3 * Class used internally by Text_Diff to actually compute the diffs.
5 * This class is implemented using native PHP code.
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
11 * More ideas are taken from: http://www.ics.uci.edu/~eppstein/161/960229.html
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
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.
21 * Copyright 2004-2010 The Horde Project (http://www.horde.org/)
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.
26 * @author Geoffrey T. Dairiki <dairiki@dairiki.org>
29 class Text_Diff_Engine_native {
31 function diff($from_lines, $to_lines)
33 array_walk($from_lines, array('Text_Diff', 'trimNewlines'));
34 array_walk($to_lines, array('Text_Diff', 'trimNewlines'));
36 $n_from = count($from_lines);
37 $n_to = count($to_lines);
39 $this->xchanged = $this->ychanged = array();
40 $this->xv = $this->yv = array();
41 $this->xind = $this->yind = array();
46 // Skip leading common lines.
47 for ($skip = 0; $skip < $n_from && $skip < $n_to; $skip++) {
48 if ($from_lines[$skip] !== $to_lines[$skip]) {
51 $this->xchanged[$skip] = $this->ychanged[$skip] = false;
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]) {
60 $this->xchanged[$xi] = $this->ychanged[$yi] = false;
63 // Ignore lines which do not exist in both files.
64 for ($xi = $skip; $xi < $n_from - $endskip; $xi++) {
65 $xhash[$from_lines[$xi]] = 1;
67 for ($yi = $skip; $yi < $n_to - $endskip; $yi++) {
68 $line = $to_lines[$yi];
69 if (($this->ychanged[$yi] = empty($xhash[$line]))) {
76 for ($xi = $skip; $xi < $n_from - $endskip; $xi++) {
77 $line = $from_lines[$xi];
78 if (($this->xchanged[$xi] = empty($yhash[$line]))) {
86 $this->_compareseq(0, count($this->xv), 0, count($this->yv));
88 // Merge edits when possible.
89 $this->_shiftBoundaries($from_lines, $this->xchanged, $this->ychanged);
90 $this->_shiftBoundaries($to_lines, $this->ychanged, $this->xchanged);
92 // Compute the edit operations.
95 while ($xi < $n_from || $yi < $n_to) {
96 assert($yi < $n_to || $this->xchanged[$xi]);
97 assert($xi < $n_from || $this->ychanged[$yi]);
99 // Skip matching "snake".
101 while ($xi < $n_from && $yi < $n_to
102 && !$this->xchanged[$xi] && !$this->ychanged[$yi]) {
103 $copy[] = $from_lines[$xi++];
107 $edits[] = new Text_Diff_Op_copy($copy);
110 // Find deletes & adds.
112 while ($xi < $n_from && $this->xchanged[$xi]) {
113 $delete[] = $from_lines[$xi++];
117 while ($yi < $n_to && $this->ychanged[$yi]) {
118 $add[] = $to_lines[$yi++];
121 if ($delete && $add) {
122 $edits[] = new Text_Diff_Op_change($delete, $add);
124 $edits[] = new Text_Diff_Op_delete($delete);
126 $edits[] = new Text_Diff_Op_add($add);
134 * Divides the Largest Common Subsequence (LCS) of the sequences (XOFF,
135 * XLIM) and (YOFF, YLIM) into NCHUNKS approximately equally sized
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).
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.
149 function _diag ($xoff, $xlim, $yoff, $ylim, $nchunks)
153 if ($xlim - $xoff > $ylim - $yoff) {
154 /* Things seems faster (I'm not sure I understand why) when the
155 * shortest sequence is in X. */
157 list ($xoff, $xlim, $yoff, $ylim)
158 = array($yoff, $ylim, $xoff, $xlim);
162 for ($i = $ylim - 1; $i >= $yoff; $i--) {
163 $ymatches[$this->xv[$i]][] = $i;
166 for ($i = $ylim - 1; $i >= $yoff; $i--) {
167 $ymatches[$this->yv[$i]][] = $i;
172 $this->seq[0]= $yoff - 1;
173 $this->in_seq = array();
176 $numer = $xlim - $xoff + $nchunks - 1;
178 for ($chunk = 0; $chunk < $nchunks; $chunk++) {
180 for ($i = 0; $i <= $this->lcs; $i++) {
181 $ymids[$i][$chunk - 1] = $this->seq[$i];
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])) {
191 $matches = $ymatches[$line];
193 while (list(, $y) = each($matches)) {
194 if (empty($this->in_seq[$y])) {
195 $k = $this->_lcsPos($y);
197 $ymids[$k] = $ymids[$k - 1];
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;
208 $this->in_seq[$y] = 1;
209 } elseif (empty($this->in_seq[$y])) {
210 $k = $this->_lcsPos($y);
212 $ymids[$k] = $ymids[$k - 1];
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);
223 $seps[] = $flip ? array($y1, $x1) : array($x1, $y1);
225 $seps[] = $flip ? array($ylim, $xlim) : array($xlim, $ylim);
227 return array($this->lcs, $seps);
230 function _lcsPos($ypos)
233 if ($end == 0 || $ypos > $this->seq[$end]) {
234 $this->seq[++$this->lcs] = $ypos;
235 $this->in_seq[$ypos] = 1;
240 while ($beg < $end) {
241 $mid = (int)(($beg + $end) / 2);
242 if ($ypos > $this->seq[$mid]) {
249 assert($ypos != $this->seq[$end]);
251 $this->in_seq[$this->seq[$end]] = false;
252 $this->seq[$end] = $ypos;
253 $this->in_seq[$ypos] = 1;
258 * Finds LCS of two sequences.
260 * The results are recorded in the vectors $this->{x,y}changed[], by
261 * storing a 1 in the element for each line that is an insertion or
262 * deletion (ie. is not in the LCS).
264 * The subsequence of file 0 is (XOFF, XLIM) and likewise for file 1.
266 * Note that XLIM, YLIM are exclusive bounds. All line numbers are
267 * origin-0 and discarded lines are not counted.
269 function _compareseq ($xoff, $xlim, $yoff, $ylim)
271 /* Slide down the bottom initial diagonal. */
272 while ($xoff < $xlim && $yoff < $ylim
273 && $this->xv[$xoff] == $this->yv[$yoff]) {
278 /* Slide up the top initial diagonal. */
279 while ($xlim > $xoff && $ylim > $yoff
280 && $this->xv[$xlim - 1] == $this->yv[$ylim - 1]) {
285 if ($xoff == $xlim || $yoff == $ylim) {
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;
293 = $this->_diag($xoff, $xlim, $yoff, $ylim, $nchunks);
297 /* X and Y sequences have no common subsequence: mark all
299 while ($yoff < $ylim) {
300 $this->ychanged[$this->yind[$yoff++]] = 1;
302 while ($xoff < $xlim) {
303 $this->xchanged[$this->xind[$xoff++]] = 1;
306 /* Use the partitions to split this problem into subproblems. */
309 while ($pt2 = next($seps)) {
310 $this->_compareseq ($pt1[0], $pt2[0], $pt1[1], $pt2[1]);
317 * Adjusts inserts/deletes of identical lines to join changes as much as
320 * We do something when a run of changed lines include a line at one end
321 * and has an excluded, identical line at the other. We are free to
322 * choose which identical line is included. `compareseq' usually chooses
323 * the one at the beginning, but usually it is cleaner to consider the
324 * following identical line to be the "change".
326 * This is extracted verbatim from analyze.c (GNU diffutils-2.7).
328 function _shiftBoundaries($lines, &$changed, $other_changed)
333 assert('count($lines) == count($changed)');
334 $len = count($lines);
335 $other_len = count($other_changed);
338 /* Scan forward to find the beginning of another run of
339 * changes. Also keep track of the corresponding point in the
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
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]) {
353 while ($i < $len && ! $changed[$i]) {
354 assert('$j < $other_len && ! $other_changed[$j]');
356 while ($j < $other_len && $other_changed[$j]) {
367 /* Find the end of this run of changes. */
368 while (++$i < $len && $changed[$i]) {
373 /* Record the length of this run of changes, so that we can
374 * later determine whether the run has grown. */
375 $runlength = $i - $start;
377 /* Move the changed region back, so long as the previous
378 * unchanged line matches the last changed one. This merges
379 * with previous changed regions. */
380 while ($start > 0 && $lines[$start - 1] == $lines[$i - 1]) {
381 $changed[--$start] = 1;
382 $changed[--$i] = false;
383 while ($start > 0 && $changed[$start - 1]) {
387 while ($other_changed[--$j]) {
390 assert('$j >= 0 && !$other_changed[$j]');
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
397 $corresponding = $j < $other_len ? $i : $len;
399 /* Move the changed region forward, so long as the first
400 * changed line matches the following unchanged one. This
401 * merges with following changed regions. Do this second, so
402 * that if there are no merges, the changed region is moved
403 * forward as far as possible. */
404 while ($i < $len && $lines[$start] == $lines[$i]) {
405 $changed[$start++] = false;
407 while ($i < $len && $changed[$i]) {
411 assert('$j < $other_len && ! $other_changed[$j]');
413 if ($j < $other_len && $other_changed[$j]) {
415 while ($j < $other_len && $other_changed[$j]) {
420 } while ($runlength != $i - $start);
422 /* If possible, move the fully-merged run of changes back to a
423 * corresponding run in the other file. */
424 while ($corresponding < $i) {
425 $changed[--$start] = 1;
428 while ($other_changed[--$j]) {
431 assert('$j >= 0 && !$other_changed[$j]');