repack: introduce repack.writeBitmaps config option
[git.git] / line-log.c
blob8b6e497b3ff187ee328cba395f30d3db7f86c0b4
1 #include "git-compat-util.h"
2 #include "line-range.h"
3 #include "cache.h"
4 #include "tag.h"
5 #include "blob.h"
6 #include "tree.h"
7 #include "diff.h"
8 #include "commit.h"
9 #include "decorate.h"
10 #include "revision.h"
11 #include "xdiff-interface.h"
12 #include "strbuf.h"
13 #include "log-tree.h"
14 #include "graph.h"
15 #include "userdiff.h"
16 #include "line-log.h"
18 static void range_set_grow(struct range_set *rs, size_t extra)
20 ALLOC_GROW(rs->ranges, rs->nr + extra, rs->alloc);
23 /* Either initialization would be fine */
24 #define RANGE_SET_INIT {0}
26 void range_set_init(struct range_set *rs, size_t prealloc)
28 rs->alloc = rs->nr = 0;
29 rs->ranges = NULL;
30 if (prealloc)
31 range_set_grow(rs, prealloc);
34 void range_set_release(struct range_set *rs)
36 free(rs->ranges);
37 rs->alloc = rs->nr = 0;
38 rs->ranges = NULL;
41 /* dst must be uninitialized! */
42 static void range_set_copy(struct range_set *dst, struct range_set *src)
44 range_set_init(dst, src->nr);
45 memcpy(dst->ranges, src->ranges, src->nr*sizeof(struct range_set));
46 dst->nr = src->nr;
48 static void range_set_move(struct range_set *dst, struct range_set *src)
50 range_set_release(dst);
51 dst->ranges = src->ranges;
52 dst->nr = src->nr;
53 dst->alloc = src->alloc;
54 src->ranges = NULL;
55 src->alloc = src->nr = 0;
58 /* tack on a _new_ range _at the end_ */
59 void range_set_append_unsafe(struct range_set *rs, long a, long b)
61 assert(a <= b);
62 range_set_grow(rs, 1);
63 rs->ranges[rs->nr].start = a;
64 rs->ranges[rs->nr].end = b;
65 rs->nr++;
68 void range_set_append(struct range_set *rs, long a, long b)
70 assert(rs->nr == 0 || rs->ranges[rs->nr-1].end <= a);
71 range_set_append_unsafe(rs, a, b);
74 static int range_cmp(const void *_r, const void *_s)
76 const struct range *r = _r;
77 const struct range *s = _s;
79 /* this could be simply 'return r.start-s.start', but for the types */
80 if (r->start == s->start)
81 return 0;
82 if (r->start < s->start)
83 return -1;
84 return 1;
88 * Check that the ranges are non-empty, sorted and non-overlapping
90 static void range_set_check_invariants(struct range_set *rs)
92 int i;
94 if (!rs)
95 return;
97 if (rs->nr)
98 assert(rs->ranges[0].start < rs->ranges[0].end);
100 for (i = 1; i < rs->nr; i++) {
101 assert(rs->ranges[i-1].end < rs->ranges[i].start);
102 assert(rs->ranges[i].start < rs->ranges[i].end);
107 * In-place pass of sorting and merging the ranges in the range set,
108 * to establish the invariants when we get the ranges from the user
110 void sort_and_merge_range_set(struct range_set *rs)
112 int i;
113 int o = 0; /* output cursor */
115 qsort(rs->ranges, rs->nr, sizeof(struct range), range_cmp);
117 for (i = 0; i < rs->nr; i++) {
118 if (rs->ranges[i].start == rs->ranges[i].end)
119 continue;
120 if (o > 0 && rs->ranges[i].start <= rs->ranges[o-1].end) {
121 if (rs->ranges[o-1].end < rs->ranges[i].end)
122 rs->ranges[o-1].end = rs->ranges[i].end;
123 } else {
124 rs->ranges[o].start = rs->ranges[i].start;
125 rs->ranges[o].end = rs->ranges[i].end;
126 o++;
129 assert(o <= rs->nr);
130 rs->nr = o;
132 range_set_check_invariants(rs);
136 * Union of range sets (i.e., sets of line numbers). Used to merge
137 * them when searches meet at a common ancestor.
139 * This is also where the ranges are consolidated into canonical form:
140 * overlapping and adjacent ranges are merged, and empty ranges are
141 * removed.
143 static void range_set_union(struct range_set *out,
144 struct range_set *a, struct range_set *b)
146 int i = 0, j = 0, o = 0;
147 struct range *ra = a->ranges;
148 struct range *rb = b->ranges;
149 /* cannot make an alias of out->ranges: it may change during grow */
151 assert(out->nr == 0);
152 while (i < a->nr || j < b->nr) {
153 struct range *new;
154 if (i < a->nr && j < b->nr) {
155 if (ra[i].start < rb[j].start)
156 new = &ra[i++];
157 else if (ra[i].start > rb[j].start)
158 new = &rb[j++];
159 else if (ra[i].end < rb[j].end)
160 new = &ra[i++];
161 else
162 new = &rb[j++];
163 } else if (i < a->nr) /* b exhausted */
164 new = &ra[i++];
165 else /* a exhausted */
166 new = &rb[j++];
167 if (new->start == new->end)
168 ; /* empty range */
169 else if (!o || out->ranges[o-1].end < new->start) {
170 range_set_grow(out, 1);
171 out->ranges[o].start = new->start;
172 out->ranges[o].end = new->end;
173 o++;
174 } else if (out->ranges[o-1].end < new->end) {
175 out->ranges[o-1].end = new->end;
178 out->nr = o;
182 * Difference of range sets (out = a \ b). Pass the "interesting"
183 * ranges as 'a' and the target side of the diff as 'b': it removes
184 * the ranges for which the commit is responsible.
186 static void range_set_difference(struct range_set *out,
187 struct range_set *a, struct range_set *b)
189 int i, j = 0;
190 for (i = 0; i < a->nr; i++) {
191 long start = a->ranges[i].start;
192 long end = a->ranges[i].end;
193 while (start < end) {
194 while (j < b->nr && start >= b->ranges[j].end)
196 * a: |-------
197 * b: ------|
199 j++;
200 if (j >= b->nr || end < b->ranges[j].start) {
202 * b exhausted, or
203 * a: ----|
204 * b: |----
206 range_set_append(out, start, end);
207 break;
209 if (start >= b->ranges[j].start) {
211 * a: |--????
212 * b: |------|
214 start = b->ranges[j].end;
215 } else if (end > b->ranges[j].start) {
217 * a: |-----|
218 * b: |--?????
220 if (start < b->ranges[j].start)
221 range_set_append(out, start, b->ranges[j].start);
222 start = b->ranges[j].end;
228 static void diff_ranges_init(struct diff_ranges *diff)
230 range_set_init(&diff->parent, 0);
231 range_set_init(&diff->target, 0);
234 static void diff_ranges_release(struct diff_ranges *diff)
236 range_set_release(&diff->parent);
237 range_set_release(&diff->target);
240 void line_log_data_init(struct line_log_data *r)
242 memset(r, 0, sizeof(struct line_log_data));
243 range_set_init(&r->ranges, 0);
246 static void line_log_data_clear(struct line_log_data *r)
248 range_set_release(&r->ranges);
249 if (r->pair)
250 diff_free_filepair(r->pair);
253 static void free_line_log_data(struct line_log_data *r)
255 while (r) {
256 struct line_log_data *next = r->next;
257 line_log_data_clear(r);
258 free(r);
259 r = next;
263 static struct line_log_data *
264 search_line_log_data(struct line_log_data *list, const char *path,
265 struct line_log_data **insertion_point)
267 struct line_log_data *p = list;
268 if (insertion_point)
269 *insertion_point = NULL;
270 while (p) {
271 int cmp = strcmp(p->path, path);
272 if (!cmp)
273 return p;
274 if (insertion_point && cmp < 0)
275 *insertion_point = p;
276 p = p->next;
278 return NULL;
282 * Note: takes ownership of 'path', which happens to be what the only
283 * caller needs.
285 static void line_log_data_insert(struct line_log_data **list,
286 char *path,
287 long begin, long end)
289 struct line_log_data *ip;
290 struct line_log_data *p = search_line_log_data(*list, path, &ip);
292 if (p) {
293 range_set_append_unsafe(&p->ranges, begin, end);
294 free(path);
295 return;
298 p = xcalloc(1, sizeof(struct line_log_data));
299 p->path = path;
300 range_set_append(&p->ranges, begin, end);
301 if (ip) {
302 p->next = ip->next;
303 ip->next = p;
304 } else {
305 p->next = *list;
306 *list = p;
310 struct collect_diff_cbdata {
311 struct diff_ranges *diff;
314 static int collect_diff_cb(long start_a, long count_a,
315 long start_b, long count_b,
316 void *data)
318 struct collect_diff_cbdata *d = data;
320 if (count_a >= 0)
321 range_set_append(&d->diff->parent, start_a, start_a + count_a);
322 if (count_b >= 0)
323 range_set_append(&d->diff->target, start_b, start_b + count_b);
325 return 0;
328 static void collect_diff(mmfile_t *parent, mmfile_t *target, struct diff_ranges *out)
330 struct collect_diff_cbdata cbdata = {NULL};
331 xpparam_t xpp;
332 xdemitconf_t xecfg;
333 xdemitcb_t ecb;
335 memset(&xpp, 0, sizeof(xpp));
336 memset(&xecfg, 0, sizeof(xecfg));
337 xecfg.ctxlen = xecfg.interhunkctxlen = 0;
339 cbdata.diff = out;
340 xecfg.hunk_func = collect_diff_cb;
341 memset(&ecb, 0, sizeof(ecb));
342 ecb.priv = &cbdata;
343 xdi_diff(parent, target, &xpp, &xecfg, &ecb);
347 * These are handy for debugging. Removing them with #if 0 silences
348 * the "unused function" warning.
350 #if 0
351 static void dump_range_set(struct range_set *rs, const char *desc)
353 int i;
354 printf("range set %s (%d items):\n", desc, rs->nr);
355 for (i = 0; i < rs->nr; i++)
356 printf("\t[%ld,%ld]\n", rs->ranges[i].start, rs->ranges[i].end);
359 static void dump_line_log_data(struct line_log_data *r)
361 char buf[4096];
362 while (r) {
363 snprintf(buf, 4096, "file %s\n", r->path);
364 dump_range_set(&r->ranges, buf);
365 r = r->next;
369 static void dump_diff_ranges(struct diff_ranges *diff, const char *desc)
371 int i;
372 assert(diff->parent.nr == diff->target.nr);
373 printf("diff ranges %s (%d items):\n", desc, diff->parent.nr);
374 printf("\tparent\ttarget\n");
375 for (i = 0; i < diff->parent.nr; i++) {
376 printf("\t[%ld,%ld]\t[%ld,%ld]\n",
377 diff->parent.ranges[i].start,
378 diff->parent.ranges[i].end,
379 diff->target.ranges[i].start,
380 diff->target.ranges[i].end);
383 #endif
386 static int ranges_overlap(struct range *a, struct range *b)
388 return !(a->end <= b->start || b->end <= a->start);
392 * Given a diff and the set of interesting ranges, determine all hunks
393 * of the diff which touch (overlap) at least one of the interesting
394 * ranges in the target.
396 static void diff_ranges_filter_touched(struct diff_ranges *out,
397 struct diff_ranges *diff,
398 struct range_set *rs)
400 int i, j = 0;
402 assert(out->target.nr == 0);
404 for (i = 0; i < diff->target.nr; i++) {
405 while (diff->target.ranges[i].start > rs->ranges[j].end) {
406 j++;
407 if (j == rs->nr)
408 return;
410 if (ranges_overlap(&diff->target.ranges[i], &rs->ranges[j])) {
411 range_set_append(&out->parent,
412 diff->parent.ranges[i].start,
413 diff->parent.ranges[i].end);
414 range_set_append(&out->target,
415 diff->target.ranges[i].start,
416 diff->target.ranges[i].end);
422 * Adjust the line counts in 'rs' to account for the lines
423 * added/removed in the diff.
425 static void range_set_shift_diff(struct range_set *out,
426 struct range_set *rs,
427 struct diff_ranges *diff)
429 int i, j = 0;
430 long offset = 0;
431 struct range *src = rs->ranges;
432 struct range *target = diff->target.ranges;
433 struct range *parent = diff->parent.ranges;
435 for (i = 0; i < rs->nr; i++) {
436 while (j < diff->target.nr && src[i].start >= target[j].start) {
437 offset += (parent[j].end-parent[j].start)
438 - (target[j].end-target[j].start);
439 j++;
441 range_set_append(out, src[i].start+offset, src[i].end+offset);
446 * Given a diff and the set of interesting ranges, map the ranges
447 * across the diff. That is: observe that the target commit takes
448 * blame for all the + (target-side) ranges. So for every pair of
449 * ranges in the diff that was touched, we remove the latter and add
450 * its parent side.
452 static void range_set_map_across_diff(struct range_set *out,
453 struct range_set *rs,
454 struct diff_ranges *diff,
455 struct diff_ranges **touched_out)
457 struct diff_ranges *touched = xmalloc(sizeof(*touched));
458 struct range_set tmp1 = RANGE_SET_INIT;
459 struct range_set tmp2 = RANGE_SET_INIT;
461 diff_ranges_init(touched);
462 diff_ranges_filter_touched(touched, diff, rs);
463 range_set_difference(&tmp1, rs, &touched->target);
464 range_set_shift_diff(&tmp2, &tmp1, diff);
465 range_set_union(out, &tmp2, &touched->parent);
466 range_set_release(&tmp1);
467 range_set_release(&tmp2);
469 *touched_out = touched;
472 static struct commit *check_single_commit(struct rev_info *revs)
474 struct object *commit = NULL;
475 int found = -1;
476 int i;
478 for (i = 0; i < revs->pending.nr; i++) {
479 struct object *obj = revs->pending.objects[i].item;
480 if (obj->flags & UNINTERESTING)
481 continue;
482 while (obj->type == OBJ_TAG)
483 obj = deref_tag(obj, NULL, 0);
484 if (obj->type != OBJ_COMMIT)
485 die("Non commit %s?", revs->pending.objects[i].name);
486 if (commit)
487 die("More than one commit to dig from: %s and %s?",
488 revs->pending.objects[i].name,
489 revs->pending.objects[found].name);
490 commit = obj;
491 found = i;
494 if (!commit)
495 die("No commit specified?");
497 return (struct commit *) commit;
500 static void fill_blob_sha1(struct commit *commit, struct diff_filespec *spec)
502 unsigned mode;
503 unsigned char sha1[20];
505 if (get_tree_entry(commit->object.sha1, spec->path,
506 sha1, &mode))
507 die("There is no path %s in the commit", spec->path);
508 fill_filespec(spec, sha1, 1, mode);
510 return;
513 static void fill_line_ends(struct diff_filespec *spec, long *lines,
514 unsigned long **line_ends)
516 int num = 0, size = 50;
517 long cur = 0;
518 unsigned long *ends = NULL;
519 char *data = NULL;
521 if (diff_populate_filespec(spec, 0))
522 die("Cannot read blob %s", sha1_to_hex(spec->sha1));
524 ends = xmalloc(size * sizeof(*ends));
525 ends[cur++] = 0;
526 data = spec->data;
527 while (num < spec->size) {
528 if (data[num] == '\n' || num == spec->size - 1) {
529 ALLOC_GROW(ends, (cur + 1), size);
530 ends[cur++] = num;
532 num++;
535 /* shrink the array to fit the elements */
536 ends = xrealloc(ends, cur * sizeof(*ends));
537 *lines = cur-1;
538 *line_ends = ends;
541 struct nth_line_cb {
542 struct diff_filespec *spec;
543 long lines;
544 unsigned long *line_ends;
547 static const char *nth_line(void *data, long line)
549 struct nth_line_cb *d = data;
550 assert(d && line <= d->lines);
551 assert(d->spec && d->spec->data);
553 if (line == 0)
554 return (char *)d->spec->data;
555 else
556 return (char *)d->spec->data + d->line_ends[line] + 1;
559 static struct line_log_data *
560 parse_lines(struct commit *commit, const char *prefix, struct string_list *args)
562 long lines = 0;
563 unsigned long *ends = NULL;
564 struct nth_line_cb cb_data;
565 struct string_list_item *item;
566 struct line_log_data *ranges = NULL;
567 struct line_log_data *p;
569 for_each_string_list_item(item, args) {
570 const char *name_part, *range_part;
571 char *full_name;
572 struct diff_filespec *spec;
573 long begin = 0, end = 0;
574 long anchor;
576 name_part = skip_range_arg(item->string);
577 if (!name_part || *name_part != ':' || !name_part[1])
578 die("-L argument '%s' not of the form start,end:file",
579 item->string);
580 range_part = xstrndup(item->string, name_part - item->string);
581 name_part++;
583 full_name = prefix_path(prefix, prefix ? strlen(prefix) : 0,
584 name_part);
586 spec = alloc_filespec(full_name);
587 fill_blob_sha1(commit, spec);
588 fill_line_ends(spec, &lines, &ends);
589 cb_data.spec = spec;
590 cb_data.lines = lines;
591 cb_data.line_ends = ends;
593 p = search_line_log_data(ranges, full_name, NULL);
594 if (p && p->ranges.nr)
595 anchor = p->ranges.ranges[p->ranges.nr - 1].end + 1;
596 else
597 anchor = 1;
599 if (parse_range_arg(range_part, nth_line, &cb_data,
600 lines, anchor, &begin, &end,
601 full_name))
602 die("malformed -L argument '%s'", range_part);
603 if (lines < end || ((lines || begin) && lines < begin))
604 die("file %s has only %lu lines", name_part, lines);
605 if (begin < 1)
606 begin = 1;
607 if (end < 1)
608 end = lines;
609 begin--;
610 line_log_data_insert(&ranges, full_name, begin, end);
612 free_filespec(spec);
613 free(ends);
614 ends = NULL;
617 for (p = ranges; p; p = p->next)
618 sort_and_merge_range_set(&p->ranges);
620 return ranges;
623 static struct line_log_data *line_log_data_copy_one(struct line_log_data *r)
625 struct line_log_data *ret = xmalloc(sizeof(*ret));
627 assert(r);
628 line_log_data_init(ret);
629 range_set_copy(&ret->ranges, &r->ranges);
631 ret->path = xstrdup(r->path);
633 return ret;
636 static struct line_log_data *
637 line_log_data_copy(struct line_log_data *r)
639 struct line_log_data *ret = NULL;
640 struct line_log_data *tmp = NULL, *prev = NULL;
642 assert(r);
643 ret = tmp = prev = line_log_data_copy_one(r);
644 r = r->next;
645 while (r) {
646 tmp = line_log_data_copy_one(r);
647 prev->next = tmp;
648 prev = tmp;
649 r = r->next;
652 return ret;
655 /* merge two range sets across files */
656 static struct line_log_data *line_log_data_merge(struct line_log_data *a,
657 struct line_log_data *b)
659 struct line_log_data *head = NULL, **pp = &head;
661 while (a || b) {
662 struct line_log_data *src;
663 struct line_log_data *src2 = NULL;
664 struct line_log_data *d;
665 int cmp;
666 if (!a)
667 cmp = 1;
668 else if (!b)
669 cmp = -1;
670 else
671 cmp = strcmp(a->path, b->path);
672 if (cmp < 0) {
673 src = a;
674 a = a->next;
675 } else if (cmp == 0) {
676 src = a;
677 a = a->next;
678 src2 = b;
679 b = b->next;
680 } else {
681 src = b;
682 b = b->next;
684 d = xmalloc(sizeof(struct line_log_data));
685 line_log_data_init(d);
686 d->path = xstrdup(src->path);
687 *pp = d;
688 pp = &d->next;
689 if (src2)
690 range_set_union(&d->ranges, &src->ranges, &src2->ranges);
691 else
692 range_set_copy(&d->ranges, &src->ranges);
695 return head;
698 static void add_line_range(struct rev_info *revs, struct commit *commit,
699 struct line_log_data *range)
701 struct line_log_data *old = NULL;
702 struct line_log_data *new = NULL;
704 old = lookup_decoration(&revs->line_log_data, &commit->object);
705 if (old && range) {
706 new = line_log_data_merge(old, range);
707 free_line_log_data(old);
708 } else if (range)
709 new = line_log_data_copy(range);
711 if (new)
712 add_decoration(&revs->line_log_data, &commit->object, new);
715 static void clear_commit_line_range(struct rev_info *revs, struct commit *commit)
717 struct line_log_data *r;
718 r = lookup_decoration(&revs->line_log_data, &commit->object);
719 if (!r)
720 return;
721 free_line_log_data(r);
722 add_decoration(&revs->line_log_data, &commit->object, NULL);
725 static struct line_log_data *lookup_line_range(struct rev_info *revs,
726 struct commit *commit)
728 struct line_log_data *ret = NULL;
729 struct line_log_data *d;
731 ret = lookup_decoration(&revs->line_log_data, &commit->object);
733 for (d = ret; d; d = d->next)
734 range_set_check_invariants(&d->ranges);
736 return ret;
739 void line_log_init(struct rev_info *rev, const char *prefix, struct string_list *args)
741 struct commit *commit = NULL;
742 struct line_log_data *range;
744 commit = check_single_commit(rev);
745 range = parse_lines(commit, prefix, args);
746 add_line_range(rev, commit, range);
748 if (!rev->diffopt.detect_rename) {
749 int i, count = 0;
750 struct line_log_data *r = range;
751 const char **paths;
752 while (r) {
753 count++;
754 r = r->next;
756 paths = xmalloc((count+1)*sizeof(char *));
757 r = range;
758 for (i = 0; i < count; i++) {
759 paths[i] = xstrdup(r->path);
760 r = r->next;
762 paths[count] = NULL;
763 parse_pathspec(&rev->diffopt.pathspec, 0, 0, "", paths);
764 free(paths);
768 static void load_tree_desc(struct tree_desc *desc, void **tree,
769 const unsigned char *sha1)
771 unsigned long size;
772 *tree = read_object_with_reference(sha1, tree_type, &size, NULL);
773 if (!*tree)
774 die("Unable to read tree (%s)", sha1_to_hex(sha1));
775 init_tree_desc(desc, *tree, size);
778 static int count_parents(struct commit *commit)
780 struct commit_list *parents = commit->parents;
781 int count = 0;
782 while (parents) {
783 count++;
784 parents = parents->next;
786 return count;
789 static void move_diff_queue(struct diff_queue_struct *dst,
790 struct diff_queue_struct *src)
792 assert(src != dst);
793 memcpy(dst, src, sizeof(struct diff_queue_struct));
794 DIFF_QUEUE_CLEAR(src);
797 static void filter_diffs_for_paths(struct line_log_data *range, int keep_deletions)
799 int i;
800 struct diff_queue_struct outq;
801 DIFF_QUEUE_CLEAR(&outq);
803 for (i = 0; i < diff_queued_diff.nr; i++) {
804 struct diff_filepair *p = diff_queued_diff.queue[i];
805 struct line_log_data *rg = NULL;
807 if (!DIFF_FILE_VALID(p->two)) {
808 if (keep_deletions)
809 diff_q(&outq, p);
810 else
811 diff_free_filepair(p);
812 continue;
814 for (rg = range; rg; rg = rg->next) {
815 if (!strcmp(rg->path, p->two->path))
816 break;
818 if (rg)
819 diff_q(&outq, p);
820 else
821 diff_free_filepair(p);
823 free(diff_queued_diff.queue);
824 diff_queued_diff = outq;
827 static inline int diff_might_be_rename(void)
829 int i;
830 for (i = 0; i < diff_queued_diff.nr; i++)
831 if (!DIFF_FILE_VALID(diff_queued_diff.queue[i]->one)) {
832 /* fprintf(stderr, "diff_might_be_rename found creation of: %s\n", */
833 /* diff_queued_diff.queue[i]->two->path); */
834 return 1;
836 return 0;
839 static void queue_diffs(struct line_log_data *range,
840 struct diff_options *opt,
841 struct diff_queue_struct *queue,
842 struct commit *commit, struct commit *parent)
844 void *tree1 = NULL, *tree2 = NULL;
845 struct tree_desc desc1, desc2;
847 assert(commit);
848 load_tree_desc(&desc2, &tree2, commit->tree->object.sha1);
849 if (parent)
850 load_tree_desc(&desc1, &tree1, parent->tree->object.sha1);
851 else
852 init_tree_desc(&desc1, "", 0);
854 DIFF_QUEUE_CLEAR(&diff_queued_diff);
855 diff_tree(&desc1, &desc2, "", opt);
856 if (opt->detect_rename) {
857 filter_diffs_for_paths(range, 1);
858 if (diff_might_be_rename())
859 diffcore_std(opt);
860 filter_diffs_for_paths(range, 0);
862 move_diff_queue(queue, &diff_queued_diff);
864 if (tree1)
865 free(tree1);
866 if (tree2)
867 free(tree2);
870 static char *get_nth_line(long line, unsigned long *ends, void *data)
872 if (line == 0)
873 return (char *)data;
874 else
875 return (char *)data + ends[line] + 1;
878 static void print_line(const char *prefix, char first,
879 long line, unsigned long *ends, void *data,
880 const char *color, const char *reset)
882 char *begin = get_nth_line(line, ends, data);
883 char *end = get_nth_line(line+1, ends, data);
884 int had_nl = 0;
886 if (end > begin && end[-1] == '\n') {
887 end--;
888 had_nl = 1;
891 fputs(prefix, stdout);
892 fputs(color, stdout);
893 putchar(first);
894 fwrite(begin, 1, end-begin, stdout);
895 fputs(reset, stdout);
896 putchar('\n');
897 if (!had_nl)
898 fputs("\\ No newline at end of file\n", stdout);
901 static char *output_prefix(struct diff_options *opt)
903 char *prefix = "";
905 if (opt->output_prefix) {
906 struct strbuf *sb = opt->output_prefix(opt, opt->output_prefix_data);
907 prefix = sb->buf;
910 return prefix;
913 static void dump_diff_hacky_one(struct rev_info *rev, struct line_log_data *range)
915 int i, j = 0;
916 long p_lines, t_lines;
917 unsigned long *p_ends = NULL, *t_ends = NULL;
918 struct diff_filepair *pair = range->pair;
919 struct diff_ranges *diff = &range->diff;
921 struct diff_options *opt = &rev->diffopt;
922 char *prefix = output_prefix(opt);
923 const char *c_reset = diff_get_color(opt->use_color, DIFF_RESET);
924 const char *c_frag = diff_get_color(opt->use_color, DIFF_FRAGINFO);
925 const char *c_meta = diff_get_color(opt->use_color, DIFF_METAINFO);
926 const char *c_old = diff_get_color(opt->use_color, DIFF_FILE_OLD);
927 const char *c_new = diff_get_color(opt->use_color, DIFF_FILE_NEW);
928 const char *c_plain = diff_get_color(opt->use_color, DIFF_PLAIN);
930 if (!pair || !diff)
931 return;
933 if (pair->one->sha1_valid)
934 fill_line_ends(pair->one, &p_lines, &p_ends);
935 fill_line_ends(pair->two, &t_lines, &t_ends);
937 printf("%s%sdiff --git a/%s b/%s%s\n", prefix, c_meta, pair->one->path, pair->two->path, c_reset);
938 printf("%s%s--- %s%s%s\n", prefix, c_meta,
939 pair->one->sha1_valid ? "a/" : "",
940 pair->one->sha1_valid ? pair->one->path : "/dev/null",
941 c_reset);
942 printf("%s%s+++ b/%s%s\n", prefix, c_meta, pair->two->path, c_reset);
943 for (i = 0; i < range->ranges.nr; i++) {
944 long p_start, p_end;
945 long t_start = range->ranges.ranges[i].start;
946 long t_end = range->ranges.ranges[i].end;
947 long t_cur = t_start;
948 int j_last;
950 while (j < diff->target.nr && diff->target.ranges[j].end < t_start)
951 j++;
952 if (j == diff->target.nr || diff->target.ranges[j].start > t_end)
953 continue;
955 /* Scan ahead to determine the last diff that falls in this range */
956 j_last = j;
957 while (j_last < diff->target.nr && diff->target.ranges[j_last].start < t_end)
958 j_last++;
959 if (j_last > j)
960 j_last--;
963 * Compute parent hunk headers: we know that the diff
964 * has the correct line numbers (but not all hunks).
965 * So it suffices to shift the start/end according to
966 * the line numbers of the first/last hunk(s) that
967 * fall in this range.
969 if (t_start < diff->target.ranges[j].start)
970 p_start = diff->parent.ranges[j].start - (diff->target.ranges[j].start-t_start);
971 else
972 p_start = diff->parent.ranges[j].start;
973 if (t_end > diff->target.ranges[j_last].end)
974 p_end = diff->parent.ranges[j_last].end + (t_end-diff->target.ranges[j_last].end);
975 else
976 p_end = diff->parent.ranges[j_last].end;
978 if (!p_start && !p_end) {
979 p_start = -1;
980 p_end = -1;
983 /* Now output a diff hunk for this range */
984 printf("%s%s@@ -%ld,%ld +%ld,%ld @@%s\n",
985 prefix, c_frag,
986 p_start+1, p_end-p_start, t_start+1, t_end-t_start,
987 c_reset);
988 while (j < diff->target.nr && diff->target.ranges[j].start < t_end) {
989 int k;
990 for (; t_cur < diff->target.ranges[j].start; t_cur++)
991 print_line(prefix, ' ', t_cur, t_ends, pair->two->data,
992 c_plain, c_reset);
993 for (k = diff->parent.ranges[j].start; k < diff->parent.ranges[j].end; k++)
994 print_line(prefix, '-', k, p_ends, pair->one->data,
995 c_old, c_reset);
996 for (; t_cur < diff->target.ranges[j].end && t_cur < t_end; t_cur++)
997 print_line(prefix, '+', t_cur, t_ends, pair->two->data,
998 c_new, c_reset);
999 j++;
1001 for (; t_cur < t_end; t_cur++)
1002 print_line(prefix, ' ', t_cur, t_ends, pair->two->data,
1003 c_plain, c_reset);
1006 free(p_ends);
1007 free(t_ends);
1011 * NEEDSWORK: manually building a diff here is not the Right
1012 * Thing(tm). log -L should be built into the diff pipeline.
1014 static void dump_diff_hacky(struct rev_info *rev, struct line_log_data *range)
1016 puts(output_prefix(&rev->diffopt));
1017 while (range) {
1018 dump_diff_hacky_one(rev, range);
1019 range = range->next;
1024 * Unlike most other functions, this destructively operates on
1025 * 'range'.
1027 static int process_diff_filepair(struct rev_info *rev,
1028 struct diff_filepair *pair,
1029 struct line_log_data *range,
1030 struct diff_ranges **diff_out)
1032 struct line_log_data *rg = range;
1033 struct range_set tmp;
1034 struct diff_ranges diff;
1035 mmfile_t file_parent, file_target;
1037 assert(pair->two->path);
1038 while (rg) {
1039 assert(rg->path);
1040 if (!strcmp(rg->path, pair->two->path))
1041 break;
1042 rg = rg->next;
1045 if (!rg)
1046 return 0;
1047 if (rg->ranges.nr == 0)
1048 return 0;
1050 assert(pair->two->sha1_valid);
1051 diff_populate_filespec(pair->two, 0);
1052 file_target.ptr = pair->two->data;
1053 file_target.size = pair->two->size;
1055 if (pair->one->sha1_valid) {
1056 diff_populate_filespec(pair->one, 0);
1057 file_parent.ptr = pair->one->data;
1058 file_parent.size = pair->one->size;
1059 } else {
1060 file_parent.ptr = "";
1061 file_parent.size = 0;
1064 diff_ranges_init(&diff);
1065 collect_diff(&file_parent, &file_target, &diff);
1067 /* NEEDSWORK should apply some heuristics to prevent mismatches */
1068 free(rg->path);
1069 rg->path = xstrdup(pair->one->path);
1071 range_set_init(&tmp, 0);
1072 range_set_map_across_diff(&tmp, &rg->ranges, &diff, diff_out);
1073 range_set_release(&rg->ranges);
1074 range_set_move(&rg->ranges, &tmp);
1076 diff_ranges_release(&diff);
1078 return ((*diff_out)->parent.nr > 0);
1081 static struct diff_filepair *diff_filepair_dup(struct diff_filepair *pair)
1083 struct diff_filepair *new = xmalloc(sizeof(struct diff_filepair));
1084 new->one = pair->one;
1085 new->two = pair->two;
1086 new->one->count++;
1087 new->two->count++;
1088 return new;
1091 static void free_diffqueues(int n, struct diff_queue_struct *dq)
1093 int i, j;
1094 for (i = 0; i < n; i++)
1095 for (j = 0; j < dq[i].nr; j++)
1096 diff_free_filepair(dq[i].queue[j]);
1097 free(dq);
1100 static int process_all_files(struct line_log_data **range_out,
1101 struct rev_info *rev,
1102 struct diff_queue_struct *queue,
1103 struct line_log_data *range)
1105 int i, changed = 0;
1107 *range_out = line_log_data_copy(range);
1109 for (i = 0; i < queue->nr; i++) {
1110 struct diff_ranges *pairdiff = NULL;
1111 struct diff_filepair *pair = queue->queue[i];
1112 if (process_diff_filepair(rev, pair, *range_out, &pairdiff)) {
1114 * Store away the diff for later output. We
1115 * tuck it in the ranges we got as _input_,
1116 * since that's the commit that caused the
1117 * diff.
1119 * NEEDSWORK not enough when we get around to
1120 * doing something interesting with merges;
1121 * currently each invocation on a merge parent
1122 * trashes the previous one's diff.
1124 * NEEDSWORK tramples over data structures not owned here
1126 struct line_log_data *rg = range;
1127 changed++;
1128 while (rg && strcmp(rg->path, pair->two->path))
1129 rg = rg->next;
1130 assert(rg);
1131 rg->pair = diff_filepair_dup(queue->queue[i]);
1132 memcpy(&rg->diff, pairdiff, sizeof(struct diff_ranges));
1136 return changed;
1139 int line_log_print(struct rev_info *rev, struct commit *commit)
1141 struct line_log_data *range = lookup_line_range(rev, commit);
1143 show_log(rev);
1144 dump_diff_hacky(rev, range);
1145 return 1;
1148 static int process_ranges_ordinary_commit(struct rev_info *rev, struct commit *commit,
1149 struct line_log_data *range)
1151 struct commit *parent = NULL;
1152 struct diff_queue_struct queue;
1153 struct line_log_data *parent_range;
1154 int changed;
1156 if (commit->parents)
1157 parent = commit->parents->item;
1159 queue_diffs(range, &rev->diffopt, &queue, commit, parent);
1160 changed = process_all_files(&parent_range, rev, &queue, range);
1161 if (parent)
1162 add_line_range(rev, parent, parent_range);
1163 return changed;
1166 static int process_ranges_merge_commit(struct rev_info *rev, struct commit *commit,
1167 struct line_log_data *range)
1169 struct diff_queue_struct *diffqueues;
1170 struct line_log_data **cand;
1171 struct commit **parents;
1172 struct commit_list *p;
1173 int i;
1174 int nparents = count_parents(commit);
1176 diffqueues = xmalloc(nparents * sizeof(*diffqueues));
1177 cand = xmalloc(nparents * sizeof(*cand));
1178 parents = xmalloc(nparents * sizeof(*parents));
1180 p = commit->parents;
1181 for (i = 0; i < nparents; i++) {
1182 parents[i] = p->item;
1183 p = p->next;
1184 queue_diffs(range, &rev->diffopt, &diffqueues[i], commit, parents[i]);
1187 for (i = 0; i < nparents; i++) {
1188 int changed;
1189 cand[i] = NULL;
1190 changed = process_all_files(&cand[i], rev, &diffqueues[i], range);
1191 if (!changed) {
1193 * This parent can take all the blame, so we
1194 * don't follow any other path in history
1196 add_line_range(rev, parents[i], cand[i]);
1197 clear_commit_line_range(rev, commit);
1198 commit->parents = xmalloc(sizeof(struct commit_list));
1199 commit->parents->item = parents[i];
1200 commit->parents->next = NULL;
1201 free(parents);
1202 free(cand);
1203 free_diffqueues(nparents, diffqueues);
1204 /* NEEDSWORK leaking like a sieve */
1205 return 0;
1210 * No single parent took the blame. We add the candidates
1211 * from the above loop to the parents.
1213 for (i = 0; i < nparents; i++) {
1214 add_line_range(rev, parents[i], cand[i]);
1217 clear_commit_line_range(rev, commit);
1218 free(parents);
1219 free(cand);
1220 free_diffqueues(nparents, diffqueues);
1221 return 1;
1223 /* NEEDSWORK evil merge detection stuff */
1224 /* NEEDSWORK leaking like a sieve */
1227 static int process_ranges_arbitrary_commit(struct rev_info *rev, struct commit *commit)
1229 struct line_log_data *range = lookup_line_range(rev, commit);
1230 int changed = 0;
1232 if (range) {
1233 if (!commit->parents || !commit->parents->next)
1234 changed = process_ranges_ordinary_commit(rev, commit, range);
1235 else
1236 changed = process_ranges_merge_commit(rev, commit, range);
1239 if (!changed)
1240 commit->object.flags |= TREESAME;
1242 return changed;
1245 static enum rewrite_result line_log_rewrite_one(struct rev_info *rev, struct commit **pp)
1247 for (;;) {
1248 struct commit *p = *pp;
1249 if (p->parents && p->parents->next)
1250 return rewrite_one_ok;
1251 if (p->object.flags & UNINTERESTING)
1252 return rewrite_one_ok;
1253 if (!(p->object.flags & TREESAME))
1254 return rewrite_one_ok;
1255 if (!p->parents)
1256 return rewrite_one_noparents;
1257 *pp = p->parents->item;
1261 int line_log_filter(struct rev_info *rev)
1263 struct commit *commit;
1264 struct commit_list *list = rev->commits;
1265 struct commit_list *out = NULL, **pp = &out;
1267 while (list) {
1268 struct commit_list *to_free = NULL;
1269 commit = list->item;
1270 if (process_ranges_arbitrary_commit(rev, commit)) {
1271 *pp = list;
1272 pp = &list->next;
1273 } else
1274 to_free = list;
1275 list = list->next;
1276 free(to_free);
1278 *pp = NULL;
1280 for (list = out; list; list = list->next)
1281 rewrite_parents(rev, list->item, line_log_rewrite_one);
1283 rev->commits = out;
1285 return 0;