[Summary view] [Print] [Text view]

   1  <?php
   2  /**
   3   * Class used internally by Text_Diff to actually compute the diffs.
   4   *
   5   * This class is implemented using native PHP code.
   6   *
   7   * The algorithm used here is mostly lifted from the perl module
   8   * Algorithm::Diff (version 1.06) by Ned Konz, which is available at:
   9   * http://www.perl.com/CPAN/authors/id/N/NE/NEDKONZ/Algorithm-Diff-1.06.zip
  10   *
  11   * More ideas are taken from: http://www.ics.uci.edu/~eppstein/161/960229.html
  12   *
  13   * Some ideas (and a bit of code) are taken from analyze.c, of GNU
  14   * diffutils-2.7, which can be found at:
  15   * ftp://gnudist.gnu.org/pub/gnu/diffutils/diffutils-2.7.tar.gz
  16   *
  17   * Some ideas (subdivision by NCHUNKS > 2, and some optimizations) are from
  18   * Geoffrey T. Dairiki <dairiki@dairiki.org>. The original PHP version of this
  19   * code was written by him, and is used/adapted with his permission.
  20   *
  21   * Copyright 2004-2010 The Horde Project (http://www.horde.org/)
  22   *
  23   * See the enclosed file COPYING for license information (LGPL). If you did
  24   * not receive this file, see http://opensource.org/licenses/lgpl-license.php.
  25   *
  26   * @author  Geoffrey T. Dairiki <dairiki@dairiki.org>
  27   * @package Text_Diff
  28   */
  29  class Text_Diff_Engine_native {
  30  
  31      function diff($from_lines, $to_lines)
  32      {
  33          array_walk($from_lines, array('Text_Diff', 'trimNewlines'));
  34          array_walk($to_lines, array('Text_Diff', 'trimNewlines'));
  35  
  36          $n_from = count($from_lines);
  37          $n_to = count($to_lines);
  38  
  39          $this->xchanged = $this->ychanged = array();
  40          $this->xv = $this->yv = array();
  41          $this->xind = $this->yind = array();
  42          unset($this->seq);
  43          unset($this->in_seq);
  44          unset($this->lcs);
  45  
  46          // Skip leading common lines.
  47          for ($skip = 0; $skip < $n_from && $skip < $n_to; $skip++) {
  48              if ($from_lines[$skip] !== $to_lines[$skip]) {
  49                  break;
  50              }
  51              $this->xchanged[$skip] = $this->ychanged[$skip] = false;
  52          }
  53  
  54          // Skip trailing common lines.
  55          $xi = $n_from; $yi = $n_to;
  56          for ($endskip = 0; --$xi > $skip && --$yi > $skip; $endskip++) {
  57              if ($from_lines[$xi] !== $to_lines[$yi]) {
  58                  break;
  59              }
  60              $this->xchanged[$xi] = $this->ychanged[$yi] = false;
  61          }
  62  
  63          // Ignore lines which do not exist in both files.
  64          for ($xi = $skip; $xi < $n_from - $endskip; $xi++) {
  65              $xhash[$from_lines[$xi]] = 1;
  66          }
  67          for ($yi = $skip; $yi < $n_to - $endskip; $yi++) {
  68              $line = $to_lines[$yi];
  69              if (($this->ychanged[$yi] = empty($xhash[$line]))) {
  70                  continue;
  71              }
  72              $yhash[$line] = 1;
  73              $this->yv[] = $line;
  74              $this->yind[] = $yi;
  75          }
  76          for ($xi = $skip; $xi < $n_from - $endskip; $xi++) {
  77              $line = $from_lines[$xi];
  78              if (($this->xchanged[$xi] = empty($yhash[$line]))) {
  79                  continue;
  80              }
  81              $this->xv[] = $line;
  82              $this->xind[] = $xi;
  83          }
  84  
  85          // Find the LCS.
  86          $this->_compareseq(0, count($this->xv), 0, count($this->yv));
  87  
  88          // Merge edits when possible.
  89          $this->_shiftBoundaries($from_lines, $this->xchanged, $this->ychanged);
  90          $this->_shiftBoundaries($to_lines, $this->ychanged, $this->xchanged);
  91  
  92          // Compute the edit operations.
  93          $edits = array();
  94          $xi = $yi = 0;
  95          while ($xi < $n_from || $yi < $n_to) {
  96              assert($yi < $n_to || $this->xchanged[$xi]);
  97              assert($xi < $n_from || $this->ychanged[$yi]);
  98  
  99              // Skip matching "snake".
 100              $copy = array();
 101              while ($xi < $n_from && $yi < $n_to
 102                     && !$this->xchanged[$xi] && !$this->ychanged[$yi]) {
 103                  $copy[] = $from_lines[$xi++];
 104                  ++$yi;
 105              }
 106              if ($copy) {
 107                  $edits[] = new Text_Diff_Op_copy($copy);
 108              }
 109  
 110              // Find deletes & adds.
 111              $delete = array();
 112              while ($xi < $n_from && $this->xchanged[$xi]) {
 113                  $delete[] = $from_lines[$xi++];
 114              }
 115  
 116              $add = array();
 117              while ($yi < $n_to && $this->ychanged[$yi]) {
 118                  $add[] = $to_lines[$yi++];
 119              }
 120  
 121              if ($delete && $add) {
 122                  $edits[] = new Text_Diff_Op_change($delete, $add);
 123              } elseif ($delete) {
 124                  $edits[] = new Text_Diff_Op_delete($delete);
 125              } elseif ($add) {
 126                  $edits[] = new Text_Diff_Op_add($add);
 127              }
 128          }
 129  
 130          return $edits;
 131      }
 132  
 133      /**
 134       * Divides the Largest Common Subsequence (LCS) of the sequences (XOFF,
 135       * XLIM) and (YOFF, YLIM) into NCHUNKS approximately equally sized
 136       * segments.
 137       *
 138       * Returns (LCS, PTS).  LCS is the length of the LCS. PTS is an array of
 139       * NCHUNKS+1 (X, Y) indexes giving the diving points between sub
 140       * sequences.  The first sub-sequence is contained in (X0, X1), (Y0, Y1),
 141       * the second in (X1, X2), (Y1, Y2) and so on.  Note that (X0, Y0) ==
 142       * (XOFF, YOFF) and (X[NCHUNKS], Y[NCHUNKS]) == (XLIM, YLIM).
 143       *
 144       * This function assumes that the first lines of the specified portions of
 145       * the two files do not match, and likewise that the last lines do not
 146       * match.  The caller must trim matching lines from the beginning and end
 147       * of the portions it is going to specify.
 148       */
 149      function _diag ($xoff, $xlim, $yoff, $ylim, $nchunks)
 150      {
 151          $flip = false;
 152  
 153          if ($xlim - $xoff > $ylim - $yoff) {
 154              /* Things seems faster (I'm not sure I understand why) when the
 155               * shortest sequence is in X. */
 156              $flip = true;
 157              list ($xoff, $xlim, $yoff, $ylim)
 158                  = array($yoff, $ylim, $xoff, $xlim);
 159          }
 160  
 161          if ($flip) {
 162              for ($i = $ylim - 1; $i >= $yoff; $i--) {
 163                  $ymatches[$this->xv[$i]][] = $i;
 164              }
 165          } else {
 166              for ($i = $ylim - 1; $i >= $yoff; $i--) {
 167                  $ymatches[$this->yv[$i]][] = $i;
 168              }
 169          }
 170  
 171          $this->lcs = 0;
 172          $this->seq[0]= $yoff - 1;
 173          $this->in_seq = array();
 174          $ymids[0] = array();
 175  
 176          $numer = $xlim - $xoff + $nchunks - 1;
 177          $x = $xoff;
 178          for ($chunk = 0; $chunk < $nchunks; $chunk++) {
 179              if ($chunk > 0) {
 180                  for ($i = 0; $i <= $this->lcs; $i++) {
 181                      $ymids[$i][$chunk - 1] = $this->seq[$i];
 182                  }
 183              }
 184  
 185              $x1 = $xoff + (int)(($numer + ($xlim - $xoff) * $chunk) / $nchunks);
 186              for (; $x < $x1; $x++) {
 187                  $line = $flip ? $this->yv[$x] : $this->xv[$x];
 188                  if (empty($ymatches[$line])) {
 189                      continue;
 190                  }
 191                  $matches = $ymatches[$line];
 192                  reset($matches);
 193                  while (list(, $y) = each($matches)) {
 194                      if (empty($this->in_seq[$y])) {
 195                          $k = $this->_lcsPos($y);
 196                          assert($k > 0);
 197                          $ymids[$k] = $ymids[$k - 1];
 198                          break;
 199                      }
 200                  }
 201                  while (list(, $y) = each($matches)) {
 202                      if ($y > $this->seq[$k - 1]) {
 203                          assert($y <= $this->seq[$k]);
 204                          /* Optimization: this is a common case: next match is
 205                           * just replacing previous match. */
 206                          $this->in_seq[$this->seq[$k]] = false;
 207                          $this->seq[$k] = $y;
 208                          $this->in_seq[$y] = 1;
 209                      } elseif (empty($this->in_seq[$y])) {
 210                          $k = $this->_lcsPos($y);
 211                          assert($k > 0);
 212                          $ymids[$k] = $ymids[$k - 1];
 213                      }
 214                  }
 215              }
 216          }
 217  
 218          $seps[] = $flip ? array($yoff, $xoff) : array($xoff, $yoff);
 219          $ymid = $ymids[$this->lcs];
 220          for ($n = 0; $n < $nchunks - 1; $n++) {
 221              $x1 = $xoff + (int)(($numer + ($xlim - $xoff) * $n) / $nchunks);
 222              $y1 = $ymid[$n] + 1;
 223              $seps[] = $flip ? array($y1, $x1) : array($x1, $y1);
 224          }
 225          $seps[] = $flip ? array($ylim, $xlim) : array($xlim, $ylim);
 226  
 227          return array($this->lcs, $seps);
 228      }
 229  
 230      function _lcsPos($ypos)
 231      {
 232          $end = $this->lcs;
 233          if ($end == 0 || $ypos > $this->seq[$end]) {
 234              $this->seq[++$this->lcs] = $ypos;
 235              $this->in_seq[$ypos] = 1;
 236              return $this->lcs;
 237          }
 238  
 239          $beg = 1;
 240          while ($beg < $end) {
 241              $mid = (int)(($beg + $end) / 2);
 242              if ($ypos > $this->seq[$mid]) {
 243                  $beg = $mid + 1;
 244              } else {
 245                  $end = $mid;
 246              }
 247          }
 248  
 249          assert($ypos != $this->seq[$end]);
 250  
 251          $this->in_seq[$this->seq[$end]] = false;
 252          $this->seq[$end] = $ypos;
 253          $this->in_seq[$ypos] = 1;
 254          return $end;
 255      }
 256  
 257      /**
 258       * Finds LCS of two sequences.
 259       *
 260       * The results are recorded in the vectors $this->{x,y}changed[], by
 261       * storing a 1 in the element for each line that is an insertion or
 262       * deletion (ie. is not in the LCS).
 263       *
 264       * The subsequence of file 0 is (XOFF, XLIM) and likewise for file 1.
 265       *
 266       * Note that XLIM, YLIM are exclusive bounds.  All line numbers are
 267       * origin-0 and discarded lines are not counted.
 268       */
 269      function _compareseq ($xoff, $xlim, $yoff, $ylim)
 270      {
 271          /* Slide down the bottom initial diagonal. */
 272          while ($xoff < $xlim && $yoff < $ylim
 273                 && $this->xv[$xoff] == $this->yv[$yoff]) {
 274              ++$xoff;
 275              ++$yoff;
 276          }
 277  
 278          /* Slide up the top initial diagonal. */
 279          while ($xlim > $xoff && $ylim > $yoff
 280                 && $this->xv[$xlim - 1] == $this->yv[$ylim - 1]) {
 281              --$xlim;
 282              --$ylim;
 283          }
 284  
 285          if ($xoff == $xlim || $yoff == $ylim) {
 286              $lcs = 0;
 287          } else {
 288              /* This is ad hoc but seems to work well.  $nchunks =
 289               * sqrt(min($xlim - $xoff, $ylim - $yoff) / 2.5); $nchunks =
 290               * max(2,min(8,(int)$nchunks)); */
 291              $nchunks = min(7, $xlim - $xoff, $ylim - $yoff) + 1;
 292              list($lcs, $seps)
 293                  = $this->_diag($xoff, $xlim, $yoff, $ylim, $nchunks);
 294          }
 295  
 296          if ($lcs == 0) {
 297              /* X and Y sequences have no common subsequence: mark all
 298               * changed. */
 299              while ($yoff < $ylim) {
 300                  $this->ychanged[$this->yind[$yoff++]] = 1;
 301              }
 302              while ($xoff < $xlim) {
 303                  $this->xchanged[$this->xind[$xoff++]] = 1;
 304              }
 305          } else {
 306              /* Use the partitions to split this problem into subproblems. */
 307              reset($seps);
 308              $pt1 = $seps[0];
 309              while ($pt2 = next($seps)) {
 310                  $this->_compareseq ($pt1[0], $pt2[0], $pt1[1], $pt2[1]);
 311                  $pt1 = $pt2;
 312              }
 313          }
 314      }
 315  
 316      /**
 317       * Adjusts inserts/deletes of identical lines to join changes as much as
 318       * possible.
 319       *
 320       * We do something when a run of changed lines include a line at one end
 321       * and has an excluded, identical line at the other.  We are free to
 322       * choose which identical line is included.  `compareseq' usually chooses
 323       * the one at the beginning, but usually it is cleaner to consider the
 324       * following identical line to be the "change".
 325       *
 326       * This is extracted verbatim from analyze.c (GNU diffutils-2.7).
 327       */
 328      function _shiftBoundaries($lines, &$changed, $other_changed)
 329      {
 330          $i = 0;
 331          $j = 0;
 332  
 333          assert('count($lines) == count($changed)');
 334          $len = count($lines);
 335          $other_len = count($other_changed);
 336  
 337          while (1) {
 338              /* Scan forward to find the beginning of another run of
 339               * changes. Also keep track of the corresponding point in the
 340               * other file.
 341               *
 342               * Throughout this code, $i and $j are adjusted together so that
 343               * the first $i elements of $changed and the first $j elements of
 344               * $other_changed both contain the same number of zeros (unchanged
 345               * lines).
 346               *
 347               * Furthermore, $j is always kept so that $j == $other_len or
 348               * $other_changed[$j] == false. */
 349              while ($j < $other_len && $other_changed[$j]) {
 350                  $j++;
 351              }
 352  
 353              while ($i < $len && ! $changed[$i]) {
 354                  assert('$j < $other_len && ! $other_changed[$j]');
 355                  $i++; $j++;
 356                  while ($j < $other_len && $other_changed[$j]) {
 357                      $j++;
 358                  }
 359              }
 360  
 361              if ($i == $len) {
 362                  break;
 363              }
 364  
 365              $start = $i;
 366  
 367              /* Find the end of this run of changes. */
 368              while (++$i < $len && $changed[$i]) {
 369                  continue;
 370              }
 371  
 372              do {
 373                  /* Record the length of this run of changes, so that we can
 374                   * later determine whether the run has grown. */
 375                  $runlength = $i - $start;
 376  
 377                  /* Move the changed region back, so long as the previous
 378                   * unchanged line matches the last changed one.  This merges
 379                   * with previous changed regions. */
 380                  while ($start > 0 && $lines[$start - 1] == $lines[$i - 1]) {
 381                      $changed[--$start] = 1;
 382                      $changed[--$i] = false;
 383                      while ($start > 0 && $changed[$start - 1]) {
 384                          $start--;
 385                      }
 386                      assert('$j > 0');
 387                      while ($other_changed[--$j]) {
 388                          continue;
 389                      }
 390                      assert('$j >= 0 && !$other_changed[$j]');
 391                  }
 392  
 393                  /* Set CORRESPONDING to the end of the changed run, at the
 394                   * last point where it corresponds to a changed run in the
 395                   * other file. CORRESPONDING == LEN means no such point has
 396                   * been found. */
 397                  $corresponding = $j < $other_len ? $i : $len;
 398  
 399                  /* Move the changed region forward, so long as the first
 400                   * changed line matches the following unchanged one.  This
 401                   * merges with following changed regions.  Do this second, so
 402                   * that if there are no merges, the changed region is moved
 403                   * forward as far as possible. */
 404                  while ($i < $len && $lines[$start] == $lines[$i]) {
 405                      $changed[$start++] = false;
 406                      $changed[$i++] = 1;
 407                      while ($i < $len && $changed[$i]) {
 408                          $i++;
 409                      }
 410  
 411                      assert('$j < $other_len && ! $other_changed[$j]');
 412                      $j++;
 413                      if ($j < $other_len && $other_changed[$j]) {
 414                          $corresponding = $i;
 415                          while ($j < $other_len && $other_changed[$j]) {
 416                              $j++;
 417                          }
 418                      }
 419                  }
 420              } while ($runlength != $i - $start);
 421  
 422              /* If possible, move the fully-merged run of changes back to a
 423               * corresponding run in the other file. */
 424              while ($corresponding < $i) {
 425                  $changed[--$start] = 1;
 426                  $changed[--$i] = 0;
 427                  assert('$j > 0');
 428                  while ($other_changed[--$j]) {
 429                      continue;
 430                  }
 431                  assert('$j >= 0 && !$other_changed[$j]');
 432              }
 433          }
 434      }
 435  
 436  }