Merge branch 'sg/test-bash-version-fix'
[git.git] / diff.c
blobdc9965e836cd49c46a01293ef1d678a619b8f997
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 "object-store.h"
17 #include "userdiff.h"
18 #include "submodule-config.h"
19 #include "submodule.h"
20 #include "hashmap.h"
21 #include "ll-merge.h"
22 #include "string-list.h"
23 #include "argv-array.h"
24 #include "graph.h"
25 #include "packfile.h"
26 #include "help.h"
28 #ifdef NO_FAST_WORKING_DIRECTORY
29 #define FAST_WORKING_DIRECTORY 0
30 #else
31 #define FAST_WORKING_DIRECTORY 1
32 #endif
34 static int diff_detect_rename_default;
35 static int diff_indent_heuristic = 1;
36 static int diff_rename_limit_default = 400;
37 static int diff_suppress_blank_empty;
38 static int diff_use_color_default = -1;
39 static int diff_color_moved_default;
40 static int diff_color_moved_ws_default;
41 static int diff_context_default = 3;
42 static int diff_interhunk_context_default;
43 static const char *diff_word_regex_cfg;
44 static const char *external_diff_cmd_cfg;
45 static const char *diff_order_file_cfg;
46 int diff_auto_refresh_index = 1;
47 static int diff_mnemonic_prefix;
48 static int diff_no_prefix;
49 static int diff_stat_graph_width;
50 static int diff_dirstat_permille_default = 30;
51 static struct diff_options default_diff_options;
52 static long diff_algorithm;
53 static unsigned ws_error_highlight_default = WSEH_NEW;
55 static char diff_colors[][COLOR_MAXLEN] = {
56 GIT_COLOR_RESET,
57 GIT_COLOR_NORMAL, /* CONTEXT */
58 GIT_COLOR_BOLD, /* METAINFO */
59 GIT_COLOR_CYAN, /* FRAGINFO */
60 GIT_COLOR_RED, /* OLD */
61 GIT_COLOR_GREEN, /* NEW */
62 GIT_COLOR_YELLOW, /* COMMIT */
63 GIT_COLOR_BG_RED, /* WHITESPACE */
64 GIT_COLOR_NORMAL, /* FUNCINFO */
65 GIT_COLOR_BOLD_MAGENTA, /* OLD_MOVED */
66 GIT_COLOR_BOLD_BLUE, /* OLD_MOVED ALTERNATIVE */
67 GIT_COLOR_FAINT, /* OLD_MOVED_DIM */
68 GIT_COLOR_FAINT_ITALIC, /* OLD_MOVED_ALTERNATIVE_DIM */
69 GIT_COLOR_BOLD_CYAN, /* NEW_MOVED */
70 GIT_COLOR_BOLD_YELLOW, /* NEW_MOVED ALTERNATIVE */
71 GIT_COLOR_FAINT, /* NEW_MOVED_DIM */
72 GIT_COLOR_FAINT_ITALIC, /* NEW_MOVED_ALTERNATIVE_DIM */
73 GIT_COLOR_FAINT, /* CONTEXT_DIM */
74 GIT_COLOR_FAINT_RED, /* OLD_DIM */
75 GIT_COLOR_FAINT_GREEN, /* NEW_DIM */
76 GIT_COLOR_BOLD, /* CONTEXT_BOLD */
77 GIT_COLOR_BOLD_RED, /* OLD_BOLD */
78 GIT_COLOR_BOLD_GREEN, /* NEW_BOLD */
81 static const char *color_diff_slots[] = {
82 [DIFF_CONTEXT] = "context",
83 [DIFF_METAINFO] = "meta",
84 [DIFF_FRAGINFO] = "frag",
85 [DIFF_FILE_OLD] = "old",
86 [DIFF_FILE_NEW] = "new",
87 [DIFF_COMMIT] = "commit",
88 [DIFF_WHITESPACE] = "whitespace",
89 [DIFF_FUNCINFO] = "func",
90 [DIFF_FILE_OLD_MOVED] = "oldMoved",
91 [DIFF_FILE_OLD_MOVED_ALT] = "oldMovedAlternative",
92 [DIFF_FILE_OLD_MOVED_DIM] = "oldMovedDimmed",
93 [DIFF_FILE_OLD_MOVED_ALT_DIM] = "oldMovedAlternativeDimmed",
94 [DIFF_FILE_NEW_MOVED] = "newMoved",
95 [DIFF_FILE_NEW_MOVED_ALT] = "newMovedAlternative",
96 [DIFF_FILE_NEW_MOVED_DIM] = "newMovedDimmed",
97 [DIFF_FILE_NEW_MOVED_ALT_DIM] = "newMovedAlternativeDimmed",
98 [DIFF_CONTEXT_DIM] = "contextDimmed",
99 [DIFF_FILE_OLD_DIM] = "oldDimmed",
100 [DIFF_FILE_NEW_DIM] = "newDimmed",
101 [DIFF_CONTEXT_BOLD] = "contextBold",
102 [DIFF_FILE_OLD_BOLD] = "oldBold",
103 [DIFF_FILE_NEW_BOLD] = "newBold",
106 static NORETURN void die_want_option(const char *option_name)
108 die(_("option '%s' requires a value"), option_name);
111 define_list_config_array_extra(color_diff_slots, {"plain"});
113 static int parse_diff_color_slot(const char *var)
115 if (!strcasecmp(var, "plain"))
116 return DIFF_CONTEXT;
117 return LOOKUP_CONFIG(color_diff_slots, var);
120 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
121 struct strbuf *errmsg)
123 char *params_copy = xstrdup(params_string);
124 struct string_list params = STRING_LIST_INIT_NODUP;
125 int ret = 0;
126 int i;
128 if (*params_copy)
129 string_list_split_in_place(&params, params_copy, ',', -1);
130 for (i = 0; i < params.nr; i++) {
131 const char *p = params.items[i].string;
132 if (!strcmp(p, "changes")) {
133 options->flags.dirstat_by_line = 0;
134 options->flags.dirstat_by_file = 0;
135 } else if (!strcmp(p, "lines")) {
136 options->flags.dirstat_by_line = 1;
137 options->flags.dirstat_by_file = 0;
138 } else if (!strcmp(p, "files")) {
139 options->flags.dirstat_by_line = 0;
140 options->flags.dirstat_by_file = 1;
141 } else if (!strcmp(p, "noncumulative")) {
142 options->flags.dirstat_cumulative = 0;
143 } else if (!strcmp(p, "cumulative")) {
144 options->flags.dirstat_cumulative = 1;
145 } else if (isdigit(*p)) {
146 char *end;
147 int permille = strtoul(p, &end, 10) * 10;
148 if (*end == '.' && isdigit(*++end)) {
149 /* only use first digit */
150 permille += *end - '0';
151 /* .. and ignore any further digits */
152 while (isdigit(*++end))
153 ; /* nothing */
155 if (!*end)
156 options->dirstat_permille = permille;
157 else {
158 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
160 ret++;
162 } else {
163 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
164 ret++;
168 string_list_clear(&params, 0);
169 free(params_copy);
170 return ret;
173 static int parse_submodule_params(struct diff_options *options, const char *value)
175 if (!strcmp(value, "log"))
176 options->submodule_format = DIFF_SUBMODULE_LOG;
177 else if (!strcmp(value, "short"))
178 options->submodule_format = DIFF_SUBMODULE_SHORT;
179 else if (!strcmp(value, "diff"))
180 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
181 else
182 return -1;
183 return 0;
186 int git_config_rename(const char *var, const char *value)
188 if (!value)
189 return DIFF_DETECT_RENAME;
190 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
191 return DIFF_DETECT_COPY;
192 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
195 long parse_algorithm_value(const char *value)
197 if (!value)
198 return -1;
199 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
200 return 0;
201 else if (!strcasecmp(value, "minimal"))
202 return XDF_NEED_MINIMAL;
203 else if (!strcasecmp(value, "patience"))
204 return XDF_PATIENCE_DIFF;
205 else if (!strcasecmp(value, "histogram"))
206 return XDF_HISTOGRAM_DIFF;
207 return -1;
210 static int parse_one_token(const char **arg, const char *token)
212 const char *rest;
213 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
214 *arg = rest;
215 return 1;
217 return 0;
220 static int parse_ws_error_highlight(const char *arg)
222 const char *orig_arg = arg;
223 unsigned val = 0;
225 while (*arg) {
226 if (parse_one_token(&arg, "none"))
227 val = 0;
228 else if (parse_one_token(&arg, "default"))
229 val = WSEH_NEW;
230 else if (parse_one_token(&arg, "all"))
231 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
232 else if (parse_one_token(&arg, "new"))
233 val |= WSEH_NEW;
234 else if (parse_one_token(&arg, "old"))
235 val |= WSEH_OLD;
236 else if (parse_one_token(&arg, "context"))
237 val |= WSEH_CONTEXT;
238 else {
239 return -1 - (int)(arg - orig_arg);
241 if (*arg)
242 arg++;
244 return val;
248 * These are to give UI layer defaults.
249 * The core-level commands such as git-diff-files should
250 * never be affected by the setting of diff.renames
251 * the user happens to have in the configuration file.
253 void init_diff_ui_defaults(void)
255 diff_detect_rename_default = DIFF_DETECT_RENAME;
258 int git_diff_heuristic_config(const char *var, const char *value, void *cb)
260 if (!strcmp(var, "diff.indentheuristic"))
261 diff_indent_heuristic = git_config_bool(var, value);
262 return 0;
265 static int parse_color_moved(const char *arg)
267 switch (git_parse_maybe_bool(arg)) {
268 case 0:
269 return COLOR_MOVED_NO;
270 case 1:
271 return COLOR_MOVED_DEFAULT;
272 default:
273 break;
276 if (!strcmp(arg, "no"))
277 return COLOR_MOVED_NO;
278 else if (!strcmp(arg, "plain"))
279 return COLOR_MOVED_PLAIN;
280 else if (!strcmp(arg, "blocks"))
281 return COLOR_MOVED_BLOCKS;
282 else if (!strcmp(arg, "zebra"))
283 return COLOR_MOVED_ZEBRA;
284 else if (!strcmp(arg, "default"))
285 return COLOR_MOVED_DEFAULT;
286 else if (!strcmp(arg, "dimmed-zebra"))
287 return COLOR_MOVED_ZEBRA_DIM;
288 else if (!strcmp(arg, "dimmed_zebra"))
289 return COLOR_MOVED_ZEBRA_DIM;
290 else
291 return error(_("color moved setting must be one of 'no', 'default', 'blocks', 'zebra', 'dimmed-zebra', 'plain'"));
294 static int parse_color_moved_ws(const char *arg)
296 int ret = 0;
297 struct string_list l = STRING_LIST_INIT_DUP;
298 struct string_list_item *i;
300 string_list_split(&l, arg, ',', -1);
302 for_each_string_list_item(i, &l) {
303 struct strbuf sb = STRBUF_INIT;
304 strbuf_addstr(&sb, i->string);
305 strbuf_trim(&sb);
307 if (!strcmp(sb.buf, "ignore-space-change"))
308 ret |= XDF_IGNORE_WHITESPACE_CHANGE;
309 else if (!strcmp(sb.buf, "ignore-space-at-eol"))
310 ret |= XDF_IGNORE_WHITESPACE_AT_EOL;
311 else if (!strcmp(sb.buf, "ignore-all-space"))
312 ret |= XDF_IGNORE_WHITESPACE;
313 else if (!strcmp(sb.buf, "allow-indentation-change"))
314 ret |= COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE;
315 else
316 error(_("ignoring unknown color-moved-ws mode '%s'"), sb.buf);
318 strbuf_release(&sb);
321 if ((ret & COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE) &&
322 (ret & XDF_WHITESPACE_FLAGS))
323 die(_("color-moved-ws: allow-indentation-change cannot be combined with other white space modes"));
325 string_list_clear(&l, 0);
327 return ret;
330 int git_diff_ui_config(const char *var, const char *value, void *cb)
332 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
333 diff_use_color_default = git_config_colorbool(var, value);
334 return 0;
336 if (!strcmp(var, "diff.colormoved")) {
337 int cm = parse_color_moved(value);
338 if (cm < 0)
339 return -1;
340 diff_color_moved_default = cm;
341 return 0;
343 if (!strcmp(var, "diff.colormovedws")) {
344 int cm = parse_color_moved_ws(value);
345 if (cm < 0)
346 return -1;
347 diff_color_moved_ws_default = cm;
348 return 0;
350 if (!strcmp(var, "diff.context")) {
351 diff_context_default = git_config_int(var, value);
352 if (diff_context_default < 0)
353 return -1;
354 return 0;
356 if (!strcmp(var, "diff.interhunkcontext")) {
357 diff_interhunk_context_default = git_config_int(var, value);
358 if (diff_interhunk_context_default < 0)
359 return -1;
360 return 0;
362 if (!strcmp(var, "diff.renames")) {
363 diff_detect_rename_default = git_config_rename(var, value);
364 return 0;
366 if (!strcmp(var, "diff.autorefreshindex")) {
367 diff_auto_refresh_index = git_config_bool(var, value);
368 return 0;
370 if (!strcmp(var, "diff.mnemonicprefix")) {
371 diff_mnemonic_prefix = git_config_bool(var, value);
372 return 0;
374 if (!strcmp(var, "diff.noprefix")) {
375 diff_no_prefix = git_config_bool(var, value);
376 return 0;
378 if (!strcmp(var, "diff.statgraphwidth")) {
379 diff_stat_graph_width = git_config_int(var, value);
380 return 0;
382 if (!strcmp(var, "diff.external"))
383 return git_config_string(&external_diff_cmd_cfg, var, value);
384 if (!strcmp(var, "diff.wordregex"))
385 return git_config_string(&diff_word_regex_cfg, var, value);
386 if (!strcmp(var, "diff.orderfile"))
387 return git_config_pathname(&diff_order_file_cfg, var, value);
389 if (!strcmp(var, "diff.ignoresubmodules"))
390 handle_ignore_submodules_arg(&default_diff_options, value);
392 if (!strcmp(var, "diff.submodule")) {
393 if (parse_submodule_params(&default_diff_options, value))
394 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
395 value);
396 return 0;
399 if (!strcmp(var, "diff.algorithm")) {
400 diff_algorithm = parse_algorithm_value(value);
401 if (diff_algorithm < 0)
402 return -1;
403 return 0;
406 if (!strcmp(var, "diff.wserrorhighlight")) {
407 int val = parse_ws_error_highlight(value);
408 if (val < 0)
409 return -1;
410 ws_error_highlight_default = val;
411 return 0;
414 if (git_color_config(var, value, cb) < 0)
415 return -1;
417 return git_diff_basic_config(var, value, cb);
420 int git_diff_basic_config(const char *var, const char *value, void *cb)
422 const char *name;
424 if (!strcmp(var, "diff.renamelimit")) {
425 diff_rename_limit_default = git_config_int(var, value);
426 return 0;
429 if (userdiff_config(var, value) < 0)
430 return -1;
432 if (skip_prefix(var, "diff.color.", &name) ||
433 skip_prefix(var, "color.diff.", &name)) {
434 int slot = parse_diff_color_slot(name);
435 if (slot < 0)
436 return 0;
437 if (!value)
438 return config_error_nonbool(var);
439 return color_parse(value, diff_colors[slot]);
442 /* like GNU diff's --suppress-blank-empty option */
443 if (!strcmp(var, "diff.suppressblankempty") ||
444 /* for backwards compatibility */
445 !strcmp(var, "diff.suppress-blank-empty")) {
446 diff_suppress_blank_empty = git_config_bool(var, value);
447 return 0;
450 if (!strcmp(var, "diff.dirstat")) {
451 struct strbuf errmsg = STRBUF_INIT;
452 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
453 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
454 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
455 errmsg.buf);
456 strbuf_release(&errmsg);
457 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
458 return 0;
461 if (git_diff_heuristic_config(var, value, cb) < 0)
462 return -1;
464 return git_default_config(var, value, cb);
467 static char *quote_two(const char *one, const char *two)
469 int need_one = quote_c_style(one, NULL, NULL, 1);
470 int need_two = quote_c_style(two, NULL, NULL, 1);
471 struct strbuf res = STRBUF_INIT;
473 if (need_one + need_two) {
474 strbuf_addch(&res, '"');
475 quote_c_style(one, &res, NULL, 1);
476 quote_c_style(two, &res, NULL, 1);
477 strbuf_addch(&res, '"');
478 } else {
479 strbuf_addstr(&res, one);
480 strbuf_addstr(&res, two);
482 return strbuf_detach(&res, NULL);
485 static const char *external_diff(void)
487 static const char *external_diff_cmd = NULL;
488 static int done_preparing = 0;
490 if (done_preparing)
491 return external_diff_cmd;
492 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
493 if (!external_diff_cmd)
494 external_diff_cmd = external_diff_cmd_cfg;
495 done_preparing = 1;
496 return external_diff_cmd;
500 * Keep track of files used for diffing. Sometimes such an entry
501 * refers to a temporary file, sometimes to an existing file, and
502 * sometimes to "/dev/null".
504 static struct diff_tempfile {
506 * filename external diff should read from, or NULL if this
507 * entry is currently not in use:
509 const char *name;
511 char hex[GIT_MAX_HEXSZ + 1];
512 char mode[10];
515 * If this diff_tempfile instance refers to a temporary file,
516 * this tempfile object is used to manage its lifetime.
518 struct tempfile *tempfile;
519 } diff_temp[2];
521 struct emit_callback {
522 int color_diff;
523 unsigned ws_rule;
524 int blank_at_eof_in_preimage;
525 int blank_at_eof_in_postimage;
526 int lno_in_preimage;
527 int lno_in_postimage;
528 const char **label_path;
529 struct diff_words_data *diff_words;
530 struct diff_options *opt;
531 struct strbuf *header;
534 static int count_lines(const char *data, int size)
536 int count, ch, completely_empty = 1, nl_just_seen = 0;
537 count = 0;
538 while (0 < size--) {
539 ch = *data++;
540 if (ch == '\n') {
541 count++;
542 nl_just_seen = 1;
543 completely_empty = 0;
545 else {
546 nl_just_seen = 0;
547 completely_empty = 0;
550 if (completely_empty)
551 return 0;
552 if (!nl_just_seen)
553 count++; /* no trailing newline */
554 return count;
557 static int fill_mmfile(struct repository *r, mmfile_t *mf,
558 struct diff_filespec *one)
560 if (!DIFF_FILE_VALID(one)) {
561 mf->ptr = (char *)""; /* does not matter */
562 mf->size = 0;
563 return 0;
565 else if (diff_populate_filespec(r, one, 0))
566 return -1;
568 mf->ptr = one->data;
569 mf->size = one->size;
570 return 0;
573 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
574 static unsigned long diff_filespec_size(struct repository *r,
575 struct diff_filespec *one)
577 if (!DIFF_FILE_VALID(one))
578 return 0;
579 diff_populate_filespec(r, one, CHECK_SIZE_ONLY);
580 return one->size;
583 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
585 char *ptr = mf->ptr;
586 long size = mf->size;
587 int cnt = 0;
589 if (!size)
590 return cnt;
591 ptr += size - 1; /* pointing at the very end */
592 if (*ptr != '\n')
593 ; /* incomplete line */
594 else
595 ptr--; /* skip the last LF */
596 while (mf->ptr < ptr) {
597 char *prev_eol;
598 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
599 if (*prev_eol == '\n')
600 break;
601 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
602 break;
603 cnt++;
604 ptr = prev_eol - 1;
606 return cnt;
609 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
610 struct emit_callback *ecbdata)
612 int l1, l2, at;
613 unsigned ws_rule = ecbdata->ws_rule;
614 l1 = count_trailing_blank(mf1, ws_rule);
615 l2 = count_trailing_blank(mf2, ws_rule);
616 if (l2 <= l1) {
617 ecbdata->blank_at_eof_in_preimage = 0;
618 ecbdata->blank_at_eof_in_postimage = 0;
619 return;
621 at = count_lines(mf1->ptr, mf1->size);
622 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
624 at = count_lines(mf2->ptr, mf2->size);
625 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
628 static void emit_line_0(struct diff_options *o,
629 const char *set_sign, const char *set, unsigned reverse, const char *reset,
630 int first, const char *line, int len)
632 int has_trailing_newline, has_trailing_carriage_return;
633 int needs_reset = 0; /* at the end of the line */
634 FILE *file = o->file;
636 fputs(diff_line_prefix(o), file);
638 has_trailing_newline = (len > 0 && line[len-1] == '\n');
639 if (has_trailing_newline)
640 len--;
642 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
643 if (has_trailing_carriage_return)
644 len--;
646 if (!len && !first)
647 goto end_of_line;
649 if (reverse && want_color(o->use_color)) {
650 fputs(GIT_COLOR_REVERSE, file);
651 needs_reset = 1;
654 if (set_sign) {
655 fputs(set_sign, file);
656 needs_reset = 1;
659 if (first)
660 fputc(first, file);
662 if (!len)
663 goto end_of_line;
665 if (set) {
666 if (set_sign && set != set_sign)
667 fputs(reset, file);
668 fputs(set, file);
669 needs_reset = 1;
671 fwrite(line, len, 1, file);
672 needs_reset = 1; /* 'line' may contain color codes. */
674 end_of_line:
675 if (needs_reset)
676 fputs(reset, file);
677 if (has_trailing_carriage_return)
678 fputc('\r', file);
679 if (has_trailing_newline)
680 fputc('\n', file);
683 static void emit_line(struct diff_options *o, const char *set, const char *reset,
684 const char *line, int len)
686 emit_line_0(o, set, NULL, 0, reset, 0, line, len);
689 enum diff_symbol {
690 DIFF_SYMBOL_BINARY_DIFF_HEADER,
691 DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
692 DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
693 DIFF_SYMBOL_BINARY_DIFF_BODY,
694 DIFF_SYMBOL_BINARY_DIFF_FOOTER,
695 DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
696 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
697 DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
698 DIFF_SYMBOL_STATS_LINE,
699 DIFF_SYMBOL_WORD_DIFF,
700 DIFF_SYMBOL_STAT_SEP,
701 DIFF_SYMBOL_SUMMARY,
702 DIFF_SYMBOL_SUBMODULE_ADD,
703 DIFF_SYMBOL_SUBMODULE_DEL,
704 DIFF_SYMBOL_SUBMODULE_UNTRACKED,
705 DIFF_SYMBOL_SUBMODULE_MODIFIED,
706 DIFF_SYMBOL_SUBMODULE_HEADER,
707 DIFF_SYMBOL_SUBMODULE_ERROR,
708 DIFF_SYMBOL_SUBMODULE_PIPETHROUGH,
709 DIFF_SYMBOL_REWRITE_DIFF,
710 DIFF_SYMBOL_BINARY_FILES,
711 DIFF_SYMBOL_HEADER,
712 DIFF_SYMBOL_FILEPAIR_PLUS,
713 DIFF_SYMBOL_FILEPAIR_MINUS,
714 DIFF_SYMBOL_WORDS_PORCELAIN,
715 DIFF_SYMBOL_WORDS,
716 DIFF_SYMBOL_CONTEXT,
717 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
718 DIFF_SYMBOL_PLUS,
719 DIFF_SYMBOL_MINUS,
720 DIFF_SYMBOL_NO_LF_EOF,
721 DIFF_SYMBOL_CONTEXT_FRAGINFO,
722 DIFF_SYMBOL_CONTEXT_MARKER,
723 DIFF_SYMBOL_SEPARATOR
726 * Flags for content lines:
727 * 0..12 are whitespace rules
728 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
729 * 16 is marking if the line is blank at EOF
731 #define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
732 #define DIFF_SYMBOL_MOVED_LINE (1<<17)
733 #define DIFF_SYMBOL_MOVED_LINE_ALT (1<<18)
734 #define DIFF_SYMBOL_MOVED_LINE_UNINTERESTING (1<<19)
735 #define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
738 * This struct is used when we need to buffer the output of the diff output.
740 * NEEDSWORK: Instead of storing a copy of the line, add an offset pointer
741 * into the pre/post image file. This pointer could be a union with the
742 * line pointer. By storing an offset into the file instead of the literal line,
743 * we can decrease the memory footprint for the buffered output. At first we
744 * may want to only have indirection for the content lines, but we could also
745 * enhance the state for emitting prefabricated lines, e.g. the similarity
746 * score line or hunk/file headers would only need to store a number or path
747 * and then the output can be constructed later on depending on state.
749 struct emitted_diff_symbol {
750 const char *line;
751 int len;
752 int flags;
753 enum diff_symbol s;
755 #define EMITTED_DIFF_SYMBOL_INIT {NULL}
757 struct emitted_diff_symbols {
758 struct emitted_diff_symbol *buf;
759 int nr, alloc;
761 #define EMITTED_DIFF_SYMBOLS_INIT {NULL, 0, 0}
763 static void append_emitted_diff_symbol(struct diff_options *o,
764 struct emitted_diff_symbol *e)
766 struct emitted_diff_symbol *f;
768 ALLOC_GROW(o->emitted_symbols->buf,
769 o->emitted_symbols->nr + 1,
770 o->emitted_symbols->alloc);
771 f = &o->emitted_symbols->buf[o->emitted_symbols->nr++];
773 memcpy(f, e, sizeof(struct emitted_diff_symbol));
774 f->line = e->line ? xmemdupz(e->line, e->len) : NULL;
777 struct moved_entry {
778 struct hashmap_entry ent;
779 const struct emitted_diff_symbol *es;
780 struct moved_entry *next_line;
784 * The struct ws_delta holds white space differences between moved lines, i.e.
785 * between '+' and '-' lines that have been detected to be a move.
786 * The string contains the difference in leading white spaces, before the
787 * rest of the line is compared using the white space config for move
788 * coloring. The current_longer indicates if the first string in the
789 * comparision is longer than the second.
791 struct ws_delta {
792 char *string;
793 unsigned int current_longer : 1;
795 #define WS_DELTA_INIT { NULL, 0 }
797 struct moved_block {
798 struct moved_entry *match;
799 struct ws_delta wsd;
802 static void moved_block_clear(struct moved_block *b)
804 FREE_AND_NULL(b->wsd.string);
805 b->match = NULL;
808 static int compute_ws_delta(const struct emitted_diff_symbol *a,
809 const struct emitted_diff_symbol *b,
810 struct ws_delta *out)
812 const struct emitted_diff_symbol *longer = a->len > b->len ? a : b;
813 const struct emitted_diff_symbol *shorter = a->len > b->len ? b : a;
814 int d = longer->len - shorter->len;
816 if (strncmp(longer->line + d, shorter->line, shorter->len))
817 return 0;
819 out->string = xmemdupz(longer->line, d);
820 out->current_longer = (a == longer);
822 return 1;
825 static int cmp_in_block_with_wsd(const struct diff_options *o,
826 const struct moved_entry *cur,
827 const struct moved_entry *match,
828 struct moved_block *pmb,
829 int n)
831 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
832 int al = cur->es->len, cl = l->len;
833 const char *a = cur->es->line,
834 *b = match->es->line,
835 *c = l->line;
837 int wslen;
840 * We need to check if 'cur' is equal to 'match'.
841 * As those are from the same (+/-) side, we do not need to adjust for
842 * indent changes. However these were found using fuzzy matching
843 * so we do have to check if they are equal.
845 if (strcmp(a, b))
846 return 1;
848 if (!pmb->wsd.string)
850 * The white space delta is not active? This can happen
851 * when we exit early in this function.
853 return 1;
856 * The indent changes of the block are known and stored in
857 * pmb->wsd; however we need to check if the indent changes of the
858 * current line are still the same as before.
860 * To do so we need to compare 'l' to 'cur', adjusting the
861 * one of them for the white spaces, depending which was longer.
864 wslen = strlen(pmb->wsd.string);
865 if (pmb->wsd.current_longer) {
866 c += wslen;
867 cl -= wslen;
868 } else {
869 a += wslen;
870 al -= wslen;
873 if (al != cl || memcmp(a, c, al))
874 return 1;
876 return 0;
879 static int moved_entry_cmp(const void *hashmap_cmp_fn_data,
880 const void *entry,
881 const void *entry_or_key,
882 const void *keydata)
884 const struct diff_options *diffopt = hashmap_cmp_fn_data;
885 const struct moved_entry *a = entry;
886 const struct moved_entry *b = entry_or_key;
887 unsigned flags = diffopt->color_moved_ws_handling
888 & XDF_WHITESPACE_FLAGS;
890 if (diffopt->color_moved_ws_handling &
891 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE)
893 * As there is not specific white space config given,
894 * we'd need to check for a new block, so ignore all
895 * white space. The setup of the white space
896 * configuration for the next block is done else where
898 flags |= XDF_IGNORE_WHITESPACE;
900 return !xdiff_compare_lines(a->es->line, a->es->len,
901 b->es->line, b->es->len,
902 flags);
905 static struct moved_entry *prepare_entry(struct diff_options *o,
906 int line_no)
908 struct moved_entry *ret = xmalloc(sizeof(*ret));
909 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[line_no];
910 unsigned flags = o->color_moved_ws_handling & XDF_WHITESPACE_FLAGS;
912 ret->ent.hash = xdiff_hash_string(l->line, l->len, flags);
913 ret->es = l;
914 ret->next_line = NULL;
916 return ret;
919 static void add_lines_to_move_detection(struct diff_options *o,
920 struct hashmap *add_lines,
921 struct hashmap *del_lines)
923 struct moved_entry *prev_line = NULL;
925 int n;
926 for (n = 0; n < o->emitted_symbols->nr; n++) {
927 struct hashmap *hm;
928 struct moved_entry *key;
930 switch (o->emitted_symbols->buf[n].s) {
931 case DIFF_SYMBOL_PLUS:
932 hm = add_lines;
933 break;
934 case DIFF_SYMBOL_MINUS:
935 hm = del_lines;
936 break;
937 default:
938 prev_line = NULL;
939 continue;
942 key = prepare_entry(o, n);
943 if (prev_line && prev_line->es->s == o->emitted_symbols->buf[n].s)
944 prev_line->next_line = key;
946 hashmap_add(hm, key);
947 prev_line = key;
951 static void pmb_advance_or_null(struct diff_options *o,
952 struct moved_entry *match,
953 struct hashmap *hm,
954 struct moved_block *pmb,
955 int pmb_nr)
957 int i;
958 for (i = 0; i < pmb_nr; i++) {
959 struct moved_entry *prev = pmb[i].match;
960 struct moved_entry *cur = (prev && prev->next_line) ?
961 prev->next_line : NULL;
962 if (cur && !hm->cmpfn(o, cur, match, NULL)) {
963 pmb[i].match = cur;
964 } else {
965 pmb[i].match = NULL;
970 static void pmb_advance_or_null_multi_match(struct diff_options *o,
971 struct moved_entry *match,
972 struct hashmap *hm,
973 struct moved_block *pmb,
974 int pmb_nr, int n)
976 int i;
977 char *got_match = xcalloc(1, pmb_nr);
979 for (; match; match = hashmap_get_next(hm, match)) {
980 for (i = 0; i < pmb_nr; i++) {
981 struct moved_entry *prev = pmb[i].match;
982 struct moved_entry *cur = (prev && prev->next_line) ?
983 prev->next_line : NULL;
984 if (!cur)
985 continue;
986 if (!cmp_in_block_with_wsd(o, cur, match, &pmb[i], n))
987 got_match[i] |= 1;
991 for (i = 0; i < pmb_nr; i++) {
992 if (got_match[i]) {
993 /* Advance to the next line */
994 pmb[i].match = pmb[i].match->next_line;
995 } else {
996 moved_block_clear(&pmb[i]);
1000 free(got_match);
1003 static int shrink_potential_moved_blocks(struct moved_block *pmb,
1004 int pmb_nr)
1006 int lp, rp;
1008 /* Shrink the set of potential block to the remaining running */
1009 for (lp = 0, rp = pmb_nr - 1; lp <= rp;) {
1010 while (lp < pmb_nr && pmb[lp].match)
1011 lp++;
1012 /* lp points at the first NULL now */
1014 while (rp > -1 && !pmb[rp].match)
1015 rp--;
1016 /* rp points at the last non-NULL */
1018 if (lp < pmb_nr && rp > -1 && lp < rp) {
1019 pmb[lp] = pmb[rp];
1020 pmb[rp].match = NULL;
1021 pmb[rp].wsd.string = NULL;
1022 rp--;
1023 lp++;
1027 /* Remember the number of running sets */
1028 return rp + 1;
1032 * If o->color_moved is COLOR_MOVED_PLAIN, this function does nothing.
1034 * Otherwise, if the last block has fewer alphanumeric characters than
1035 * COLOR_MOVED_MIN_ALNUM_COUNT, unset DIFF_SYMBOL_MOVED_LINE on all lines in
1036 * that block.
1038 * The last block consists of the (n - block_length)'th line up to but not
1039 * including the nth line.
1041 * NEEDSWORK: This uses the same heuristic as blame_entry_score() in blame.c.
1042 * Think of a way to unify them.
1044 static void adjust_last_block(struct diff_options *o, int n, int block_length)
1046 int i, alnum_count = 0;
1047 if (o->color_moved == COLOR_MOVED_PLAIN)
1048 return;
1049 for (i = 1; i < block_length + 1; i++) {
1050 const char *c = o->emitted_symbols->buf[n - i].line;
1051 for (; *c; c++) {
1052 if (!isalnum(*c))
1053 continue;
1054 alnum_count++;
1055 if (alnum_count >= COLOR_MOVED_MIN_ALNUM_COUNT)
1056 return;
1059 for (i = 1; i < block_length + 1; i++)
1060 o->emitted_symbols->buf[n - i].flags &= ~DIFF_SYMBOL_MOVED_LINE;
1063 /* Find blocks of moved code, delegate actual coloring decision to helper */
1064 static void mark_color_as_moved(struct diff_options *o,
1065 struct hashmap *add_lines,
1066 struct hashmap *del_lines)
1068 struct moved_block *pmb = NULL; /* potentially moved blocks */
1069 int pmb_nr = 0, pmb_alloc = 0;
1070 int n, flipped_block = 1, block_length = 0;
1073 for (n = 0; n < o->emitted_symbols->nr; n++) {
1074 struct hashmap *hm = NULL;
1075 struct moved_entry *key;
1076 struct moved_entry *match = NULL;
1077 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
1079 switch (l->s) {
1080 case DIFF_SYMBOL_PLUS:
1081 hm = del_lines;
1082 key = prepare_entry(o, n);
1083 match = hashmap_get(hm, key, NULL);
1084 free(key);
1085 break;
1086 case DIFF_SYMBOL_MINUS:
1087 hm = add_lines;
1088 key = prepare_entry(o, n);
1089 match = hashmap_get(hm, key, NULL);
1090 free(key);
1091 break;
1092 default:
1093 flipped_block = 1;
1096 if (!match) {
1097 int i;
1099 adjust_last_block(o, n, block_length);
1100 for(i = 0; i < pmb_nr; i++)
1101 moved_block_clear(&pmb[i]);
1102 pmb_nr = 0;
1103 block_length = 0;
1104 continue;
1107 l->flags |= DIFF_SYMBOL_MOVED_LINE;
1109 if (o->color_moved == COLOR_MOVED_PLAIN)
1110 continue;
1112 if (o->color_moved_ws_handling &
1113 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE)
1114 pmb_advance_or_null_multi_match(o, match, hm, pmb, pmb_nr, n);
1115 else
1116 pmb_advance_or_null(o, match, hm, pmb, pmb_nr);
1118 pmb_nr = shrink_potential_moved_blocks(pmb, pmb_nr);
1120 if (pmb_nr == 0) {
1122 * The current line is the start of a new block.
1123 * Setup the set of potential blocks.
1125 for (; match; match = hashmap_get_next(hm, match)) {
1126 ALLOC_GROW(pmb, pmb_nr + 1, pmb_alloc);
1127 if (o->color_moved_ws_handling &
1128 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE) {
1129 if (compute_ws_delta(l, match->es,
1130 &pmb[pmb_nr].wsd))
1131 pmb[pmb_nr++].match = match;
1132 } else {
1133 pmb[pmb_nr].wsd.string = NULL;
1134 pmb[pmb_nr++].match = match;
1138 flipped_block = (flipped_block + 1) % 2;
1140 adjust_last_block(o, n, block_length);
1141 block_length = 0;
1144 block_length++;
1146 if (flipped_block && o->color_moved != COLOR_MOVED_BLOCKS)
1147 l->flags |= DIFF_SYMBOL_MOVED_LINE_ALT;
1149 adjust_last_block(o, n, block_length);
1151 for(n = 0; n < pmb_nr; n++)
1152 moved_block_clear(&pmb[n]);
1153 free(pmb);
1156 #define DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK \
1157 (DIFF_SYMBOL_MOVED_LINE | DIFF_SYMBOL_MOVED_LINE_ALT)
1158 static void dim_moved_lines(struct diff_options *o)
1160 int n;
1161 for (n = 0; n < o->emitted_symbols->nr; n++) {
1162 struct emitted_diff_symbol *prev = (n != 0) ?
1163 &o->emitted_symbols->buf[n - 1] : NULL;
1164 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
1165 struct emitted_diff_symbol *next =
1166 (n < o->emitted_symbols->nr - 1) ?
1167 &o->emitted_symbols->buf[n + 1] : NULL;
1169 /* Not a plus or minus line? */
1170 if (l->s != DIFF_SYMBOL_PLUS && l->s != DIFF_SYMBOL_MINUS)
1171 continue;
1173 /* Not a moved line? */
1174 if (!(l->flags & DIFF_SYMBOL_MOVED_LINE))
1175 continue;
1178 * If prev or next are not a plus or minus line,
1179 * pretend they don't exist
1181 if (prev && prev->s != DIFF_SYMBOL_PLUS &&
1182 prev->s != DIFF_SYMBOL_MINUS)
1183 prev = NULL;
1184 if (next && next->s != DIFF_SYMBOL_PLUS &&
1185 next->s != DIFF_SYMBOL_MINUS)
1186 next = NULL;
1188 /* Inside a block? */
1189 if ((prev &&
1190 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
1191 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK)) &&
1192 (next &&
1193 (next->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
1194 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK))) {
1195 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1196 continue;
1199 /* Check if we are at an interesting bound: */
1200 if (prev && (prev->flags & DIFF_SYMBOL_MOVED_LINE) &&
1201 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1202 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1203 continue;
1204 if (next && (next->flags & DIFF_SYMBOL_MOVED_LINE) &&
1205 (next->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1206 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1207 continue;
1210 * The boundary to prev and next are not interesting,
1211 * so this line is not interesting as a whole
1213 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1217 static void emit_line_ws_markup(struct diff_options *o,
1218 const char *set_sign, const char *set,
1219 const char *reset,
1220 int sign_index, const char *line, int len,
1221 unsigned ws_rule, int blank_at_eof)
1223 const char *ws = NULL;
1224 int sign = o->output_indicators[sign_index];
1226 if (o->ws_error_highlight & ws_rule) {
1227 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
1228 if (!*ws)
1229 ws = NULL;
1232 if (!ws && !set_sign)
1233 emit_line_0(o, set, NULL, 0, reset, sign, line, len);
1234 else if (!ws) {
1235 emit_line_0(o, set_sign, set, !!set_sign, reset, sign, line, len);
1236 } else if (blank_at_eof)
1237 /* Blank line at EOF - paint '+' as well */
1238 emit_line_0(o, ws, NULL, 0, reset, sign, line, len);
1239 else {
1240 /* Emit just the prefix, then the rest. */
1241 emit_line_0(o, set_sign ? set_sign : set, NULL, !!set_sign, reset,
1242 sign, "", 0);
1243 ws_check_emit(line, len, ws_rule,
1244 o->file, set, reset, ws);
1248 static void emit_diff_symbol_from_struct(struct diff_options *o,
1249 struct emitted_diff_symbol *eds)
1251 static const char *nneof = " No newline at end of file\n";
1252 const char *context, *reset, *set, *set_sign, *meta, *fraginfo;
1253 struct strbuf sb = STRBUF_INIT;
1255 enum diff_symbol s = eds->s;
1256 const char *line = eds->line;
1257 int len = eds->len;
1258 unsigned flags = eds->flags;
1260 switch (s) {
1261 case DIFF_SYMBOL_NO_LF_EOF:
1262 context = diff_get_color_opt(o, DIFF_CONTEXT);
1263 reset = diff_get_color_opt(o, DIFF_RESET);
1264 putc('\n', o->file);
1265 emit_line_0(o, context, NULL, 0, reset, '\\',
1266 nneof, strlen(nneof));
1267 break;
1268 case DIFF_SYMBOL_SUBMODULE_HEADER:
1269 case DIFF_SYMBOL_SUBMODULE_ERROR:
1270 case DIFF_SYMBOL_SUBMODULE_PIPETHROUGH:
1271 case DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES:
1272 case DIFF_SYMBOL_SUMMARY:
1273 case DIFF_SYMBOL_STATS_LINE:
1274 case DIFF_SYMBOL_BINARY_DIFF_BODY:
1275 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
1276 emit_line(o, "", "", line, len);
1277 break;
1278 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
1279 case DIFF_SYMBOL_CONTEXT_MARKER:
1280 context = diff_get_color_opt(o, DIFF_CONTEXT);
1281 reset = diff_get_color_opt(o, DIFF_RESET);
1282 emit_line(o, context, reset, line, len);
1283 break;
1284 case DIFF_SYMBOL_SEPARATOR:
1285 fprintf(o->file, "%s%c",
1286 diff_line_prefix(o),
1287 o->line_termination);
1288 break;
1289 case DIFF_SYMBOL_CONTEXT:
1290 set = diff_get_color_opt(o, DIFF_CONTEXT);
1291 reset = diff_get_color_opt(o, DIFF_RESET);
1292 set_sign = NULL;
1293 if (o->flags.dual_color_diffed_diffs) {
1294 char c = !len ? 0 : line[0];
1296 if (c == '+')
1297 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1298 else if (c == '@')
1299 set = diff_get_color_opt(o, DIFF_FRAGINFO);
1300 else if (c == '-')
1301 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1303 emit_line_ws_markup(o, set_sign, set, reset,
1304 OUTPUT_INDICATOR_CONTEXT, line, len,
1305 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
1306 break;
1307 case DIFF_SYMBOL_PLUS:
1308 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1309 DIFF_SYMBOL_MOVED_LINE_ALT |
1310 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1311 case DIFF_SYMBOL_MOVED_LINE |
1312 DIFF_SYMBOL_MOVED_LINE_ALT |
1313 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1314 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT_DIM);
1315 break;
1316 case DIFF_SYMBOL_MOVED_LINE |
1317 DIFF_SYMBOL_MOVED_LINE_ALT:
1318 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT);
1319 break;
1320 case DIFF_SYMBOL_MOVED_LINE |
1321 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1322 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_DIM);
1323 break;
1324 case DIFF_SYMBOL_MOVED_LINE:
1325 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED);
1326 break;
1327 default:
1328 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1330 reset = diff_get_color_opt(o, DIFF_RESET);
1331 if (!o->flags.dual_color_diffed_diffs)
1332 set_sign = NULL;
1333 else {
1334 char c = !len ? 0 : line[0];
1336 set_sign = set;
1337 if (c == '-')
1338 set = diff_get_color_opt(o, DIFF_FILE_OLD_BOLD);
1339 else if (c == '@')
1340 set = diff_get_color_opt(o, DIFF_FRAGINFO);
1341 else if (c == '+')
1342 set = diff_get_color_opt(o, DIFF_FILE_NEW_BOLD);
1343 else
1344 set = diff_get_color_opt(o, DIFF_CONTEXT_BOLD);
1345 flags &= ~DIFF_SYMBOL_CONTENT_WS_MASK;
1347 emit_line_ws_markup(o, set_sign, set, reset,
1348 OUTPUT_INDICATOR_NEW, line, len,
1349 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
1350 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
1351 break;
1352 case DIFF_SYMBOL_MINUS:
1353 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1354 DIFF_SYMBOL_MOVED_LINE_ALT |
1355 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1356 case DIFF_SYMBOL_MOVED_LINE |
1357 DIFF_SYMBOL_MOVED_LINE_ALT |
1358 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1359 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT_DIM);
1360 break;
1361 case DIFF_SYMBOL_MOVED_LINE |
1362 DIFF_SYMBOL_MOVED_LINE_ALT:
1363 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT);
1364 break;
1365 case DIFF_SYMBOL_MOVED_LINE |
1366 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1367 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_DIM);
1368 break;
1369 case DIFF_SYMBOL_MOVED_LINE:
1370 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED);
1371 break;
1372 default:
1373 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1375 reset = diff_get_color_opt(o, DIFF_RESET);
1376 if (!o->flags.dual_color_diffed_diffs)
1377 set_sign = NULL;
1378 else {
1379 char c = !len ? 0 : line[0];
1381 set_sign = set;
1382 if (c == '+')
1383 set = diff_get_color_opt(o, DIFF_FILE_NEW_DIM);
1384 else if (c == '@')
1385 set = diff_get_color_opt(o, DIFF_FRAGINFO);
1386 else if (c == '-')
1387 set = diff_get_color_opt(o, DIFF_FILE_OLD_DIM);
1388 else
1389 set = diff_get_color_opt(o, DIFF_CONTEXT_DIM);
1391 emit_line_ws_markup(o, set_sign, set, reset,
1392 OUTPUT_INDICATOR_OLD, line, len,
1393 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
1394 break;
1395 case DIFF_SYMBOL_WORDS_PORCELAIN:
1396 context = diff_get_color_opt(o, DIFF_CONTEXT);
1397 reset = diff_get_color_opt(o, DIFF_RESET);
1398 emit_line(o, context, reset, line, len);
1399 fputs("~\n", o->file);
1400 break;
1401 case DIFF_SYMBOL_WORDS:
1402 context = diff_get_color_opt(o, DIFF_CONTEXT);
1403 reset = diff_get_color_opt(o, DIFF_RESET);
1405 * Skip the prefix character, if any. With
1406 * diff_suppress_blank_empty, there may be
1407 * none.
1409 if (line[0] != '\n') {
1410 line++;
1411 len--;
1413 emit_line(o, context, reset, line, len);
1414 break;
1415 case DIFF_SYMBOL_FILEPAIR_PLUS:
1416 meta = diff_get_color_opt(o, DIFF_METAINFO);
1417 reset = diff_get_color_opt(o, DIFF_RESET);
1418 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
1419 line, reset,
1420 strchr(line, ' ') ? "\t" : "");
1421 break;
1422 case DIFF_SYMBOL_FILEPAIR_MINUS:
1423 meta = diff_get_color_opt(o, DIFF_METAINFO);
1424 reset = diff_get_color_opt(o, DIFF_RESET);
1425 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
1426 line, reset,
1427 strchr(line, ' ') ? "\t" : "");
1428 break;
1429 case DIFF_SYMBOL_BINARY_FILES:
1430 case DIFF_SYMBOL_HEADER:
1431 fprintf(o->file, "%s", line);
1432 break;
1433 case DIFF_SYMBOL_BINARY_DIFF_HEADER:
1434 fprintf(o->file, "%sGIT binary patch\n", diff_line_prefix(o));
1435 break;
1436 case DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA:
1437 fprintf(o->file, "%sdelta %s\n", diff_line_prefix(o), line);
1438 break;
1439 case DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL:
1440 fprintf(o->file, "%sliteral %s\n", diff_line_prefix(o), line);
1441 break;
1442 case DIFF_SYMBOL_BINARY_DIFF_FOOTER:
1443 fputs(diff_line_prefix(o), o->file);
1444 fputc('\n', o->file);
1445 break;
1446 case DIFF_SYMBOL_REWRITE_DIFF:
1447 fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
1448 reset = diff_get_color_opt(o, DIFF_RESET);
1449 emit_line(o, fraginfo, reset, line, len);
1450 break;
1451 case DIFF_SYMBOL_SUBMODULE_ADD:
1452 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1453 reset = diff_get_color_opt(o, DIFF_RESET);
1454 emit_line(o, set, reset, line, len);
1455 break;
1456 case DIFF_SYMBOL_SUBMODULE_DEL:
1457 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1458 reset = diff_get_color_opt(o, DIFF_RESET);
1459 emit_line(o, set, reset, line, len);
1460 break;
1461 case DIFF_SYMBOL_SUBMODULE_UNTRACKED:
1462 fprintf(o->file, "%sSubmodule %s contains untracked content\n",
1463 diff_line_prefix(o), line);
1464 break;
1465 case DIFF_SYMBOL_SUBMODULE_MODIFIED:
1466 fprintf(o->file, "%sSubmodule %s contains modified content\n",
1467 diff_line_prefix(o), line);
1468 break;
1469 case DIFF_SYMBOL_STATS_SUMMARY_NO_FILES:
1470 emit_line(o, "", "", " 0 files changed\n",
1471 strlen(" 0 files changed\n"));
1472 break;
1473 case DIFF_SYMBOL_STATS_SUMMARY_ABBREV:
1474 emit_line(o, "", "", " ...\n", strlen(" ...\n"));
1475 break;
1476 case DIFF_SYMBOL_WORD_DIFF:
1477 fprintf(o->file, "%.*s", len, line);
1478 break;
1479 case DIFF_SYMBOL_STAT_SEP:
1480 fputs(o->stat_sep, o->file);
1481 break;
1482 default:
1483 BUG("unknown diff symbol");
1485 strbuf_release(&sb);
1488 static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
1489 const char *line, int len, unsigned flags)
1491 struct emitted_diff_symbol e = {line, len, flags, s};
1493 if (o->emitted_symbols)
1494 append_emitted_diff_symbol(o, &e);
1495 else
1496 emit_diff_symbol_from_struct(o, &e);
1499 void diff_emit_submodule_del(struct diff_options *o, const char *line)
1501 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_DEL, line, strlen(line), 0);
1504 void diff_emit_submodule_add(struct diff_options *o, const char *line)
1506 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ADD, line, strlen(line), 0);
1509 void diff_emit_submodule_untracked(struct diff_options *o, const char *path)
1511 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_UNTRACKED,
1512 path, strlen(path), 0);
1515 void diff_emit_submodule_modified(struct diff_options *o, const char *path)
1517 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_MODIFIED,
1518 path, strlen(path), 0);
1521 void diff_emit_submodule_header(struct diff_options *o, const char *header)
1523 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_HEADER,
1524 header, strlen(header), 0);
1527 void diff_emit_submodule_error(struct diff_options *o, const char *err)
1529 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ERROR, err, strlen(err), 0);
1532 void diff_emit_submodule_pipethrough(struct diff_options *o,
1533 const char *line, int len)
1535 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_PIPETHROUGH, line, len, 0);
1538 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
1540 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
1541 ecbdata->blank_at_eof_in_preimage &&
1542 ecbdata->blank_at_eof_in_postimage &&
1543 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
1544 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
1545 return 0;
1546 return ws_blank_line(line, len, ecbdata->ws_rule);
1549 static void emit_add_line(const char *reset,
1550 struct emit_callback *ecbdata,
1551 const char *line, int len)
1553 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
1554 if (new_blank_line_at_eof(ecbdata, line, len))
1555 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
1557 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
1560 static void emit_del_line(const char *reset,
1561 struct emit_callback *ecbdata,
1562 const char *line, int len)
1564 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
1565 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
1568 static void emit_context_line(const char *reset,
1569 struct emit_callback *ecbdata,
1570 const char *line, int len)
1572 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
1573 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
1576 static void emit_hunk_header(struct emit_callback *ecbdata,
1577 const char *line, int len)
1579 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1580 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
1581 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
1582 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1583 const char *reverse = ecbdata->color_diff ? GIT_COLOR_REVERSE : "";
1584 static const char atat[2] = { '@', '@' };
1585 const char *cp, *ep;
1586 struct strbuf msgbuf = STRBUF_INIT;
1587 int org_len = len;
1588 int i = 1;
1591 * As a hunk header must begin with "@@ -<old>, +<new> @@",
1592 * it always is at least 10 bytes long.
1594 if (len < 10 ||
1595 memcmp(line, atat, 2) ||
1596 !(ep = memmem(line + 2, len - 2, atat, 2))) {
1597 emit_diff_symbol(ecbdata->opt,
1598 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
1599 return;
1601 ep += 2; /* skip over @@ */
1603 /* The hunk header in fraginfo color */
1604 if (ecbdata->opt->flags.dual_color_diffed_diffs)
1605 strbuf_addstr(&msgbuf, reverse);
1606 strbuf_addstr(&msgbuf, frag);
1607 strbuf_add(&msgbuf, line, ep - line);
1608 strbuf_addstr(&msgbuf, reset);
1611 * trailing "\r\n"
1613 for ( ; i < 3; i++)
1614 if (line[len - i] == '\r' || line[len - i] == '\n')
1615 len--;
1617 /* blank before the func header */
1618 for (cp = ep; ep - line < len; ep++)
1619 if (*ep != ' ' && *ep != '\t')
1620 break;
1621 if (ep != cp) {
1622 strbuf_addstr(&msgbuf, context);
1623 strbuf_add(&msgbuf, cp, ep - cp);
1624 strbuf_addstr(&msgbuf, reset);
1627 if (ep < line + len) {
1628 strbuf_addstr(&msgbuf, func);
1629 strbuf_add(&msgbuf, ep, line + len - ep);
1630 strbuf_addstr(&msgbuf, reset);
1633 strbuf_add(&msgbuf, line + len, org_len - len);
1634 strbuf_complete_line(&msgbuf);
1635 emit_diff_symbol(ecbdata->opt,
1636 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
1637 strbuf_release(&msgbuf);
1640 static struct diff_tempfile *claim_diff_tempfile(void) {
1641 int i;
1642 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
1643 if (!diff_temp[i].name)
1644 return diff_temp + i;
1645 BUG("diff is failing to clean up its tempfiles");
1648 static void remove_tempfile(void)
1650 int i;
1651 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
1652 if (is_tempfile_active(diff_temp[i].tempfile))
1653 delete_tempfile(&diff_temp[i].tempfile);
1654 diff_temp[i].name = NULL;
1658 static void add_line_count(struct strbuf *out, int count)
1660 switch (count) {
1661 case 0:
1662 strbuf_addstr(out, "0,0");
1663 break;
1664 case 1:
1665 strbuf_addstr(out, "1");
1666 break;
1667 default:
1668 strbuf_addf(out, "1,%d", count);
1669 break;
1673 static void emit_rewrite_lines(struct emit_callback *ecb,
1674 int prefix, const char *data, int size)
1676 const char *endp = NULL;
1677 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
1679 while (0 < size) {
1680 int len;
1682 endp = memchr(data, '\n', size);
1683 len = endp ? (endp - data + 1) : size;
1684 if (prefix != '+') {
1685 ecb->lno_in_preimage++;
1686 emit_del_line(reset, ecb, data, len);
1687 } else {
1688 ecb->lno_in_postimage++;
1689 emit_add_line(reset, ecb, data, len);
1691 size -= len;
1692 data += len;
1694 if (!endp)
1695 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
1698 static void emit_rewrite_diff(const char *name_a,
1699 const char *name_b,
1700 struct diff_filespec *one,
1701 struct diff_filespec *two,
1702 struct userdiff_driver *textconv_one,
1703 struct userdiff_driver *textconv_two,
1704 struct diff_options *o)
1706 int lc_a, lc_b;
1707 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
1708 const char *a_prefix, *b_prefix;
1709 char *data_one, *data_two;
1710 size_t size_one, size_two;
1711 struct emit_callback ecbdata;
1712 struct strbuf out = STRBUF_INIT;
1714 if (diff_mnemonic_prefix && o->flags.reverse_diff) {
1715 a_prefix = o->b_prefix;
1716 b_prefix = o->a_prefix;
1717 } else {
1718 a_prefix = o->a_prefix;
1719 b_prefix = o->b_prefix;
1722 name_a += (*name_a == '/');
1723 name_b += (*name_b == '/');
1725 strbuf_reset(&a_name);
1726 strbuf_reset(&b_name);
1727 quote_two_c_style(&a_name, a_prefix, name_a, 0);
1728 quote_two_c_style(&b_name, b_prefix, name_b, 0);
1730 size_one = fill_textconv(o->repo, textconv_one, one, &data_one);
1731 size_two = fill_textconv(o->repo, textconv_two, two, &data_two);
1733 memset(&ecbdata, 0, sizeof(ecbdata));
1734 ecbdata.color_diff = want_color(o->use_color);
1735 ecbdata.ws_rule = whitespace_rule(o->repo->index, name_b);
1736 ecbdata.opt = o;
1737 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
1738 mmfile_t mf1, mf2;
1739 mf1.ptr = (char *)data_one;
1740 mf2.ptr = (char *)data_two;
1741 mf1.size = size_one;
1742 mf2.size = size_two;
1743 check_blank_at_eof(&mf1, &mf2, &ecbdata);
1745 ecbdata.lno_in_preimage = 1;
1746 ecbdata.lno_in_postimage = 1;
1748 lc_a = count_lines(data_one, size_one);
1749 lc_b = count_lines(data_two, size_two);
1751 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1752 a_name.buf, a_name.len, 0);
1753 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1754 b_name.buf, b_name.len, 0);
1756 strbuf_addstr(&out, "@@ -");
1757 if (!o->irreversible_delete)
1758 add_line_count(&out, lc_a);
1759 else
1760 strbuf_addstr(&out, "?,?");
1761 strbuf_addstr(&out, " +");
1762 add_line_count(&out, lc_b);
1763 strbuf_addstr(&out, " @@\n");
1764 emit_diff_symbol(o, DIFF_SYMBOL_REWRITE_DIFF, out.buf, out.len, 0);
1765 strbuf_release(&out);
1767 if (lc_a && !o->irreversible_delete)
1768 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
1769 if (lc_b)
1770 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
1771 if (textconv_one)
1772 free((char *)data_one);
1773 if (textconv_two)
1774 free((char *)data_two);
1777 struct diff_words_buffer {
1778 mmfile_t text;
1779 unsigned long alloc;
1780 struct diff_words_orig {
1781 const char *begin, *end;
1782 } *orig;
1783 int orig_nr, orig_alloc;
1786 static void diff_words_append(char *line, unsigned long len,
1787 struct diff_words_buffer *buffer)
1789 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
1790 line++;
1791 len--;
1792 memcpy(buffer->text.ptr + buffer->text.size, line, len);
1793 buffer->text.size += len;
1794 buffer->text.ptr[buffer->text.size] = '\0';
1797 struct diff_words_style_elem {
1798 const char *prefix;
1799 const char *suffix;
1800 const char *color; /* NULL; filled in by the setup code if
1801 * color is enabled */
1804 struct diff_words_style {
1805 enum diff_words_type type;
1806 struct diff_words_style_elem new_word, old_word, ctx;
1807 const char *newline;
1810 static struct diff_words_style diff_words_styles[] = {
1811 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
1812 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
1813 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
1816 struct diff_words_data {
1817 struct diff_words_buffer minus, plus;
1818 const char *current_plus;
1819 int last_minus;
1820 struct diff_options *opt;
1821 regex_t *word_regex;
1822 enum diff_words_type type;
1823 struct diff_words_style *style;
1826 static int fn_out_diff_words_write_helper(struct diff_options *o,
1827 struct diff_words_style_elem *st_el,
1828 const char *newline,
1829 size_t count, const char *buf)
1831 int print = 0;
1832 struct strbuf sb = STRBUF_INIT;
1834 while (count) {
1835 char *p = memchr(buf, '\n', count);
1836 if (print)
1837 strbuf_addstr(&sb, diff_line_prefix(o));
1839 if (p != buf) {
1840 const char *reset = st_el->color && *st_el->color ?
1841 GIT_COLOR_RESET : NULL;
1842 if (st_el->color && *st_el->color)
1843 strbuf_addstr(&sb, st_el->color);
1844 strbuf_addstr(&sb, st_el->prefix);
1845 strbuf_add(&sb, buf, p ? p - buf : count);
1846 strbuf_addstr(&sb, st_el->suffix);
1847 if (reset)
1848 strbuf_addstr(&sb, reset);
1850 if (!p)
1851 goto out;
1853 strbuf_addstr(&sb, newline);
1854 count -= p + 1 - buf;
1855 buf = p + 1;
1856 print = 1;
1857 if (count) {
1858 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1859 sb.buf, sb.len, 0);
1860 strbuf_reset(&sb);
1864 out:
1865 if (sb.len)
1866 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1867 sb.buf, sb.len, 0);
1868 strbuf_release(&sb);
1869 return 0;
1873 * '--color-words' algorithm can be described as:
1875 * 1. collect the minus/plus lines of a diff hunk, divided into
1876 * minus-lines and plus-lines;
1878 * 2. break both minus-lines and plus-lines into words and
1879 * place them into two mmfile_t with one word for each line;
1881 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1883 * And for the common parts of the both file, we output the plus side text.
1884 * diff_words->current_plus is used to trace the current position of the plus file
1885 * which printed. diff_words->last_minus is used to trace the last minus word
1886 * printed.
1888 * For '--graph' to work with '--color-words', we need to output the graph prefix
1889 * on each line of color words output. Generally, there are two conditions on
1890 * which we should output the prefix.
1892 * 1. diff_words->last_minus == 0 &&
1893 * diff_words->current_plus == diff_words->plus.text.ptr
1895 * that is: the plus text must start as a new line, and if there is no minus
1896 * word printed, a graph prefix must be printed.
1898 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1899 * *(diff_words->current_plus - 1) == '\n'
1901 * that is: a graph prefix must be printed following a '\n'
1903 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1905 if ((diff_words->last_minus == 0 &&
1906 diff_words->current_plus == diff_words->plus.text.ptr) ||
1907 (diff_words->current_plus > diff_words->plus.text.ptr &&
1908 *(diff_words->current_plus - 1) == '\n')) {
1909 return 1;
1910 } else {
1911 return 0;
1915 static void fn_out_diff_words_aux(void *priv,
1916 long minus_first, long minus_len,
1917 long plus_first, long plus_len,
1918 const char *func, long funclen)
1920 struct diff_words_data *diff_words = priv;
1921 struct diff_words_style *style = diff_words->style;
1922 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1923 struct diff_options *opt = diff_words->opt;
1924 const char *line_prefix;
1926 assert(opt);
1927 line_prefix = diff_line_prefix(opt);
1929 /* POSIX requires that first be decremented by one if len == 0... */
1930 if (minus_len) {
1931 minus_begin = diff_words->minus.orig[minus_first].begin;
1932 minus_end =
1933 diff_words->minus.orig[minus_first + minus_len - 1].end;
1934 } else
1935 minus_begin = minus_end =
1936 diff_words->minus.orig[minus_first].end;
1938 if (plus_len) {
1939 plus_begin = diff_words->plus.orig[plus_first].begin;
1940 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1941 } else
1942 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1944 if (color_words_output_graph_prefix(diff_words)) {
1945 fputs(line_prefix, diff_words->opt->file);
1947 if (diff_words->current_plus != plus_begin) {
1948 fn_out_diff_words_write_helper(diff_words->opt,
1949 &style->ctx, style->newline,
1950 plus_begin - diff_words->current_plus,
1951 diff_words->current_plus);
1953 if (minus_begin != minus_end) {
1954 fn_out_diff_words_write_helper(diff_words->opt,
1955 &style->old_word, style->newline,
1956 minus_end - minus_begin, minus_begin);
1958 if (plus_begin != plus_end) {
1959 fn_out_diff_words_write_helper(diff_words->opt,
1960 &style->new_word, style->newline,
1961 plus_end - plus_begin, plus_begin);
1964 diff_words->current_plus = plus_end;
1965 diff_words->last_minus = minus_first;
1968 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1969 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1970 int *begin, int *end)
1972 if (word_regex && *begin < buffer->size) {
1973 regmatch_t match[1];
1974 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1975 buffer->size - *begin, 1, match, 0)) {
1976 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1977 '\n', match[0].rm_eo - match[0].rm_so);
1978 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1979 *begin += match[0].rm_so;
1980 return *begin >= *end;
1982 return -1;
1985 /* find the next word */
1986 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1987 (*begin)++;
1988 if (*begin >= buffer->size)
1989 return -1;
1991 /* find the end of the word */
1992 *end = *begin + 1;
1993 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1994 (*end)++;
1996 return 0;
2000 * This function splits the words in buffer->text, stores the list with
2001 * newline separator into out, and saves the offsets of the original words
2002 * in buffer->orig.
2004 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
2005 regex_t *word_regex)
2007 int i, j;
2008 long alloc = 0;
2010 out->size = 0;
2011 out->ptr = NULL;
2013 /* fake an empty "0th" word */
2014 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
2015 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
2016 buffer->orig_nr = 1;
2018 for (i = 0; i < buffer->text.size; i++) {
2019 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
2020 return;
2022 /* store original boundaries */
2023 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
2024 buffer->orig_alloc);
2025 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
2026 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
2027 buffer->orig_nr++;
2029 /* store one word */
2030 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
2031 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
2032 out->ptr[out->size + j - i] = '\n';
2033 out->size += j - i + 1;
2035 i = j - 1;
2039 /* this executes the word diff on the accumulated buffers */
2040 static void diff_words_show(struct diff_words_data *diff_words)
2042 xpparam_t xpp;
2043 xdemitconf_t xecfg;
2044 mmfile_t minus, plus;
2045 struct diff_words_style *style = diff_words->style;
2047 struct diff_options *opt = diff_words->opt;
2048 const char *line_prefix;
2050 assert(opt);
2051 line_prefix = diff_line_prefix(opt);
2053 /* special case: only removal */
2054 if (!diff_words->plus.text.size) {
2055 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
2056 line_prefix, strlen(line_prefix), 0);
2057 fn_out_diff_words_write_helper(diff_words->opt,
2058 &style->old_word, style->newline,
2059 diff_words->minus.text.size,
2060 diff_words->minus.text.ptr);
2061 diff_words->minus.text.size = 0;
2062 return;
2065 diff_words->current_plus = diff_words->plus.text.ptr;
2066 diff_words->last_minus = 0;
2068 memset(&xpp, 0, sizeof(xpp));
2069 memset(&xecfg, 0, sizeof(xecfg));
2070 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
2071 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
2072 xpp.flags = 0;
2073 /* as only the hunk header will be parsed, we need a 0-context */
2074 xecfg.ctxlen = 0;
2075 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, NULL,
2076 diff_words, &xpp, &xecfg))
2077 die("unable to generate word diff");
2078 free(minus.ptr);
2079 free(plus.ptr);
2080 if (diff_words->current_plus != diff_words->plus.text.ptr +
2081 diff_words->plus.text.size) {
2082 if (color_words_output_graph_prefix(diff_words))
2083 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
2084 line_prefix, strlen(line_prefix), 0);
2085 fn_out_diff_words_write_helper(diff_words->opt,
2086 &style->ctx, style->newline,
2087 diff_words->plus.text.ptr + diff_words->plus.text.size
2088 - diff_words->current_plus, diff_words->current_plus);
2090 diff_words->minus.text.size = diff_words->plus.text.size = 0;
2093 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
2094 static void diff_words_flush(struct emit_callback *ecbdata)
2096 struct diff_options *wo = ecbdata->diff_words->opt;
2098 if (ecbdata->diff_words->minus.text.size ||
2099 ecbdata->diff_words->plus.text.size)
2100 diff_words_show(ecbdata->diff_words);
2102 if (wo->emitted_symbols) {
2103 struct diff_options *o = ecbdata->opt;
2104 struct emitted_diff_symbols *wol = wo->emitted_symbols;
2105 int i;
2108 * NEEDSWORK:
2109 * Instead of appending each, concat all words to a line?
2111 for (i = 0; i < wol->nr; i++)
2112 append_emitted_diff_symbol(o, &wol->buf[i]);
2114 for (i = 0; i < wol->nr; i++)
2115 free((void *)wol->buf[i].line);
2117 wol->nr = 0;
2121 static void diff_filespec_load_driver(struct diff_filespec *one,
2122 struct index_state *istate)
2124 /* Use already-loaded driver */
2125 if (one->driver)
2126 return;
2128 if (S_ISREG(one->mode))
2129 one->driver = userdiff_find_by_path(istate, one->path);
2131 /* Fallback to default settings */
2132 if (!one->driver)
2133 one->driver = userdiff_find_by_name("default");
2136 static const char *userdiff_word_regex(struct diff_filespec *one,
2137 struct index_state *istate)
2139 diff_filespec_load_driver(one, istate);
2140 return one->driver->word_regex;
2143 static void init_diff_words_data(struct emit_callback *ecbdata,
2144 struct diff_options *orig_opts,
2145 struct diff_filespec *one,
2146 struct diff_filespec *two)
2148 int i;
2149 struct diff_options *o = xmalloc(sizeof(struct diff_options));
2150 memcpy(o, orig_opts, sizeof(struct diff_options));
2152 ecbdata->diff_words =
2153 xcalloc(1, sizeof(struct diff_words_data));
2154 ecbdata->diff_words->type = o->word_diff;
2155 ecbdata->diff_words->opt = o;
2157 if (orig_opts->emitted_symbols)
2158 o->emitted_symbols =
2159 xcalloc(1, sizeof(struct emitted_diff_symbols));
2161 if (!o->word_regex)
2162 o->word_regex = userdiff_word_regex(one, o->repo->index);
2163 if (!o->word_regex)
2164 o->word_regex = userdiff_word_regex(two, o->repo->index);
2165 if (!o->word_regex)
2166 o->word_regex = diff_word_regex_cfg;
2167 if (o->word_regex) {
2168 ecbdata->diff_words->word_regex = (regex_t *)
2169 xmalloc(sizeof(regex_t));
2170 if (regcomp(ecbdata->diff_words->word_regex,
2171 o->word_regex,
2172 REG_EXTENDED | REG_NEWLINE))
2173 die("invalid regular expression: %s",
2174 o->word_regex);
2176 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
2177 if (o->word_diff == diff_words_styles[i].type) {
2178 ecbdata->diff_words->style =
2179 &diff_words_styles[i];
2180 break;
2183 if (want_color(o->use_color)) {
2184 struct diff_words_style *st = ecbdata->diff_words->style;
2185 st->old_word.color = diff_get_color_opt(o, DIFF_FILE_OLD);
2186 st->new_word.color = diff_get_color_opt(o, DIFF_FILE_NEW);
2187 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
2191 static void free_diff_words_data(struct emit_callback *ecbdata)
2193 if (ecbdata->diff_words) {
2194 diff_words_flush(ecbdata);
2195 free (ecbdata->diff_words->opt->emitted_symbols);
2196 free (ecbdata->diff_words->opt);
2197 free (ecbdata->diff_words->minus.text.ptr);
2198 free (ecbdata->diff_words->minus.orig);
2199 free (ecbdata->diff_words->plus.text.ptr);
2200 free (ecbdata->diff_words->plus.orig);
2201 if (ecbdata->diff_words->word_regex) {
2202 regfree(ecbdata->diff_words->word_regex);
2203 free(ecbdata->diff_words->word_regex);
2205 FREE_AND_NULL(ecbdata->diff_words);
2209 const char *diff_get_color(int diff_use_color, enum color_diff ix)
2211 if (want_color(diff_use_color))
2212 return diff_colors[ix];
2213 return "";
2216 const char *diff_line_prefix(struct diff_options *opt)
2218 struct strbuf *msgbuf;
2219 if (!opt->output_prefix)
2220 return "";
2222 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
2223 return msgbuf->buf;
2226 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
2228 const char *cp;
2229 unsigned long allot;
2230 size_t l = len;
2232 cp = line;
2233 allot = l;
2234 while (0 < l) {
2235 (void) utf8_width(&cp, &l);
2236 if (!cp)
2237 break; /* truncated in the middle? */
2239 return allot - l;
2242 static void find_lno(const char *line, struct emit_callback *ecbdata)
2244 const char *p;
2245 ecbdata->lno_in_preimage = 0;
2246 ecbdata->lno_in_postimage = 0;
2247 p = strchr(line, '-');
2248 if (!p)
2249 return; /* cannot happen */
2250 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
2251 p = strchr(p, '+');
2252 if (!p)
2253 return; /* cannot happen */
2254 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
2257 static void fn_out_consume(void *priv, char *line, unsigned long len)
2259 struct emit_callback *ecbdata = priv;
2260 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
2261 struct diff_options *o = ecbdata->opt;
2263 o->found_changes = 1;
2265 if (ecbdata->header) {
2266 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2267 ecbdata->header->buf, ecbdata->header->len, 0);
2268 strbuf_reset(ecbdata->header);
2269 ecbdata->header = NULL;
2272 if (ecbdata->label_path[0]) {
2273 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
2274 ecbdata->label_path[0],
2275 strlen(ecbdata->label_path[0]), 0);
2276 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
2277 ecbdata->label_path[1],
2278 strlen(ecbdata->label_path[1]), 0);
2279 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
2282 if (diff_suppress_blank_empty
2283 && len == 2 && line[0] == ' ' && line[1] == '\n') {
2284 line[0] = '\n';
2285 len = 1;
2288 if (line[0] == '@') {
2289 if (ecbdata->diff_words)
2290 diff_words_flush(ecbdata);
2291 len = sane_truncate_line(ecbdata, line, len);
2292 find_lno(line, ecbdata);
2293 emit_hunk_header(ecbdata, line, len);
2294 return;
2297 if (ecbdata->diff_words) {
2298 enum diff_symbol s =
2299 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
2300 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
2301 if (line[0] == '-') {
2302 diff_words_append(line, len,
2303 &ecbdata->diff_words->minus);
2304 return;
2305 } else if (line[0] == '+') {
2306 diff_words_append(line, len,
2307 &ecbdata->diff_words->plus);
2308 return;
2309 } else if (starts_with(line, "\\ ")) {
2311 * Eat the "no newline at eof" marker as if we
2312 * saw a "+" or "-" line with nothing on it,
2313 * and return without diff_words_flush() to
2314 * defer processing. If this is the end of
2315 * preimage, more "+" lines may come after it.
2317 return;
2319 diff_words_flush(ecbdata);
2320 emit_diff_symbol(o, s, line, len, 0);
2321 return;
2324 switch (line[0]) {
2325 case '+':
2326 ecbdata->lno_in_postimage++;
2327 emit_add_line(reset, ecbdata, line + 1, len - 1);
2328 break;
2329 case '-':
2330 ecbdata->lno_in_preimage++;
2331 emit_del_line(reset, ecbdata, line + 1, len - 1);
2332 break;
2333 case ' ':
2334 ecbdata->lno_in_postimage++;
2335 ecbdata->lno_in_preimage++;
2336 emit_context_line(reset, ecbdata, line + 1, len - 1);
2337 break;
2338 default:
2339 /* incomplete line at the end */
2340 ecbdata->lno_in_preimage++;
2341 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
2342 line, len, 0);
2343 break;
2347 static void pprint_rename(struct strbuf *name, const char *a, const char *b)
2349 const char *old_name = a;
2350 const char *new_name = b;
2351 int pfx_length, sfx_length;
2352 int pfx_adjust_for_slash;
2353 int len_a = strlen(a);
2354 int len_b = strlen(b);
2355 int a_midlen, b_midlen;
2356 int qlen_a = quote_c_style(a, NULL, NULL, 0);
2357 int qlen_b = quote_c_style(b, NULL, NULL, 0);
2359 if (qlen_a || qlen_b) {
2360 quote_c_style(a, name, NULL, 0);
2361 strbuf_addstr(name, " => ");
2362 quote_c_style(b, name, NULL, 0);
2363 return;
2366 /* Find common prefix */
2367 pfx_length = 0;
2368 while (*old_name && *new_name && *old_name == *new_name) {
2369 if (*old_name == '/')
2370 pfx_length = old_name - a + 1;
2371 old_name++;
2372 new_name++;
2375 /* Find common suffix */
2376 old_name = a + len_a;
2377 new_name = b + len_b;
2378 sfx_length = 0;
2380 * If there is a common prefix, it must end in a slash. In
2381 * that case we let this loop run 1 into the prefix to see the
2382 * same slash.
2384 * If there is no common prefix, we cannot do this as it would
2385 * underrun the input strings.
2387 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
2388 while (a + pfx_length - pfx_adjust_for_slash <= old_name &&
2389 b + pfx_length - pfx_adjust_for_slash <= new_name &&
2390 *old_name == *new_name) {
2391 if (*old_name == '/')
2392 sfx_length = len_a - (old_name - a);
2393 old_name--;
2394 new_name--;
2398 * pfx{mid-a => mid-b}sfx
2399 * {pfx-a => pfx-b}sfx
2400 * pfx{sfx-a => sfx-b}
2401 * name-a => name-b
2403 a_midlen = len_a - pfx_length - sfx_length;
2404 b_midlen = len_b - pfx_length - sfx_length;
2405 if (a_midlen < 0)
2406 a_midlen = 0;
2407 if (b_midlen < 0)
2408 b_midlen = 0;
2410 strbuf_grow(name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
2411 if (pfx_length + sfx_length) {
2412 strbuf_add(name, a, pfx_length);
2413 strbuf_addch(name, '{');
2415 strbuf_add(name, a + pfx_length, a_midlen);
2416 strbuf_addstr(name, " => ");
2417 strbuf_add(name, b + pfx_length, b_midlen);
2418 if (pfx_length + sfx_length) {
2419 strbuf_addch(name, '}');
2420 strbuf_add(name, a + len_a - sfx_length, sfx_length);
2424 struct diffstat_t {
2425 int nr;
2426 int alloc;
2427 struct diffstat_file {
2428 char *from_name;
2429 char *name;
2430 char *print_name;
2431 const char *comments;
2432 unsigned is_unmerged:1;
2433 unsigned is_binary:1;
2434 unsigned is_renamed:1;
2435 unsigned is_interesting:1;
2436 uintmax_t added, deleted;
2437 } **files;
2440 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
2441 const char *name_a,
2442 const char *name_b)
2444 struct diffstat_file *x;
2445 x = xcalloc(1, sizeof(*x));
2446 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
2447 diffstat->files[diffstat->nr++] = x;
2448 if (name_b) {
2449 x->from_name = xstrdup(name_a);
2450 x->name = xstrdup(name_b);
2451 x->is_renamed = 1;
2453 else {
2454 x->from_name = NULL;
2455 x->name = xstrdup(name_a);
2457 return x;
2460 static void diffstat_consume(void *priv, char *line, unsigned long len)
2462 struct diffstat_t *diffstat = priv;
2463 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
2465 if (line[0] == '+')
2466 x->added++;
2467 else if (line[0] == '-')
2468 x->deleted++;
2471 const char mime_boundary_leader[] = "------------";
2473 static int scale_linear(int it, int width, int max_change)
2475 if (!it)
2476 return 0;
2478 * make sure that at least one '-' or '+' is printed if
2479 * there is any change to this path. The easiest way is to
2480 * scale linearly as if the alloted width is one column shorter
2481 * than it is, and then add 1 to the result.
2483 return 1 + (it * (width - 1) / max_change);
2486 static void show_graph(struct strbuf *out, char ch, int cnt,
2487 const char *set, const char *reset)
2489 if (cnt <= 0)
2490 return;
2491 strbuf_addstr(out, set);
2492 strbuf_addchars(out, ch, cnt);
2493 strbuf_addstr(out, reset);
2496 static void fill_print_name(struct diffstat_file *file)
2498 struct strbuf pname = STRBUF_INIT;
2500 if (file->print_name)
2501 return;
2503 if (file->is_renamed)
2504 pprint_rename(&pname, file->from_name, file->name);
2505 else
2506 quote_c_style(file->name, &pname, NULL, 0);
2508 if (file->comments)
2509 strbuf_addf(&pname, " (%s)", file->comments);
2511 file->print_name = strbuf_detach(&pname, NULL);
2514 static void print_stat_summary_inserts_deletes(struct diff_options *options,
2515 int files, int insertions, int deletions)
2517 struct strbuf sb = STRBUF_INIT;
2519 if (!files) {
2520 assert(insertions == 0 && deletions == 0);
2521 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
2522 NULL, 0, 0);
2523 return;
2526 strbuf_addf(&sb,
2527 (files == 1) ? " %d file changed" : " %d files changed",
2528 files);
2531 * For binary diff, the caller may want to print "x files
2532 * changed" with insertions == 0 && deletions == 0.
2534 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
2535 * is probably less confusing (i.e skip over "2 files changed
2536 * but nothing about added/removed lines? Is this a bug in Git?").
2538 if (insertions || deletions == 0) {
2539 strbuf_addf(&sb,
2540 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
2541 insertions);
2544 if (deletions || insertions == 0) {
2545 strbuf_addf(&sb,
2546 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
2547 deletions);
2549 strbuf_addch(&sb, '\n');
2550 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
2551 sb.buf, sb.len, 0);
2552 strbuf_release(&sb);
2555 void print_stat_summary(FILE *fp, int files,
2556 int insertions, int deletions)
2558 struct diff_options o;
2559 memset(&o, 0, sizeof(o));
2560 o.file = fp;
2562 print_stat_summary_inserts_deletes(&o, files, insertions, deletions);
2565 static void show_stats(struct diffstat_t *data, struct diff_options *options)
2567 int i, len, add, del, adds = 0, dels = 0;
2568 uintmax_t max_change = 0, max_len = 0;
2569 int total_files = data->nr, count;
2570 int width, name_width, graph_width, number_width = 0, bin_width = 0;
2571 const char *reset, *add_c, *del_c;
2572 int extra_shown = 0;
2573 const char *line_prefix = diff_line_prefix(options);
2574 struct strbuf out = STRBUF_INIT;
2576 if (data->nr == 0)
2577 return;
2579 count = options->stat_count ? options->stat_count : data->nr;
2581 reset = diff_get_color_opt(options, DIFF_RESET);
2582 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
2583 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
2586 * Find the longest filename and max number of changes
2588 for (i = 0; (i < count) && (i < data->nr); i++) {
2589 struct diffstat_file *file = data->files[i];
2590 uintmax_t change = file->added + file->deleted;
2592 if (!file->is_interesting && (change == 0)) {
2593 count++; /* not shown == room for one more */
2594 continue;
2596 fill_print_name(file);
2597 len = strlen(file->print_name);
2598 if (max_len < len)
2599 max_len = len;
2601 if (file->is_unmerged) {
2602 /* "Unmerged" is 8 characters */
2603 bin_width = bin_width < 8 ? 8 : bin_width;
2604 continue;
2606 if (file->is_binary) {
2607 /* "Bin XXX -> YYY bytes" */
2608 int w = 14 + decimal_width(file->added)
2609 + decimal_width(file->deleted);
2610 bin_width = bin_width < w ? w : bin_width;
2611 /* Display change counts aligned with "Bin" */
2612 number_width = 3;
2613 continue;
2616 if (max_change < change)
2617 max_change = change;
2619 count = i; /* where we can stop scanning in data->files[] */
2622 * We have width = stat_width or term_columns() columns total.
2623 * We want a maximum of min(max_len, stat_name_width) for the name part.
2624 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
2625 * We also need 1 for " " and 4 + decimal_width(max_change)
2626 * for " | NNNN " and one the empty column at the end, altogether
2627 * 6 + decimal_width(max_change).
2629 * If there's not enough space, we will use the smaller of
2630 * stat_name_width (if set) and 5/8*width for the filename,
2631 * and the rest for constant elements + graph part, but no more
2632 * than stat_graph_width for the graph part.
2633 * (5/8 gives 50 for filename and 30 for the constant parts + graph
2634 * for the standard terminal size).
2636 * In other words: stat_width limits the maximum width, and
2637 * stat_name_width fixes the maximum width of the filename,
2638 * and is also used to divide available columns if there
2639 * aren't enough.
2641 * Binary files are displayed with "Bin XXX -> YYY bytes"
2642 * instead of the change count and graph. This part is treated
2643 * similarly to the graph part, except that it is not
2644 * "scaled". If total width is too small to accommodate the
2645 * guaranteed minimum width of the filename part and the
2646 * separators and this message, this message will "overflow"
2647 * making the line longer than the maximum width.
2650 if (options->stat_width == -1)
2651 width = term_columns() - strlen(line_prefix);
2652 else
2653 width = options->stat_width ? options->stat_width : 80;
2654 number_width = decimal_width(max_change) > number_width ?
2655 decimal_width(max_change) : number_width;
2657 if (options->stat_graph_width == -1)
2658 options->stat_graph_width = diff_stat_graph_width;
2661 * Guarantee 3/8*16==6 for the graph part
2662 * and 5/8*16==10 for the filename part
2664 if (width < 16 + 6 + number_width)
2665 width = 16 + 6 + number_width;
2668 * First assign sizes that are wanted, ignoring available width.
2669 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
2670 * starting from "XXX" should fit in graph_width.
2672 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
2673 if (options->stat_graph_width &&
2674 options->stat_graph_width < graph_width)
2675 graph_width = options->stat_graph_width;
2677 name_width = (options->stat_name_width > 0 &&
2678 options->stat_name_width < max_len) ?
2679 options->stat_name_width : max_len;
2682 * Adjust adjustable widths not to exceed maximum width
2684 if (name_width + number_width + 6 + graph_width > width) {
2685 if (graph_width > width * 3/8 - number_width - 6) {
2686 graph_width = width * 3/8 - number_width - 6;
2687 if (graph_width < 6)
2688 graph_width = 6;
2691 if (options->stat_graph_width &&
2692 graph_width > options->stat_graph_width)
2693 graph_width = options->stat_graph_width;
2694 if (name_width > width - number_width - 6 - graph_width)
2695 name_width = width - number_width - 6 - graph_width;
2696 else
2697 graph_width = width - number_width - 6 - name_width;
2701 * From here name_width is the width of the name area,
2702 * and graph_width is the width of the graph area.
2703 * max_change is used to scale graph properly.
2705 for (i = 0; i < count; i++) {
2706 const char *prefix = "";
2707 struct diffstat_file *file = data->files[i];
2708 char *name = file->print_name;
2709 uintmax_t added = file->added;
2710 uintmax_t deleted = file->deleted;
2711 int name_len;
2713 if (!file->is_interesting && (added + deleted == 0))
2714 continue;
2717 * "scale" the filename
2719 len = name_width;
2720 name_len = strlen(name);
2721 if (name_width < name_len) {
2722 char *slash;
2723 prefix = "...";
2724 len -= 3;
2725 name += name_len - len;
2726 slash = strchr(name, '/');
2727 if (slash)
2728 name = slash;
2731 if (file->is_binary) {
2732 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2733 strbuf_addf(&out, " %*s", number_width, "Bin");
2734 if (!added && !deleted) {
2735 strbuf_addch(&out, '\n');
2736 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2737 out.buf, out.len, 0);
2738 strbuf_reset(&out);
2739 continue;
2741 strbuf_addf(&out, " %s%"PRIuMAX"%s",
2742 del_c, deleted, reset);
2743 strbuf_addstr(&out, " -> ");
2744 strbuf_addf(&out, "%s%"PRIuMAX"%s",
2745 add_c, added, reset);
2746 strbuf_addstr(&out, " bytes\n");
2747 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2748 out.buf, out.len, 0);
2749 strbuf_reset(&out);
2750 continue;
2752 else if (file->is_unmerged) {
2753 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2754 strbuf_addstr(&out, " Unmerged\n");
2755 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2756 out.buf, out.len, 0);
2757 strbuf_reset(&out);
2758 continue;
2762 * scale the add/delete
2764 add = added;
2765 del = deleted;
2767 if (graph_width <= max_change) {
2768 int total = scale_linear(add + del, graph_width, max_change);
2769 if (total < 2 && add && del)
2770 /* width >= 2 due to the sanity check */
2771 total = 2;
2772 if (add < del) {
2773 add = scale_linear(add, graph_width, max_change);
2774 del = total - add;
2775 } else {
2776 del = scale_linear(del, graph_width, max_change);
2777 add = total - del;
2780 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2781 strbuf_addf(&out, " %*"PRIuMAX"%s",
2782 number_width, added + deleted,
2783 added + deleted ? " " : "");
2784 show_graph(&out, '+', add, add_c, reset);
2785 show_graph(&out, '-', del, del_c, reset);
2786 strbuf_addch(&out, '\n');
2787 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2788 out.buf, out.len, 0);
2789 strbuf_reset(&out);
2792 for (i = 0; i < data->nr; i++) {
2793 struct diffstat_file *file = data->files[i];
2794 uintmax_t added = file->added;
2795 uintmax_t deleted = file->deleted;
2797 if (file->is_unmerged ||
2798 (!file->is_interesting && (added + deleted == 0))) {
2799 total_files--;
2800 continue;
2803 if (!file->is_binary) {
2804 adds += added;
2805 dels += deleted;
2807 if (i < count)
2808 continue;
2809 if (!extra_shown)
2810 emit_diff_symbol(options,
2811 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
2812 NULL, 0, 0);
2813 extra_shown = 1;
2816 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2817 strbuf_release(&out);
2820 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
2822 int i, adds = 0, dels = 0, total_files = data->nr;
2824 if (data->nr == 0)
2825 return;
2827 for (i = 0; i < data->nr; i++) {
2828 int added = data->files[i]->added;
2829 int deleted = data->files[i]->deleted;
2831 if (data->files[i]->is_unmerged ||
2832 (!data->files[i]->is_interesting && (added + deleted == 0))) {
2833 total_files--;
2834 } else if (!data->files[i]->is_binary) { /* don't count bytes */
2835 adds += added;
2836 dels += deleted;
2839 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2842 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
2844 int i;
2846 if (data->nr == 0)
2847 return;
2849 for (i = 0; i < data->nr; i++) {
2850 struct diffstat_file *file = data->files[i];
2852 fprintf(options->file, "%s", diff_line_prefix(options));
2854 if (file->is_binary)
2855 fprintf(options->file, "-\t-\t");
2856 else
2857 fprintf(options->file,
2858 "%"PRIuMAX"\t%"PRIuMAX"\t",
2859 file->added, file->deleted);
2860 if (options->line_termination) {
2861 fill_print_name(file);
2862 if (!file->is_renamed)
2863 write_name_quoted(file->name, options->file,
2864 options->line_termination);
2865 else {
2866 fputs(file->print_name, options->file);
2867 putc(options->line_termination, options->file);
2869 } else {
2870 if (file->is_renamed) {
2871 putc('\0', options->file);
2872 write_name_quoted(file->from_name, options->file, '\0');
2874 write_name_quoted(file->name, options->file, '\0');
2879 struct dirstat_file {
2880 const char *name;
2881 unsigned long changed;
2884 struct dirstat_dir {
2885 struct dirstat_file *files;
2886 int alloc, nr, permille, cumulative;
2889 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2890 unsigned long changed, const char *base, int baselen)
2892 unsigned long sum_changes = 0;
2893 unsigned int sources = 0;
2894 const char *line_prefix = diff_line_prefix(opt);
2896 while (dir->nr) {
2897 struct dirstat_file *f = dir->files;
2898 int namelen = strlen(f->name);
2899 unsigned long changes;
2900 char *slash;
2902 if (namelen < baselen)
2903 break;
2904 if (memcmp(f->name, base, baselen))
2905 break;
2906 slash = strchr(f->name + baselen, '/');
2907 if (slash) {
2908 int newbaselen = slash + 1 - f->name;
2909 changes = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2910 sources++;
2911 } else {
2912 changes = f->changed;
2913 dir->files++;
2914 dir->nr--;
2915 sources += 2;
2917 sum_changes += changes;
2921 * We don't report dirstat's for
2922 * - the top level
2923 * - or cases where everything came from a single directory
2924 * under this directory (sources == 1).
2926 if (baselen && sources != 1) {
2927 if (sum_changes) {
2928 int permille = sum_changes * 1000 / changed;
2929 if (permille >= dir->permille) {
2930 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2931 permille / 10, permille % 10, baselen, base);
2932 if (!dir->cumulative)
2933 return 0;
2937 return sum_changes;
2940 static int dirstat_compare(const void *_a, const void *_b)
2942 const struct dirstat_file *a = _a;
2943 const struct dirstat_file *b = _b;
2944 return strcmp(a->name, b->name);
2947 static void show_dirstat(struct diff_options *options)
2949 int i;
2950 unsigned long changed;
2951 struct dirstat_dir dir;
2952 struct diff_queue_struct *q = &diff_queued_diff;
2954 dir.files = NULL;
2955 dir.alloc = 0;
2956 dir.nr = 0;
2957 dir.permille = options->dirstat_permille;
2958 dir.cumulative = options->flags.dirstat_cumulative;
2960 changed = 0;
2961 for (i = 0; i < q->nr; i++) {
2962 struct diff_filepair *p = q->queue[i];
2963 const char *name;
2964 unsigned long copied, added, damage;
2966 name = p->two->path ? p->two->path : p->one->path;
2968 if (p->one->oid_valid && p->two->oid_valid &&
2969 oideq(&p->one->oid, &p->two->oid)) {
2971 * The SHA1 has not changed, so pre-/post-content is
2972 * identical. We can therefore skip looking at the
2973 * file contents altogether.
2975 damage = 0;
2976 goto found_damage;
2979 if (options->flags.dirstat_by_file) {
2981 * In --dirstat-by-file mode, we don't really need to
2982 * look at the actual file contents at all.
2983 * The fact that the SHA1 changed is enough for us to
2984 * add this file to the list of results
2985 * (with each file contributing equal damage).
2987 damage = 1;
2988 goto found_damage;
2991 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2992 diff_populate_filespec(options->repo, p->one, 0);
2993 diff_populate_filespec(options->repo, p->two, 0);
2994 diffcore_count_changes(options->repo,
2995 p->one, p->two, NULL, NULL,
2996 &copied, &added);
2997 diff_free_filespec_data(p->one);
2998 diff_free_filespec_data(p->two);
2999 } else if (DIFF_FILE_VALID(p->one)) {
3000 diff_populate_filespec(options->repo, p->one, CHECK_SIZE_ONLY);
3001 copied = added = 0;
3002 diff_free_filespec_data(p->one);
3003 } else if (DIFF_FILE_VALID(p->two)) {
3004 diff_populate_filespec(options->repo, p->two, CHECK_SIZE_ONLY);
3005 copied = 0;
3006 added = p->two->size;
3007 diff_free_filespec_data(p->two);
3008 } else
3009 continue;
3012 * Original minus copied is the removed material,
3013 * added is the new material. They are both damages
3014 * made to the preimage.
3015 * If the resulting damage is zero, we know that
3016 * diffcore_count_changes() considers the two entries to
3017 * be identical, but since the oid changed, we
3018 * know that there must have been _some_ kind of change,
3019 * so we force all entries to have damage > 0.
3021 damage = (p->one->size - copied) + added;
3022 if (!damage)
3023 damage = 1;
3025 found_damage:
3026 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
3027 dir.files[dir.nr].name = name;
3028 dir.files[dir.nr].changed = damage;
3029 changed += damage;
3030 dir.nr++;
3033 /* This can happen even with many files, if everything was renames */
3034 if (!changed)
3035 return;
3037 /* Show all directories with more than x% of the changes */
3038 QSORT(dir.files, dir.nr, dirstat_compare);
3039 gather_dirstat(options, &dir, changed, "", 0);
3042 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
3044 int i;
3045 unsigned long changed;
3046 struct dirstat_dir dir;
3048 if (data->nr == 0)
3049 return;
3051 dir.files = NULL;
3052 dir.alloc = 0;
3053 dir.nr = 0;
3054 dir.permille = options->dirstat_permille;
3055 dir.cumulative = options->flags.dirstat_cumulative;
3057 changed = 0;
3058 for (i = 0; i < data->nr; i++) {
3059 struct diffstat_file *file = data->files[i];
3060 unsigned long damage = file->added + file->deleted;
3061 if (file->is_binary)
3063 * binary files counts bytes, not lines. Must find some
3064 * way to normalize binary bytes vs. textual lines.
3065 * The following heuristic assumes that there are 64
3066 * bytes per "line".
3067 * This is stupid and ugly, but very cheap...
3069 damage = DIV_ROUND_UP(damage, 64);
3070 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
3071 dir.files[dir.nr].name = file->name;
3072 dir.files[dir.nr].changed = damage;
3073 changed += damage;
3074 dir.nr++;
3077 /* This can happen even with many files, if everything was renames */
3078 if (!changed)
3079 return;
3081 /* Show all directories with more than x% of the changes */
3082 QSORT(dir.files, dir.nr, dirstat_compare);
3083 gather_dirstat(options, &dir, changed, "", 0);
3086 static void free_diffstat_info(struct diffstat_t *diffstat)
3088 int i;
3089 for (i = 0; i < diffstat->nr; i++) {
3090 struct diffstat_file *f = diffstat->files[i];
3091 free(f->print_name);
3092 free(f->name);
3093 free(f->from_name);
3094 free(f);
3096 free(diffstat->files);
3099 struct checkdiff_t {
3100 const char *filename;
3101 int lineno;
3102 int conflict_marker_size;
3103 struct diff_options *o;
3104 unsigned ws_rule;
3105 unsigned status;
3108 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
3110 char firstchar;
3111 int cnt;
3113 if (len < marker_size + 1)
3114 return 0;
3115 firstchar = line[0];
3116 switch (firstchar) {
3117 case '=': case '>': case '<': case '|':
3118 break;
3119 default:
3120 return 0;
3122 for (cnt = 1; cnt < marker_size; cnt++)
3123 if (line[cnt] != firstchar)
3124 return 0;
3125 /* line[1] thru line[marker_size-1] are same as firstchar */
3126 if (len < marker_size + 1 || !isspace(line[marker_size]))
3127 return 0;
3128 return 1;
3131 static void checkdiff_consume_hunk(void *priv,
3132 long ob, long on, long nb, long nn,
3133 const char *func, long funclen)
3136 struct checkdiff_t *data = priv;
3137 data->lineno = nb - 1;
3140 static void checkdiff_consume(void *priv, char *line, unsigned long len)
3142 struct checkdiff_t *data = priv;
3143 int marker_size = data->conflict_marker_size;
3144 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
3145 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
3146 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
3147 char *err;
3148 const char *line_prefix;
3150 assert(data->o);
3151 line_prefix = diff_line_prefix(data->o);
3153 if (line[0] == '+') {
3154 unsigned bad;
3155 data->lineno++;
3156 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
3157 data->status |= 1;
3158 fprintf(data->o->file,
3159 "%s%s:%d: leftover conflict marker\n",
3160 line_prefix, data->filename, data->lineno);
3162 bad = ws_check(line + 1, len - 1, data->ws_rule);
3163 if (!bad)
3164 return;
3165 data->status |= bad;
3166 err = whitespace_error_string(bad);
3167 fprintf(data->o->file, "%s%s:%d: %s.\n",
3168 line_prefix, data->filename, data->lineno, err);
3169 free(err);
3170 emit_line(data->o, set, reset, line, 1);
3171 ws_check_emit(line + 1, len - 1, data->ws_rule,
3172 data->o->file, set, reset, ws);
3173 } else if (line[0] == ' ') {
3174 data->lineno++;
3178 static unsigned char *deflate_it(char *data,
3179 unsigned long size,
3180 unsigned long *result_size)
3182 int bound;
3183 unsigned char *deflated;
3184 git_zstream stream;
3186 git_deflate_init(&stream, zlib_compression_level);
3187 bound = git_deflate_bound(&stream, size);
3188 deflated = xmalloc(bound);
3189 stream.next_out = deflated;
3190 stream.avail_out = bound;
3192 stream.next_in = (unsigned char *)data;
3193 stream.avail_in = size;
3194 while (git_deflate(&stream, Z_FINISH) == Z_OK)
3195 ; /* nothing */
3196 git_deflate_end(&stream);
3197 *result_size = stream.total_out;
3198 return deflated;
3201 static void emit_binary_diff_body(struct diff_options *o,
3202 mmfile_t *one, mmfile_t *two)
3204 void *cp;
3205 void *delta;
3206 void *deflated;
3207 void *data;
3208 unsigned long orig_size;
3209 unsigned long delta_size;
3210 unsigned long deflate_size;
3211 unsigned long data_size;
3213 /* We could do deflated delta, or we could do just deflated two,
3214 * whichever is smaller.
3216 delta = NULL;
3217 deflated = deflate_it(two->ptr, two->size, &deflate_size);
3218 if (one->size && two->size) {
3219 delta = diff_delta(one->ptr, one->size,
3220 two->ptr, two->size,
3221 &delta_size, deflate_size);
3222 if (delta) {
3223 void *to_free = delta;
3224 orig_size = delta_size;
3225 delta = deflate_it(delta, delta_size, &delta_size);
3226 free(to_free);
3230 if (delta && delta_size < deflate_size) {
3231 char *s = xstrfmt("%"PRIuMAX , (uintmax_t)orig_size);
3232 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
3233 s, strlen(s), 0);
3234 free(s);
3235 free(deflated);
3236 data = delta;
3237 data_size = delta_size;
3238 } else {
3239 char *s = xstrfmt("%lu", two->size);
3240 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
3241 s, strlen(s), 0);
3242 free(s);
3243 free(delta);
3244 data = deflated;
3245 data_size = deflate_size;
3248 /* emit data encoded in base85 */
3249 cp = data;
3250 while (data_size) {
3251 int len;
3252 int bytes = (52 < data_size) ? 52 : data_size;
3253 char line[71];
3254 data_size -= bytes;
3255 if (bytes <= 26)
3256 line[0] = bytes + 'A' - 1;
3257 else
3258 line[0] = bytes - 26 + 'a' - 1;
3259 encode_85(line + 1, cp, bytes);
3260 cp = (char *) cp + bytes;
3262 len = strlen(line);
3263 line[len++] = '\n';
3264 line[len] = '\0';
3266 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_BODY,
3267 line, len, 0);
3269 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_FOOTER, NULL, 0, 0);
3270 free(data);
3273 static void emit_binary_diff(struct diff_options *o,
3274 mmfile_t *one, mmfile_t *two)
3276 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER, NULL, 0, 0);
3277 emit_binary_diff_body(o, one, two);
3278 emit_binary_diff_body(o, two, one);
3281 int diff_filespec_is_binary(struct repository *r,
3282 struct diff_filespec *one)
3284 if (one->is_binary == -1) {
3285 diff_filespec_load_driver(one, r->index);
3286 if (one->driver->binary != -1)
3287 one->is_binary = one->driver->binary;
3288 else {
3289 if (!one->data && DIFF_FILE_VALID(one))
3290 diff_populate_filespec(r, one, CHECK_BINARY);
3291 if (one->is_binary == -1 && one->data)
3292 one->is_binary = buffer_is_binary(one->data,
3293 one->size);
3294 if (one->is_binary == -1)
3295 one->is_binary = 0;
3298 return one->is_binary;
3301 static const struct userdiff_funcname *
3302 diff_funcname_pattern(struct diff_options *o, struct diff_filespec *one)
3304 diff_filespec_load_driver(one, o->repo->index);
3305 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
3308 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
3310 if (!options->a_prefix)
3311 options->a_prefix = a;
3312 if (!options->b_prefix)
3313 options->b_prefix = b;
3316 struct userdiff_driver *get_textconv(struct index_state *istate,
3317 struct diff_filespec *one)
3319 if (!DIFF_FILE_VALID(one))
3320 return NULL;
3322 diff_filespec_load_driver(one, istate);
3323 return userdiff_get_textconv(one->driver);
3326 static void builtin_diff(const char *name_a,
3327 const char *name_b,
3328 struct diff_filespec *one,
3329 struct diff_filespec *two,
3330 const char *xfrm_msg,
3331 int must_show_header,
3332 struct diff_options *o,
3333 int complete_rewrite)
3335 mmfile_t mf1, mf2;
3336 const char *lbl[2];
3337 char *a_one, *b_two;
3338 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
3339 const char *reset = diff_get_color_opt(o, DIFF_RESET);
3340 const char *a_prefix, *b_prefix;
3341 struct userdiff_driver *textconv_one = NULL;
3342 struct userdiff_driver *textconv_two = NULL;
3343 struct strbuf header = STRBUF_INIT;
3344 const char *line_prefix = diff_line_prefix(o);
3346 diff_set_mnemonic_prefix(o, "a/", "b/");
3347 if (o->flags.reverse_diff) {
3348 a_prefix = o->b_prefix;
3349 b_prefix = o->a_prefix;
3350 } else {
3351 a_prefix = o->a_prefix;
3352 b_prefix = o->b_prefix;
3355 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
3356 (!one->mode || S_ISGITLINK(one->mode)) &&
3357 (!two->mode || S_ISGITLINK(two->mode))) {
3358 show_submodule_summary(o, one->path ? one->path : two->path,
3359 &one->oid, &two->oid,
3360 two->dirty_submodule);
3361 return;
3362 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
3363 (!one->mode || S_ISGITLINK(one->mode)) &&
3364 (!two->mode || S_ISGITLINK(two->mode))) {
3365 show_submodule_inline_diff(o, one->path ? one->path : two->path,
3366 &one->oid, &two->oid,
3367 two->dirty_submodule);
3368 return;
3371 if (o->flags.allow_textconv) {
3372 textconv_one = get_textconv(o->repo->index, one);
3373 textconv_two = get_textconv(o->repo->index, two);
3376 /* Never use a non-valid filename anywhere if at all possible */
3377 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
3378 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
3380 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
3381 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
3382 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
3383 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
3384 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
3385 if (lbl[0][0] == '/') {
3386 /* /dev/null */
3387 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
3388 if (xfrm_msg)
3389 strbuf_addstr(&header, xfrm_msg);
3390 must_show_header = 1;
3392 else if (lbl[1][0] == '/') {
3393 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
3394 if (xfrm_msg)
3395 strbuf_addstr(&header, xfrm_msg);
3396 must_show_header = 1;
3398 else {
3399 if (one->mode != two->mode) {
3400 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
3401 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
3402 must_show_header = 1;
3404 if (xfrm_msg)
3405 strbuf_addstr(&header, xfrm_msg);
3408 * we do not run diff between different kind
3409 * of objects.
3411 if ((one->mode ^ two->mode) & S_IFMT)
3412 goto free_ab_and_return;
3413 if (complete_rewrite &&
3414 (textconv_one || !diff_filespec_is_binary(o->repo, one)) &&
3415 (textconv_two || !diff_filespec_is_binary(o->repo, two))) {
3416 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3417 header.buf, header.len, 0);
3418 strbuf_reset(&header);
3419 emit_rewrite_diff(name_a, name_b, one, two,
3420 textconv_one, textconv_two, o);
3421 o->found_changes = 1;
3422 goto free_ab_and_return;
3426 if (o->irreversible_delete && lbl[1][0] == '/') {
3427 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
3428 header.len, 0);
3429 strbuf_reset(&header);
3430 goto free_ab_and_return;
3431 } else if (!o->flags.text &&
3432 ( (!textconv_one && diff_filespec_is_binary(o->repo, one)) ||
3433 (!textconv_two && diff_filespec_is_binary(o->repo, two)) )) {
3434 struct strbuf sb = STRBUF_INIT;
3435 if (!one->data && !two->data &&
3436 S_ISREG(one->mode) && S_ISREG(two->mode) &&
3437 !o->flags.binary) {
3438 if (oideq(&one->oid, &two->oid)) {
3439 if (must_show_header)
3440 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3441 header.buf, header.len,
3443 goto free_ab_and_return;
3445 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3446 header.buf, header.len, 0);
3447 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3448 diff_line_prefix(o), lbl[0], lbl[1]);
3449 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3450 sb.buf, sb.len, 0);
3451 strbuf_release(&sb);
3452 goto free_ab_and_return;
3454 if (fill_mmfile(o->repo, &mf1, one) < 0 ||
3455 fill_mmfile(o->repo, &mf2, two) < 0)
3456 die("unable to read files to diff");
3457 /* Quite common confusing case */
3458 if (mf1.size == mf2.size &&
3459 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
3460 if (must_show_header)
3461 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3462 header.buf, header.len, 0);
3463 goto free_ab_and_return;
3465 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
3466 strbuf_reset(&header);
3467 if (o->flags.binary)
3468 emit_binary_diff(o, &mf1, &mf2);
3469 else {
3470 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3471 diff_line_prefix(o), lbl[0], lbl[1]);
3472 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3473 sb.buf, sb.len, 0);
3474 strbuf_release(&sb);
3476 o->found_changes = 1;
3477 } else {
3478 /* Crazy xdl interfaces.. */
3479 const char *diffopts = getenv("GIT_DIFF_OPTS");
3480 const char *v;
3481 xpparam_t xpp;
3482 xdemitconf_t xecfg;
3483 struct emit_callback ecbdata;
3484 const struct userdiff_funcname *pe;
3486 if (must_show_header) {
3487 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3488 header.buf, header.len, 0);
3489 strbuf_reset(&header);
3492 mf1.size = fill_textconv(o->repo, textconv_one, one, &mf1.ptr);
3493 mf2.size = fill_textconv(o->repo, textconv_two, two, &mf2.ptr);
3495 pe = diff_funcname_pattern(o, one);
3496 if (!pe)
3497 pe = diff_funcname_pattern(o, two);
3499 memset(&xpp, 0, sizeof(xpp));
3500 memset(&xecfg, 0, sizeof(xecfg));
3501 memset(&ecbdata, 0, sizeof(ecbdata));
3502 if (o->flags.suppress_diff_headers)
3503 lbl[0] = NULL;
3504 ecbdata.label_path = lbl;
3505 ecbdata.color_diff = want_color(o->use_color);
3506 ecbdata.ws_rule = whitespace_rule(o->repo->index, name_b);
3507 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
3508 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3509 ecbdata.opt = o;
3510 if (header.len && !o->flags.suppress_diff_headers)
3511 ecbdata.header = &header;
3512 xpp.flags = o->xdl_opts;
3513 xpp.anchors = o->anchors;
3514 xpp.anchors_nr = o->anchors_nr;
3515 xecfg.ctxlen = o->context;
3516 xecfg.interhunkctxlen = o->interhunkcontext;
3517 xecfg.flags = XDL_EMIT_FUNCNAMES;
3518 if (o->flags.funccontext)
3519 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
3520 if (pe)
3521 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
3522 if (!diffopts)
3524 else if (skip_prefix(diffopts, "--unified=", &v))
3525 xecfg.ctxlen = strtoul(v, NULL, 10);
3526 else if (skip_prefix(diffopts, "-u", &v))
3527 xecfg.ctxlen = strtoul(v, NULL, 10);
3528 if (o->word_diff)
3529 init_diff_words_data(&ecbdata, o, one, two);
3530 if (xdi_diff_outf(&mf1, &mf2, NULL, fn_out_consume,
3531 &ecbdata, &xpp, &xecfg))
3532 die("unable to generate diff for %s", one->path);
3533 if (o->word_diff)
3534 free_diff_words_data(&ecbdata);
3535 if (textconv_one)
3536 free(mf1.ptr);
3537 if (textconv_two)
3538 free(mf2.ptr);
3539 xdiff_clear_find_func(&xecfg);
3542 free_ab_and_return:
3543 strbuf_release(&header);
3544 diff_free_filespec_data(one);
3545 diff_free_filespec_data(two);
3546 free(a_one);
3547 free(b_two);
3548 return;
3551 static char *get_compact_summary(const struct diff_filepair *p, int is_renamed)
3553 if (!is_renamed) {
3554 if (p->status == DIFF_STATUS_ADDED) {
3555 if (S_ISLNK(p->two->mode))
3556 return "new +l";
3557 else if ((p->two->mode & 0777) == 0755)
3558 return "new +x";
3559 else
3560 return "new";
3561 } else if (p->status == DIFF_STATUS_DELETED)
3562 return "gone";
3564 if (S_ISLNK(p->one->mode) && !S_ISLNK(p->two->mode))
3565 return "mode -l";
3566 else if (!S_ISLNK(p->one->mode) && S_ISLNK(p->two->mode))
3567 return "mode +l";
3568 else if ((p->one->mode & 0777) == 0644 &&
3569 (p->two->mode & 0777) == 0755)
3570 return "mode +x";
3571 else if ((p->one->mode & 0777) == 0755 &&
3572 (p->two->mode & 0777) == 0644)
3573 return "mode -x";
3574 return NULL;
3577 static void builtin_diffstat(const char *name_a, const char *name_b,
3578 struct diff_filespec *one,
3579 struct diff_filespec *two,
3580 struct diffstat_t *diffstat,
3581 struct diff_options *o,
3582 struct diff_filepair *p)
3584 mmfile_t mf1, mf2;
3585 struct diffstat_file *data;
3586 int same_contents;
3587 int complete_rewrite = 0;
3589 if (!DIFF_PAIR_UNMERGED(p)) {
3590 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3591 complete_rewrite = 1;
3594 data = diffstat_add(diffstat, name_a, name_b);
3595 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
3596 if (o->flags.stat_with_summary)
3597 data->comments = get_compact_summary(p, data->is_renamed);
3599 if (!one || !two) {
3600 data->is_unmerged = 1;
3601 return;
3604 same_contents = oideq(&one->oid, &two->oid);
3606 if (diff_filespec_is_binary(o->repo, one) ||
3607 diff_filespec_is_binary(o->repo, two)) {
3608 data->is_binary = 1;
3609 if (same_contents) {
3610 data->added = 0;
3611 data->deleted = 0;
3612 } else {
3613 data->added = diff_filespec_size(o->repo, two);
3614 data->deleted = diff_filespec_size(o->repo, one);
3618 else if (complete_rewrite) {
3619 diff_populate_filespec(o->repo, one, 0);
3620 diff_populate_filespec(o->repo, two, 0);
3621 data->deleted = count_lines(one->data, one->size);
3622 data->added = count_lines(two->data, two->size);
3625 else if (!same_contents) {
3626 /* Crazy xdl interfaces.. */
3627 xpparam_t xpp;
3628 xdemitconf_t xecfg;
3630 if (fill_mmfile(o->repo, &mf1, one) < 0 ||
3631 fill_mmfile(o->repo, &mf2, two) < 0)
3632 die("unable to read files to diff");
3634 memset(&xpp, 0, sizeof(xpp));
3635 memset(&xecfg, 0, sizeof(xecfg));
3636 xpp.flags = o->xdl_opts;
3637 xpp.anchors = o->anchors;
3638 xpp.anchors_nr = o->anchors_nr;
3639 xecfg.ctxlen = o->context;
3640 xecfg.interhunkctxlen = o->interhunkcontext;
3641 if (xdi_diff_outf(&mf1, &mf2, discard_hunk_line,
3642 diffstat_consume, diffstat, &xpp, &xecfg))
3643 die("unable to generate diffstat for %s", one->path);
3646 diff_free_filespec_data(one);
3647 diff_free_filespec_data(two);
3650 static void builtin_checkdiff(const char *name_a, const char *name_b,
3651 const char *attr_path,
3652 struct diff_filespec *one,
3653 struct diff_filespec *two,
3654 struct diff_options *o)
3656 mmfile_t mf1, mf2;
3657 struct checkdiff_t data;
3659 if (!two)
3660 return;
3662 memset(&data, 0, sizeof(data));
3663 data.filename = name_b ? name_b : name_a;
3664 data.lineno = 0;
3665 data.o = o;
3666 data.ws_rule = whitespace_rule(o->repo->index, attr_path);
3667 data.conflict_marker_size = ll_merge_marker_size(o->repo->index, attr_path);
3669 if (fill_mmfile(o->repo, &mf1, one) < 0 ||
3670 fill_mmfile(o->repo, &mf2, two) < 0)
3671 die("unable to read files to diff");
3674 * All the other codepaths check both sides, but not checking
3675 * the "old" side here is deliberate. We are checking the newly
3676 * introduced changes, and as long as the "new" side is text, we
3677 * can and should check what it introduces.
3679 if (diff_filespec_is_binary(o->repo, two))
3680 goto free_and_return;
3681 else {
3682 /* Crazy xdl interfaces.. */
3683 xpparam_t xpp;
3684 xdemitconf_t xecfg;
3686 memset(&xpp, 0, sizeof(xpp));
3687 memset(&xecfg, 0, sizeof(xecfg));
3688 xecfg.ctxlen = 1; /* at least one context line */
3689 xpp.flags = 0;
3690 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume_hunk,
3691 checkdiff_consume, &data,
3692 &xpp, &xecfg))
3693 die("unable to generate checkdiff for %s", one->path);
3695 if (data.ws_rule & WS_BLANK_AT_EOF) {
3696 struct emit_callback ecbdata;
3697 int blank_at_eof;
3699 ecbdata.ws_rule = data.ws_rule;
3700 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3701 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
3703 if (blank_at_eof) {
3704 static char *err;
3705 if (!err)
3706 err = whitespace_error_string(WS_BLANK_AT_EOF);
3707 fprintf(o->file, "%s:%d: %s.\n",
3708 data.filename, blank_at_eof, err);
3709 data.status = 1; /* report errors */
3713 free_and_return:
3714 diff_free_filespec_data(one);
3715 diff_free_filespec_data(two);
3716 if (data.status)
3717 o->flags.check_failed = 1;
3720 struct diff_filespec *alloc_filespec(const char *path)
3722 struct diff_filespec *spec;
3724 FLEXPTR_ALLOC_STR(spec, path, path);
3725 spec->count = 1;
3726 spec->is_binary = -1;
3727 return spec;
3730 void free_filespec(struct diff_filespec *spec)
3732 if (!--spec->count) {
3733 diff_free_filespec_data(spec);
3734 free(spec);
3738 void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
3739 int oid_valid, unsigned short mode)
3741 if (mode) {
3742 spec->mode = canon_mode(mode);
3743 oidcpy(&spec->oid, oid);
3744 spec->oid_valid = oid_valid;
3749 * Given a name and sha1 pair, if the index tells us the file in
3750 * the work tree has that object contents, return true, so that
3751 * prepare_temp_file() does not have to inflate and extract.
3753 static int reuse_worktree_file(struct index_state *istate,
3754 const char *name,
3755 const struct object_id *oid,
3756 int want_file)
3758 const struct cache_entry *ce;
3759 struct stat st;
3760 int pos, len;
3763 * We do not read the cache ourselves here, because the
3764 * benchmark with my previous version that always reads cache
3765 * shows that it makes things worse for diff-tree comparing
3766 * two linux-2.6 kernel trees in an already checked out work
3767 * tree. This is because most diff-tree comparisons deal with
3768 * only a small number of files, while reading the cache is
3769 * expensive for a large project, and its cost outweighs the
3770 * savings we get by not inflating the object to a temporary
3771 * file. Practically, this code only helps when we are used
3772 * by diff-cache --cached, which does read the cache before
3773 * calling us.
3775 if (!istate->cache)
3776 return 0;
3778 /* We want to avoid the working directory if our caller
3779 * doesn't need the data in a normal file, this system
3780 * is rather slow with its stat/open/mmap/close syscalls,
3781 * and the object is contained in a pack file. The pack
3782 * is probably already open and will be faster to obtain
3783 * the data through than the working directory. Loose
3784 * objects however would tend to be slower as they need
3785 * to be individually opened and inflated.
3787 if (!FAST_WORKING_DIRECTORY && !want_file && has_object_pack(oid))
3788 return 0;
3791 * Similarly, if we'd have to convert the file contents anyway, that
3792 * makes the optimization not worthwhile.
3794 if (!want_file && would_convert_to_git(istate, name))
3795 return 0;
3797 len = strlen(name);
3798 pos = index_name_pos(istate, name, len);
3799 if (pos < 0)
3800 return 0;
3801 ce = istate->cache[pos];
3804 * This is not the sha1 we are looking for, or
3805 * unreusable because it is not a regular file.
3807 if (!oideq(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
3808 return 0;
3811 * If ce is marked as "assume unchanged", there is no
3812 * guarantee that work tree matches what we are looking for.
3814 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
3815 return 0;
3818 * If ce matches the file in the work tree, we can reuse it.
3820 if (ce_uptodate(ce) ||
3821 (!lstat(name, &st) && !ie_match_stat(istate, ce, &st, 0)))
3822 return 1;
3824 return 0;
3827 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
3829 struct strbuf buf = STRBUF_INIT;
3830 char *dirty = "";
3832 /* Are we looking at the work tree? */
3833 if (s->dirty_submodule)
3834 dirty = "-dirty";
3836 strbuf_addf(&buf, "Subproject commit %s%s\n",
3837 oid_to_hex(&s->oid), dirty);
3838 s->size = buf.len;
3839 if (size_only) {
3840 s->data = NULL;
3841 strbuf_release(&buf);
3842 } else {
3843 s->data = strbuf_detach(&buf, NULL);
3844 s->should_free = 1;
3846 return 0;
3850 * While doing rename detection and pickaxe operation, we may need to
3851 * grab the data for the blob (or file) for our own in-core comparison.
3852 * diff_filespec has data and size fields for this purpose.
3854 int diff_populate_filespec(struct repository *r,
3855 struct diff_filespec *s,
3856 unsigned int flags)
3858 int size_only = flags & CHECK_SIZE_ONLY;
3859 int err = 0;
3860 int conv_flags = global_conv_flags_eol;
3862 * demote FAIL to WARN to allow inspecting the situation
3863 * instead of refusing.
3865 if (conv_flags & CONV_EOL_RNDTRP_DIE)
3866 conv_flags = CONV_EOL_RNDTRP_WARN;
3868 if (!DIFF_FILE_VALID(s))
3869 die("internal error: asking to populate invalid file.");
3870 if (S_ISDIR(s->mode))
3871 return -1;
3873 if (s->data)