1 /* $NetBSD: analyze.c,v 1.1.1.1 2016/01/13 03:15:30 christos Exp $ */
2
3 /* Analyze file differences for GNU DIFF.
4
5 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1998, 2001, 2002
6 Free Software Foundation, Inc.
7
8 This file is part of GNU DIFF.
9
10 GNU DIFF is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
13 any later version.
14
15 GNU DIFF is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
19
20 You should have received a copy of the GNU General Public License
21 along with this program; see the file COPYING.
22 If not, write to the Free Software Foundation,
23 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
24
25 /* The basic algorithm is described in:
26 "An O(ND) Difference Algorithm and its Variations", Eugene Myers,
27 Algorithmica Vol. 1 No. 2, 1986, pp. 251-266;
28 see especially section 4.2, which describes the variation used below.
29 Unless the --minimal option is specified, this code uses the TOO_EXPENSIVE
30 heuristic, by Paul Eggert, to limit the cost to O(N**1.5 log N)
31 at the price of producing suboptimal output for large inputs with
32 many differences.
33
34 The basic algorithm was independently discovered as described in:
35 "Algorithms for Approximate String Matching", E. Ukkonen,
36 Information and Control Vol. 64, 1985, pp. 100-118. */
37
38 #include "diff.h"
39 #include <cmpbuf.h>
40 #include <error.h>
41 #include <regex.h>
42 #include <xalloc.h>
43
44 static lin *xvec, *yvec; /* Vectors being compared. */
45 static lin *fdiag; /* Vector, indexed by diagonal, containing
46 1 + the X coordinate of the point furthest
47 along the given diagonal in the forward
48 search of the edit matrix. */
49 static lin *bdiag; /* Vector, indexed by diagonal, containing
50 the X coordinate of the point furthest
51 along the given diagonal in the backward
52 search of the edit matrix. */
53 static lin too_expensive; /* Edit scripts longer than this are too
54 expensive to compute. */
55
56 #define SNAKE_LIMIT 20 /* Snakes bigger than this are considered `big'. */
57
58 struct partition
59 {
60 lin xmid, ymid; /* Midpoints of this partition. */
61 bool lo_minimal; /* Nonzero if low half will be analyzed minimally. */
62 bool hi_minimal; /* Likewise for high half. */
63 };
64
65 /* Find the midpoint of the shortest edit script for a specified
66 portion of the two files.
67
68 Scan from the beginnings of the files, and simultaneously from the ends,
69 doing a breadth-first search through the space of edit-sequence.
70 When the two searches meet, we have found the midpoint of the shortest
71 edit sequence.
72
73 If FIND_MINIMAL is nonzero, find the minimal edit script regardless
74 of expense. Otherwise, if the search is too expensive, use
75 heuristics to stop the search and report a suboptimal answer.
76
77 Set PART->(xmid,ymid) to the midpoint (XMID,YMID). The diagonal number
78 XMID - YMID equals the number of inserted lines minus the number
79 of deleted lines (counting only lines before the midpoint).
80 Return the approximate edit cost; this is the total number of
81 lines inserted or deleted (counting only lines before the midpoint),
82 unless a heuristic is used to terminate the search prematurely.
83
84 Set PART->lo_minimal to true iff the minimal edit script for the
85 left half of the partition is known; similarly for PART->hi_minimal.
86
87 This function assumes that the first lines of the specified portions
88 of the two files do not match, and likewise that the last lines do not
89 match. The caller must trim matching lines from the beginning and end
90 of the portions it is going to specify.
91
92 If we return the "wrong" partitions,
93 the worst this can do is cause suboptimal diff output.
94 It cannot cause incorrect diff output. */
95
96 static lin
diag(lin xoff,lin xlim,lin yoff,lin ylim,bool find_minimal,struct partition * part)97 diag (lin xoff, lin xlim, lin yoff, lin ylim, bool find_minimal,
98 struct partition *part)
99 {
100 lin *const fd = fdiag; /* Give the compiler a chance. */
101 lin *const bd = bdiag; /* Additional help for the compiler. */
102 lin const *const xv = xvec; /* Still more help for the compiler. */
103 lin const *const yv = yvec; /* And more and more . . . */
104 lin const dmin = xoff - ylim; /* Minimum valid diagonal. */
105 lin const dmax = xlim - yoff; /* Maximum valid diagonal. */
106 lin const fmid = xoff - yoff; /* Center diagonal of top-down search. */
107 lin const bmid = xlim - ylim; /* Center diagonal of bottom-up search. */
108 lin fmin = fmid, fmax = fmid; /* Limits of top-down search. */
109 lin bmin = bmid, bmax = bmid; /* Limits of bottom-up search. */
110 lin c; /* Cost. */
111 bool odd = (fmid - bmid) & 1; /* True if southeast corner is on an odd
112 diagonal with respect to the northwest. */
113
114 fd[fmid] = xoff;
115 bd[bmid] = xlim;
116
117 for (c = 1;; ++c)
118 {
119 lin d; /* Active diagonal. */
120 bool big_snake = 0;
121
122 /* Extend the top-down search by an edit step in each diagonal. */
123 fmin > dmin ? fd[--fmin - 1] = -1 : ++fmin;
124 fmax < dmax ? fd[++fmax + 1] = -1 : --fmax;
125 for (d = fmax; d >= fmin; d -= 2)
126 {
127 lin x, y, oldx, tlo = fd[d - 1], thi = fd[d + 1];
128
129 if (tlo >= thi)
130 x = tlo + 1;
131 else
132 x = thi;
133 oldx = x;
134 y = x - d;
135 while (x < xlim && y < ylim && xv[x] == yv[y])
136 ++x, ++y;
137 if (x - oldx > SNAKE_LIMIT)
138 big_snake = 1;
139 fd[d] = x;
140 if (odd && bmin <= d && d <= bmax && bd[d] <= x)
141 {
142 part->xmid = x;
143 part->ymid = y;
144 part->lo_minimal = part->hi_minimal = 1;
145 return 2 * c - 1;
146 }
147 }
148
149 /* Similarly extend the bottom-up search. */
150 bmin > dmin ? bd[--bmin - 1] = LIN_MAX : ++bmin;
151 bmax < dmax ? bd[++bmax + 1] = LIN_MAX : --bmax;
152 for (d = bmax; d >= bmin; d -= 2)
153 {
154 lin x, y, oldx, tlo = bd[d - 1], thi = bd[d + 1];
155
156 if (tlo < thi)
157 x = tlo;
158 else
159 x = thi - 1;
160 oldx = x;
161 y = x - d;
162 while (x > xoff && y > yoff && xv[x - 1] == yv[y - 1])
163 --x, --y;
164 if (oldx - x > SNAKE_LIMIT)
165 big_snake = 1;
166 bd[d] = x;
167 if (!odd && fmin <= d && d <= fmax && x <= fd[d])
168 {
169 part->xmid = x;
170 part->ymid = y;
171 part->lo_minimal = part->hi_minimal = 1;
172 return 2 * c;
173 }
174 }
175
176 if (find_minimal)
177 continue;
178
179 /* Heuristic: check occasionally for a diagonal that has made
180 lots of progress compared with the edit distance.
181 If we have any such, find the one that has made the most
182 progress and return it as if it had succeeded.
183
184 With this heuristic, for files with a constant small density
185 of changes, the algorithm is linear in the file size. */
186
187 if (200 < c && big_snake && speed_large_files)
188 {
189 lin best;
190
191 best = 0;
192 for (d = fmax; d >= fmin; d -= 2)
193 {
194 lin dd = d - fmid;
195 lin x = fd[d];
196 lin y = x - d;
197 lin v = (x - xoff) * 2 - dd;
198 if (v > 12 * (c + (dd < 0 ? -dd : dd)))
199 {
200 if (v > best
201 && xoff + SNAKE_LIMIT <= x && x < xlim
202 && yoff + SNAKE_LIMIT <= y && y < ylim)
203 {
204 /* We have a good enough best diagonal;
205 now insist that it end with a significant snake. */
206 int k;
207
208 for (k = 1; xv[x - k] == yv[y - k]; k++)
209 if (k == SNAKE_LIMIT)
210 {
211 best = v;
212 part->xmid = x;
213 part->ymid = y;
214 break;
215 }
216 }
217 }
218 }
219 if (best > 0)
220 {
221 part->lo_minimal = 1;
222 part->hi_minimal = 0;
223 return 2 * c - 1;
224 }
225
226 best = 0;
227 for (d = bmax; d >= bmin; d -= 2)
228 {
229 lin dd = d - bmid;
230 lin x = bd[d];
231 lin y = x - d;
232 lin v = (xlim - x) * 2 + dd;
233 if (v > 12 * (c + (dd < 0 ? -dd : dd)))
234 {
235 if (v > best
236 && xoff < x && x <= xlim - SNAKE_LIMIT
237 && yoff < y && y <= ylim - SNAKE_LIMIT)
238 {
239 /* We have a good enough best diagonal;
240 now insist that it end with a significant snake. */
241 int k;
242
243 for (k = 0; xv[x + k] == yv[y + k]; k++)
244 if (k == SNAKE_LIMIT - 1)
245 {
246 best = v;
247 part->xmid = x;
248 part->ymid = y;
249 break;
250 }
251 }
252 }
253 }
254 if (best > 0)
255 {
256 part->lo_minimal = 0;
257 part->hi_minimal = 1;
258 return 2 * c - 1;
259 }
260 }
261
262 /* Heuristic: if we've gone well beyond the call of duty,
263 give up and report halfway between our best results so far. */
264 if (c >= too_expensive)
265 {
266 lin fxybest, fxbest;
267 lin bxybest, bxbest;
268
269 fxbest = bxbest = 0; /* Pacify `gcc -Wall'. */
270
271 /* Find forward diagonal that maximizes X + Y. */
272 fxybest = -1;
273 for (d = fmax; d >= fmin; d -= 2)
274 {
275 lin x = MIN (fd[d], xlim);
276 lin y = x - d;
277 if (ylim < y)
278 x = ylim + d, y = ylim;
279 if (fxybest < x + y)
280 {
281 fxybest = x + y;
282 fxbest = x;
283 }
284 }
285
286 /* Find backward diagonal that minimizes X + Y. */
287 bxybest = LIN_MAX;
288 for (d = bmax; d >= bmin; d -= 2)
289 {
290 lin x = MAX (xoff, bd[d]);
291 lin y = x - d;
292 if (y < yoff)
293 x = yoff + d, y = yoff;
294 if (x + y < bxybest)
295 {
296 bxybest = x + y;
297 bxbest = x;
298 }
299 }
300
301 /* Use the better of the two diagonals. */
302 if ((xlim + ylim) - bxybest < fxybest - (xoff + yoff))
303 {
304 part->xmid = fxbest;
305 part->ymid = fxybest - fxbest;
306 part->lo_minimal = 1;
307 part->hi_minimal = 0;
308 }
309 else
310 {
311 part->xmid = bxbest;
312 part->ymid = bxybest - bxbest;
313 part->lo_minimal = 0;
314 part->hi_minimal = 1;
315 }
316 return 2 * c - 1;
317 }
318 }
319 }
320
321 /* Compare in detail contiguous subsequences of the two files
322 which are known, as a whole, to match each other.
323
324 The results are recorded in the vectors files[N].changed, by
325 storing 1 in the element for each line that is an insertion or deletion.
326
327 The subsequence of file 0 is [XOFF, XLIM) and likewise for file 1.
328
329 Note that XLIM, YLIM are exclusive bounds.
330 All line numbers are origin-0 and discarded lines are not counted.
331
332 If FIND_MINIMAL, find a minimal difference no matter how
333 expensive it is. */
334
335 static void
compareseq(lin xoff,lin xlim,lin yoff,lin ylim,bool find_minimal)336 compareseq (lin xoff, lin xlim, lin yoff, lin ylim, bool find_minimal)
337 {
338 lin * const xv = xvec; /* Help the compiler. */
339 lin * const yv = yvec;
340
341 /* Slide down the bottom initial diagonal. */
342 while (xoff < xlim && yoff < ylim && xv[xoff] == yv[yoff])
343 ++xoff, ++yoff;
344 /* Slide up the top initial diagonal. */
345 while (xlim > xoff && ylim > yoff && xv[xlim - 1] == yv[ylim - 1])
346 --xlim, --ylim;
347
348 /* Handle simple cases. */
349 if (xoff == xlim)
350 while (yoff < ylim)
351 files[1].changed[files[1].realindexes[yoff++]] = 1;
352 else if (yoff == ylim)
353 while (xoff < xlim)
354 files[0].changed[files[0].realindexes[xoff++]] = 1;
355 else
356 {
357 lin c;
358 struct partition part;
359
360 /* Find a point of correspondence in the middle of the files. */
361
362 c = diag (xoff, xlim, yoff, ylim, find_minimal, &part);
363
364 if (c == 1)
365 {
366 /* This should be impossible, because it implies that
367 one of the two subsequences is empty,
368 and that case was handled above without calling `diag'.
369 Let's verify that this is true. */
370 abort ();
371 #if 0
372 /* The two subsequences differ by a single insert or delete;
373 record it and we are done. */
374 if (part.xmid - part.ymid < xoff - yoff)
375 files[1].changed[files[1].realindexes[part.ymid - 1]] = 1;
376 else
377 files[0].changed[files[0].realindexes[part.xmid]] = 1;
378 #endif
379 }
380 else
381 {
382 /* Use the partitions to split this problem into subproblems. */
383 compareseq (xoff, part.xmid, yoff, part.ymid, part.lo_minimal);
384 compareseq (part.xmid, xlim, part.ymid, ylim, part.hi_minimal);
385 }
386 }
387 }
388
389 /* Discard lines from one file that have no matches in the other file.
390
391 A line which is discarded will not be considered by the actual
392 comparison algorithm; it will be as if that line were not in the file.
393 The file's `realindexes' table maps virtual line numbers
394 (which don't count the discarded lines) into real line numbers;
395 this is how the actual comparison algorithm produces results
396 that are comprehensible when the discarded lines are counted.
397
398 When we discard a line, we also mark it as a deletion or insertion
399 so that it will be printed in the output. */
400
401 static void
discard_confusing_lines(struct file_data filevec[])402 discard_confusing_lines (struct file_data filevec[])
403 {
404 int f;
405 lin i;
406 char *discarded[2];
407 lin *equiv_count[2];
408 lin *p;
409
410 /* Allocate our results. */
411 p = xmalloc ((filevec[0].buffered_lines + filevec[1].buffered_lines)
412 * (2 * sizeof *p));
413 for (f = 0; f < 2; f++)
414 {
415 filevec[f].undiscarded = p; p += filevec[f].buffered_lines;
416 filevec[f].realindexes = p; p += filevec[f].buffered_lines;
417 }
418
419 /* Set up equiv_count[F][I] as the number of lines in file F
420 that fall in equivalence class I. */
421
422 p = zalloc (filevec[0].equiv_max * (2 * sizeof *p));
423 equiv_count[0] = p;
424 equiv_count[1] = p + filevec[0].equiv_max;
425
426 for (i = 0; i < filevec[0].buffered_lines; ++i)
427 ++equiv_count[0][filevec[0].equivs[i]];
428 for (i = 0; i < filevec[1].buffered_lines; ++i)
429 ++equiv_count[1][filevec[1].equivs[i]];
430
431 /* Set up tables of which lines are going to be discarded. */
432
433 discarded[0] = zalloc (filevec[0].buffered_lines
434 + filevec[1].buffered_lines);
435 discarded[1] = discarded[0] + filevec[0].buffered_lines;
436
437 /* Mark to be discarded each line that matches no line of the other file.
438 If a line matches many lines, mark it as provisionally discardable. */
439
440 for (f = 0; f < 2; f++)
441 {
442 size_t end = filevec[f].buffered_lines;
443 char *discards = discarded[f];
444 lin *counts = equiv_count[1 - f];
445 lin *equivs = filevec[f].equivs;
446 size_t many = 5;
447 size_t tem = end / 64;
448
449 /* Multiply MANY by approximate square root of number of lines.
450 That is the threshold for provisionally discardable lines. */
451 while ((tem = tem >> 2) > 0)
452 many *= 2;
453
454 for (i = 0; i < end; i++)
455 {
456 lin nmatch;
457 if (equivs[i] == 0)
458 continue;
459 nmatch = counts[equivs[i]];
460 if (nmatch == 0)
461 discards[i] = 1;
462 else if (nmatch > many)
463 discards[i] = 2;
464 }
465 }
466
467 /* Don't really discard the provisional lines except when they occur
468 in a run of discardables, with nonprovisionals at the beginning
469 and end. */
470
471 for (f = 0; f < 2; f++)
472 {
473 lin end = filevec[f].buffered_lines;
474 register char *discards = discarded[f];
475
476 for (i = 0; i < end; i++)
477 {
478 /* Cancel provisional discards not in middle of run of discards. */
479 if (discards[i] == 2)
480 discards[i] = 0;
481 else if (discards[i] != 0)
482 {
483 /* We have found a nonprovisional discard. */
484 register lin j;
485 lin length;
486 lin provisional = 0;
487
488 /* Find end of this run of discardable lines.
489 Count how many are provisionally discardable. */
490 for (j = i; j < end; j++)
491 {
492 if (discards[j] == 0)
493 break;
494 if (discards[j] == 2)
495 ++provisional;
496 }
497
498 /* Cancel provisional discards at end, and shrink the run. */
499 while (j > i && discards[j - 1] == 2)
500 discards[--j] = 0, --provisional;
501
502 /* Now we have the length of a run of discardable lines
503 whose first and last are not provisional. */
504 length = j - i;
505
506 /* If 1/4 of the lines in the run are provisional,
507 cancel discarding of all provisional lines in the run. */
508 if (provisional * 4 > length)
509 {
510 while (j > i)
511 if (discards[--j] == 2)
512 discards[j] = 0;
513 }
514 else
515 {
516 register lin consec;
517 lin minimum = 1;
518 lin tem = length >> 2;
519
520 /* MINIMUM is approximate square root of LENGTH/4.
521 A subrun of two or more provisionals can stand
522 when LENGTH is at least 16.
523 A subrun of 4 or more can stand when LENGTH >= 64. */
524 while (0 < (tem >>= 2))
525 minimum <<= 1;
526 minimum++;
527
528 /* Cancel any subrun of MINIMUM or more provisionals
529 within the larger run. */
530 for (j = 0, consec = 0; j < length; j++)
531 if (discards[i + j] != 2)
532 consec = 0;
533 else if (minimum == ++consec)
534 /* Back up to start of subrun, to cancel it all. */
535 j -= consec;
536 else if (minimum < consec)
537 discards[i + j] = 0;
538
539 /* Scan from beginning of run
540 until we find 3 or more nonprovisionals in a row
541 or until the first nonprovisional at least 8 lines in.
542 Until that point, cancel any provisionals. */
543 for (j = 0, consec = 0; j < length; j++)
544 {
545 if (j >= 8 && discards[i + j] == 1)
546 break;
547 if (discards[i + j] == 2)
548 consec = 0, discards[i + j] = 0;
549 else if (discards[i + j] == 0)
550 consec = 0;
551 else
552 consec++;
553 if (consec == 3)
554 break;
555 }
556
557 /* I advances to the last line of the run. */
558 i += length - 1;
559
560 /* Same thing, from end. */
561 for (j = 0, consec = 0; j < length; j++)
562 {
563 if (j >= 8 && discards[i - j] == 1)
564 break;
565 if (discards[i - j] == 2)
566 consec = 0, discards[i - j] = 0;
567 else if (discards[i - j] == 0)
568 consec = 0;
569 else
570 consec++;
571 if (consec == 3)
572 break;
573 }
574 }
575 }
576 }
577 }
578
579 /* Actually discard the lines. */
580 for (f = 0; f < 2; f++)
581 {
582 char *discards = discarded[f];
583 lin end = filevec[f].buffered_lines;
584 lin j = 0;
585 for (i = 0; i < end; ++i)
586 if (minimal || discards[i] == 0)
587 {
588 filevec[f].undiscarded[j] = filevec[f].equivs[i];
589 filevec[f].realindexes[j++] = i;
590 }
591 else
592 filevec[f].changed[i] = 1;
593 filevec[f].nondiscarded_lines = j;
594 }
595
596 free (discarded[0]);
597 free (equiv_count[0]);
598 }
599
600 /* Adjust inserts/deletes of identical lines to join changes
601 as much as possible.
602
603 We do something when a run of changed lines include a
604 line at one end and have an excluded, identical line at the other.
605 We are free to choose which identical line is included.
606 `compareseq' usually chooses the one at the beginning,
607 but usually it is cleaner to consider the following identical line
608 to be the "change". */
609
610 static void
shift_boundaries(struct file_data filevec[])611 shift_boundaries (struct file_data filevec[])
612 {
613 int f;
614
615 for (f = 0; f < 2; f++)
616 {
617 bool *changed = filevec[f].changed;
618 bool const *other_changed = filevec[1 - f].changed;
619 lin const *equivs = filevec[f].equivs;
620 lin i = 0;
621 lin j = 0;
622 lin i_end = filevec[f].buffered_lines;
623
624 while (1)
625 {
626 lin runlength, start, corresponding;
627
628 /* Scan forwards to find beginning of another run of changes.
629 Also keep track of the corresponding point in the other file. */
630
631 while (i < i_end && !changed[i])
632 {
633 while (other_changed[j++])
634 continue;
635 i++;
636 }
637
638 if (i == i_end)
639 break;
640
641 start = i;
642
643 /* Find the end of this run of changes. */
644
645 while (changed[++i])
646 continue;
647 while (other_changed[j])
648 j++;
649
650 do
651 {
652 /* Record the length of this run of changes, so that
653 we can later determine whether the run has grown. */
654 runlength = i - start;
655
656 /* Move the changed region back, so long as the
657 previous unchanged line matches the last changed one.
658 This merges with previous changed regions. */
659
660 while (start && equivs[start - 1] == equivs[i - 1])
661 {
662 changed[--start] = 1;
663 changed[--i] = 0;
664 while (changed[start - 1])
665 start--;
666 while (other_changed[--j])
667 continue;
668 }
669
670 /* Set CORRESPONDING to the end of the changed run, at the last
671 point where it corresponds to a changed run in the other file.
672 CORRESPONDING == I_END means no such point has been found. */
673 corresponding = other_changed[j - 1] ? i : i_end;
674
675 /* Move the changed region forward, so long as the
676 first changed line matches the following unchanged one.
677 This merges with following changed regions.
678 Do this second, so that if there are no merges,
679 the changed region is moved forward as far as possible. */
680
681 while (i != i_end && equivs[start] == equivs[i])
682 {
683 changed[start++] = 0;
684 changed[i++] = 1;
685 while (changed[i])
686 i++;
687 while (other_changed[++j])
688 corresponding = i;
689 }
690 }
691 while (runlength != i - start);
692
693 /* If possible, move the fully-merged run of changes
694 back to a corresponding run in the other file. */
695
696 while (corresponding < i)
697 {
698 changed[--start] = 1;
699 changed[--i] = 0;
700 while (other_changed[--j])
701 continue;
702 }
703 }
704 }
705 }
706
707 /* Cons an additional entry onto the front of an edit script OLD.
708 LINE0 and LINE1 are the first affected lines in the two files (origin 0).
709 DELETED is the number of lines deleted here from file 0.
710 INSERTED is the number of lines inserted here in file 1.
711
712 If DELETED is 0 then LINE0 is the number of the line before
713 which the insertion was done; vice versa for INSERTED and LINE1. */
714
715 static struct change *
add_change(lin line0,lin line1,lin deleted,lin inserted,struct change * old)716 add_change (lin line0, lin line1, lin deleted, lin inserted,
717 struct change *old)
718 {
719 struct change *new = xmalloc (sizeof *new);
720
721 new->line0 = line0;
722 new->line1 = line1;
723 new->inserted = inserted;
724 new->deleted = deleted;
725 new->link = old;
726 return new;
727 }
728
729 /* Scan the tables of which lines are inserted and deleted,
730 producing an edit script in reverse order. */
731
732 static struct change *
build_reverse_script(struct file_data const filevec[])733 build_reverse_script (struct file_data const filevec[])
734 {
735 struct change *script = 0;
736 bool *changed0 = filevec[0].changed;
737 bool *changed1 = filevec[1].changed;
738 lin len0 = filevec[0].buffered_lines;
739 lin len1 = filevec[1].buffered_lines;
740
741 /* Note that changedN[len0] does exist, and is 0. */
742
743 lin i0 = 0, i1 = 0;
744
745 while (i0 < len0 || i1 < len1)
746 {
747 if (changed0[i0] | changed1[i1])
748 {
749 lin line0 = i0, line1 = i1;
750
751 /* Find # lines changed here in each file. */
752 while (changed0[i0]) ++i0;
753 while (changed1[i1]) ++i1;
754
755 /* Record this change. */
756 script = add_change (line0, line1, i0 - line0, i1 - line1, script);
757 }
758
759 /* We have reached lines in the two files that match each other. */
760 i0++, i1++;
761 }
762
763 return script;
764 }
765
766 /* Scan the tables of which lines are inserted and deleted,
767 producing an edit script in forward order. */
768
769 static struct change *
build_script(struct file_data const filevec[])770 build_script (struct file_data const filevec[])
771 {
772 struct change *script = 0;
773 bool *changed0 = filevec[0].changed;
774 bool *changed1 = filevec[1].changed;
775 lin i0 = filevec[0].buffered_lines, i1 = filevec[1].buffered_lines;
776
777 /* Note that changedN[-1] does exist, and is 0. */
778
779 while (i0 >= 0 || i1 >= 0)
780 {
781 if (changed0[i0 - 1] | changed1[i1 - 1])
782 {
783 lin line0 = i0, line1 = i1;
784
785 /* Find # lines changed here in each file. */
786 while (changed0[i0 - 1]) --i0;
787 while (changed1[i1 - 1]) --i1;
788
789 /* Record this change. */
790 script = add_change (i0, i1, line0 - i0, line1 - i1, script);
791 }
792
793 /* We have reached lines in the two files that match each other. */
794 i0--, i1--;
795 }
796
797 return script;
798 }
799
800 /* If CHANGES, briefly report that two files differed.
801 Return 2 if trouble, CHANGES otherwise. */
802 static int
briefly_report(int changes,struct file_data const filevec[])803 briefly_report (int changes, struct file_data const filevec[])
804 {
805 if (changes)
806 {
807 char const *label0 = file_label[0] ? file_label[0] : filevec[0].name;
808 char const *label1 = file_label[1] ? file_label[1] : filevec[1].name;
809
810 if (brief)
811 message ("Files %s and %s differ\n", label0, label1);
812 else
813 {
814 message ("Binary files %s and %s differ\n", label0, label1);
815 changes = 2;
816 }
817 }
818
819 return changes;
820 }
821
822 /* Report the differences of two files. */
823 int
diff_2_files(struct comparison * cmp)824 diff_2_files (struct comparison *cmp)
825 {
826 lin diags;
827 int f;
828 struct change *e, *p;
829 struct change *script;
830 int changes;
831
832
833 /* If we have detected that either file is binary,
834 compare the two files as binary. This can happen
835 only when the first chunk is read.
836 Also, --brief without any --ignore-* options means
837 we can speed things up by treating the files as binary. */
838
839 if (read_files (cmp->file, files_can_be_treated_as_binary))
840 {
841 /* Files with different lengths must be different. */
842 if (cmp->file[0].stat.st_size != cmp->file[1].stat.st_size
843 && (cmp->file[0].desc < 0 || S_ISREG (cmp->file[0].stat.st_mode))
844 && (cmp->file[1].desc < 0 || S_ISREG (cmp->file[1].stat.st_mode)))
845 changes = 1;
846
847 /* Standard input equals itself. */
848 else if (cmp->file[0].desc == cmp->file[1].desc)
849 changes = 0;
850
851 else
852 /* Scan both files, a buffer at a time, looking for a difference. */
853 {
854 /* Allocate same-sized buffers for both files. */
855 size_t lcm_max = PTRDIFF_MAX - 1;
856 size_t buffer_size =
857 buffer_lcm (sizeof (word),
858 buffer_lcm (STAT_BLOCKSIZE (cmp->file[0].stat),
859 STAT_BLOCKSIZE (cmp->file[1].stat),
860 lcm_max),
861 lcm_max);
862 for (f = 0; f < 2; f++)
863 cmp->file[f].buffer = xrealloc (cmp->file[f].buffer, buffer_size);
864
865 for (;; cmp->file[0].buffered = cmp->file[1].buffered = 0)
866 {
867 /* Read a buffer's worth from both files. */
868 for (f = 0; f < 2; f++)
869 if (0 <= cmp->file[f].desc)
870 file_block_read (&cmp->file[f],
871 buffer_size - cmp->file[f].buffered);
872
873 /* If the buffers differ, the files differ. */
874 if (cmp->file[0].buffered != cmp->file[1].buffered
875 || memcmp (cmp->file[0].buffer,
876 cmp->file[1].buffer,
877 cmp->file[0].buffered))
878 {
879 changes = 1;
880 break;
881 }
882
883 /* If we reach end of file, the files are the same. */
884 if (cmp->file[0].buffered != buffer_size)
885 {
886 changes = 0;
887 break;
888 }
889 }
890 }
891
892 changes = briefly_report (changes, cmp->file);
893 }
894 else
895 {
896 /* Allocate vectors for the results of comparison:
897 a flag for each line of each file, saying whether that line
898 is an insertion or deletion.
899 Allocate an extra element, always 0, at each end of each vector. */
900
901 size_t s = cmp->file[0].buffered_lines + cmp->file[1].buffered_lines + 4;
902 bool *flag_space = zalloc (s * sizeof *flag_space);
903 cmp->file[0].changed = flag_space + 1;
904 cmp->file[1].changed = flag_space + cmp->file[0].buffered_lines + 3;
905
906 /* Some lines are obviously insertions or deletions
907 because they don't match anything. Detect them now, and
908 avoid even thinking about them in the main comparison algorithm. */
909
910 discard_confusing_lines (cmp->file);
911
912 /* Now do the main comparison algorithm, considering just the
913 undiscarded lines. */
914
915 xvec = cmp->file[0].undiscarded;
916 yvec = cmp->file[1].undiscarded;
917 diags = (cmp->file[0].nondiscarded_lines
918 + cmp->file[1].nondiscarded_lines + 3);
919 fdiag = xmalloc (diags * (2 * sizeof *fdiag));
920 bdiag = fdiag + diags;
921 fdiag += cmp->file[1].nondiscarded_lines + 1;
922 bdiag += cmp->file[1].nondiscarded_lines + 1;
923
924 /* Set TOO_EXPENSIVE to be approximate square root of input size,
925 bounded below by 256. */
926 too_expensive = 1;
927 for (; diags != 0; diags >>= 2)
928 too_expensive <<= 1;
929 too_expensive = MAX (256, too_expensive);
930
931 files[0] = cmp->file[0];
932 files[1] = cmp->file[1];
933
934 compareseq (0, cmp->file[0].nondiscarded_lines,
935 0, cmp->file[1].nondiscarded_lines, minimal);
936
937 free (fdiag - (cmp->file[1].nondiscarded_lines + 1));
938
939 /* Modify the results slightly to make them prettier
940 in cases where that can validly be done. */
941
942 shift_boundaries (cmp->file);
943
944 /* Get the results of comparison in the form of a chain
945 of `struct change's -- an edit script. */
946
947 if (output_style == OUTPUT_ED)
948 script = build_reverse_script (cmp->file);
949 else
950 script = build_script (cmp->file);
951
952 /* Set CHANGES if we had any diffs.
953 If some changes are ignored, we must scan the script to decide. */
954 if (ignore_blank_lines || ignore_regexp.fastmap)
955 {
956 struct change *next = script;
957 changes = 0;
958
959 while (next && changes == 0)
960 {
961 struct change *this, *end;
962 lin first0, last0, first1, last1;
963
964 /* Find a set of changes that belong together. */
965 this = next;
966 end = find_change (next);
967
968 /* Disconnect them from the rest of the changes, making them
969 a hunk, and remember the rest for next iteration. */
970 next = end->link;
971 end->link = 0;
972
973 /* Determine whether this hunk is really a difference. */
974 if (analyze_hunk (this, &first0, &last0, &first1, &last1))
975 changes = 1;
976
977 /* Reconnect the script so it will all be freed properly. */
978 end->link = next;
979 }
980 }
981 else
982 changes = (script != 0);
983
984 if (brief)
985 changes = briefly_report (changes, cmp->file);
986 else
987 {
988 if (changes | !no_diff_means_no_output)
989 {
990 /* Record info for starting up output,
991 to be used if and when we have some output to print. */
992 setup_output (file_label[0] ? file_label[0] : cmp->file[0].name,
993 file_label[1] ? file_label[1] : cmp->file[1].name,
994 cmp->parent != 0);
995
996 switch (output_style)
997 {
998 case OUTPUT_CONTEXT:
999 print_context_script (script, 0);
1000 break;
1001
1002 case OUTPUT_UNIFIED:
1003 print_context_script (script, 1);
1004 break;
1005
1006 case OUTPUT_ED:
1007 print_ed_script (script);
1008 break;
1009
1010 case OUTPUT_FORWARD_ED:
1011 pr_forward_ed_script (script);
1012 break;
1013
1014 case OUTPUT_RCS:
1015 print_rcs_script (script);
1016 break;
1017
1018 case OUTPUT_NORMAL:
1019 print_normal_script (script);
1020 break;
1021
1022 case OUTPUT_IFDEF:
1023 print_ifdef_script (script);
1024 break;
1025
1026 case OUTPUT_SDIFF:
1027 print_sdiff_script (script);
1028 break;
1029
1030 default:
1031 abort ();
1032 }
1033
1034 finish_output ();
1035 }
1036 }
1037
1038 free (cmp->file[0].undiscarded);
1039
1040 free (flag_space);
1041
1042 for (f = 0; f < 2; f++)
1043 {
1044 free (cmp->file[f].equivs);
1045 free (cmp->file[f].linbuf + cmp->file[f].linbuf_base);
1046 }
1047
1048 for (e = script; e; e = p)
1049 {
1050 p = e->link;
1051 free (e);
1052 }
1053
1054 if (! ROBUST_OUTPUT_STYLE (output_style))
1055 for (f = 0; f < 2; ++f)
1056 if (cmp->file[f].missing_newline)
1057 {
1058 error (0, 0, "%s: %s\n",
1059 file_label[f] ? file_label[f] : cmp->file[f].name,
1060 _("No newline at end of file"));
1061 changes = 2;
1062 }
1063 }
1064
1065 if (cmp->file[0].buffer != cmp->file[1].buffer)
1066 free (cmp->file[0].buffer);
1067 free (cmp->file[1].buffer);
1068
1069 return changes;
1070 }
1071