($line) > $this::MAX_XREF_LENGTH) {
Chris@0:       return md5($line);
Chris@0:     }
Chris@0:     else {
Chris@0:       return $line;
Chris@0:     }
Chris@0:   }
Chris@0: 
Chris@0: 
Chris@0:   /**
Chris@0:    * Divide the Largest Common Subsequence (LCS) of the sequences
Chris@0:    * [XOFF, XLIM) and [YOFF, YLIM) into NCHUNKS approximately equally
Chris@0:    * sized segments.
Chris@0:    *
Chris@0:    * Returns (LCS, PTS).  LCS is the length of the LCS. PTS is an
Chris@0:    * array of NCHUNKS+1 (X, Y) indexes giving the diving points between
Chris@0:    * sub sequences.  The first sub-sequence is contained in [X0, X1),
Chris@0:    * [Y0, Y1), the second in [X1, X2), [Y1, Y2) and so on.  Note
Chris@0:    * that (X0, Y0) == (XOFF, YOFF) and
Chris@0:    * (X[NCHUNKS], Y[NCHUNKS]) == (XLIM, YLIM).
Chris@0:    *
Chris@0:    * This function assumes that the first lines of the specified portions
Chris@0:    * of the two files do not match, and likewise that the last lines do not
Chris@0:    * match.  The caller must trim matching lines from the beginning and end
Chris@0:    * of the portions it is going to specify.
Chris@0:    */
Chris@0:   protected function _diag($xoff, $xlim, $yoff, $ylim, $nchunks) {
Chris@0:     $flip = FALSE;
Chris@0: 
Chris@0:     if ($xlim - $xoff > $ylim - $yoff) {
Chris@0:       // Things seems faster (I'm not sure I understand why)
Chris@0:       // when the shortest sequence in X.
Chris@0:       $flip = TRUE;
Chris@0:       list($xoff, $xlim, $yoff, $ylim) = [$yoff, $ylim, $xoff, $xlim];
Chris@0:     }
Chris@0: 
Chris@0:     if ($flip) {
Chris@0:       for ($i = $ylim - 1; $i >= $yoff; $i--) {
Chris@0:         $ymatches[$this->xv[$i]][] = $i;
Chris@0:       }
Chris@0:     }
Chris@0:     else {
Chris@0:       for ($i = $ylim - 1; $i >= $yoff; $i--) {
Chris@0:         $ymatches[$this->yv[$i]][] = $i;
Chris@0:       }
Chris@0:     }
Chris@0:     $this->lcs = 0;
Chris@0:     $this->seq[0] = $yoff - 1;
Chris@0:     $this->in_seq = [];
Chris@0:     $ymids[0] = [];
Chris@0: 
Chris@0:     $numer = $xlim - $xoff + $nchunks - 1;
Chris@0:     $x = $xoff;
Chris@0:     for ($chunk = 0; $chunk < $nchunks; $chunk++) {
Chris@0:       if ($chunk > 0) {
Chris@0:         for ($i = 0; $i <= $this->lcs; $i++) {
Chris@0:           $ymids[$i][$chunk - 1] = $this->seq[$i];
Chris@0:         }
Chris@0:       }
Chris@0: 
Chris@0:       $x1 = $xoff + (int)(($numer + ($xlim - $xoff) * $chunk) / $nchunks);
Chris@0:       for (; $x < $x1; $x++) {
Chris@0:         $line = $flip ? $this->yv[$x] : $this->xv[$x];
Chris@0:         if (empty($ymatches[$line])) {
Chris@0:           continue;
Chris@0:         }
Chris@0:         $matches = $ymatches[$line];
Chris@14:         $found_empty = FALSE;
Chris@14:         foreach ($matches as $y) {
Chris@14:           if (!$found_empty) {
Chris@14:             if (empty($this->in_seq[$y])) {
Chris@14:               $k = $this->_lcs_pos($y);
Chris@14:               $this::USE_ASSERTS && assert($k > 0);
Chris@14:               $ymids[$k] = $ymids[$k - 1];
Chris@14:               $found_empty = TRUE;
Chris@14:             }
Chris@0:           }
Chris@14:           else {
Chris@14:             if ($y > $this->seq[$k - 1]) {
Chris@14:               $this::USE_ASSERTS && assert($y < $this->seq[$k]);
Chris@14:               // Optimization: this is a common case:
Chris@14:               // next match is just replacing previous match.
Chris@14:               $this->in_seq[$this->seq[$k]] = FALSE;
Chris@14:               $this->seq[$k] = $y;
Chris@14:               $this->in_seq[$y] = 1;
Chris@14:             }
Chris@14:             elseif (empty($this->in_seq[$y])) {
Chris@14:               $k = $this->_lcs_pos($y);
Chris@14:               $this::USE_ASSERTS && assert($k > 0);
Chris@14:               $ymids[$k] = $ymids[$k - 1];
Chris@14:             }
Chris@0:           }
Chris@0:         }
Chris@0:       }
Chris@0:     }
Chris@0: 
Chris@0:     $seps[] = $flip ? [$yoff, $xoff] : [$xoff, $yoff];
Chris@0:     $ymid = $ymids[$this->lcs];
Chris@0:     for ($n = 0; $n < $nchunks - 1; $n++) {
Chris@0:       $x1 = $xoff + (int)(($numer + ($xlim - $xoff) * $n) / $nchunks);
Chris@0:       $y1 = $ymid[$n] + 1;
Chris@0:       $seps[] = $flip ? [$y1, $x1] : [$x1, $y1];
Chris@0:     }
Chris@0:     $seps[] = $flip ? [$ylim, $xlim] : [$xlim, $ylim];
Chris@0: 
Chris@0:     return [$this->lcs, $seps];
Chris@0:   }
Chris@0: 
Chris@0:   protected function _lcs_pos($ypos) {
Chris@0: 
Chris@0:     $end = $this->lcs;
Chris@0:     if ($end == 0 || $ypos > $this->seq[$end]) {
Chris@0:       $this->seq[++$this->lcs] = $ypos;
Chris@0:       $this->in_seq[$ypos] = 1;
Chris@0:       return $this->lcs;
Chris@0:     }
Chris@0: 
Chris@0:     $beg = 1;
Chris@0:     while ($beg < $end) {
Chris@0:       $mid = (int)(($beg + $end) / 2);
Chris@0:       if ($ypos > $this->seq[$mid]) {
Chris@0:         $beg = $mid + 1;
Chris@0:       }
Chris@0:       else {
Chris@0:         $end = $mid;
Chris@0:       }
Chris@0:     }
Chris@0: 
Chris@0:     $this::USE_ASSERTS && assert($ypos != $this->seq[$end]);
Chris@0: 
Chris@0:     $this->in_seq[$this->seq[$end]] = FALSE;
Chris@0:     $this->seq[$end] = $ypos;
Chris@0:     $this->in_seq[$ypos] = 1;
Chris@0:     return $end;
Chris@0:   }
Chris@0: 
Chris@0:   /**
Chris@0:    * Find LCS of two sequences.
Chris@0:    *
Chris@0:    * The results are recorded in the vectors $this->{x,y}changed[], by
Chris@0:    * storing a 1 in the element for each line that is an insertion
Chris@0:    * or deletion (ie. is not in the LCS).
Chris@0:    *
Chris@0:    * The subsequence of file 0 is [XOFF, XLIM) and likewise for file 1.
Chris@0:    *
Chris@0:    * Note that XLIM, YLIM are exclusive bounds.
Chris@0:    * All line numbers are origin-0 and discarded lines are not counted.
Chris@0:    */
Chris@0:   protected function _compareseq($xoff, $xlim, $yoff, $ylim) {
Chris@0: 
Chris@0:     // Slide down the bottom initial diagonal.
Chris@0:     while ($xoff < $xlim && $yoff < $ylim && $this->xv[$xoff] == $this->yv[$yoff]) {
Chris@0:       ++$xoff;
Chris@0:       ++$yoff;
Chris@0:     }
Chris@0: 
Chris@0:     // Slide up the top initial diagonal.
Chris@0:     while ($xlim > $xoff && $ylim > $yoff && $this->xv[$xlim - 1] == $this->yv[$ylim - 1]) {
Chris@0:       --$xlim;
Chris@0:       --$ylim;
Chris@0:     }
Chris@0: 
Chris@0:     if ($xoff == $xlim || $yoff == $ylim) {
Chris@0:       $lcs = 0;
Chris@0:     }
Chris@0:     else {
Chris@0:       // This is ad hoc but seems to work well.
Chris@0:       //$nchunks = sqrt(min($xlim - $xoff, $ylim - $yoff) / 2.5);
Chris@0:       //$nchunks = max(2, min(8, (int)$nchunks));
Chris@0:       $nchunks = min(7, $xlim - $xoff, $ylim - $yoff) + 1;
Chris@0:       list($lcs, $seps) = $this->_diag($xoff, $xlim, $yoff, $ylim, $nchunks);
Chris@0:     }
Chris@0: 
Chris@0:     if ($lcs == 0) {
Chris@0:       // X and Y sequences have no common subsequence:
Chris@0:       // mark all changed.
Chris@0:       while ($yoff < $ylim) {
Chris@0:         $this->ychanged[$this->yind[$yoff++]] = 1;
Chris@0:       }
Chris@0:       while ($xoff < $xlim) {
Chris@0:         $this->xchanged[$this->xind[$xoff++]] = 1;
Chris@0:       }
Chris@0:     }
Chris@0:     else {
Chris@0:       // Use the partitions to split this problem into subproblems.
Chris@0:       reset($seps);
Chris@0:       $pt1 = $seps[0];
Chris@0:       while ($pt2 = next($seps)) {
Chris@0:         $this->_compareseq ($pt1[0], $pt2[0], $pt1[1], $pt2[1]);
Chris@0:         $pt1 = $pt2;
Chris@0:       }
Chris@0:     }
Chris@0:   }
Chris@0: 
Chris@0:   /**
Chris@0:    * Adjust inserts/deletes of identical lines to join changes
Chris@0:    * as much as possible.
Chris@0:    *
Chris@0:    * We do something when a run of changed lines include a
Chris@0:    * line at one end and has an excluded, identical line at the other.
Chris@0:    * We are free to choose which identical line is included.
Chris@0:    * `compareseq' usually chooses the one at the beginning,
Chris@0:    * but usually it is cleaner to consider the following identical line
Chris@0:    * to be the "change".
Chris@0:    *
Chris@0:    * This is extracted verbatim from analyze.c (GNU diffutils-2.7).
Chris@0:    */
Chris@0:   protected function _shift_boundaries($lines, &$changed, $other_changed) {
Chris@0:     $i = 0;
Chris@0:     $j = 0;
Chris@0: 
Chris@14:     $this::USE_ASSERTS && assert(sizeof($lines) == sizeof($changed));
Chris@0:     $len = sizeof($lines);
Chris@0:     $other_len = sizeof($other_changed);
Chris@0: 
Chris@0:     while (1) {
Chris@0:       /*
Chris@0:        * Scan forwards to find beginning of another run of changes.
Chris@0:        * Also keep track of the corresponding point in the other file.
Chris@0:        *
Chris@0:        * Throughout this code, $i and $j are adjusted together so that
Chris@0:        * the first $i elements of $changed and the first $j elements
Chris@0:        * of $other_changed both contain the same number of zeros
Chris@0:        * (unchanged lines).
Chris@0:        * Furthermore, $j is always kept so that $j == $other_len or
Chris@0:        * $other_changed[$j] == FALSE.
Chris@0:        */
Chris@0:       while ($j < $other_len && $other_changed[$j]) {
Chris@0:         $j++;
Chris@0:       }
Chris@0:       while ($i < $len && !$changed[$i]) {
Chris@14:         $this::USE_ASSERTS && assert($j < $other_len && ! $other_changed[$j]);
Chris@0:         $i++;
Chris@0:         $j++;
Chris@0:         while ($j < $other_len && $other_changed[$j]) {
Chris@0:           $j++;
Chris@0:         }
Chris@0:       }
Chris@0: 
Chris@0:       if ($i == $len) {
Chris@0:         break;
Chris@0:       }
Chris@0:       $start = $i;
Chris@0: 
Chris@0:       // Find the end of this run of changes.
Chris@0:       while (++$i < $len && $changed[$i]) {
Chris@0:         continue;
Chris@0:       }
Chris@0: 
Chris@0:       do {
Chris@0:         /*
Chris@0:          * Record the length of this run of changes, so that
Chris@0:          * we can later determine whether the run has grown.
Chris@0:          */
Chris@0:         $runlength = $i - $start;
Chris@0: 
Chris@0:         /*
Chris@0:          * Move the changed region back, so long as the
Chris@0:          * previous unchanged line matches the last changed one.
Chris@0:          * This merges with previous changed regions.
Chris@0:          */
Chris@0:         while ($start > 0 && $lines[$start - 1] == $lines[$i - 1]) {
Chris@0:           $changed[--$start] = 1;
Chris@0:           $changed[--$i] = FALSE;
Chris@0:           while ($start > 0 && $changed[$start - 1]) {
Chris@0:             $start--;
Chris@0:           }
Chris@14:           $this::USE_ASSERTS && assert($j > 0);
Chris@0:           while ($other_changed[--$j]) {
Chris@0:             continue;
Chris@0:           }
Chris@14:           $this::USE_ASSERTS && assert($j >= 0 && !$other_changed[$j]);
Chris@0:         }
Chris@0: 
Chris@0:         /*
Chris@0:          * Set CORRESPONDING to the end of the changed run, at the last
Chris@0:          * point where it corresponds to a changed run in the other file.
Chris@0:          * CORRESPONDING == LEN means no such point has been found.
Chris@0:          */
Chris@0:         $corresponding = $j < $other_len ? $i : $len;
Chris@0: 
Chris@0:         /*
Chris@0:          * Move the changed region forward, so long as the
Chris@0:          * first changed line matches the following unchanged one.
Chris@0:          * This merges with following changed regions.
Chris@0:          * Do this second, so that if there are no merges,
Chris@0:          * the changed region is moved forward as far as possible.
Chris@0:          */
Chris@0:         while ($i < $len && $lines[$start] == $lines[$i]) {
Chris@0:           $changed[$start++] = FALSE;
Chris@0:           $changed[$i++] = 1;
Chris@0:           while ($i < $len && $changed[$i]) {
Chris@0:             $i++;
Chris@0:           }
Chris@14:           $this::USE_ASSERTS && assert($j < $other_len && ! $other_changed[$j]);
Chris@0:           $j++;
Chris@0:           if ($j < $other_len && $other_changed[$j]) {
Chris@0:             $corresponding = $i;
Chris@0:             while ($j < $other_len && $other_changed[$j]) {
Chris@0:               $j++;
Chris@0:             }
Chris@0:           }
Chris@0:         }
Chris@0:       } while ($runlength != $i - $start);
Chris@0: 
Chris@0:       /*
Chris@0:        * If possible, move the fully-merged run of changes
Chris@0:        * back to a corresponding run in the other file.
Chris@0:        */
Chris@0:       while ($corresponding < $i) {
Chris@0:         $changed[--$start] = 1;
Chris@0:         $changed[--$i] = 0;
Chris@14:         $this::USE_ASSERTS && assert($j > 0);
Chris@0:         while ($other_changed[--$j]) {
Chris@0:           continue;
Chris@0:         }
Chris@14:         $this::USE_ASSERTS && assert($j >= 0 && !$other_changed[$j]);
Chris@0:       }
Chris@0:     }
Chris@0:   }
Chris@0: 
Chris@0: }