Merge branch 'jt/http-redact-cookies' into maint
[git.git] / diff.c
blobe9d0e38123a59f2ca54674115df4140836b98f18
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "config.h"
6 #include "tempfile.h"
7 #include "quote.h"
8 #include "diff.h"
9 #include "diffcore.h"
10 #include "delta.h"
11 #include "xdiff-interface.h"
12 #include "color.h"
13 #include "attr.h"
14 #include "run-command.h"
15 #include "utf8.h"
16 #include "userdiff.h"
17 #include "submodule-config.h"
18 #include "submodule.h"
19 #include "hashmap.h"
20 #include "ll-merge.h"
21 #include "string-list.h"
22 #include "argv-array.h"
23 #include "graph.h"
24 #include "packfile.h"
26 #ifdef NO_FAST_WORKING_DIRECTORY
27 #define FAST_WORKING_DIRECTORY 0
28 #else
29 #define FAST_WORKING_DIRECTORY 1
30 #endif
32 static int diff_detect_rename_default;
33 static int diff_indent_heuristic = 1;
34 static int diff_rename_limit_default = 400;
35 static int diff_suppress_blank_empty;
36 static int diff_use_color_default = -1;
37 static int diff_color_moved_default;
38 static int diff_context_default = 3;
39 static int diff_interhunk_context_default;
40 static const char *diff_word_regex_cfg;
41 static const char *external_diff_cmd_cfg;
42 static const char *diff_order_file_cfg;
43 int diff_auto_refresh_index = 1;
44 static int diff_mnemonic_prefix;
45 static int diff_no_prefix;
46 static int diff_stat_graph_width;
47 static int diff_dirstat_permille_default = 30;
48 static struct diff_options default_diff_options;
49 static long diff_algorithm;
50 static unsigned ws_error_highlight_default = WSEH_NEW;
52 static char diff_colors[][COLOR_MAXLEN] = {
53 GIT_COLOR_RESET,
54 GIT_COLOR_NORMAL, /* CONTEXT */
55 GIT_COLOR_BOLD, /* METAINFO */
56 GIT_COLOR_CYAN, /* FRAGINFO */
57 GIT_COLOR_RED, /* OLD */
58 GIT_COLOR_GREEN, /* NEW */
59 GIT_COLOR_YELLOW, /* COMMIT */
60 GIT_COLOR_BG_RED, /* WHITESPACE */
61 GIT_COLOR_NORMAL, /* FUNCINFO */
62 GIT_COLOR_BOLD_MAGENTA, /* OLD_MOVED */
63 GIT_COLOR_BOLD_BLUE, /* OLD_MOVED ALTERNATIVE */
64 GIT_COLOR_FAINT, /* OLD_MOVED_DIM */
65 GIT_COLOR_FAINT_ITALIC, /* OLD_MOVED_ALTERNATIVE_DIM */
66 GIT_COLOR_BOLD_CYAN, /* NEW_MOVED */
67 GIT_COLOR_BOLD_YELLOW, /* NEW_MOVED ALTERNATIVE */
68 GIT_COLOR_FAINT, /* NEW_MOVED_DIM */
69 GIT_COLOR_FAINT_ITALIC, /* NEW_MOVED_ALTERNATIVE_DIM */
72 static NORETURN void die_want_option(const char *option_name)
74 die(_("option '%s' requires a value"), option_name);
77 static int parse_diff_color_slot(const char *var)
79 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
80 return DIFF_CONTEXT;
81 if (!strcasecmp(var, "meta"))
82 return DIFF_METAINFO;
83 if (!strcasecmp(var, "frag"))
84 return DIFF_FRAGINFO;
85 if (!strcasecmp(var, "old"))
86 return DIFF_FILE_OLD;
87 if (!strcasecmp(var, "new"))
88 return DIFF_FILE_NEW;
89 if (!strcasecmp(var, "commit"))
90 return DIFF_COMMIT;
91 if (!strcasecmp(var, "whitespace"))
92 return DIFF_WHITESPACE;
93 if (!strcasecmp(var, "func"))
94 return DIFF_FUNCINFO;
95 if (!strcasecmp(var, "oldmoved"))
96 return DIFF_FILE_OLD_MOVED;
97 if (!strcasecmp(var, "oldmovedalternative"))
98 return DIFF_FILE_OLD_MOVED_ALT;
99 if (!strcasecmp(var, "oldmoveddimmed"))
100 return DIFF_FILE_OLD_MOVED_DIM;
101 if (!strcasecmp(var, "oldmovedalternativedimmed"))
102 return DIFF_FILE_OLD_MOVED_ALT_DIM;
103 if (!strcasecmp(var, "newmoved"))
104 return DIFF_FILE_NEW_MOVED;
105 if (!strcasecmp(var, "newmovedalternative"))
106 return DIFF_FILE_NEW_MOVED_ALT;
107 if (!strcasecmp(var, "newmoveddimmed"))
108 return DIFF_FILE_NEW_MOVED_DIM;
109 if (!strcasecmp(var, "newmovedalternativedimmed"))
110 return DIFF_FILE_NEW_MOVED_ALT_DIM;
111 return -1;
114 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
115 struct strbuf *errmsg)
117 char *params_copy = xstrdup(params_string);
118 struct string_list params = STRING_LIST_INIT_NODUP;
119 int ret = 0;
120 int i;
122 if (*params_copy)
123 string_list_split_in_place(&params, params_copy, ',', -1);
124 for (i = 0; i < params.nr; i++) {
125 const char *p = params.items[i].string;
126 if (!strcmp(p, "changes")) {
127 options->flags.dirstat_by_line = 0;
128 options->flags.dirstat_by_file = 0;
129 } else if (!strcmp(p, "lines")) {
130 options->flags.dirstat_by_line = 1;
131 options->flags.dirstat_by_file = 0;
132 } else if (!strcmp(p, "files")) {
133 options->flags.dirstat_by_line = 0;
134 options->flags.dirstat_by_file = 1;
135 } else if (!strcmp(p, "noncumulative")) {
136 options->flags.dirstat_cumulative = 0;
137 } else if (!strcmp(p, "cumulative")) {
138 options->flags.dirstat_cumulative = 1;
139 } else if (isdigit(*p)) {
140 char *end;
141 int permille = strtoul(p, &end, 10) * 10;
142 if (*end == '.' && isdigit(*++end)) {
143 /* only use first digit */
144 permille += *end - '0';
145 /* .. and ignore any further digits */
146 while (isdigit(*++end))
147 ; /* nothing */
149 if (!*end)
150 options->dirstat_permille = permille;
151 else {
152 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
154 ret++;
156 } else {
157 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
158 ret++;
162 string_list_clear(&params, 0);
163 free(params_copy);
164 return ret;
167 static int parse_submodule_params(struct diff_options *options, const char *value)
169 if (!strcmp(value, "log"))
170 options->submodule_format = DIFF_SUBMODULE_LOG;
171 else if (!strcmp(value, "short"))
172 options->submodule_format = DIFF_SUBMODULE_SHORT;
173 else if (!strcmp(value, "diff"))
174 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
175 else
176 return -1;
177 return 0;
180 static int git_config_rename(const char *var, const char *value)
182 if (!value)
183 return DIFF_DETECT_RENAME;
184 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
185 return DIFF_DETECT_COPY;
186 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
189 long parse_algorithm_value(const char *value)
191 if (!value)
192 return -1;
193 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
194 return 0;
195 else if (!strcasecmp(value, "minimal"))
196 return XDF_NEED_MINIMAL;
197 else if (!strcasecmp(value, "patience"))
198 return XDF_PATIENCE_DIFF;
199 else if (!strcasecmp(value, "histogram"))
200 return XDF_HISTOGRAM_DIFF;
201 return -1;
204 static int parse_one_token(const char **arg, const char *token)
206 const char *rest;
207 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
208 *arg = rest;
209 return 1;
211 return 0;
214 static int parse_ws_error_highlight(const char *arg)
216 const char *orig_arg = arg;
217 unsigned val = 0;
219 while (*arg) {
220 if (parse_one_token(&arg, "none"))
221 val = 0;
222 else if (parse_one_token(&arg, "default"))
223 val = WSEH_NEW;
224 else if (parse_one_token(&arg, "all"))
225 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
226 else if (parse_one_token(&arg, "new"))
227 val |= WSEH_NEW;
228 else if (parse_one_token(&arg, "old"))
229 val |= WSEH_OLD;
230 else if (parse_one_token(&arg, "context"))
231 val |= WSEH_CONTEXT;
232 else {
233 return -1 - (int)(arg - orig_arg);
235 if (*arg)
236 arg++;
238 return val;
242 * These are to give UI layer defaults.
243 * The core-level commands such as git-diff-files should
244 * never be affected by the setting of diff.renames
245 * the user happens to have in the configuration file.
247 void init_diff_ui_defaults(void)
249 diff_detect_rename_default = DIFF_DETECT_RENAME;
252 int git_diff_heuristic_config(const char *var, const char *value, void *cb)
254 if (!strcmp(var, "diff.indentheuristic"))
255 diff_indent_heuristic = git_config_bool(var, value);
256 return 0;
259 static int parse_color_moved(const char *arg)
261 switch (git_parse_maybe_bool(arg)) {
262 case 0:
263 return COLOR_MOVED_NO;
264 case 1:
265 return COLOR_MOVED_DEFAULT;
266 default:
267 break;
270 if (!strcmp(arg, "no"))
271 return COLOR_MOVED_NO;
272 else if (!strcmp(arg, "plain"))
273 return COLOR_MOVED_PLAIN;
274 else if (!strcmp(arg, "zebra"))
275 return COLOR_MOVED_ZEBRA;
276 else if (!strcmp(arg, "default"))
277 return COLOR_MOVED_DEFAULT;
278 else if (!strcmp(arg, "dimmed_zebra"))
279 return COLOR_MOVED_ZEBRA_DIM;
280 else
281 return error(_("color moved setting must be one of 'no', 'default', 'zebra', 'dimmed_zebra', 'plain'"));
284 int git_diff_ui_config(const char *var, const char *value, void *cb)
286 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
287 diff_use_color_default = git_config_colorbool(var, value);
288 return 0;
290 if (!strcmp(var, "diff.colormoved")) {
291 int cm = parse_color_moved(value);
292 if (cm < 0)
293 return -1;
294 diff_color_moved_default = cm;
295 return 0;
297 if (!strcmp(var, "diff.context")) {
298 diff_context_default = git_config_int(var, value);
299 if (diff_context_default < 0)
300 return -1;
301 return 0;
303 if (!strcmp(var, "diff.interhunkcontext")) {
304 diff_interhunk_context_default = git_config_int(var, value);
305 if (diff_interhunk_context_default < 0)
306 return -1;
307 return 0;
309 if (!strcmp(var, "diff.renames")) {
310 diff_detect_rename_default = git_config_rename(var, value);
311 return 0;
313 if (!strcmp(var, "diff.autorefreshindex")) {
314 diff_auto_refresh_index = git_config_bool(var, value);
315 return 0;
317 if (!strcmp(var, "diff.mnemonicprefix")) {
318 diff_mnemonic_prefix = git_config_bool(var, value);
319 return 0;
321 if (!strcmp(var, "diff.noprefix")) {
322 diff_no_prefix = git_config_bool(var, value);
323 return 0;
325 if (!strcmp(var, "diff.statgraphwidth")) {
326 diff_stat_graph_width = git_config_int(var, value);
327 return 0;
329 if (!strcmp(var, "diff.external"))
330 return git_config_string(&external_diff_cmd_cfg, var, value);
331 if (!strcmp(var, "diff.wordregex"))
332 return git_config_string(&diff_word_regex_cfg, var, value);
333 if (!strcmp(var, "diff.orderfile"))
334 return git_config_pathname(&diff_order_file_cfg, var, value);
336 if (!strcmp(var, "diff.ignoresubmodules"))
337 handle_ignore_submodules_arg(&default_diff_options, value);
339 if (!strcmp(var, "diff.submodule")) {
340 if (parse_submodule_params(&default_diff_options, value))
341 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
342 value);
343 return 0;
346 if (!strcmp(var, "diff.algorithm")) {
347 diff_algorithm = parse_algorithm_value(value);
348 if (diff_algorithm < 0)
349 return -1;
350 return 0;
353 if (!strcmp(var, "diff.wserrorhighlight")) {
354 int val = parse_ws_error_highlight(value);
355 if (val < 0)
356 return -1;
357 ws_error_highlight_default = val;
358 return 0;
361 if (git_color_config(var, value, cb) < 0)
362 return -1;
364 return git_diff_basic_config(var, value, cb);
367 int git_diff_basic_config(const char *var, const char *value, void *cb)
369 const char *name;
371 if (!strcmp(var, "diff.renamelimit")) {
372 diff_rename_limit_default = git_config_int(var, value);
373 return 0;
376 if (userdiff_config(var, value) < 0)
377 return -1;
379 if (skip_prefix(var, "diff.color.", &name) ||
380 skip_prefix(var, "color.diff.", &name)) {
381 int slot = parse_diff_color_slot(name);
382 if (slot < 0)
383 return 0;
384 if (!value)
385 return config_error_nonbool(var);
386 return color_parse(value, diff_colors[slot]);
389 /* like GNU diff's --suppress-blank-empty option */
390 if (!strcmp(var, "diff.suppressblankempty") ||
391 /* for backwards compatibility */
392 !strcmp(var, "diff.suppress-blank-empty")) {
393 diff_suppress_blank_empty = git_config_bool(var, value);
394 return 0;
397 if (!strcmp(var, "diff.dirstat")) {
398 struct strbuf errmsg = STRBUF_INIT;
399 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
400 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
401 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
402 errmsg.buf);
403 strbuf_release(&errmsg);
404 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
405 return 0;
408 if (git_diff_heuristic_config(var, value, cb) < 0)
409 return -1;
411 return git_default_config(var, value, cb);
414 static char *quote_two(const char *one, const char *two)
416 int need_one = quote_c_style(one, NULL, NULL, 1);
417 int need_two = quote_c_style(two, NULL, NULL, 1);
418 struct strbuf res = STRBUF_INIT;
420 if (need_one + need_two) {
421 strbuf_addch(&res, '"');
422 quote_c_style(one, &res, NULL, 1);
423 quote_c_style(two, &res, NULL, 1);
424 strbuf_addch(&res, '"');
425 } else {
426 strbuf_addstr(&res, one);
427 strbuf_addstr(&res, two);
429 return strbuf_detach(&res, NULL);
432 static const char *external_diff(void)
434 static const char *external_diff_cmd = NULL;
435 static int done_preparing = 0;
437 if (done_preparing)
438 return external_diff_cmd;
439 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
440 if (!external_diff_cmd)
441 external_diff_cmd = external_diff_cmd_cfg;
442 done_preparing = 1;
443 return external_diff_cmd;
447 * Keep track of files used for diffing. Sometimes such an entry
448 * refers to a temporary file, sometimes to an existing file, and
449 * sometimes to "/dev/null".
451 static struct diff_tempfile {
453 * filename external diff should read from, or NULL if this
454 * entry is currently not in use:
456 const char *name;
458 char hex[GIT_MAX_HEXSZ + 1];
459 char mode[10];
462 * If this diff_tempfile instance refers to a temporary file,
463 * this tempfile object is used to manage its lifetime.
465 struct tempfile *tempfile;
466 } diff_temp[2];
468 struct emit_callback {
469 int color_diff;
470 unsigned ws_rule;
471 int blank_at_eof_in_preimage;
472 int blank_at_eof_in_postimage;
473 int lno_in_preimage;
474 int lno_in_postimage;
475 const char **label_path;
476 struct diff_words_data *diff_words;
477 struct diff_options *opt;
478 struct strbuf *header;
481 static int count_lines(const char *data, int size)
483 int count, ch, completely_empty = 1, nl_just_seen = 0;
484 count = 0;
485 while (0 < size--) {
486 ch = *data++;
487 if (ch == '\n') {
488 count++;
489 nl_just_seen = 1;
490 completely_empty = 0;
492 else {
493 nl_just_seen = 0;
494 completely_empty = 0;
497 if (completely_empty)
498 return 0;
499 if (!nl_just_seen)
500 count++; /* no trailing newline */
501 return count;
504 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
506 if (!DIFF_FILE_VALID(one)) {
507 mf->ptr = (char *)""; /* does not matter */
508 mf->size = 0;
509 return 0;
511 else if (diff_populate_filespec(one, 0))
512 return -1;
514 mf->ptr = one->data;
515 mf->size = one->size;
516 return 0;
519 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
520 static unsigned long diff_filespec_size(struct diff_filespec *one)
522 if (!DIFF_FILE_VALID(one))
523 return 0;
524 diff_populate_filespec(one, CHECK_SIZE_ONLY);
525 return one->size;
528 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
530 char *ptr = mf->ptr;
531 long size = mf->size;
532 int cnt = 0;
534 if (!size)
535 return cnt;
536 ptr += size - 1; /* pointing at the very end */
537 if (*ptr != '\n')
538 ; /* incomplete line */
539 else
540 ptr--; /* skip the last LF */
541 while (mf->ptr < ptr) {
542 char *prev_eol;
543 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
544 if (*prev_eol == '\n')
545 break;
546 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
547 break;
548 cnt++;
549 ptr = prev_eol - 1;
551 return cnt;
554 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
555 struct emit_callback *ecbdata)
557 int l1, l2, at;
558 unsigned ws_rule = ecbdata->ws_rule;
559 l1 = count_trailing_blank(mf1, ws_rule);
560 l2 = count_trailing_blank(mf2, ws_rule);
561 if (l2 <= l1) {
562 ecbdata->blank_at_eof_in_preimage = 0;
563 ecbdata->blank_at_eof_in_postimage = 0;
564 return;
566 at = count_lines(mf1->ptr, mf1->size);
567 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
569 at = count_lines(mf2->ptr, mf2->size);
570 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
573 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
574 int first, const char *line, int len)
576 int has_trailing_newline, has_trailing_carriage_return;
577 int nofirst;
578 FILE *file = o->file;
580 fputs(diff_line_prefix(o), file);
582 if (len == 0) {
583 has_trailing_newline = (first == '\n');
584 has_trailing_carriage_return = (!has_trailing_newline &&
585 (first == '\r'));
586 nofirst = has_trailing_newline || has_trailing_carriage_return;
587 } else {
588 has_trailing_newline = (len > 0 && line[len-1] == '\n');
589 if (has_trailing_newline)
590 len--;
591 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
592 if (has_trailing_carriage_return)
593 len--;
594 nofirst = 0;
597 if (len || !nofirst) {
598 fputs(set, file);
599 if (!nofirst)
600 fputc(first, file);
601 fwrite(line, len, 1, file);
602 fputs(reset, file);
604 if (has_trailing_carriage_return)
605 fputc('\r', file);
606 if (has_trailing_newline)
607 fputc('\n', file);
610 static void emit_line(struct diff_options *o, const char *set, const char *reset,
611 const char *line, int len)
613 emit_line_0(o, set, reset, line[0], line+1, len-1);
616 enum diff_symbol {
617 DIFF_SYMBOL_BINARY_DIFF_HEADER,
618 DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
619 DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
620 DIFF_SYMBOL_BINARY_DIFF_BODY,
621 DIFF_SYMBOL_BINARY_DIFF_FOOTER,
622 DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
623 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
624 DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
625 DIFF_SYMBOL_STATS_LINE,
626 DIFF_SYMBOL_WORD_DIFF,
627 DIFF_SYMBOL_STAT_SEP,
628 DIFF_SYMBOL_SUMMARY,
629 DIFF_SYMBOL_SUBMODULE_ADD,
630 DIFF_SYMBOL_SUBMODULE_DEL,
631 DIFF_SYMBOL_SUBMODULE_UNTRACKED,
632 DIFF_SYMBOL_SUBMODULE_MODIFIED,
633 DIFF_SYMBOL_SUBMODULE_HEADER,
634 DIFF_SYMBOL_SUBMODULE_ERROR,
635 DIFF_SYMBOL_SUBMODULE_PIPETHROUGH,
636 DIFF_SYMBOL_REWRITE_DIFF,
637 DIFF_SYMBOL_BINARY_FILES,
638 DIFF_SYMBOL_HEADER,
639 DIFF_SYMBOL_FILEPAIR_PLUS,
640 DIFF_SYMBOL_FILEPAIR_MINUS,
641 DIFF_SYMBOL_WORDS_PORCELAIN,
642 DIFF_SYMBOL_WORDS,
643 DIFF_SYMBOL_CONTEXT,
644 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
645 DIFF_SYMBOL_PLUS,
646 DIFF_SYMBOL_MINUS,
647 DIFF_SYMBOL_NO_LF_EOF,
648 DIFF_SYMBOL_CONTEXT_FRAGINFO,
649 DIFF_SYMBOL_CONTEXT_MARKER,
650 DIFF_SYMBOL_SEPARATOR
653 * Flags for content lines:
654 * 0..12 are whitespace rules
655 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
656 * 16 is marking if the line is blank at EOF
658 #define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
659 #define DIFF_SYMBOL_MOVED_LINE (1<<17)
660 #define DIFF_SYMBOL_MOVED_LINE_ALT (1<<18)
661 #define DIFF_SYMBOL_MOVED_LINE_UNINTERESTING (1<<19)
662 #define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
665 * This struct is used when we need to buffer the output of the diff output.
667 * NEEDSWORK: Instead of storing a copy of the line, add an offset pointer
668 * into the pre/post image file. This pointer could be a union with the
669 * line pointer. By storing an offset into the file instead of the literal line,
670 * we can decrease the memory footprint for the buffered output. At first we
671 * may want to only have indirection for the content lines, but we could also
672 * enhance the state for emitting prefabricated lines, e.g. the similarity
673 * score line or hunk/file headers would only need to store a number or path
674 * and then the output can be constructed later on depending on state.
676 struct emitted_diff_symbol {
677 const char *line;
678 int len;
679 int flags;
680 enum diff_symbol s;
682 #define EMITTED_DIFF_SYMBOL_INIT {NULL}
684 struct emitted_diff_symbols {
685 struct emitted_diff_symbol *buf;
686 int nr, alloc;
688 #define EMITTED_DIFF_SYMBOLS_INIT {NULL, 0, 0}
690 static void append_emitted_diff_symbol(struct diff_options *o,
691 struct emitted_diff_symbol *e)
693 struct emitted_diff_symbol *f;
695 ALLOC_GROW(o->emitted_symbols->buf,
696 o->emitted_symbols->nr + 1,
697 o->emitted_symbols->alloc);
698 f = &o->emitted_symbols->buf[o->emitted_symbols->nr++];
700 memcpy(f, e, sizeof(struct emitted_diff_symbol));
701 f->line = e->line ? xmemdupz(e->line, e->len) : NULL;
704 struct moved_entry {
705 struct hashmap_entry ent;
706 const struct emitted_diff_symbol *es;
707 struct moved_entry *next_line;
710 static int moved_entry_cmp(const struct diff_options *diffopt,
711 const struct moved_entry *a,
712 const struct moved_entry *b,
713 const void *keydata)
715 return !xdiff_compare_lines(a->es->line, a->es->len,
716 b->es->line, b->es->len,
717 diffopt->xdl_opts);
720 static struct moved_entry *prepare_entry(struct diff_options *o,
721 int line_no)
723 struct moved_entry *ret = xmalloc(sizeof(*ret));
724 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[line_no];
726 ret->ent.hash = xdiff_hash_string(l->line, l->len, o->xdl_opts);
727 ret->es = l;
728 ret->next_line = NULL;
730 return ret;
733 static void add_lines_to_move_detection(struct diff_options *o,
734 struct hashmap *add_lines,
735 struct hashmap *del_lines)
737 struct moved_entry *prev_line = NULL;
739 int n;
740 for (n = 0; n < o->emitted_symbols->nr; n++) {
741 struct hashmap *hm;
742 struct moved_entry *key;
744 switch (o->emitted_symbols->buf[n].s) {
745 case DIFF_SYMBOL_PLUS:
746 hm = add_lines;
747 break;
748 case DIFF_SYMBOL_MINUS:
749 hm = del_lines;
750 break;
751 default:
752 prev_line = NULL;
753 continue;
756 key = prepare_entry(o, n);
757 if (prev_line && prev_line->es->s == o->emitted_symbols->buf[n].s)
758 prev_line->next_line = key;
760 hashmap_add(hm, key);
761 prev_line = key;
765 static int shrink_potential_moved_blocks(struct moved_entry **pmb,
766 int pmb_nr)
768 int lp, rp;
770 /* Shrink the set of potential block to the remaining running */
771 for (lp = 0, rp = pmb_nr - 1; lp <= rp;) {
772 while (lp < pmb_nr && pmb[lp])
773 lp++;
774 /* lp points at the first NULL now */
776 while (rp > -1 && !pmb[rp])
777 rp--;
778 /* rp points at the last non-NULL */
780 if (lp < pmb_nr && rp > -1 && lp < rp) {
781 pmb[lp] = pmb[rp];
782 pmb[rp] = NULL;
783 rp--;
784 lp++;
788 /* Remember the number of running sets */
789 return rp + 1;
793 * If o->color_moved is COLOR_MOVED_PLAIN, this function does nothing.
795 * Otherwise, if the last block has fewer alphanumeric characters than
796 * COLOR_MOVED_MIN_ALNUM_COUNT, unset DIFF_SYMBOL_MOVED_LINE on all lines in
797 * that block.
799 * The last block consists of the (n - block_length)'th line up to but not
800 * including the nth line.
802 * NEEDSWORK: This uses the same heuristic as blame_entry_score() in blame.c.
803 * Think of a way to unify them.
805 static void adjust_last_block(struct diff_options *o, int n, int block_length)
807 int i, alnum_count = 0;
808 if (o->color_moved == COLOR_MOVED_PLAIN)
809 return;
810 for (i = 1; i < block_length + 1; i++) {
811 const char *c = o->emitted_symbols->buf[n - i].line;
812 for (; *c; c++) {
813 if (!isalnum(*c))
814 continue;
815 alnum_count++;
816 if (alnum_count >= COLOR_MOVED_MIN_ALNUM_COUNT)
817 return;
820 for (i = 1; i < block_length + 1; i++)
821 o->emitted_symbols->buf[n - i].flags &= ~DIFF_SYMBOL_MOVED_LINE;
824 /* Find blocks of moved code, delegate actual coloring decision to helper */
825 static void mark_color_as_moved(struct diff_options *o,
826 struct hashmap *add_lines,
827 struct hashmap *del_lines)
829 struct moved_entry **pmb = NULL; /* potentially moved blocks */
830 int pmb_nr = 0, pmb_alloc = 0;
831 int n, flipped_block = 1, block_length = 0;
834 for (n = 0; n < o->emitted_symbols->nr; n++) {
835 struct hashmap *hm = NULL;
836 struct moved_entry *key;
837 struct moved_entry *match = NULL;
838 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
839 int i;
841 switch (l->s) {
842 case DIFF_SYMBOL_PLUS:
843 hm = del_lines;
844 key = prepare_entry(o, n);
845 match = hashmap_get(hm, key, o);
846 free(key);
847 break;
848 case DIFF_SYMBOL_MINUS:
849 hm = add_lines;
850 key = prepare_entry(o, n);
851 match = hashmap_get(hm, key, o);
852 free(key);
853 break;
854 default:
855 flipped_block = 1;
858 if (!match) {
859 adjust_last_block(o, n, block_length);
860 pmb_nr = 0;
861 block_length = 0;
862 continue;
865 l->flags |= DIFF_SYMBOL_MOVED_LINE;
867 if (o->color_moved == COLOR_MOVED_PLAIN)
868 continue;
870 /* Check any potential block runs, advance each or nullify */
871 for (i = 0; i < pmb_nr; i++) {
872 struct moved_entry *p = pmb[i];
873 struct moved_entry *pnext = (p && p->next_line) ?
874 p->next_line : NULL;
875 if (pnext && !hm->cmpfn(o, pnext, match, NULL)) {
876 pmb[i] = p->next_line;
877 } else {
878 pmb[i] = NULL;
882 pmb_nr = shrink_potential_moved_blocks(pmb, pmb_nr);
884 if (pmb_nr == 0) {
886 * The current line is the start of a new block.
887 * Setup the set of potential blocks.
889 for (; match; match = hashmap_get_next(hm, match)) {
890 ALLOC_GROW(pmb, pmb_nr + 1, pmb_alloc);
891 pmb[pmb_nr++] = match;
894 flipped_block = (flipped_block + 1) % 2;
896 adjust_last_block(o, n, block_length);
897 block_length = 0;
900 block_length++;
902 if (flipped_block)
903 l->flags |= DIFF_SYMBOL_MOVED_LINE_ALT;
905 adjust_last_block(o, n, block_length);
907 free(pmb);
910 #define DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK \
911 (DIFF_SYMBOL_MOVED_LINE | DIFF_SYMBOL_MOVED_LINE_ALT)
912 static void dim_moved_lines(struct diff_options *o)
914 int n;
915 for (n = 0; n < o->emitted_symbols->nr; n++) {
916 struct emitted_diff_symbol *prev = (n != 0) ?
917 &o->emitted_symbols->buf[n - 1] : NULL;
918 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
919 struct emitted_diff_symbol *next =
920 (n < o->emitted_symbols->nr - 1) ?
921 &o->emitted_symbols->buf[n + 1] : NULL;
923 /* Not a plus or minus line? */
924 if (l->s != DIFF_SYMBOL_PLUS && l->s != DIFF_SYMBOL_MINUS)
925 continue;
927 /* Not a moved line? */
928 if (!(l->flags & DIFF_SYMBOL_MOVED_LINE))
929 continue;
932 * If prev or next are not a plus or minus line,
933 * pretend they don't exist
935 if (prev && prev->s != DIFF_SYMBOL_PLUS &&
936 prev->s != DIFF_SYMBOL_MINUS)
937 prev = NULL;
938 if (next && next->s != DIFF_SYMBOL_PLUS &&
939 next->s != DIFF_SYMBOL_MINUS)
940 next = NULL;
942 /* Inside a block? */
943 if ((prev &&
944 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
945 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK)) &&
946 (next &&
947 (next->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
948 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK))) {
949 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
950 continue;
953 /* Check if we are at an interesting bound: */
954 if (prev && (prev->flags & DIFF_SYMBOL_MOVED_LINE) &&
955 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
956 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
957 continue;
958 if (next && (next->flags & DIFF_SYMBOL_MOVED_LINE) &&
959 (next->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
960 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
961 continue;
964 * The boundary to prev and next are not interesting,
965 * so this line is not interesting as a whole
967 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
971 static void emit_line_ws_markup(struct diff_options *o,
972 const char *set, const char *reset,
973 const char *line, int len, char sign,
974 unsigned ws_rule, int blank_at_eof)
976 const char *ws = NULL;
978 if (o->ws_error_highlight & ws_rule) {
979 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
980 if (!*ws)
981 ws = NULL;
984 if (!ws)
985 emit_line_0(o, set, reset, sign, line, len);
986 else if (blank_at_eof)
987 /* Blank line at EOF - paint '+' as well */
988 emit_line_0(o, ws, reset, sign, line, len);
989 else {
990 /* Emit just the prefix, then the rest. */
991 emit_line_0(o, set, reset, sign, "", 0);
992 ws_check_emit(line, len, ws_rule,
993 o->file, set, reset, ws);
997 static void emit_diff_symbol_from_struct(struct diff_options *o,
998 struct emitted_diff_symbol *eds)
1000 static const char *nneof = " No newline at end of file\n";
1001 const char *context, *reset, *set, *meta, *fraginfo;
1002 struct strbuf sb = STRBUF_INIT;
1004 enum diff_symbol s = eds->s;
1005 const char *line = eds->line;
1006 int len = eds->len;
1007 unsigned flags = eds->flags;
1009 switch (s) {
1010 case DIFF_SYMBOL_NO_LF_EOF:
1011 context = diff_get_color_opt(o, DIFF_CONTEXT);
1012 reset = diff_get_color_opt(o, DIFF_RESET);
1013 putc('\n', o->file);
1014 emit_line_0(o, context, reset, '\\',
1015 nneof, strlen(nneof));
1016 break;
1017 case DIFF_SYMBOL_SUBMODULE_HEADER:
1018 case DIFF_SYMBOL_SUBMODULE_ERROR:
1019 case DIFF_SYMBOL_SUBMODULE_PIPETHROUGH:
1020 case DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES:
1021 case DIFF_SYMBOL_SUMMARY:
1022 case DIFF_SYMBOL_STATS_LINE:
1023 case DIFF_SYMBOL_BINARY_DIFF_BODY:
1024 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
1025 emit_line(o, "", "", line, len);
1026 break;
1027 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
1028 case DIFF_SYMBOL_CONTEXT_MARKER:
1029 context = diff_get_color_opt(o, DIFF_CONTEXT);
1030 reset = diff_get_color_opt(o, DIFF_RESET);
1031 emit_line(o, context, reset, line, len);
1032 break;
1033 case DIFF_SYMBOL_SEPARATOR:
1034 fprintf(o->file, "%s%c",
1035 diff_line_prefix(o),
1036 o->line_termination);
1037 break;
1038 case DIFF_SYMBOL_CONTEXT:
1039 set = diff_get_color_opt(o, DIFF_CONTEXT);
1040 reset = diff_get_color_opt(o, DIFF_RESET);
1041 emit_line_ws_markup(o, set, reset, line, len, ' ',
1042 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
1043 break;
1044 case DIFF_SYMBOL_PLUS:
1045 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1046 DIFF_SYMBOL_MOVED_LINE_ALT |
1047 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1048 case DIFF_SYMBOL_MOVED_LINE |
1049 DIFF_SYMBOL_MOVED_LINE_ALT |
1050 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1051 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT_DIM);
1052 break;
1053 case DIFF_SYMBOL_MOVED_LINE |
1054 DIFF_SYMBOL_MOVED_LINE_ALT:
1055 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT);
1056 break;
1057 case DIFF_SYMBOL_MOVED_LINE |
1058 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1059 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_DIM);
1060 break;
1061 case DIFF_SYMBOL_MOVED_LINE:
1062 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED);
1063 break;
1064 default:
1065 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1067 reset = diff_get_color_opt(o, DIFF_RESET);
1068 emit_line_ws_markup(o, set, reset, line, len, '+',
1069 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
1070 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
1071 break;
1072 case DIFF_SYMBOL_MINUS:
1073 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1074 DIFF_SYMBOL_MOVED_LINE_ALT |
1075 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1076 case DIFF_SYMBOL_MOVED_LINE |
1077 DIFF_SYMBOL_MOVED_LINE_ALT |
1078 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1079 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT_DIM);
1080 break;
1081 case DIFF_SYMBOL_MOVED_LINE |
1082 DIFF_SYMBOL_MOVED_LINE_ALT:
1083 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT);
1084 break;
1085 case DIFF_SYMBOL_MOVED_LINE |
1086 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1087 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_DIM);
1088 break;
1089 case DIFF_SYMBOL_MOVED_LINE:
1090 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED);
1091 break;
1092 default:
1093 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1095 reset = diff_get_color_opt(o, DIFF_RESET);
1096 emit_line_ws_markup(o, set, reset, line, len, '-',
1097 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
1098 break;
1099 case DIFF_SYMBOL_WORDS_PORCELAIN:
1100 context = diff_get_color_opt(o, DIFF_CONTEXT);
1101 reset = diff_get_color_opt(o, DIFF_RESET);
1102 emit_line(o, context, reset, line, len);
1103 fputs("~\n", o->file);
1104 break;
1105 case DIFF_SYMBOL_WORDS:
1106 context = diff_get_color_opt(o, DIFF_CONTEXT);
1107 reset = diff_get_color_opt(o, DIFF_RESET);
1109 * Skip the prefix character, if any. With
1110 * diff_suppress_blank_empty, there may be
1111 * none.
1113 if (line[0] != '\n') {
1114 line++;
1115 len--;
1117 emit_line(o, context, reset, line, len);
1118 break;
1119 case DIFF_SYMBOL_FILEPAIR_PLUS:
1120 meta = diff_get_color_opt(o, DIFF_METAINFO);
1121 reset = diff_get_color_opt(o, DIFF_RESET);
1122 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
1123 line, reset,
1124 strchr(line, ' ') ? "\t" : "");
1125 break;
1126 case DIFF_SYMBOL_FILEPAIR_MINUS:
1127 meta = diff_get_color_opt(o, DIFF_METAINFO);
1128 reset = diff_get_color_opt(o, DIFF_RESET);
1129 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
1130 line, reset,
1131 strchr(line, ' ') ? "\t" : "");
1132 break;
1133 case DIFF_SYMBOL_BINARY_FILES:
1134 case DIFF_SYMBOL_HEADER:
1135 fprintf(o->file, "%s", line);
1136 break;
1137 case DIFF_SYMBOL_BINARY_DIFF_HEADER:
1138 fprintf(o->file, "%sGIT binary patch\n", diff_line_prefix(o));
1139 break;
1140 case DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA:
1141 fprintf(o->file, "%sdelta %s\n", diff_line_prefix(o), line);
1142 break;
1143 case DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL:
1144 fprintf(o->file, "%sliteral %s\n", diff_line_prefix(o), line);
1145 break;
1146 case DIFF_SYMBOL_BINARY_DIFF_FOOTER:
1147 fputs(diff_line_prefix(o), o->file);
1148 fputc('\n', o->file);
1149 break;
1150 case DIFF_SYMBOL_REWRITE_DIFF:
1151 fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
1152 reset = diff_get_color_opt(o, DIFF_RESET);
1153 emit_line(o, fraginfo, reset, line, len);
1154 break;
1155 case DIFF_SYMBOL_SUBMODULE_ADD:
1156 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1157 reset = diff_get_color_opt(o, DIFF_RESET);
1158 emit_line(o, set, reset, line, len);
1159 break;
1160 case DIFF_SYMBOL_SUBMODULE_DEL:
1161 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1162 reset = diff_get_color_opt(o, DIFF_RESET);
1163 emit_line(o, set, reset, line, len);
1164 break;
1165 case DIFF_SYMBOL_SUBMODULE_UNTRACKED:
1166 fprintf(o->file, "%sSubmodule %s contains untracked content\n",
1167 diff_line_prefix(o), line);
1168 break;
1169 case DIFF_SYMBOL_SUBMODULE_MODIFIED:
1170 fprintf(o->file, "%sSubmodule %s contains modified content\n",
1171 diff_line_prefix(o), line);
1172 break;
1173 case DIFF_SYMBOL_STATS_SUMMARY_NO_FILES:
1174 emit_line(o, "", "", " 0 files changed\n",
1175 strlen(" 0 files changed\n"));
1176 break;
1177 case DIFF_SYMBOL_STATS_SUMMARY_ABBREV:
1178 emit_line(o, "", "", " ...\n", strlen(" ...\n"));
1179 break;
1180 case DIFF_SYMBOL_WORD_DIFF:
1181 fprintf(o->file, "%.*s", len, line);
1182 break;
1183 case DIFF_SYMBOL_STAT_SEP:
1184 fputs(o->stat_sep, o->file);
1185 break;
1186 default:
1187 die("BUG: unknown diff symbol");
1189 strbuf_release(&sb);
1192 static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
1193 const char *line, int len, unsigned flags)
1195 struct emitted_diff_symbol e = {line, len, flags, s};
1197 if (o->emitted_symbols)
1198 append_emitted_diff_symbol(o, &e);
1199 else
1200 emit_diff_symbol_from_struct(o, &e);
1203 void diff_emit_submodule_del(struct diff_options *o, const char *line)
1205 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_DEL, line, strlen(line), 0);
1208 void diff_emit_submodule_add(struct diff_options *o, const char *line)
1210 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ADD, line, strlen(line), 0);
1213 void diff_emit_submodule_untracked(struct diff_options *o, const char *path)
1215 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_UNTRACKED,
1216 path, strlen(path), 0);
1219 void diff_emit_submodule_modified(struct diff_options *o, const char *path)
1221 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_MODIFIED,
1222 path, strlen(path), 0);
1225 void diff_emit_submodule_header(struct diff_options *o, const char *header)
1227 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_HEADER,
1228 header, strlen(header), 0);
1231 void diff_emit_submodule_error(struct diff_options *o, const char *err)
1233 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ERROR, err, strlen(err), 0);
1236 void diff_emit_submodule_pipethrough(struct diff_options *o,
1237 const char *line, int len)
1239 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_PIPETHROUGH, line, len, 0);
1242 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
1244 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
1245 ecbdata->blank_at_eof_in_preimage &&
1246 ecbdata->blank_at_eof_in_postimage &&
1247 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
1248 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
1249 return 0;
1250 return ws_blank_line(line, len, ecbdata->ws_rule);
1253 static void emit_add_line(const char *reset,
1254 struct emit_callback *ecbdata,
1255 const char *line, int len)
1257 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
1258 if (new_blank_line_at_eof(ecbdata, line, len))
1259 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
1261 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
1264 static void emit_del_line(const char *reset,
1265 struct emit_callback *ecbdata,
1266 const char *line, int len)
1268 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
1269 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
1272 static void emit_context_line(const char *reset,
1273 struct emit_callback *ecbdata,
1274 const char *line, int len)
1276 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
1277 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
1280 static void emit_hunk_header(struct emit_callback *ecbdata,
1281 const char *line, int len)
1283 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1284 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
1285 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
1286 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1287 static const char atat[2] = { '@', '@' };
1288 const char *cp, *ep;
1289 struct strbuf msgbuf = STRBUF_INIT;
1290 int org_len = len;
1291 int i = 1;
1294 * As a hunk header must begin with "@@ -<old>, +<new> @@",
1295 * it always is at least 10 bytes long.
1297 if (len < 10 ||
1298 memcmp(line, atat, 2) ||
1299 !(ep = memmem(line + 2, len - 2, atat, 2))) {
1300 emit_diff_symbol(ecbdata->opt,
1301 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
1302 return;
1304 ep += 2; /* skip over @@ */
1306 /* The hunk header in fraginfo color */
1307 strbuf_addstr(&msgbuf, frag);
1308 strbuf_add(&msgbuf, line, ep - line);
1309 strbuf_addstr(&msgbuf, reset);
1312 * trailing "\r\n"
1314 for ( ; i < 3; i++)
1315 if (line[len - i] == '\r' || line[len - i] == '\n')
1316 len--;
1318 /* blank before the func header */
1319 for (cp = ep; ep - line < len; ep++)
1320 if (*ep != ' ' && *ep != '\t')
1321 break;
1322 if (ep != cp) {
1323 strbuf_addstr(&msgbuf, context);
1324 strbuf_add(&msgbuf, cp, ep - cp);
1325 strbuf_addstr(&msgbuf, reset);
1328 if (ep < line + len) {
1329 strbuf_addstr(&msgbuf, func);
1330 strbuf_add(&msgbuf, ep, line + len - ep);
1331 strbuf_addstr(&msgbuf, reset);
1334 strbuf_add(&msgbuf, line + len, org_len - len);
1335 strbuf_complete_line(&msgbuf);
1336 emit_diff_symbol(ecbdata->opt,
1337 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
1338 strbuf_release(&msgbuf);
1341 static struct diff_tempfile *claim_diff_tempfile(void) {
1342 int i;
1343 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
1344 if (!diff_temp[i].name)
1345 return diff_temp + i;
1346 die("BUG: diff is failing to clean up its tempfiles");
1349 static void remove_tempfile(void)
1351 int i;
1352 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
1353 if (is_tempfile_active(diff_temp[i].tempfile))
1354 delete_tempfile(&diff_temp[i].tempfile);
1355 diff_temp[i].name = NULL;
1359 static void add_line_count(struct strbuf *out, int count)
1361 switch (count) {
1362 case 0:
1363 strbuf_addstr(out, "0,0");
1364 break;
1365 case 1:
1366 strbuf_addstr(out, "1");
1367 break;
1368 default:
1369 strbuf_addf(out, "1,%d", count);
1370 break;
1374 static void emit_rewrite_lines(struct emit_callback *ecb,
1375 int prefix, const char *data, int size)
1377 const char *endp = NULL;
1378 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
1380 while (0 < size) {
1381 int len;
1383 endp = memchr(data, '\n', size);
1384 len = endp ? (endp - data + 1) : size;
1385 if (prefix != '+') {
1386 ecb->lno_in_preimage++;
1387 emit_del_line(reset, ecb, data, len);
1388 } else {
1389 ecb->lno_in_postimage++;
1390 emit_add_line(reset, ecb, data, len);
1392 size -= len;
1393 data += len;
1395 if (!endp)
1396 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
1399 static void emit_rewrite_diff(const char *name_a,
1400 const char *name_b,
1401 struct diff_filespec *one,
1402 struct diff_filespec *two,
1403 struct userdiff_driver *textconv_one,
1404 struct userdiff_driver *textconv_two,
1405 struct diff_options *o)
1407 int lc_a, lc_b;
1408 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
1409 const char *a_prefix, *b_prefix;
1410 char *data_one, *data_two;
1411 size_t size_one, size_two;
1412 struct emit_callback ecbdata;
1413 struct strbuf out = STRBUF_INIT;
1415 if (diff_mnemonic_prefix && o->flags.reverse_diff) {
1416 a_prefix = o->b_prefix;
1417 b_prefix = o->a_prefix;
1418 } else {
1419 a_prefix = o->a_prefix;
1420 b_prefix = o->b_prefix;
1423 name_a += (*name_a == '/');
1424 name_b += (*name_b == '/');
1426 strbuf_reset(&a_name);
1427 strbuf_reset(&b_name);
1428 quote_two_c_style(&a_name, a_prefix, name_a, 0);
1429 quote_two_c_style(&b_name, b_prefix, name_b, 0);
1431 size_one = fill_textconv(textconv_one, one, &data_one);
1432 size_two = fill_textconv(textconv_two, two, &data_two);
1434 memset(&ecbdata, 0, sizeof(ecbdata));
1435 ecbdata.color_diff = want_color(o->use_color);
1436 ecbdata.ws_rule = whitespace_rule(name_b);
1437 ecbdata.opt = o;
1438 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
1439 mmfile_t mf1, mf2;
1440 mf1.ptr = (char *)data_one;
1441 mf2.ptr = (char *)data_two;
1442 mf1.size = size_one;
1443 mf2.size = size_two;
1444 check_blank_at_eof(&mf1, &mf2, &ecbdata);
1446 ecbdata.lno_in_preimage = 1;
1447 ecbdata.lno_in_postimage = 1;
1449 lc_a = count_lines(data_one, size_one);
1450 lc_b = count_lines(data_two, size_two);
1452 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1453 a_name.buf, a_name.len, 0);
1454 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1455 b_name.buf, b_name.len, 0);
1457 strbuf_addstr(&out, "@@ -");
1458 if (!o->irreversible_delete)
1459 add_line_count(&out, lc_a);
1460 else
1461 strbuf_addstr(&out, "?,?");
1462 strbuf_addstr(&out, " +");
1463 add_line_count(&out, lc_b);
1464 strbuf_addstr(&out, " @@\n");
1465 emit_diff_symbol(o, DIFF_SYMBOL_REWRITE_DIFF, out.buf, out.len, 0);
1466 strbuf_release(&out);
1468 if (lc_a && !o->irreversible_delete)
1469 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
1470 if (lc_b)
1471 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
1472 if (textconv_one)
1473 free((char *)data_one);
1474 if (textconv_two)
1475 free((char *)data_two);
1478 struct diff_words_buffer {
1479 mmfile_t text;
1480 unsigned long alloc;
1481 struct diff_words_orig {
1482 const char *begin, *end;
1483 } *orig;
1484 int orig_nr, orig_alloc;
1487 static void diff_words_append(char *line, unsigned long len,
1488 struct diff_words_buffer *buffer)
1490 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
1491 line++;
1492 len--;
1493 memcpy(buffer->text.ptr + buffer->text.size, line, len);
1494 buffer->text.size += len;
1495 buffer->text.ptr[buffer->text.size] = '\0';
1498 struct diff_words_style_elem {
1499 const char *prefix;
1500 const char *suffix;
1501 const char *color; /* NULL; filled in by the setup code if
1502 * color is enabled */
1505 struct diff_words_style {
1506 enum diff_words_type type;
1507 struct diff_words_style_elem new, old, ctx;
1508 const char *newline;
1511 static struct diff_words_style diff_words_styles[] = {
1512 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
1513 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
1514 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
1517 struct diff_words_data {
1518 struct diff_words_buffer minus, plus;
1519 const char *current_plus;
1520 int last_minus;
1521 struct diff_options *opt;
1522 regex_t *word_regex;
1523 enum diff_words_type type;
1524 struct diff_words_style *style;
1527 static int fn_out_diff_words_write_helper(struct diff_options *o,
1528 struct diff_words_style_elem *st_el,
1529 const char *newline,
1530 size_t count, const char *buf)
1532 int print = 0;
1533 struct strbuf sb = STRBUF_INIT;
1535 while (count) {
1536 char *p = memchr(buf, '\n', count);
1537 if (print)
1538 strbuf_addstr(&sb, diff_line_prefix(o));
1540 if (p != buf) {
1541 const char *reset = st_el->color && *st_el->color ?
1542 GIT_COLOR_RESET : NULL;
1543 if (st_el->color && *st_el->color)
1544 strbuf_addstr(&sb, st_el->color);
1545 strbuf_addstr(&sb, st_el->prefix);
1546 strbuf_add(&sb, buf, p ? p - buf : count);
1547 strbuf_addstr(&sb, st_el->suffix);
1548 if (reset)
1549 strbuf_addstr(&sb, reset);
1551 if (!p)
1552 goto out;
1554 strbuf_addstr(&sb, newline);
1555 count -= p + 1 - buf;
1556 buf = p + 1;
1557 print = 1;
1558 if (count) {
1559 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1560 sb.buf, sb.len, 0);
1561 strbuf_reset(&sb);
1565 out:
1566 if (sb.len)
1567 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1568 sb.buf, sb.len, 0);
1569 strbuf_release(&sb);
1570 return 0;
1574 * '--color-words' algorithm can be described as:
1576 * 1. collect the minus/plus lines of a diff hunk, divided into
1577 * minus-lines and plus-lines;
1579 * 2. break both minus-lines and plus-lines into words and
1580 * place them into two mmfile_t with one word for each line;
1582 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1584 * And for the common parts of the both file, we output the plus side text.
1585 * diff_words->current_plus is used to trace the current position of the plus file
1586 * which printed. diff_words->last_minus is used to trace the last minus word
1587 * printed.
1589 * For '--graph' to work with '--color-words', we need to output the graph prefix
1590 * on each line of color words output. Generally, there are two conditions on
1591 * which we should output the prefix.
1593 * 1. diff_words->last_minus == 0 &&
1594 * diff_words->current_plus == diff_words->plus.text.ptr
1596 * that is: the plus text must start as a new line, and if there is no minus
1597 * word printed, a graph prefix must be printed.
1599 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1600 * *(diff_words->current_plus - 1) == '\n'
1602 * that is: a graph prefix must be printed following a '\n'
1604 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1606 if ((diff_words->last_minus == 0 &&
1607 diff_words->current_plus == diff_words->plus.text.ptr) ||
1608 (diff_words->current_plus > diff_words->plus.text.ptr &&
1609 *(diff_words->current_plus - 1) == '\n')) {
1610 return 1;
1611 } else {
1612 return 0;
1616 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1618 struct diff_words_data *diff_words = priv;
1619 struct diff_words_style *style = diff_words->style;
1620 int minus_first, minus_len, plus_first, plus_len;
1621 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1622 struct diff_options *opt = diff_words->opt;
1623 const char *line_prefix;
1625 if (line[0] != '@' || parse_hunk_header(line, len,
1626 &minus_first, &minus_len, &plus_first, &plus_len))
1627 return;
1629 assert(opt);
1630 line_prefix = diff_line_prefix(opt);
1632 /* POSIX requires that first be decremented by one if len == 0... */
1633 if (minus_len) {
1634 minus_begin = diff_words->minus.orig[minus_first].begin;
1635 minus_end =
1636 diff_words->minus.orig[minus_first + minus_len - 1].end;
1637 } else
1638 minus_begin = minus_end =
1639 diff_words->minus.orig[minus_first].end;
1641 if (plus_len) {
1642 plus_begin = diff_words->plus.orig[plus_first].begin;
1643 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1644 } else
1645 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1647 if (color_words_output_graph_prefix(diff_words)) {
1648 fputs(line_prefix, diff_words->opt->file);
1650 if (diff_words->current_plus != plus_begin) {
1651 fn_out_diff_words_write_helper(diff_words->opt,
1652 &style->ctx, style->newline,
1653 plus_begin - diff_words->current_plus,
1654 diff_words->current_plus);
1656 if (minus_begin != minus_end) {
1657 fn_out_diff_words_write_helper(diff_words->opt,
1658 &style->old, style->newline,
1659 minus_end - minus_begin, minus_begin);
1661 if (plus_begin != plus_end) {
1662 fn_out_diff_words_write_helper(diff_words->opt,
1663 &style->new, style->newline,
1664 plus_end - plus_begin, plus_begin);
1667 diff_words->current_plus = plus_end;
1668 diff_words->last_minus = minus_first;
1671 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1672 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1673 int *begin, int *end)
1675 if (word_regex && *begin < buffer->size) {
1676 regmatch_t match[1];
1677 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1678 buffer->size - *begin, 1, match, 0)) {
1679 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1680 '\n', match[0].rm_eo - match[0].rm_so);
1681 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1682 *begin += match[0].rm_so;
1683 return *begin >= *end;
1685 return -1;
1688 /* find the next word */
1689 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1690 (*begin)++;
1691 if (*begin >= buffer->size)
1692 return -1;
1694 /* find the end of the word */
1695 *end = *begin + 1;
1696 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1697 (*end)++;
1699 return 0;
1703 * This function splits the words in buffer->text, stores the list with
1704 * newline separator into out, and saves the offsets of the original words
1705 * in buffer->orig.
1707 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1708 regex_t *word_regex)
1710 int i, j;
1711 long alloc = 0;
1713 out->size = 0;
1714 out->ptr = NULL;
1716 /* fake an empty "0th" word */
1717 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1718 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1719 buffer->orig_nr = 1;
1721 for (i = 0; i < buffer->text.size; i++) {
1722 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1723 return;
1725 /* store original boundaries */
1726 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1727 buffer->orig_alloc);
1728 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1729 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1730 buffer->orig_nr++;
1732 /* store one word */
1733 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1734 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1735 out->ptr[out->size + j - i] = '\n';
1736 out->size += j - i + 1;
1738 i = j - 1;
1742 /* this executes the word diff on the accumulated buffers */
1743 static void diff_words_show(struct diff_words_data *diff_words)
1745 xpparam_t xpp;
1746 xdemitconf_t xecfg;
1747 mmfile_t minus, plus;
1748 struct diff_words_style *style = diff_words->style;
1750 struct diff_options *opt = diff_words->opt;
1751 const char *line_prefix;
1753 assert(opt);
1754 line_prefix = diff_line_prefix(opt);
1756 /* special case: only removal */
1757 if (!diff_words->plus.text.size) {
1758 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
1759 line_prefix, strlen(line_prefix), 0);
1760 fn_out_diff_words_write_helper(diff_words->opt,
1761 &style->old, style->newline,
1762 diff_words->minus.text.size,
1763 diff_words->minus.text.ptr);
1764 diff_words->minus.text.size = 0;
1765 return;
1768 diff_words->current_plus = diff_words->plus.text.ptr;
1769 diff_words->last_minus = 0;
1771 memset(&xpp, 0, sizeof(xpp));
1772 memset(&xecfg, 0, sizeof(xecfg));
1773 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1774 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1775 xpp.flags = 0;
1776 /* as only the hunk header will be parsed, we need a 0-context */
1777 xecfg.ctxlen = 0;
1778 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1779 &xpp, &xecfg))
1780 die("unable to generate word diff");
1781 free(minus.ptr);
1782 free(plus.ptr);
1783 if (diff_words->current_plus != diff_words->plus.text.ptr +
1784 diff_words->plus.text.size) {
1785 if (color_words_output_graph_prefix(diff_words))
1786 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
1787 line_prefix, strlen(line_prefix), 0);
1788 fn_out_diff_words_write_helper(diff_words->opt,
1789 &style->ctx, style->newline,
1790 diff_words->plus.text.ptr + diff_words->plus.text.size
1791 - diff_words->current_plus, diff_words->current_plus);
1793 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1796 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1797 static void diff_words_flush(struct emit_callback *ecbdata)
1799 struct diff_options *wo = ecbdata->diff_words->opt;
1801 if (ecbdata->diff_words->minus.text.size ||
1802 ecbdata->diff_words->plus.text.size)
1803 diff_words_show(ecbdata->diff_words);
1805 if (wo->emitted_symbols) {
1806 struct diff_options *o = ecbdata->opt;
1807 struct emitted_diff_symbols *wol = wo->emitted_symbols;
1808 int i;
1811 * NEEDSWORK:
1812 * Instead of appending each, concat all words to a line?
1814 for (i = 0; i < wol->nr; i++)
1815 append_emitted_diff_symbol(o, &wol->buf[i]);
1817 for (i = 0; i < wol->nr; i++)
1818 free((void *)wol->buf[i].line);
1820 wol->nr = 0;
1824 static void diff_filespec_load_driver(struct diff_filespec *one)
1826 /* Use already-loaded driver */
1827 if (one->driver)
1828 return;
1830 if (S_ISREG(one->mode))
1831 one->driver = userdiff_find_by_path(one->path);
1833 /* Fallback to default settings */
1834 if (!one->driver)
1835 one->driver = userdiff_find_by_name("default");
1838 static const char *userdiff_word_regex(struct diff_filespec *one)
1840 diff_filespec_load_driver(one);
1841 return one->driver->word_regex;
1844 static void init_diff_words_data(struct emit_callback *ecbdata,
1845 struct diff_options *orig_opts,
1846 struct diff_filespec *one,
1847 struct diff_filespec *two)
1849 int i;
1850 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1851 memcpy(o, orig_opts, sizeof(struct diff_options));
1853 ecbdata->diff_words =
1854 xcalloc(1, sizeof(struct diff_words_data));
1855 ecbdata->diff_words->type = o->word_diff;
1856 ecbdata->diff_words->opt = o;
1858 if (orig_opts->emitted_symbols)
1859 o->emitted_symbols =
1860 xcalloc(1, sizeof(struct emitted_diff_symbols));
1862 if (!o->word_regex)
1863 o->word_regex = userdiff_word_regex(one);
1864 if (!o->word_regex)
1865 o->word_regex = userdiff_word_regex(two);
1866 if (!o->word_regex)
1867 o->word_regex = diff_word_regex_cfg;
1868 if (o->word_regex) {
1869 ecbdata->diff_words->word_regex = (regex_t *)
1870 xmalloc(sizeof(regex_t));
1871 if (regcomp(ecbdata->diff_words->word_regex,
1872 o->word_regex,
1873 REG_EXTENDED | REG_NEWLINE))
1874 die ("Invalid regular expression: %s",
1875 o->word_regex);
1877 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1878 if (o->word_diff == diff_words_styles[i].type) {
1879 ecbdata->diff_words->style =
1880 &diff_words_styles[i];
1881 break;
1884 if (want_color(o->use_color)) {
1885 struct diff_words_style *st = ecbdata->diff_words->style;
1886 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1887 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1888 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1892 static void free_diff_words_data(struct emit_callback *ecbdata)
1894 if (ecbdata->diff_words) {
1895 diff_words_flush(ecbdata);
1896 free (ecbdata->diff_words->opt->emitted_symbols);
1897 free (ecbdata->diff_words->opt);
1898 free (ecbdata->diff_words->minus.text.ptr);
1899 free (ecbdata->diff_words->minus.orig);
1900 free (ecbdata->diff_words->plus.text.ptr);
1901 free (ecbdata->diff_words->plus.orig);
1902 if (ecbdata->diff_words->word_regex) {
1903 regfree(ecbdata->diff_words->word_regex);
1904 free(ecbdata->diff_words->word_regex);
1906 FREE_AND_NULL(ecbdata->diff_words);
1910 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1912 if (want_color(diff_use_color))
1913 return diff_colors[ix];
1914 return "";
1917 const char *diff_line_prefix(struct diff_options *opt)
1919 struct strbuf *msgbuf;
1920 if (!opt->output_prefix)
1921 return "";
1923 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1924 return msgbuf->buf;
1927 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1929 const char *cp;
1930 unsigned long allot;
1931 size_t l = len;
1933 cp = line;
1934 allot = l;
1935 while (0 < l) {
1936 (void) utf8_width(&cp, &l);
1937 if (!cp)
1938 break; /* truncated in the middle? */
1940 return allot - l;
1943 static void find_lno(const char *line, struct emit_callback *ecbdata)
1945 const char *p;
1946 ecbdata->lno_in_preimage = 0;
1947 ecbdata->lno_in_postimage = 0;
1948 p = strchr(line, '-');
1949 if (!p)
1950 return; /* cannot happen */
1951 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1952 p = strchr(p, '+');
1953 if (!p)
1954 return; /* cannot happen */
1955 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1958 static void fn_out_consume(void *priv, char *line, unsigned long len)
1960 struct emit_callback *ecbdata = priv;
1961 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1962 struct diff_options *o = ecbdata->opt;
1964 o->found_changes = 1;
1966 if (ecbdata->header) {
1967 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
1968 ecbdata->header->buf, ecbdata->header->len, 0);
1969 strbuf_reset(ecbdata->header);
1970 ecbdata->header = NULL;
1973 if (ecbdata->label_path[0]) {
1974 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1975 ecbdata->label_path[0],
1976 strlen(ecbdata->label_path[0]), 0);
1977 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1978 ecbdata->label_path[1],
1979 strlen(ecbdata->label_path[1]), 0);
1980 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1983 if (diff_suppress_blank_empty
1984 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1985 line[0] = '\n';
1986 len = 1;
1989 if (line[0] == '@') {
1990 if (ecbdata->diff_words)
1991 diff_words_flush(ecbdata);
1992 len = sane_truncate_line(ecbdata, line, len);
1993 find_lno(line, ecbdata);
1994 emit_hunk_header(ecbdata, line, len);
1995 return;
1998 if (ecbdata->diff_words) {
1999 enum diff_symbol s =
2000 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
2001 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
2002 if (line[0] == '-') {
2003 diff_words_append(line, len,
2004 &ecbdata->diff_words->minus);
2005 return;
2006 } else if (line[0] == '+') {
2007 diff_words_append(line, len,
2008 &ecbdata->diff_words->plus);
2009 return;
2010 } else if (starts_with(line, "\\ ")) {
2012 * Eat the "no newline at eof" marker as if we
2013 * saw a "+" or "-" line with nothing on it,
2014 * and return without diff_words_flush() to
2015 * defer processing. If this is the end of
2016 * preimage, more "+" lines may come after it.
2018 return;
2020 diff_words_flush(ecbdata);
2021 emit_diff_symbol(o, s, line, len, 0);
2022 return;
2025 switch (line[0]) {
2026 case '+':
2027 ecbdata->lno_in_postimage++;
2028 emit_add_line(reset, ecbdata, line + 1, len - 1);
2029 break;
2030 case '-':
2031 ecbdata->lno_in_preimage++;
2032 emit_del_line(reset, ecbdata, line + 1, len - 1);
2033 break;
2034 case ' ':
2035 ecbdata->lno_in_postimage++;
2036 ecbdata->lno_in_preimage++;
2037 emit_context_line(reset, ecbdata, line + 1, len - 1);
2038 break;
2039 default:
2040 /* incomplete line at the end */
2041 ecbdata->lno_in_preimage++;
2042 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
2043 line, len, 0);
2044 break;
2048 static char *pprint_rename(const char *a, const char *b)
2050 const char *old = a;
2051 const char *new = b;
2052 struct strbuf name = STRBUF_INIT;
2053 int pfx_length, sfx_length;
2054 int pfx_adjust_for_slash;
2055 int len_a = strlen(a);
2056 int len_b = strlen(b);
2057 int a_midlen, b_midlen;
2058 int qlen_a = quote_c_style(a, NULL, NULL, 0);
2059 int qlen_b = quote_c_style(b, NULL, NULL, 0);
2061 if (qlen_a || qlen_b) {
2062 quote_c_style(a, &name, NULL, 0);
2063 strbuf_addstr(&name, " => ");
2064 quote_c_style(b, &name, NULL, 0);
2065 return strbuf_detach(&name, NULL);
2068 /* Find common prefix */
2069 pfx_length = 0;
2070 while (*old && *new && *old == *new) {
2071 if (*old == '/')
2072 pfx_length = old - a + 1;
2073 old++;
2074 new++;
2077 /* Find common suffix */
2078 old = a + len_a;
2079 new = b + len_b;
2080 sfx_length = 0;
2082 * If there is a common prefix, it must end in a slash. In
2083 * that case we let this loop run 1 into the prefix to see the
2084 * same slash.
2086 * If there is no common prefix, we cannot do this as it would
2087 * underrun the input strings.
2089 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
2090 while (a + pfx_length - pfx_adjust_for_slash <= old &&
2091 b + pfx_length - pfx_adjust_for_slash <= new &&
2092 *old == *new) {
2093 if (*old == '/')
2094 sfx_length = len_a - (old - a);
2095 old--;
2096 new--;
2100 * pfx{mid-a => mid-b}sfx
2101 * {pfx-a => pfx-b}sfx
2102 * pfx{sfx-a => sfx-b}
2103 * name-a => name-b
2105 a_midlen = len_a - pfx_length - sfx_length;
2106 b_midlen = len_b - pfx_length - sfx_length;
2107 if (a_midlen < 0)
2108 a_midlen = 0;
2109 if (b_midlen < 0)
2110 b_midlen = 0;
2112 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
2113 if (pfx_length + sfx_length) {
2114 strbuf_add(&name, a, pfx_length);
2115 strbuf_addch(&name, '{');
2117 strbuf_add(&name, a + pfx_length, a_midlen);
2118 strbuf_addstr(&name, " => ");
2119 strbuf_add(&name, b + pfx_length, b_midlen);
2120 if (pfx_length + sfx_length) {
2121 strbuf_addch(&name, '}');
2122 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
2124 return strbuf_detach(&name, NULL);
2127 struct diffstat_t {
2128 int nr;
2129 int alloc;
2130 struct diffstat_file {
2131 char *from_name;
2132 char *name;
2133 char *print_name;
2134 unsigned is_unmerged:1;
2135 unsigned is_binary:1;
2136 unsigned is_renamed:1;
2137 unsigned is_interesting:1;
2138 uintmax_t added, deleted;
2139 } **files;
2142 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
2143 const char *name_a,
2144 const char *name_b)
2146 struct diffstat_file *x;
2147 x = xcalloc(1, sizeof(*x));
2148 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
2149 diffstat->files[diffstat->nr++] = x;
2150 if (name_b) {
2151 x->from_name = xstrdup(name_a);
2152 x->name = xstrdup(name_b);
2153 x->is_renamed = 1;
2155 else {
2156 x->from_name = NULL;
2157 x->name = xstrdup(name_a);
2159 return x;
2162 static void diffstat_consume(void *priv, char *line, unsigned long len)
2164 struct diffstat_t *diffstat = priv;
2165 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
2167 if (line[0] == '+')
2168 x->added++;
2169 else if (line[0] == '-')
2170 x->deleted++;
2173 const char mime_boundary_leader[] = "------------";
2175 static int scale_linear(int it, int width, int max_change)
2177 if (!it)
2178 return 0;
2180 * make sure that at least one '-' or '+' is printed if
2181 * there is any change to this path. The easiest way is to
2182 * scale linearly as if the alloted width is one column shorter
2183 * than it is, and then add 1 to the result.
2185 return 1 + (it * (width - 1) / max_change);
2188 static void show_graph(struct strbuf *out, char ch, int cnt,
2189 const char *set, const char *reset)
2191 if (cnt <= 0)
2192 return;
2193 strbuf_addstr(out, set);
2194 strbuf_addchars(out, ch, cnt);
2195 strbuf_addstr(out, reset);
2198 static void fill_print_name(struct diffstat_file *file)
2200 char *pname;
2202 if (file->print_name)
2203 return;
2205 if (!file->is_renamed) {
2206 struct strbuf buf = STRBUF_INIT;
2207 if (quote_c_style(file->name, &buf, NULL, 0)) {
2208 pname = strbuf_detach(&buf, NULL);
2209 } else {
2210 pname = file->name;
2211 strbuf_release(&buf);
2213 } else {
2214 pname = pprint_rename(file->from_name, file->name);
2216 file->print_name = pname;
2219 static void print_stat_summary_inserts_deletes(struct diff_options *options,
2220 int files, int insertions, int deletions)
2222 struct strbuf sb = STRBUF_INIT;
2224 if (!files) {
2225 assert(insertions == 0 && deletions == 0);
2226 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
2227 NULL, 0, 0);
2228 return;
2231 strbuf_addf(&sb,
2232 (files == 1) ? " %d file changed" : " %d files changed",
2233 files);
2236 * For binary diff, the caller may want to print "x files
2237 * changed" with insertions == 0 && deletions == 0.
2239 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
2240 * is probably less confusing (i.e skip over "2 files changed
2241 * but nothing about added/removed lines? Is this a bug in Git?").
2243 if (insertions || deletions == 0) {
2244 strbuf_addf(&sb,
2245 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
2246 insertions);
2249 if (deletions || insertions == 0) {
2250 strbuf_addf(&sb,
2251 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
2252 deletions);
2254 strbuf_addch(&sb, '\n');
2255 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
2256 sb.buf, sb.len, 0);
2257 strbuf_release(&sb);
2260 void print_stat_summary(FILE *fp, int files,
2261 int insertions, int deletions)
2263 struct diff_options o;
2264 memset(&o, 0, sizeof(o));
2265 o.file = fp;
2267 print_stat_summary_inserts_deletes(&o, files, insertions, deletions);
2270 static void show_stats(struct diffstat_t *data, struct diff_options *options)
2272 int i, len, add, del, adds = 0, dels = 0;
2273 uintmax_t max_change = 0, max_len = 0;
2274 int total_files = data->nr, count;
2275 int width, name_width, graph_width, number_width = 0, bin_width = 0;
2276 const char *reset, *add_c, *del_c;
2277 int extra_shown = 0;
2278 const char *line_prefix = diff_line_prefix(options);
2279 struct strbuf out = STRBUF_INIT;
2281 if (data->nr == 0)
2282 return;
2284 count = options->stat_count ? options->stat_count : data->nr;
2286 reset = diff_get_color_opt(options, DIFF_RESET);
2287 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
2288 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
2291 * Find the longest filename and max number of changes
2293 for (i = 0; (i < count) && (i < data->nr); i++) {
2294 struct diffstat_file *file = data->files[i];
2295 uintmax_t change = file->added + file->deleted;
2297 if (!file->is_interesting && (change == 0)) {
2298 count++; /* not shown == room for one more */
2299 continue;
2301 fill_print_name(file);
2302 len = strlen(file->print_name);
2303 if (max_len < len)
2304 max_len = len;
2306 if (file->is_unmerged) {
2307 /* "Unmerged" is 8 characters */
2308 bin_width = bin_width < 8 ? 8 : bin_width;
2309 continue;
2311 if (file->is_binary) {
2312 /* "Bin XXX -> YYY bytes" */
2313 int w = 14 + decimal_width(file->added)
2314 + decimal_width(file->deleted);
2315 bin_width = bin_width < w ? w : bin_width;
2316 /* Display change counts aligned with "Bin" */
2317 number_width = 3;
2318 continue;
2321 if (max_change < change)
2322 max_change = change;
2324 count = i; /* where we can stop scanning in data->files[] */
2327 * We have width = stat_width or term_columns() columns total.
2328 * We want a maximum of min(max_len, stat_name_width) for the name part.
2329 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
2330 * We also need 1 for " " and 4 + decimal_width(max_change)
2331 * for " | NNNN " and one the empty column at the end, altogether
2332 * 6 + decimal_width(max_change).
2334 * If there's not enough space, we will use the smaller of
2335 * stat_name_width (if set) and 5/8*width for the filename,
2336 * and the rest for constant elements + graph part, but no more
2337 * than stat_graph_width for the graph part.
2338 * (5/8 gives 50 for filename and 30 for the constant parts + graph
2339 * for the standard terminal size).
2341 * In other words: stat_width limits the maximum width, and
2342 * stat_name_width fixes the maximum width of the filename,
2343 * and is also used to divide available columns if there
2344 * aren't enough.
2346 * Binary files are displayed with "Bin XXX -> YYY bytes"
2347 * instead of the change count and graph. This part is treated
2348 * similarly to the graph part, except that it is not
2349 * "scaled". If total width is too small to accommodate the
2350 * guaranteed minimum width of the filename part and the
2351 * separators and this message, this message will "overflow"
2352 * making the line longer than the maximum width.
2355 if (options->stat_width == -1)
2356 width = term_columns() - strlen(line_prefix);
2357 else
2358 width = options->stat_width ? options->stat_width : 80;
2359 number_width = decimal_width(max_change) > number_width ?
2360 decimal_width(max_change) : number_width;
2362 if (options->stat_graph_width == -1)
2363 options->stat_graph_width = diff_stat_graph_width;
2366 * Guarantee 3/8*16==6 for the graph part
2367 * and 5/8*16==10 for the filename part
2369 if (width < 16 + 6 + number_width)
2370 width = 16 + 6 + number_width;
2373 * First assign sizes that are wanted, ignoring available width.
2374 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
2375 * starting from "XXX" should fit in graph_width.
2377 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
2378 if (options->stat_graph_width &&
2379 options->stat_graph_width < graph_width)
2380 graph_width = options->stat_graph_width;
2382 name_width = (options->stat_name_width > 0 &&
2383 options->stat_name_width < max_len) ?
2384 options->stat_name_width : max_len;
2387 * Adjust adjustable widths not to exceed maximum width
2389 if (name_width + number_width + 6 + graph_width > width) {
2390 if (graph_width > width * 3/8 - number_width - 6) {
2391 graph_width = width * 3/8 - number_width - 6;
2392 if (graph_width < 6)
2393 graph_width = 6;
2396 if (options->stat_graph_width &&
2397 graph_width > options->stat_graph_width)
2398 graph_width = options->stat_graph_width;
2399 if (name_width > width - number_width - 6 - graph_width)
2400 name_width = width - number_width - 6 - graph_width;
2401 else
2402 graph_width = width - number_width - 6 - name_width;
2406 * From here name_width is the width of the name area,
2407 * and graph_width is the width of the graph area.
2408 * max_change is used to scale graph properly.
2410 for (i = 0; i < count; i++) {
2411 const char *prefix = "";
2412 struct diffstat_file *file = data->files[i];
2413 char *name = file->print_name;
2414 uintmax_t added = file->added;
2415 uintmax_t deleted = file->deleted;
2416 int name_len;
2418 if (!file->is_interesting && (added + deleted == 0))
2419 continue;
2422 * "scale" the filename
2424 len = name_width;
2425 name_len = strlen(name);
2426 if (name_width < name_len) {
2427 char *slash;
2428 prefix = "...";
2429 len -= 3;
2430 name += name_len - len;
2431 slash = strchr(name, '/');
2432 if (slash)
2433 name = slash;
2436 if (file->is_binary) {
2437 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2438 strbuf_addf(&out, " %*s", number_width, "Bin");
2439 if (!added && !deleted) {
2440 strbuf_addch(&out, '\n');
2441 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2442 out.buf, out.len, 0);
2443 strbuf_reset(&out);
2444 continue;
2446 strbuf_addf(&out, " %s%"PRIuMAX"%s",
2447 del_c, deleted, reset);
2448 strbuf_addstr(&out, " -> ");
2449 strbuf_addf(&out, "%s%"PRIuMAX"%s",
2450 add_c, added, reset);
2451 strbuf_addstr(&out, " bytes\n");
2452 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2453 out.buf, out.len, 0);
2454 strbuf_reset(&out);
2455 continue;
2457 else if (file->is_unmerged) {
2458 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2459 strbuf_addstr(&out, " Unmerged\n");
2460 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2461 out.buf, out.len, 0);
2462 strbuf_reset(&out);
2463 continue;
2467 * scale the add/delete
2469 add = added;
2470 del = deleted;
2472 if (graph_width <= max_change) {
2473 int total = scale_linear(add + del, graph_width, max_change);
2474 if (total < 2 && add && del)
2475 /* width >= 2 due to the sanity check */
2476 total = 2;
2477 if (add < del) {
2478 add = scale_linear(add, graph_width, max_change);
2479 del = total - add;
2480 } else {
2481 del = scale_linear(del, graph_width, max_change);
2482 add = total - del;
2485 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2486 strbuf_addf(&out, " %*"PRIuMAX"%s",
2487 number_width, added + deleted,
2488 added + deleted ? " " : "");
2489 show_graph(&out, '+', add, add_c, reset);
2490 show_graph(&out, '-', del, del_c, reset);
2491 strbuf_addch(&out, '\n');
2492 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2493 out.buf, out.len, 0);
2494 strbuf_reset(&out);
2497 for (i = 0; i < data->nr; i++) {
2498 struct diffstat_file *file = data->files[i];
2499 uintmax_t added = file->added;
2500 uintmax_t deleted = file->deleted;
2502 if (file->is_unmerged ||
2503 (!file->is_interesting && (added + deleted == 0))) {
2504 total_files--;
2505 continue;
2508 if (!file->is_binary) {
2509 adds += added;
2510 dels += deleted;
2512 if (i < count)
2513 continue;
2514 if (!extra_shown)
2515 emit_diff_symbol(options,
2516 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
2517 NULL, 0, 0);
2518 extra_shown = 1;
2521 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2522 strbuf_release(&out);
2525 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
2527 int i, adds = 0, dels = 0, total_files = data->nr;
2529 if (data->nr == 0)
2530 return;
2532 for (i = 0; i < data->nr; i++) {
2533 int added = data->files[i]->added;
2534 int deleted = data->files[i]->deleted;
2536 if (data->files[i]->is_unmerged ||
2537 (!data->files[i]->is_interesting && (added + deleted == 0))) {
2538 total_files--;
2539 } else if (!data->files[i]->is_binary) { /* don't count bytes */
2540 adds += added;
2541 dels += deleted;
2544 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2547 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
2549 int i;
2551 if (data->nr == 0)
2552 return;
2554 for (i = 0; i < data->nr; i++) {
2555 struct diffstat_file *file = data->files[i];
2557 fprintf(options->file, "%s", diff_line_prefix(options));
2559 if (file->is_binary)
2560 fprintf(options->file, "-\t-\t");
2561 else
2562 fprintf(options->file,
2563 "%"PRIuMAX"\t%"PRIuMAX"\t",
2564 file->added, file->deleted);
2565 if (options->line_termination) {
2566 fill_print_name(file);
2567 if (!file->is_renamed)
2568 write_name_quoted(file->name, options->file,
2569 options->line_termination);
2570 else {
2571 fputs(file->print_name, options->file);
2572 putc(options->line_termination, options->file);
2574 } else {
2575 if (file->is_renamed) {
2576 putc('\0', options->file);
2577 write_name_quoted(file->from_name, options->file, '\0');
2579 write_name_quoted(file->name, options->file, '\0');
2584 struct dirstat_file {
2585 const char *name;
2586 unsigned long changed;
2589 struct dirstat_dir {
2590 struct dirstat_file *files;
2591 int alloc, nr, permille, cumulative;
2594 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2595 unsigned long changed, const char *base, int baselen)
2597 unsigned long this_dir = 0;
2598 unsigned int sources = 0;
2599 const char *line_prefix = diff_line_prefix(opt);
2601 while (dir->nr) {
2602 struct dirstat_file *f = dir->files;
2603 int namelen = strlen(f->name);
2604 unsigned long this;
2605 char *slash;
2607 if (namelen < baselen)
2608 break;
2609 if (memcmp(f->name, base, baselen))
2610 break;
2611 slash = strchr(f->name + baselen, '/');
2612 if (slash) {
2613 int newbaselen = slash + 1 - f->name;
2614 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2615 sources++;
2616 } else {
2617 this = f->changed;
2618 dir->files++;
2619 dir->nr--;
2620 sources += 2;
2622 this_dir += this;
2626 * We don't report dirstat's for
2627 * - the top level
2628 * - or cases where everything came from a single directory
2629 * under this directory (sources == 1).
2631 if (baselen && sources != 1) {
2632 if (this_dir) {
2633 int permille = this_dir * 1000 / changed;
2634 if (permille >= dir->permille) {
2635 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2636 permille / 10, permille % 10, baselen, base);
2637 if (!dir->cumulative)
2638 return 0;
2642 return this_dir;
2645 static int dirstat_compare(const void *_a, const void *_b)
2647 const struct dirstat_file *a = _a;
2648 const struct dirstat_file *b = _b;
2649 return strcmp(a->name, b->name);
2652 static void show_dirstat(struct diff_options *options)
2654 int i;
2655 unsigned long changed;
2656 struct dirstat_dir dir;
2657 struct diff_queue_struct *q = &diff_queued_diff;
2659 dir.files = NULL;
2660 dir.alloc = 0;
2661 dir.nr = 0;
2662 dir.permille = options->dirstat_permille;
2663 dir.cumulative = options->flags.dirstat_cumulative;
2665 changed = 0;
2666 for (i = 0; i < q->nr; i++) {
2667 struct diff_filepair *p = q->queue[i];
2668 const char *name;
2669 unsigned long copied, added, damage;
2670 int content_changed;
2672 name = p->two->path ? p->two->path : p->one->path;
2674 if (p->one->oid_valid && p->two->oid_valid)
2675 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2676 else
2677 content_changed = 1;
2679 if (!content_changed) {
2681 * The SHA1 has not changed, so pre-/post-content is
2682 * identical. We can therefore skip looking at the
2683 * file contents altogether.
2685 damage = 0;
2686 goto found_damage;
2689 if (options->flags.dirstat_by_file) {
2691 * In --dirstat-by-file mode, we don't really need to
2692 * look at the actual file contents at all.
2693 * The fact that the SHA1 changed is enough for us to
2694 * add this file to the list of results
2695 * (with each file contributing equal damage).
2697 damage = 1;
2698 goto found_damage;
2701 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2702 diff_populate_filespec(p->one, 0);
2703 diff_populate_filespec(p->two, 0);
2704 diffcore_count_changes(p->one, p->two, NULL, NULL,
2705 &copied, &added);
2706 diff_free_filespec_data(p->one);
2707 diff_free_filespec_data(p->two);
2708 } else if (DIFF_FILE_VALID(p->one)) {
2709 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2710 copied = added = 0;
2711 diff_free_filespec_data(p->one);
2712 } else if (DIFF_FILE_VALID(p->two)) {
2713 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2714 copied = 0;
2715 added = p->two->size;
2716 diff_free_filespec_data(p->two);
2717 } else
2718 continue;
2721 * Original minus copied is the removed material,
2722 * added is the new material. They are both damages
2723 * made to the preimage.
2724 * If the resulting damage is zero, we know that
2725 * diffcore_count_changes() considers the two entries to
2726 * be identical, but since content_changed is true, we
2727 * know that there must have been _some_ kind of change,
2728 * so we force all entries to have damage > 0.
2730 damage = (p->one->size - copied) + added;
2731 if (!damage)
2732 damage = 1;
2734 found_damage:
2735 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2736 dir.files[dir.nr].name = name;
2737 dir.files[dir.nr].changed = damage;
2738 changed += damage;
2739 dir.nr++;
2742 /* This can happen even with many files, if everything was renames */
2743 if (!changed)
2744 return;
2746 /* Show all directories with more than x% of the changes */
2747 QSORT(dir.files, dir.nr, dirstat_compare);
2748 gather_dirstat(options, &dir, changed, "", 0);
2751 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2753 int i;
2754 unsigned long changed;
2755 struct dirstat_dir dir;
2757 if (data->nr == 0)
2758 return;
2760 dir.files = NULL;
2761 dir.alloc = 0;
2762 dir.nr = 0;
2763 dir.permille = options->dirstat_permille;
2764 dir.cumulative = options->flags.dirstat_cumulative;
2766 changed = 0;
2767 for (i = 0; i < data->nr; i++) {
2768 struct diffstat_file *file = data->files[i];
2769 unsigned long damage = file->added + file->deleted;
2770 if (file->is_binary)
2772 * binary files counts bytes, not lines. Must find some
2773 * way to normalize binary bytes vs. textual lines.
2774 * The following heuristic assumes that there are 64
2775 * bytes per "line".
2776 * This is stupid and ugly, but very cheap...
2778 damage = DIV_ROUND_UP(damage, 64);
2779 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2780 dir.files[dir.nr].name = file->name;
2781 dir.files[dir.nr].changed = damage;
2782 changed += damage;
2783 dir.nr++;
2786 /* This can happen even with many files, if everything was renames */
2787 if (!changed)
2788 return;
2790 /* Show all directories with more than x% of the changes */
2791 QSORT(dir.files, dir.nr, dirstat_compare);
2792 gather_dirstat(options, &dir, changed, "", 0);
2795 static void free_diffstat_info(struct diffstat_t *diffstat)
2797 int i;
2798 for (i = 0; i < diffstat->nr; i++) {
2799 struct diffstat_file *f = diffstat->files[i];
2800 if (f->name != f->print_name)
2801 free(f->print_name);
2802 free(f->name);
2803 free(f->from_name);
2804 free(f);
2806 free(diffstat->files);
2809 struct checkdiff_t {
2810 const char *filename;
2811 int lineno;
2812 int conflict_marker_size;
2813 struct diff_options *o;
2814 unsigned ws_rule;
2815 unsigned status;
2818 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2820 char firstchar;
2821 int cnt;
2823 if (len < marker_size + 1)
2824 return 0;
2825 firstchar = line[0];
2826 switch (firstchar) {
2827 case '=': case '>': case '<': case '|':
2828 break;
2829 default:
2830 return 0;
2832 for (cnt = 1; cnt < marker_size; cnt++)
2833 if (line[cnt] != firstchar)
2834 return 0;
2835 /* line[1] thru line[marker_size-1] are same as firstchar */
2836 if (len < marker_size + 1 || !isspace(line[marker_size]))
2837 return 0;
2838 return 1;
2841 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2843 struct checkdiff_t *data = priv;
2844 int marker_size = data->conflict_marker_size;
2845 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2846 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2847 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2848 char *err;
2849 const char *line_prefix;
2851 assert(data->o);
2852 line_prefix = diff_line_prefix(data->o);
2854 if (line[0] == '+') {
2855 unsigned bad;
2856 data->lineno++;
2857 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2858 data->status |= 1;
2859 fprintf(data->o->file,
2860 "%s%s:%d: leftover conflict marker\n",
2861 line_prefix, data->filename, data->lineno);
2863 bad = ws_check(line + 1, len - 1, data->ws_rule);
2864 if (!bad)
2865 return;
2866 data->status |= bad;
2867 err = whitespace_error_string(bad);
2868 fprintf(data->o->file, "%s%s:%d: %s.\n",
2869 line_prefix, data->filename, data->lineno, err);
2870 free(err);
2871 emit_line(data->o, set, reset, line, 1);
2872 ws_check_emit(line + 1, len - 1, data->ws_rule,
2873 data->o->file, set, reset, ws);
2874 } else if (line[0] == ' ') {
2875 data->lineno++;
2876 } else if (line[0] == '@') {
2877 char *plus = strchr(line, '+');
2878 if (plus)
2879 data->lineno = strtol(plus, NULL, 10) - 1;
2880 else
2881 die("invalid diff");
2885 static unsigned char *deflate_it(char *data,
2886 unsigned long size,
2887 unsigned long *result_size)
2889 int bound;
2890 unsigned char *deflated;
2891 git_zstream stream;
2893 git_deflate_init(&stream, zlib_compression_level);
2894 bound = git_deflate_bound(&stream, size);
2895 deflated = xmalloc(bound);
2896 stream.next_out = deflated;
2897 stream.avail_out = bound;
2899 stream.next_in = (unsigned char *)data;
2900 stream.avail_in = size;
2901 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2902 ; /* nothing */
2903 git_deflate_end(&stream);
2904 *result_size = stream.total_out;
2905 return deflated;
2908 static void emit_binary_diff_body(struct diff_options *o,
2909 mmfile_t *one, mmfile_t *two)
2911 void *cp;
2912 void *delta;
2913 void *deflated;
2914 void *data;
2915 unsigned long orig_size;
2916 unsigned long delta_size;
2917 unsigned long deflate_size;
2918 unsigned long data_size;
2920 /* We could do deflated delta, or we could do just deflated two,
2921 * whichever is smaller.
2923 delta = NULL;
2924 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2925 if (one->size && two->size) {
2926 delta = diff_delta(one->ptr, one->size,
2927 two->ptr, two->size,
2928 &delta_size, deflate_size);
2929 if (delta) {
2930 void *to_free = delta;
2931 orig_size = delta_size;
2932 delta = deflate_it(delta, delta_size, &delta_size);
2933 free(to_free);
2937 if (delta && delta_size < deflate_size) {
2938 char *s = xstrfmt("%lu", orig_size);
2939 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
2940 s, strlen(s), 0);
2941 free(s);
2942 free(deflated);
2943 data = delta;
2944 data_size = delta_size;
2945 } else {
2946 char *s = xstrfmt("%lu", two->size);
2947 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
2948 s, strlen(s), 0);
2949 free(s);
2950 free(delta);
2951 data = deflated;
2952 data_size = deflate_size;
2955 /* emit data encoded in base85 */
2956 cp = data;
2957 while (data_size) {
2958 int len;
2959 int bytes = (52 < data_size) ? 52 : data_size;
2960 char line[71];
2961 data_size -= bytes;
2962 if (bytes <= 26)
2963 line[0] = bytes + 'A' - 1;
2964 else
2965 line[0] = bytes - 26 + 'a' - 1;
2966 encode_85(line + 1, cp, bytes);
2967 cp = (char *) cp + bytes;
2969 len = strlen(line);
2970 line[len++] = '\n';
2971 line[len] = '\0';
2973 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_BODY,
2974 line, len, 0);
2976 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_FOOTER, NULL, 0, 0);
2977 free(data);
2980 static void emit_binary_diff(struct diff_options *o,
2981 mmfile_t *one, mmfile_t *two)
2983 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER, NULL, 0, 0);
2984 emit_binary_diff_body(o, one, two);
2985 emit_binary_diff_body(o, two, one);
2988 int diff_filespec_is_binary(struct diff_filespec *one)
2990 if (one->is_binary == -1) {
2991 diff_filespec_load_driver(one);
2992 if (one->driver->binary != -1)
2993 one->is_binary = one->driver->binary;
2994 else {
2995 if (!one->data && DIFF_FILE_VALID(one))
2996 diff_populate_filespec(one, CHECK_BINARY);
2997 if (one->is_binary == -1 && one->data)
2998 one->is_binary = buffer_is_binary(one->data,
2999 one->size);
3000 if (one->is_binary == -1)
3001 one->is_binary = 0;
3004 return one->is_binary;
3007 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
3009 diff_filespec_load_driver(one);
3010 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
3013 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
3015 if (!options->a_prefix)
3016 options->a_prefix = a;
3017 if (!options->b_prefix)
3018 options->b_prefix = b;
3021 struct userdiff_driver *get_textconv(struct diff_filespec *one)
3023 if (!DIFF_FILE_VALID(one))
3024 return NULL;
3026 diff_filespec_load_driver(one);
3027 return userdiff_get_textconv(one->driver);
3030 static void builtin_diff(const char *name_a,
3031 const char *name_b,
3032 struct diff_filespec *one,
3033 struct diff_filespec *two,
3034 const char *xfrm_msg,
3035 int must_show_header,
3036 struct diff_options *o,
3037 int complete_rewrite)
3039 mmfile_t mf1, mf2;
3040 const char *lbl[2];
3041 char *a_one, *b_two;
3042 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
3043 const char *reset = diff_get_color_opt(o, DIFF_RESET);
3044 const char *a_prefix, *b_prefix;
3045 struct userdiff_driver *textconv_one = NULL;
3046 struct userdiff_driver *textconv_two = NULL;
3047 struct strbuf header = STRBUF_INIT;
3048 const char *line_prefix = diff_line_prefix(o);
3050 diff_set_mnemonic_prefix(o, "a/", "b/");
3051 if (o->flags.reverse_diff) {
3052 a_prefix = o->b_prefix;
3053 b_prefix = o->a_prefix;
3054 } else {
3055 a_prefix = o->a_prefix;
3056 b_prefix = o->b_prefix;
3059 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
3060 (!one->mode || S_ISGITLINK(one->mode)) &&
3061 (!two->mode || S_ISGITLINK(two->mode))) {
3062 show_submodule_summary(o, one->path ? one->path : two->path,
3063 &one->oid, &two->oid,
3064 two->dirty_submodule);
3065 return;
3066 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
3067 (!one->mode || S_ISGITLINK(one->mode)) &&
3068 (!two->mode || S_ISGITLINK(two->mode))) {
3069 show_submodule_inline_diff(o, one->path ? one->path : two->path,
3070 &one->oid, &two->oid,
3071 two->dirty_submodule);
3072 return;
3075 if (o->flags.allow_textconv) {
3076 textconv_one = get_textconv(one);
3077 textconv_two = get_textconv(two);
3080 /* Never use a non-valid filename anywhere if at all possible */
3081 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
3082 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
3084 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
3085 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
3086 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
3087 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
3088 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
3089 if (lbl[0][0] == '/') {
3090 /* /dev/null */
3091 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
3092 if (xfrm_msg)
3093 strbuf_addstr(&header, xfrm_msg);
3094 must_show_header = 1;
3096 else if (lbl[1][0] == '/') {
3097 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
3098 if (xfrm_msg)
3099 strbuf_addstr(&header, xfrm_msg);
3100 must_show_header = 1;
3102 else {
3103 if (one->mode != two->mode) {
3104 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
3105 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
3106 must_show_header = 1;
3108 if (xfrm_msg)
3109 strbuf_addstr(&header, xfrm_msg);
3112 * we do not run diff between different kind
3113 * of objects.
3115 if ((one->mode ^ two->mode) & S_IFMT)
3116 goto free_ab_and_return;
3117 if (complete_rewrite &&
3118 (textconv_one || !diff_filespec_is_binary(one)) &&
3119 (textconv_two || !diff_filespec_is_binary(two))) {
3120 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3121 header.buf, header.len, 0);
3122 strbuf_reset(&header);
3123 emit_rewrite_diff(name_a, name_b, one, two,
3124 textconv_one, textconv_two, o);
3125 o->found_changes = 1;
3126 goto free_ab_and_return;
3130 if (o->irreversible_delete && lbl[1][0] == '/') {
3131 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
3132 header.len, 0);
3133 strbuf_reset(&header);
3134 goto free_ab_and_return;
3135 } else if (!o->flags.text &&
3136 ( (!textconv_one && diff_filespec_is_binary(one)) ||
3137 (!textconv_two && diff_filespec_is_binary(two)) )) {
3138 struct strbuf sb = STRBUF_INIT;
3139 if (!one->data && !two->data &&
3140 S_ISREG(one->mode) && S_ISREG(two->mode) &&
3141 !o->flags.binary) {
3142 if (!oidcmp(&one->oid, &two->oid)) {
3143 if (must_show_header)
3144 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3145 header.buf, header.len,
3147 goto free_ab_and_return;
3149 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3150 header.buf, header.len, 0);
3151 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3152 diff_line_prefix(o), lbl[0], lbl[1]);
3153 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3154 sb.buf, sb.len, 0);
3155 strbuf_release(&sb);
3156 goto free_ab_and_return;
3158 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3159 die("unable to read files to diff");
3160 /* Quite common confusing case */
3161 if (mf1.size == mf2.size &&
3162 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
3163 if (must_show_header)
3164 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3165 header.buf, header.len, 0);
3166 goto free_ab_and_return;
3168 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
3169 strbuf_reset(&header);
3170 if (o->flags.binary)
3171 emit_binary_diff(o, &mf1, &mf2);
3172 else {
3173 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3174 diff_line_prefix(o), lbl[0], lbl[1]);
3175 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3176 sb.buf, sb.len, 0);
3177 strbuf_release(&sb);
3179 o->found_changes = 1;
3180 } else {
3181 /* Crazy xdl interfaces.. */
3182 const char *diffopts = getenv("GIT_DIFF_OPTS");
3183 const char *v;
3184 xpparam_t xpp;
3185 xdemitconf_t xecfg;
3186 struct emit_callback ecbdata;
3187 const struct userdiff_funcname *pe;
3189 if (must_show_header) {
3190 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3191 header.buf, header.len, 0);
3192 strbuf_reset(&header);
3195 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
3196 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
3198 pe = diff_funcname_pattern(one);
3199 if (!pe)
3200 pe = diff_funcname_pattern(two);
3202 memset(&xpp, 0, sizeof(xpp));
3203 memset(&xecfg, 0, sizeof(xecfg));
3204 memset(&ecbdata, 0, sizeof(ecbdata));
3205 ecbdata.label_path = lbl;
3206 ecbdata.color_diff = want_color(o->use_color);
3207 ecbdata.ws_rule = whitespace_rule(name_b);
3208 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
3209 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3210 ecbdata.opt = o;
3211 ecbdata.header = header.len ? &header : NULL;
3212 xpp.flags = o->xdl_opts;
3213 xpp.anchors = o->anchors;
3214 xpp.anchors_nr = o->anchors_nr;
3215 xecfg.ctxlen = o->context;
3216 xecfg.interhunkctxlen = o->interhunkcontext;
3217 xecfg.flags = XDL_EMIT_FUNCNAMES;
3218 if (o->flags.funccontext)
3219 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
3220 if (pe)
3221 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
3222 if (!diffopts)
3224 else if (skip_prefix(diffopts, "--unified=", &v))
3225 xecfg.ctxlen = strtoul(v, NULL, 10);
3226 else if (skip_prefix(diffopts, "-u", &v))
3227 xecfg.ctxlen = strtoul(v, NULL, 10);
3228 if (o->word_diff)
3229 init_diff_words_data(&ecbdata, o, one, two);
3230 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
3231 &xpp, &xecfg))
3232 die("unable to generate diff for %s", one->path);
3233 if (o->word_diff)
3234 free_diff_words_data(&ecbdata);
3235 if (textconv_one)
3236 free(mf1.ptr);
3237 if (textconv_two)
3238 free(mf2.ptr);
3239 xdiff_clear_find_func(&xecfg);
3242 free_ab_and_return:
3243 strbuf_release(&header);
3244 diff_free_filespec_data(one);
3245 diff_free_filespec_data(two);
3246 free(a_one);
3247 free(b_two);
3248 return;
3251 static void builtin_diffstat(const char *name_a, const char *name_b,
3252 struct diff_filespec *one,
3253 struct diff_filespec *two,
3254 struct diffstat_t *diffstat,
3255 struct diff_options *o,
3256 struct diff_filepair *p)
3258 mmfile_t mf1, mf2;
3259 struct diffstat_file *data;
3260 int same_contents;
3261 int complete_rewrite = 0;
3263 if (!DIFF_PAIR_UNMERGED(p)) {
3264 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3265 complete_rewrite = 1;
3268 data = diffstat_add(diffstat, name_a, name_b);
3269 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
3271 if (!one || !two) {
3272 data->is_unmerged = 1;
3273 return;
3276 same_contents = !oidcmp(&one->oid, &two->oid);
3278 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
3279 data->is_binary = 1;
3280 if (same_contents) {
3281 data->added = 0;
3282 data->deleted = 0;
3283 } else {
3284 data->added = diff_filespec_size(two);
3285 data->deleted = diff_filespec_size(one);
3289 else if (complete_rewrite) {
3290 diff_populate_filespec(one, 0);
3291 diff_populate_filespec(two, 0);
3292 data->deleted = count_lines(one->data, one->size);
3293 data->added = count_lines(two->data, two->size);
3296 else if (!same_contents) {
3297 /* Crazy xdl interfaces.. */
3298 xpparam_t xpp;
3299 xdemitconf_t xecfg;
3301 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3302 die("unable to read files to diff");
3304 memset(&xpp, 0, sizeof(xpp));
3305 memset(&xecfg, 0, sizeof(xecfg));
3306 xpp.flags = o->xdl_opts;
3307 xpp.anchors = o->anchors;
3308 xpp.anchors_nr = o->anchors_nr;
3309 xecfg.ctxlen = o->context;
3310 xecfg.interhunkctxlen = o->interhunkcontext;
3311 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
3312 &xpp, &xecfg))
3313 die("unable to generate diffstat for %s", one->path);
3316 diff_free_filespec_data(one);
3317 diff_free_filespec_data(two);
3320 static void builtin_checkdiff(const char *name_a, const char *name_b,
3321 const char *attr_path,
3322 struct diff_filespec *one,
3323 struct diff_filespec *two,
3324 struct diff_options *o)
3326 mmfile_t mf1, mf2;
3327 struct checkdiff_t data;
3329 if (!two)
3330 return;
3332 memset(&data, 0, sizeof(data));
3333 data.filename = name_b ? name_b : name_a;
3334 data.lineno = 0;
3335 data.o = o;
3336 data.ws_rule = whitespace_rule(attr_path);
3337 data.conflict_marker_size = ll_merge_marker_size(attr_path);
3339 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3340 die("unable to read files to diff");
3343 * All the other codepaths check both sides, but not checking
3344 * the "old" side here is deliberate. We are checking the newly
3345 * introduced changes, and as long as the "new" side is text, we
3346 * can and should check what it introduces.
3348 if (diff_filespec_is_binary(two))
3349 goto free_and_return;
3350 else {
3351 /* Crazy xdl interfaces.. */
3352 xpparam_t xpp;
3353 xdemitconf_t xecfg;
3355 memset(&xpp, 0, sizeof(xpp));
3356 memset(&xecfg, 0, sizeof(xecfg));
3357 xecfg.ctxlen = 1; /* at least one context line */
3358 xpp.flags = 0;
3359 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
3360 &xpp, &xecfg))
3361 die("unable to generate checkdiff for %s", one->path);
3363 if (data.ws_rule & WS_BLANK_AT_EOF) {
3364 struct emit_callback ecbdata;
3365 int blank_at_eof;
3367 ecbdata.ws_rule = data.ws_rule;
3368 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3369 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
3371 if (blank_at_eof) {
3372 static char *err;
3373 if (!err)
3374 err = whitespace_error_string(WS_BLANK_AT_EOF);
3375 fprintf(o->file, "%s:%d: %s.\n",
3376 data.filename, blank_at_eof, err);
3377 data.status = 1; /* report errors */
3381 free_and_return:
3382 diff_free_filespec_data(one);
3383 diff_free_filespec_data(two);
3384 if (data.status)
3385 o->flags.check_failed = 1;
3388 struct diff_filespec *alloc_filespec(const char *path)
3390 struct diff_filespec *spec;
3392 FLEXPTR_ALLOC_STR(spec, path, path);
3393 spec->count = 1;
3394 spec->is_binary = -1;
3395 return spec;
3398 void free_filespec(struct diff_filespec *spec)
3400 if (!--spec->count) {
3401 diff_free_filespec_data(spec);
3402 free(spec);
3406 void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
3407 int oid_valid, unsigned short mode)
3409 if (mode) {
3410 spec->mode = canon_mode(mode);
3411 oidcpy(&spec->oid, oid);
3412 spec->oid_valid = oid_valid;
3417 * Given a name and sha1 pair, if the index tells us the file in
3418 * the work tree has that object contents, return true, so that
3419 * prepare_temp_file() does not have to inflate and extract.
3421 static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
3423 const struct cache_entry *ce;
3424 struct stat st;
3425 int pos, len;
3428 * We do not read the cache ourselves here, because the
3429 * benchmark with my previous version that always reads cache
3430 * shows that it makes things worse for diff-tree comparing
3431 * two linux-2.6 kernel trees in an already checked out work
3432 * tree. This is because most diff-tree comparisons deal with
3433 * only a small number of files, while reading the cache is
3434 * expensive for a large project, and its cost outweighs the
3435 * savings we get by not inflating the object to a temporary
3436 * file. Practically, this code only helps when we are used
3437 * by diff-cache --cached, which does read the cache before
3438 * calling us.
3440 if (!active_cache)
3441 return 0;
3443 /* We want to avoid the working directory if our caller
3444 * doesn't need the data in a normal file, this system
3445 * is rather slow with its stat/open/mmap/close syscalls,
3446 * and the object is contained in a pack file. The pack
3447 * is probably already open and will be faster to obtain
3448 * the data through than the working directory. Loose
3449 * objects however would tend to be slower as they need
3450 * to be individually opened and inflated.
3452 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
3453 return 0;
3456 * Similarly, if we'd have to convert the file contents anyway, that
3457 * makes the optimization not worthwhile.
3459 if (!want_file && would_convert_to_git(&the_index, name))
3460 return 0;
3462 len = strlen(name);
3463 pos = cache_name_pos(name, len);
3464 if (pos < 0)
3465 return 0;
3466 ce = active_cache[pos];
3469 * This is not the sha1 we are looking for, or
3470 * unreusable because it is not a regular file.
3472 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
3473 return 0;
3476 * If ce is marked as "assume unchanged", there is no
3477 * guarantee that work tree matches what we are looking for.
3479 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
3480 return 0;
3483 * If ce matches the file in the work tree, we can reuse it.
3485 if (ce_uptodate(ce) ||
3486 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
3487 return 1;
3489 return 0;
3492 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
3494 struct strbuf buf = STRBUF_INIT;
3495 char *dirty = "";
3497 /* Are we looking at the work tree? */
3498 if (s->dirty_submodule)
3499 dirty = "-dirty";
3501 strbuf_addf(&buf, "Subproject commit %s%s\n",
3502 oid_to_hex(&s->oid), dirty);
3503 s->size = buf.len;
3504 if (size_only) {
3505 s->data = NULL;
3506 strbuf_release(&buf);
3507 } else {
3508 s->data = strbuf_detach(&buf, NULL);
3509 s->should_free = 1;
3511 return 0;
3515 * While doing rename detection and pickaxe operation, we may need to
3516 * grab the data for the blob (or file) for our own in-core comparison.
3517 * diff_filespec has data and size fields for this purpose.
3519 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
3521 int size_only = flags & CHECK_SIZE_ONLY;
3522 int err = 0;
3524 * demote FAIL to WARN to allow inspecting the situation
3525 * instead of refusing.
3527 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
3528 ? SAFE_CRLF_WARN
3529 : safe_crlf);
3531 if (!DIFF_FILE_VALID(s))
3532 die("internal error: asking to populate invalid file.");
3533 if (S_ISDIR(s->mode))
3534 return -1;
3536 if (s->data)
3537 return 0;
3539 if (size_only && 0 < s->size)
3540 return 0;
3542 if (S_ISGITLINK(s->mode))
3543 return diff_populate_gitlink(s, size_only);
3545 if (!s->oid_valid ||
3546 reuse_worktree_file(s->path, &s->oid, 0)) {
3547 struct strbuf buf = STRBUF_INIT;
3548 struct stat st;
3549 int fd;
3551 if (lstat(s->path, &st) < 0) {
3552 err_empty:
3553 err = -1;
3554 empty:
3555 s->data = (char *)"";
3556 s->size = 0;
3557 return err;
3559 s->size = xsize_t(st.st_size);
3560 if (!s->size)
3561 goto empty;
3562 if (S_ISLNK(st.st_mode)) {
3563 struct strbuf sb = STRBUF_INIT;
3565 if (strbuf_readlink(&sb, s->path, s->size))
3566 goto err_empty;
3567 s->size = sb.len;
3568 s->data = strbuf_detach(&sb, NULL);
3569 s->should_free = 1;
3570 return 0;
3574 * Even if the caller would be happy with getting
3575 * only the size, we cannot return early at this
3576 * point if the path requires us to run the content
3577 * conversion.
3579 if (size_only && !would_convert_to_git(&the_index, s->path))
3580 return 0;
3583 * Note: this check uses xsize_t(st.st_size) that may
3584 * not be the true size of the blob after it goes
3585 * through convert_to_git(). This may not strictly be
3586 * correct, but the whole point of big_file_threshold
3587 * and is_binary check being that we want to avoid
3588 * opening the file and inspecting the contents, this
3589 * is probably fine.
3591 if ((flags & CHECK_BINARY) &&
3592 s->size > big_file_threshold && s->is_binary == -1) {
3593 s->is_binary = 1;
3594 return 0;
3596 fd = open(s->path, O_RDONLY);
3597 if (fd < 0)
3598 goto err_empty;
3599 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
3600 close(fd);
3601 s->should_munmap = 1;
3604 * Convert from working tree format to canonical git format
3606 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
3607 size_t size = 0;
3608 munmap(s->data, s->size);
3609 s->should_munmap = 0;
3610 s->data = strbuf_detach(&buf, &size);
3611 s->size = size;
3612 s->should_free = 1;
3615 else {
3616 enum object_type type;
3617 if (size_only || (flags & CHECK_BINARY)) {
3618 type = sha1_object_info(s->oid.hash, &s->size);
3619 if (type < 0)
3620 die("unable to read %s",
3621 oid_to_hex(&s->oid));
3622 if (size_only)
3623 return 0;
3624 if (s->size > big_file_threshold && s->is_binary == -1) {
3625 s->is_binary = 1;
3626 return 0;
3629 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
3630 if (!s->data)
3631 die("unable to read %s", oid_to_hex(&s->oid));
3632 s->should_free = 1;
3634 return 0;
3637 void diff_free_filespec_blob(struct diff_filespec *s)
3639 if (s->should_free)
3640 free(s->data);
3641 else if (s->should_munmap)
3642 munmap(s->data, s->size);
3644 if (s->should_free || s->should_munmap) {
3645 s->should_free = s->should_munmap = 0;
3646 s->data = NULL;
3650 void diff_free_filespec_data(struct diff_filespec *s)
3652 diff_free_filespec_blob(s);
3653 FREE_AND_NULL(s->cnt_data);
3656 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
3657 void *blob,
3658 unsigned long size,
3659 const struct object_id *oid,
3660 int mode)
3662 struct strbuf buf = STRBUF_INIT;
3663 struct strbuf template = STRBUF_INIT;
3664 char *path_dup = xstrdup(path);
3665 const char *base = basename(path_dup);
3667 /* Generate "XXXXXX_basename.ext" */
3668 strbuf_addstr(&template, "XXXXXX_");
3669 strbuf_addstr(&template, base);
3671 temp->tempfile = mks_tempfile_ts(template.buf, strlen(base) + 1);
3672 if (!temp->tempfile)
3673 die_errno("unable to create temp-file");
3674 if (convert_to_working_tree(path,
3675 (const char *)blob, (size_t)size, &buf)) {
3676 blob = buf.buf;
3677 size = buf.len;
3679 if (write_in_full(temp->tempfile->fd, blob, size) < 0 ||
3680 close_tempfile_gently(temp->tempfile))
3681 die_errno("unable to write temp-file");
3682 temp->name = get_tempfile_path(temp->tempfile);
3683 oid_to_hex_r(temp->hex, oid);
3684 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3685 strbuf_release(&buf);
3686 strbuf_release(&template);
3687 free(path_dup);
3690 static struct diff_tempfile *prepare_temp_file(const char *name,
3691 struct diff_filespec *one)
3693 struct diff_tempfile *temp = claim_diff_tempfile();
3695 if (!DIFF_FILE_VALID(one)) {
3696 not_a_valid_file:
3697 /* A '-' entry produces this for file-2, and
3698 * a '+' entry produces this for file-1.
3700 temp->name = "/dev/null";
3701 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3702 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3703 return temp;
3706 if (!S_ISGITLINK(one->mode) &&
3707 (!one->oid_valid ||
3708 reuse_worktree_file(name, &one->oid, 1))) {
3709 struct stat st;
3710 if (lstat(name, &st) < 0) {
3711 if (errno == ENOENT)
3712 goto not_a_valid_file;
3713 die_errno("stat(%s)", name);
3715 if (S_ISLNK(st.st_mode)) {
3716 struct strbuf sb = STRBUF_INIT;
3717 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3718 die_errno("readlink(%s)", name);
3719 prep_temp_blob(name, temp, sb.buf, sb.len,
3720 (one->oid_valid ?
3721 &one->oid : &null_oid),
3722 (one->oid_valid ?
3723 one->mode : S_IFLNK));
3724 strbuf_release(&sb);
3726 else {
3727 /* we can borrow from the file in the work tree */
3728 temp->name = name;
3729 if (!one->oid_valid)
3730 oid_to_hex_r(temp->hex, &null_oid);
3731 else
3732 oid_to_hex_r(temp->hex, &one->oid);
3733 /* Even though we may sometimes borrow the
3734 * contents from the work tree, we always want
3735 * one->mode. mode is trustworthy even when
3736 * !(one->oid_valid), as long as
3737 * DIFF_FILE_VALID(one).
3739 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3741 return temp;
3743 else {
3744 if (diff_populate_filespec(one, 0))
3745 die("cannot read data blob for %s", one->path);
3746 prep_temp_blob(name, temp, one->data, one->size,
3747 &one->oid, one->mode);
3749 return temp;
3752 static void add_external_diff_name(struct argv_array *argv,
3753 const char *name,
3754 struct diff_filespec *df)
3756 struct diff_tempfile *temp = prepare_temp_file(name, df);
3757 argv_array_push(argv, temp->name);
3758 argv_array_push(argv, temp->hex);
3759 argv_array_push(argv, temp->mode);
3762 /* An external diff command takes:
3764 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3765 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3768 static void run_external_diff(const char *pgm,
3769 const char *name,
3770 const char *other,
3771 struct diff_filespec *one,
3772 struct diff_filespec *two,
3773 const char *xfrm_msg,
3774 int complete_rewrite,
3775 struct diff_options *o)
3777 struct argv_array argv = ARGV_ARRAY_INIT;
3778 struct argv_array env = ARGV_ARRAY_INIT;
3779 struct diff_queue_struct *q = &diff_queued_diff;
3781 argv_array_push(&argv, pgm);
3782 argv_array_push(&argv, name);
3784 if (one && two) {
3785 add_external_diff_name(&argv, name, one);
3786 if (!other)
3787 add_external_diff_name(&argv, name, two);
3788 else {
3789 add_external_diff_name(&argv, other, two);
3790 argv_array_push(&argv, other);
3791 argv_array_push(&argv, xfrm_msg);
3795 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3796 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3798 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3799 die(_("external diff died, stopping at %s"), name);
3801 remove_tempfile();
3802 argv_array_clear(&argv);
3803 argv_array_clear(&env);
3806 static int similarity_index(struct diff_filepair *p)
3808 return p->score * 100 / MAX_SCORE;
3811 static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3813 if (startup_info->have_repository)
3814 return find_unique_abbrev(oid->hash, abbrev);
3815 else {
3816 char *hex = oid_to_hex(oid);
3817 if (abbrev < 0)
3818 abbrev = FALLBACK_DEFAULT_ABBREV;
3819 if (abbrev > GIT_SHA1_HEXSZ)
3820 die("BUG: oid abbreviation out of range: %d", abbrev);
3821 if (abbrev)
3822 hex[abbrev] = '\0';
3823 return hex;
3827 static void fill_metainfo(struct strbuf *msg,
3828 const char *name,
3829 const char *other,
3830 struct diff_filespec *one,
3831 struct diff_filespec *two,
3832 struct diff_options *o,
3833 struct diff_filepair *p,
3834 int *must_show_header,
3835 int use_color)
3837 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3838 const char *reset = diff_get_color(use_color, DIFF_RESET);
3839 const char *line_prefix = diff_line_prefix(o);
3841 *must_show_header = 1;
3842 strbuf_init(msg, PATH_MAX * 2 + 300);
3843 switch (p->status) {
3844 case DIFF_STATUS_COPIED:
3845 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3846 line_prefix, set, similarity_index(p));
3847 strbuf_addf(msg, "%s\n%s%scopy from ",
3848 reset, line_prefix, set);
3849 quote_c_style(name, msg, NULL, 0);
3850 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3851 quote_c_style(other, msg, NULL, 0);
3852 strbuf_addf(msg, "%s\n", reset);
3853 break;
3854 case DIFF_STATUS_RENAMED:
3855 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3856 line_prefix, set, similarity_index(p));
3857 strbuf_addf(msg, "%s\n%s%srename from ",
3858 reset, line_prefix, set);
3859 quote_c_style(name, msg, NULL, 0);
3860 strbuf_addf(msg, "%s\n%s%srename to ",
3861 reset, line_prefix, set);
3862 quote_c_style(other, msg, NULL, 0);
3863 strbuf_addf(msg, "%s\n", reset);
3864 break;
3865 case DIFF_STATUS_MODIFIED: