[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   * https://cpan.metacpan.org/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 https://opensource.org/license/lgpl-2-1/.
  25   *
  26   * @author  Geoffrey T. Dairiki <dairiki@dairiki.org>
  27   * @package Text_Diff
  28   */
  29  class Text_Diff_Engine_native {
  30  
  31      public $xchanged;
  32      public $ychanged;
  33      public $xv;
  34      public $yv;
  35      public $xind;
  36      public $yind;
  37      public $seq;
  38      public $in_seq;
  39      public $lcs;
  40  
  41      function diff($from_lines, $to_lines)
  42      {
  43          array_walk($from_lines, array('Text_Diff', 'trimNewlines'));
  44          array_walk($to_lines, array('Text_Diff', 'trimNewlines'));
  45  
  46          $n_from = count($from_lines);
  47          $n_to = count($to_lines);
  48  
  49          $this->xchanged = $this->ychanged = array();
  50          $this->xv = $this->yv = array();
  51          $this->xind = $this->yind = array();
  52          unset($this->seq);
  53          unset($this->in_seq);
  54          unset($this->lcs);
  55  
  56          // Skip leading common lines.
  57          for ($skip = 0; $skip < $n_from && $skip < $n_to; $skip++) {
  58              if ($from_lines[$skip] !== $to_lines[$skip]) {
  59                  break;
  60              }
  61              $this->xchanged[$skip] = $this->ychanged[$skip] = false;
  62          }
  63  
  64          // Skip trailing common lines.
  65          $xi = $n_from; $yi = $n_to;
  66          for ($endskip = 0; --$xi > $skip && --$yi > $skip; $endskip++) {
  67              if ($from_lines[$xi] !== $to_lines[$yi]) {
  68                  break;
  69              }
  70              $this->xchanged[$xi] = $this->ychanged[$yi] = false;
  71          }
  72  
  73          // Ignore lines which do not exist in both files.
  74          for ($xi = $skip; $xi < $n_from - $endskip; $xi++) {
  75              $xhash[$from_lines[$xi]] = 1;
  76          }
  77          for ($yi = $skip; $yi < $n_to - $endskip; $yi++) {
  78              $line = $to_lines[$yi];
  79              if (($this->ychanged[$yi] = empty($xhash[$line]))) {
  80                  continue;
  81              }
  82              $yhash[$line] = 1;
  83              $this->yv[] = $line;
  84              $this->yind[] = $yi;
  85          }
  86          for ($xi = $skip; $xi < $n_from - $endskip; $xi++) {
  87              $line = $from_lines[$xi];
  88              if (($this->xchanged[$xi] = empty($yhash[$line]))) {
  89                  continue;
  90              }
  91              $this->xv[] = $line;
  92              $this->xind[] = $xi;
  93          }
  94  
  95          // Find the LCS.
  96          $this->_compareseq(0, count($this->xv), 0, count($this->yv));
  97  
  98          // Merge edits when possible.
  99          $this->_shiftBoundaries($from_lines, $this->xchanged, $this->ychanged);
 100          $this->_shiftBoundaries($to_lines, $this->ychanged, $this->xchanged);
 101  
 102          // Compute the edit operations.
 103          $edits = array();
 104          $xi = $yi = 0;
 105          while ($xi < $n_from || $yi < $n_to) {
 106              assert($yi < $n_to || $this->xchanged[$xi]);
 107              assert($xi < $n_from || $this->ychanged[$yi]);
 108  
 109              // Skip matching "snake".
 110              $copy = array();
 111              while ($xi < $n_from && $yi < $n_to
 112                     && !$this->xchanged[$xi] && !$this->ychanged[$yi]) {
 113                  $copy[] = $from_lines[$xi++];
 114                  ++$yi;
 115              }
 116              if ($copy) {
 117                  $edits[] = new Text_Diff_Op_copy($copy);
 118              }
 119  
 120              // Find deletes & adds.
 121              $delete = array();
 122              while ($xi < $n_from && $this->xchanged[$xi]) {
 123                  $delete[] = $from_lines[$xi++];
 124              }
 125  
 126              $add = array();
 127              while ($yi < $n_to && $this->ychanged[$yi]) {
 128                  $add[] = $to_lines[$yi++];
 129              }
 130  
 131              if ($delete && $add) {
 132                  $edits[] = new Text_Diff_Op_change($delete, $add);
 133              } elseif ($delete) {
 134                  $edits[] = new Text_Diff_Op_delete($delete);
 135              } elseif ($add) {
 136                  $edits[] = new Text_Diff_Op_add($add);
 137              }
 138          }
 139  
 140          return $edits;
 141      }
 142  
 143      /**
 144       * Divides the Largest Common Subsequence (LCS) of the sequences (XOFF,
 145       * XLIM) and (YOFF, YLIM) into NCHUNKS approximately equally sized
 146       * segments.
 147       *
 148       * Returns (LCS, PTS).  LCS is the length of the LCS. PTS is an array of
 149       * NCHUNKS+1 (X, Y) indexes giving the diving points between sub
 150       * sequences.  The first sub-sequence is contained in (X0, X1), (Y0, Y1),
 151       * the second in (X1, X2), (Y1, Y2) and so on.  Note that (X0, Y0) ==
 152       * (XOFF, YOFF) and (X[NCHUNKS], Y[NCHUNKS]) == (XLIM, YLIM).
 153       *
 154       * This function assumes that the first lines of the specified portions of
 155       * the two files do not match, and likewise that the last lines do not
 156       * match.  The caller must trim matching lines from the beginning and end
 157       * of the portions it is going to specify.
 158       */
 159      function _diag ($xoff, $xlim, $yoff, $ylim, $nchunks)
 160      {
 161          $flip = false;
 162  
 163          if ($xlim - $xoff > $ylim - $yoff) {
 164              /* Things seems faster (I'm not sure I understand why) when the
 165               * shortest sequence is in X. */
 166              $flip = true;
 167              list ($xoff, $xlim, $yoff, $ylim)
 168                  = array($yoff, $ylim, $xoff, $xlim);
 169          }
 170  
 171          if ($flip) {
 172              for ($i = $ylim - 1; $i >= $yoff; $i--) {
 173                  $ymatches[$this->xv[$i]][] = $i;
 174              }
 175          } else {
 176              for ($i = $ylim - 1; $i >= $yoff; $i--) {
 177                  $ymatches[$this->yv[$i]][] = $i;
 178              }
 179          }
 180  
 181          $this->lcs = 0;
 182          $this->seq[0]= $yoff - 1;
 183          $this->in_seq = array();
 184          $ymids[0] = array();
 185  
 186          $numer = $xlim - $xoff + $nchunks - 1;
 187          $x = $xoff;
 188          for ($chunk = 0; $chunk < $nchunks; $chunk++) {
 189              if ($chunk > 0) {
 190                  for ($i = 0; $i <= $this->lcs; $i++) {
 191                      $ymids[$i][$chunk - 1] = $this->seq[$i];
 192                  }
 193              }
 194  
 195              $x1 = $xoff + (int)(($numer + ($xlim - $xoff) * $chunk) / $nchunks);
 196              for (; $x < $x1; $x++) {
 197                  $line = $flip ? $this->yv[$x] : $this->xv[$x];
 198                  if (empty($ymatches[$line])) {
 199                      continue;
 200                  }
 201                  $matches = $ymatches[$line];
 202                  reset($matches);
 203                  while ($y = current($matches)) {
 204                      if (empty($this->in_seq[$y])) {
 205                          $k = $this->_lcsPos($y);
 206                          assert($k > 0);
 207                          $ymids[$k] = $ymids[$k - 1];
 208                          break;
 209                      }
 210                      next($matches);
 211                  }
 212                  while ($y = current($matches)) {
 213                      if ($y > $this->seq[$k - 1]) {
 214                          assert($y <= $this->seq[$k]);
 215                          /* Optimization: this is a common case: next match is
 216                           * just replacing previous match. */
 217                          $this->in_seq[$this->seq[$k]] = false;
 218                          $this->seq[$k] = $y;
 219                          $this->in_seq[$y] = 1;
 220                      } elseif (empty($this->in_seq[$y])) {
 221                          $k = $this->_lcsPos($y);
 222                          assert($k > 0);
 223                          $ymids[$k] = $ymids[$k - 1];
 224                      }
 225                      next($matches);
 226                  }
 227              }
 228          }
 229  
 230          $seps[] = $flip ? array($yoff, $xoff) : array($xoff, $yoff);
 231          $ymid = $ymids[$this->lcs];
 232          for ($n = 0; $n < $nchunks - 1; $n++) {
 233              $x1 = $xoff + (int)(($numer + ($xlim - $xoff) * $n) / $nchunks);
 234              $y1 = $ymid[$n] + 1;
 235              $seps[] = $flip ? array($y1, $x1) : array($x1, $y1);
 236          }
 237          $seps[] = $flip ? array($ylim, $xlim) : array($xlim, $ylim);
 238  
 239          return array($this->lcs, $seps);
 240      }
 241  
 242      function _lcsPos($ypos)
 243      {
 244          $end = $this->lcs;
 245          if ($end == 0 || $ypos > $this->seq[$end]) {
 246              $this->seq[++$this->lcs] = $ypos;
 247              $this->in_seq[$ypos] = 1;
 248              return $this->lcs;
 249          }
 250  
 251          $beg = 1;
 252          while ($beg < $end) {
 253              $mid = (int)(($beg + $end) / 2);
 254              if ($ypos > $this->seq[$mid]) {
 255                  $beg = $mid + 1;
 256              } else {
 257                  $end = $mid;
 258              }
 259          }
 260  
 261          assert($ypos != $this->seq[$end]);
 262  
 263          $this->in_seq[$this->seq[$end]] = false;
 264          $this->seq[$end] = $ypos;
 265          $this->in_seq[$ypos] = 1;
 266          return $end;
 267      }
 268  
 269      /**
 270       * Finds LCS of two sequences.
 271       *
 272       * The results are recorded in the vectors $this->{x,y}changed[], by
 273       * storing a 1 in the element for each line that is an insertion or
 274       * deletion (ie. is not in the LCS).
 275       *
 276       * The subsequence of file 0 is (XOFF, XLIM) and likewise for file 1.
 277       *
 278       * Note that XLIM, YLIM are exclusive bounds.  All line numbers are
 279       * origin-0 and discarded lines are not counted.
 280       */
 281      function _compareseq ($xoff, $xlim, $yoff, $ylim)
 282      {
 283          /* Slide down the bottom initial diagonal. */
 284          while ($xoff < $xlim && $yoff < $ylim
 285                 && $this->xv[$xoff] == $this->yv[$yoff]) {
 286              ++$xoff;
 287              ++$yoff;
 288          }
 289  
 290          /* Slide up the top initial diagonal. */
 291          while ($xlim > $xoff && $ylim > $yoff
 292                 && $this->xv[$xlim - 1] == $this->yv[$ylim - 1]) {
 293              --$xlim;
 294              --$ylim;
 295          }
 296  
 297          if ($xoff == $xlim || $yoff == $ylim) {
 298              $lcs = 0;
 299          } else {
 300              /* This is ad hoc but seems to work well.  $nchunks =
 301               * sqrt(min($xlim - $xoff, $ylim - $yoff) / 2.5); $nchunks =
 302               * max(2,min(8,(int)$nchunks)); */
 303              $nchunks = min(7, $xlim - $xoff, $ylim - $yoff) + 1;
 304              list($lcs, $seps)
 305                  = $this->_diag($xoff, $xlim, $yoff, $ylim, $nchunks);
 306          }
 307  
 308          if ($lcs == 0) {
 309              /* X and Y sequences have no common subsequence: mark all
 310               * changed. */
 311              while ($yoff < $ylim) {
 312                  $this->ychanged[$this->yind[$yoff++]] = 1;
 313              }
 314              while ($xoff < $xlim) {
 315                  $this->xchanged[$this->xind[$xoff++]] = 1;
 316              }
 317          } else {
 318              /* Use the partitions to split this problem into subproblems. */
 319              reset($seps);
 320              $pt1 = $seps[0];
 321              while ($pt2 = next($seps)) {
 322                  $this->_compareseq ($pt1[0], $pt2[0], $pt1[1], $pt2[1]);
 323                  $pt1 = $pt2;
 324              }
 325          }
 326      }
 327  
 328      /**
 329       * Adjusts inserts/deletes of identical lines to join changes as much as
 330       * possible.
 331       *
 332       * We do something when a run of changed lines include a line at one end
 333       * and has an excluded, identical line at the other.  We are free to
 334       * choose which identical line is included.  `compareseq' usually chooses
 335       * the one at the beginning, but usually it is cleaner to consider the
 336       * following identical line to be the "change".
 337       *
 338       * This is extracted verbatim from analyze.c (GNU diffutils-2.7).
 339       */
 340      function _shiftBoundaries($lines, &$changed, $other_changed)
 341      {
 342          $i = 0;
 343          $j = 0;
 344  
 345          assert(count($lines) == count($changed));
 346          $len = count($lines);
 347          $other_len = count($other_changed);
 348  
 349          while (1) {
 350              /* Scan forward to find the beginning of another run of
 351               * changes. Also keep track of the corresponding point in the
 352               * other file.
 353               *
 354               * Throughout this code, $i and $j are adjusted together so that
 355               * the first $i elements of $changed and the first $j elements of
 356               * $other_changed both contain the same number of zeros (unchanged
 357               * lines).
 358               *
 359               * Furthermore, $j is always kept so that $j == $other_len or
 360               * $other_changed[$j] == false. */
 361              while ($j < $other_len && $other_changed[$j]) {
 362                  $j++;
 363              }
 364  
 365              while ($i < $len && ! $changed[$i]) {
 366                  assert($j < $other_len && ! $other_changed[$j]);
 367                  $i++; $j++;
 368                  while ($j < $other_len && $other_changed[$j]) {
 369                      $j++;
 370                  }
 371              }
 372  
 373              if ($i == $len) {
 374                  break;
 375              }
 376  
 377              $start = $i;
 378  
 379              /* Find the end of this run of changes. */
 380              while (++$i < $len && $changed[$i]) {
 381                  continue;
 382              }
 383  
 384              do {
 385                  /* Record the length of this run of changes, so that we can
 386                   * later determine whether the run has grown. */
 387                  $runlength = $i - $start;
 388  
 389                  /* Move the changed region back, so long as the previous
 390                   * unchanged line matches the last changed one.  This merges
 391                   * with previous changed regions. */
 392                  while ($start > 0 && $lines[$start - 1] == $lines[$i - 1]) {
 393                      $changed[--$start] = 1;
 394                      $changed[--$i] = false;
 395                      while ($start > 0 && $changed[$start - 1]) {
 396                          $start--;
 397                      }
 398                      assert($j > 0);
 399                      while ($other_changed[--$j]) {
 400                          continue;
 401                      }
 402                      assert($j >= 0 && !$other_changed[$j]);
 403                  }
 404  
 405                  /* Set CORRESPONDING to the end of the changed run, at the
 406                   * last point where it corresponds to a changed run in the
 407                   * other file. CORRESPONDING == LEN means no such point has
 408                   * been found. */
 409                  $corresponding = $j < $other_len ? $i : $len;
 410  
 411                  /* Move the changed region forward, so long as the first
 412                   * changed line matches the following unchanged one.  This
 413                   * merges with following changed regions.  Do this second, so
 414                   * that if there are no merges, the changed region is moved
 415                   * forward as far as possible. */
 416                  while ($i < $len && $lines[$start] == $lines[$i]) {
 417                      $changed[$start++] = false;
 418                      $changed[$i++] = 1;
 419                      while ($i < $len && $changed[$i]) {
 420                          $i++;
 421                      }
 422  
 423                      assert($j < $other_len && ! $other_changed[$j]);
 424                      $j++;
 425                      if ($j < $other_len && $other_changed[$j]) {
 426                          $corresponding = $i;
 427                          while ($j < $other_len && $other_changed[$j]) {
 428                              $j++;
 429                          }
 430                      }
 431                  }
 432              } while ($runlength != $i - $start);
 433  
 434              /* If possible, move the fully-merged run of changes back to a
 435               * corresponding run in the other file. */
 436              while ($corresponding < $i) {
 437                  $changed[--$start] = 1;
 438                  $changed[--$i] = 0;
 439                  assert($j > 0);
 440                  while ($other_changed[--$j]) {
 441                      continue;
 442                  }
 443                  assert($j >= 0 && !$other_changed[$j]);
 444              }
 445          }
 446      }
 447  
 448  }