stash: remove now superfluos help for "stash push"
[git.git] / diff.c
blob6fd288420bc55b5213ad6dfea4a4f411543d6dbf
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "config.h"
6 #include "tempfile.h"
7 #include "quote.h"
8 #include "diff.h"
9 #include "diffcore.h"
10 #include "delta.h"
11 #include "xdiff-interface.h"
12 #include "color.h"
13 #include "attr.h"
14 #include "run-command.h"
15 #include "utf8.h"
16 #include "userdiff.h"
17 #include "submodule-config.h"
18 #include "submodule.h"
19 #include "hashmap.h"
20 #include "ll-merge.h"
21 #include "string-list.h"
22 #include "argv-array.h"
23 #include "graph.h"
24 #include "packfile.h"
26 #ifdef NO_FAST_WORKING_DIRECTORY
27 #define FAST_WORKING_DIRECTORY 0
28 #else
29 #define FAST_WORKING_DIRECTORY 1
30 #endif
32 static int diff_detect_rename_default;
33 static int diff_indent_heuristic = 1;
34 static int diff_rename_limit_default = 400;
35 static int diff_suppress_blank_empty;
36 static int diff_use_color_default = -1;
37 static int diff_color_moved_default;
38 static int diff_context_default = 3;
39 static int diff_interhunk_context_default;
40 static const char *diff_word_regex_cfg;
41 static const char *external_diff_cmd_cfg;
42 static const char *diff_order_file_cfg;
43 int diff_auto_refresh_index = 1;
44 static int diff_mnemonic_prefix;
45 static int diff_no_prefix;
46 static int diff_stat_graph_width;
47 static int diff_dirstat_permille_default = 30;
48 static struct diff_options default_diff_options;
49 static long diff_algorithm;
50 static unsigned ws_error_highlight_default = WSEH_NEW;
52 static char diff_colors[][COLOR_MAXLEN] = {
53 GIT_COLOR_RESET,
54 GIT_COLOR_NORMAL, /* CONTEXT */
55 GIT_COLOR_BOLD, /* METAINFO */
56 GIT_COLOR_CYAN, /* FRAGINFO */
57 GIT_COLOR_RED, /* OLD */
58 GIT_COLOR_GREEN, /* NEW */
59 GIT_COLOR_YELLOW, /* COMMIT */
60 GIT_COLOR_BG_RED, /* WHITESPACE */
61 GIT_COLOR_NORMAL, /* FUNCINFO */
62 GIT_COLOR_BOLD_MAGENTA, /* OLD_MOVED */
63 GIT_COLOR_BOLD_BLUE, /* OLD_MOVED ALTERNATIVE */
64 GIT_COLOR_FAINT, /* OLD_MOVED_DIM */
65 GIT_COLOR_FAINT_ITALIC, /* OLD_MOVED_ALTERNATIVE_DIM */
66 GIT_COLOR_BOLD_CYAN, /* NEW_MOVED */
67 GIT_COLOR_BOLD_YELLOW, /* NEW_MOVED ALTERNATIVE */
68 GIT_COLOR_FAINT, /* NEW_MOVED_DIM */
69 GIT_COLOR_FAINT_ITALIC, /* NEW_MOVED_ALTERNATIVE_DIM */
72 static NORETURN void die_want_option(const char *option_name)
74 die(_("option '%s' requires a value"), option_name);
77 static int parse_diff_color_slot(const char *var)
79 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
80 return DIFF_CONTEXT;
81 if (!strcasecmp(var, "meta"))
82 return DIFF_METAINFO;
83 if (!strcasecmp(var, "frag"))
84 return DIFF_FRAGINFO;
85 if (!strcasecmp(var, "old"))
86 return DIFF_FILE_OLD;
87 if (!strcasecmp(var, "new"))
88 return DIFF_FILE_NEW;
89 if (!strcasecmp(var, "commit"))
90 return DIFF_COMMIT;
91 if (!strcasecmp(var, "whitespace"))
92 return DIFF_WHITESPACE;
93 if (!strcasecmp(var, "func"))
94 return DIFF_FUNCINFO;
95 if (!strcasecmp(var, "oldmoved"))
96 return DIFF_FILE_OLD_MOVED;
97 if (!strcasecmp(var, "oldmovedalternative"))
98 return DIFF_FILE_OLD_MOVED_ALT;
99 if (!strcasecmp(var, "oldmoveddimmed"))
100 return DIFF_FILE_OLD_MOVED_DIM;
101 if (!strcasecmp(var, "oldmovedalternativedimmed"))
102 return DIFF_FILE_OLD_MOVED_ALT_DIM;
103 if (!strcasecmp(var, "newmoved"))
104 return DIFF_FILE_NEW_MOVED;
105 if (!strcasecmp(var, "newmovedalternative"))
106 return DIFF_FILE_NEW_MOVED_ALT;
107 if (!strcasecmp(var, "newmoveddimmed"))
108 return DIFF_FILE_NEW_MOVED_DIM;
109 if (!strcasecmp(var, "newmovedalternativedimmed"))
110 return DIFF_FILE_NEW_MOVED_ALT_DIM;
111 return -1;
114 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
115 struct strbuf *errmsg)
117 char *params_copy = xstrdup(params_string);
118 struct string_list params = STRING_LIST_INIT_NODUP;
119 int ret = 0;
120 int i;
122 if (*params_copy)
123 string_list_split_in_place(&params, params_copy, ',', -1);
124 for (i = 0; i < params.nr; i++) {
125 const char *p = params.items[i].string;
126 if (!strcmp(p, "changes")) {
127 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
128 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
129 } else if (!strcmp(p, "lines")) {
130 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
131 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
132 } else if (!strcmp(p, "files")) {
133 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
134 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
135 } else if (!strcmp(p, "noncumulative")) {
136 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
137 } else if (!strcmp(p, "cumulative")) {
138 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
139 } else if (isdigit(*p)) {
140 char *end;
141 int permille = strtoul(p, &end, 10) * 10;
142 if (*end == '.' && isdigit(*++end)) {
143 /* only use first digit */
144 permille += *end - '0';
145 /* .. and ignore any further digits */
146 while (isdigit(*++end))
147 ; /* nothing */
149 if (!*end)
150 options->dirstat_permille = permille;
151 else {
152 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
154 ret++;
156 } else {
157 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
158 ret++;
162 string_list_clear(&params, 0);
163 free(params_copy);
164 return ret;
167 static int parse_submodule_params(struct diff_options *options, const char *value)
169 if (!strcmp(value, "log"))
170 options->submodule_format = DIFF_SUBMODULE_LOG;
171 else if (!strcmp(value, "short"))
172 options->submodule_format = DIFF_SUBMODULE_SHORT;
173 else if (!strcmp(value, "diff"))
174 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
175 else
176 return -1;
177 return 0;
180 static int git_config_rename(const char *var, const char *value)
182 if (!value)
183 return DIFF_DETECT_RENAME;
184 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
185 return DIFF_DETECT_COPY;
186 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
189 long parse_algorithm_value(const char *value)
191 if (!value)
192 return -1;
193 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
194 return 0;
195 else if (!strcasecmp(value, "minimal"))
196 return XDF_NEED_MINIMAL;
197 else if (!strcasecmp(value, "patience"))
198 return XDF_PATIENCE_DIFF;
199 else if (!strcasecmp(value, "histogram"))
200 return XDF_HISTOGRAM_DIFF;
201 return -1;
204 static int parse_one_token(const char **arg, const char *token)
206 const char *rest;
207 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
208 *arg = rest;
209 return 1;
211 return 0;
214 static int parse_ws_error_highlight(const char *arg)
216 const char *orig_arg = arg;
217 unsigned val = 0;
219 while (*arg) {
220 if (parse_one_token(&arg, "none"))
221 val = 0;
222 else if (parse_one_token(&arg, "default"))
223 val = WSEH_NEW;
224 else if (parse_one_token(&arg, "all"))
225 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
226 else if (parse_one_token(&arg, "new"))
227 val |= WSEH_NEW;
228 else if (parse_one_token(&arg, "old"))
229 val |= WSEH_OLD;
230 else if (parse_one_token(&arg, "context"))
231 val |= WSEH_CONTEXT;
232 else {
233 return -1 - (int)(arg - orig_arg);
235 if (*arg)
236 arg++;
238 return val;
242 * These are to give UI layer defaults.
243 * The core-level commands such as git-diff-files should
244 * never be affected by the setting of diff.renames
245 * the user happens to have in the configuration file.
247 void init_diff_ui_defaults(void)
249 diff_detect_rename_default = 1;
252 int git_diff_heuristic_config(const char *var, const char *value, void *cb)
254 if (!strcmp(var, "diff.indentheuristic"))
255 diff_indent_heuristic = git_config_bool(var, value);
256 return 0;
259 static int parse_color_moved(const char *arg)
261 switch (git_parse_maybe_bool(arg)) {
262 case 0:
263 return COLOR_MOVED_NO;
264 case 1:
265 return COLOR_MOVED_DEFAULT;
266 default:
267 break;
270 if (!strcmp(arg, "no"))
271 return COLOR_MOVED_NO;
272 else if (!strcmp(arg, "plain"))
273 return COLOR_MOVED_PLAIN;
274 else if (!strcmp(arg, "zebra"))
275 return COLOR_MOVED_ZEBRA;
276 else if (!strcmp(arg, "default"))
277 return COLOR_MOVED_DEFAULT;
278 else if (!strcmp(arg, "dimmed_zebra"))
279 return COLOR_MOVED_ZEBRA_DIM;
280 else
281 return error(_("color moved setting must be one of 'no', 'default', 'zebra', 'dimmed_zebra', 'plain'"));
284 int git_diff_ui_config(const char *var, const char *value, void *cb)
286 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
287 diff_use_color_default = git_config_colorbool(var, value);
288 return 0;
290 if (!strcmp(var, "diff.colormoved")) {
291 int cm = parse_color_moved(value);
292 if (cm < 0)
293 return -1;
294 diff_color_moved_default = cm;
295 return 0;
297 if (!strcmp(var, "diff.context")) {
298 diff_context_default = git_config_int(var, value);
299 if (diff_context_default < 0)
300 return -1;
301 return 0;
303 if (!strcmp(var, "diff.interhunkcontext")) {
304 diff_interhunk_context_default = git_config_int(var, value);
305 if (diff_interhunk_context_default < 0)
306 return -1;
307 return 0;
309 if (!strcmp(var, "diff.renames")) {
310 diff_detect_rename_default = git_config_rename(var, value);
311 return 0;
313 if (!strcmp(var, "diff.autorefreshindex")) {
314 diff_auto_refresh_index = git_config_bool(var, value);
315 return 0;
317 if (!strcmp(var, "diff.mnemonicprefix")) {
318 diff_mnemonic_prefix = git_config_bool(var, value);
319 return 0;
321 if (!strcmp(var, "diff.noprefix")) {
322 diff_no_prefix = git_config_bool(var, value);
323 return 0;
325 if (!strcmp(var, "diff.statgraphwidth")) {
326 diff_stat_graph_width = git_config_int(var, value);
327 return 0;
329 if (!strcmp(var, "diff.external"))
330 return git_config_string(&external_diff_cmd_cfg, var, value);
331 if (!strcmp(var, "diff.wordregex"))
332 return git_config_string(&diff_word_regex_cfg, var, value);
333 if (!strcmp(var, "diff.orderfile"))
334 return git_config_pathname(&diff_order_file_cfg, var, value);
336 if (!strcmp(var, "diff.ignoresubmodules"))
337 handle_ignore_submodules_arg(&default_diff_options, value);
339 if (!strcmp(var, "diff.submodule")) {
340 if (parse_submodule_params(&default_diff_options, value))
341 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
342 value);
343 return 0;
346 if (!strcmp(var, "diff.algorithm")) {
347 diff_algorithm = parse_algorithm_value(value);
348 if (diff_algorithm < 0)
349 return -1;
350 return 0;
353 if (!strcmp(var, "diff.wserrorhighlight")) {
354 int val = parse_ws_error_highlight(value);
355 if (val < 0)
356 return -1;
357 ws_error_highlight_default = val;
358 return 0;
361 if (git_color_config(var, value, cb) < 0)
362 return -1;
364 return git_diff_basic_config(var, value, cb);
367 int git_diff_basic_config(const char *var, const char *value, void *cb)
369 const char *name;
371 if (!strcmp(var, "diff.renamelimit")) {
372 diff_rename_limit_default = git_config_int(var, value);
373 return 0;
376 if (userdiff_config(var, value) < 0)
377 return -1;
379 if (skip_prefix(var, "diff.color.", &name) ||
380 skip_prefix(var, "color.diff.", &name)) {
381 int slot = parse_diff_color_slot(name);
382 if (slot < 0)
383 return 0;
384 if (!value)
385 return config_error_nonbool(var);
386 return color_parse(value, diff_colors[slot]);
389 /* like GNU diff's --suppress-blank-empty option */
390 if (!strcmp(var, "diff.suppressblankempty") ||
391 /* for backwards compatibility */
392 !strcmp(var, "diff.suppress-blank-empty")) {
393 diff_suppress_blank_empty = git_config_bool(var, value);
394 return 0;
397 if (!strcmp(var, "diff.dirstat")) {
398 struct strbuf errmsg = STRBUF_INIT;
399 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
400 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
401 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
402 errmsg.buf);
403 strbuf_release(&errmsg);
404 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
405 return 0;
408 if (git_diff_heuristic_config(var, value, cb) < 0)
409 return -1;
411 return git_default_config(var, value, cb);
414 static char *quote_two(const char *one, const char *two)
416 int need_one = quote_c_style(one, NULL, NULL, 1);
417 int need_two = quote_c_style(two, NULL, NULL, 1);
418 struct strbuf res = STRBUF_INIT;
420 if (need_one + need_two) {
421 strbuf_addch(&res, '"');
422 quote_c_style(one, &res, NULL, 1);
423 quote_c_style(two, &res, NULL, 1);
424 strbuf_addch(&res, '"');
425 } else {
426 strbuf_addstr(&res, one);
427 strbuf_addstr(&res, two);
429 return strbuf_detach(&res, NULL);
432 static const char *external_diff(void)
434 static const char *external_diff_cmd = NULL;
435 static int done_preparing = 0;
437 if (done_preparing)
438 return external_diff_cmd;
439 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
440 if (!external_diff_cmd)
441 external_diff_cmd = external_diff_cmd_cfg;
442 done_preparing = 1;
443 return external_diff_cmd;
447 * Keep track of files used for diffing. Sometimes such an entry
448 * refers to a temporary file, sometimes to an existing file, and
449 * sometimes to "/dev/null".
451 static struct diff_tempfile {
453 * filename external diff should read from, or NULL if this
454 * entry is currently not in use:
456 const char *name;
458 char hex[GIT_MAX_HEXSZ + 1];
459 char mode[10];
462 * If this diff_tempfile instance refers to a temporary file,
463 * this tempfile object is used to manage its lifetime.
465 struct tempfile *tempfile;
466 } diff_temp[2];
468 struct emit_callback {
469 int color_diff;
470 unsigned ws_rule;
471 int blank_at_eof_in_preimage;
472 int blank_at_eof_in_postimage;
473 int lno_in_preimage;
474 int lno_in_postimage;
475 const char **label_path;
476 struct diff_words_data *diff_words;
477 struct diff_options *opt;
478 struct strbuf *header;
481 static int count_lines(const char *data, int size)
483 int count, ch, completely_empty = 1, nl_just_seen = 0;
484 count = 0;
485 while (0 < size--) {
486 ch = *data++;
487 if (ch == '\n') {
488 count++;
489 nl_just_seen = 1;
490 completely_empty = 0;
492 else {
493 nl_just_seen = 0;
494 completely_empty = 0;
497 if (completely_empty)
498 return 0;
499 if (!nl_just_seen)
500 count++; /* no trailing newline */
501 return count;
504 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
506 if (!DIFF_FILE_VALID(one)) {
507 mf->ptr = (char *)""; /* does not matter */
508 mf->size = 0;
509 return 0;
511 else if (diff_populate_filespec(one, 0))
512 return -1;
514 mf->ptr = one->data;
515 mf->size = one->size;
516 return 0;
519 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
520 static unsigned long diff_filespec_size(struct diff_filespec *one)
522 if (!DIFF_FILE_VALID(one))
523 return 0;
524 diff_populate_filespec(one, CHECK_SIZE_ONLY);
525 return one->size;
528 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
530 char *ptr = mf->ptr;
531 long size = mf->size;
532 int cnt = 0;
534 if (!size)
535 return cnt;
536 ptr += size - 1; /* pointing at the very end */
537 if (*ptr != '\n')
538 ; /* incomplete line */
539 else
540 ptr--; /* skip the last LF */
541 while (mf->ptr < ptr) {
542 char *prev_eol;
543 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
544 if (*prev_eol == '\n')
545 break;
546 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
547 break;
548 cnt++;
549 ptr = prev_eol - 1;
551 return cnt;
554 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
555 struct emit_callback *ecbdata)
557 int l1, l2, at;
558 unsigned ws_rule = ecbdata->ws_rule;
559 l1 = count_trailing_blank(mf1, ws_rule);
560 l2 = count_trailing_blank(mf2, ws_rule);
561 if (l2 <= l1) {
562 ecbdata->blank_at_eof_in_preimage = 0;
563 ecbdata->blank_at_eof_in_postimage = 0;
564 return;
566 at = count_lines(mf1->ptr, mf1->size);
567 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
569 at = count_lines(mf2->ptr, mf2->size);
570 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
573 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
574 int first, const char *line, int len)
576 int has_trailing_newline, has_trailing_carriage_return;
577 int nofirst;
578 FILE *file = o->file;
580 fputs(diff_line_prefix(o), file);
582 if (len == 0) {
583 has_trailing_newline = (first == '\n');
584 has_trailing_carriage_return = (!has_trailing_newline &&
585 (first == '\r'));
586 nofirst = has_trailing_newline || has_trailing_carriage_return;
587 } else {
588 has_trailing_newline = (len > 0 && line[len-1] == '\n');
589 if (has_trailing_newline)
590 len--;
591 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
592 if (has_trailing_carriage_return)
593 len--;
594 nofirst = 0;
597 if (len || !nofirst) {
598 fputs(set, file);
599 if (!nofirst)
600 fputc(first, file);
601 fwrite(line, len, 1, file);
602 fputs(reset, file);
604 if (has_trailing_carriage_return)
605 fputc('\r', file);
606 if (has_trailing_newline)
607 fputc('\n', file);
610 static void emit_line(struct diff_options *o, const char *set, const char *reset,
611 const char *line, int len)
613 emit_line_0(o, set, reset, line[0], line+1, len-1);
616 enum diff_symbol {
617 DIFF_SYMBOL_BINARY_DIFF_HEADER,
618 DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
619 DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
620 DIFF_SYMBOL_BINARY_DIFF_BODY,
621 DIFF_SYMBOL_BINARY_DIFF_FOOTER,
622 DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
623 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
624 DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
625 DIFF_SYMBOL_STATS_LINE,
626 DIFF_SYMBOL_WORD_DIFF,
627 DIFF_SYMBOL_STAT_SEP,
628 DIFF_SYMBOL_SUMMARY,
629 DIFF_SYMBOL_SUBMODULE_ADD,
630 DIFF_SYMBOL_SUBMODULE_DEL,
631 DIFF_SYMBOL_SUBMODULE_UNTRACKED,
632 DIFF_SYMBOL_SUBMODULE_MODIFIED,
633 DIFF_SYMBOL_SUBMODULE_HEADER,
634 DIFF_SYMBOL_SUBMODULE_ERROR,
635 DIFF_SYMBOL_SUBMODULE_PIPETHROUGH,
636 DIFF_SYMBOL_REWRITE_DIFF,
637 DIFF_SYMBOL_BINARY_FILES,
638 DIFF_SYMBOL_HEADER,
639 DIFF_SYMBOL_FILEPAIR_PLUS,
640 DIFF_SYMBOL_FILEPAIR_MINUS,
641 DIFF_SYMBOL_WORDS_PORCELAIN,
642 DIFF_SYMBOL_WORDS,
643 DIFF_SYMBOL_CONTEXT,
644 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
645 DIFF_SYMBOL_PLUS,
646 DIFF_SYMBOL_MINUS,
647 DIFF_SYMBOL_NO_LF_EOF,
648 DIFF_SYMBOL_CONTEXT_FRAGINFO,
649 DIFF_SYMBOL_CONTEXT_MARKER,
650 DIFF_SYMBOL_SEPARATOR
653 * Flags for content lines:
654 * 0..12 are whitespace rules
655 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
656 * 16 is marking if the line is blank at EOF
658 #define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
659 #define DIFF_SYMBOL_MOVED_LINE (1<<17)
660 #define DIFF_SYMBOL_MOVED_LINE_ALT (1<<18)
661 #define DIFF_SYMBOL_MOVED_LINE_UNINTERESTING (1<<19)
662 #define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
665 * This struct is used when we need to buffer the output of the diff output.
667 * NEEDSWORK: Instead of storing a copy of the line, add an offset pointer
668 * into the pre/post image file. This pointer could be a union with the
669 * line pointer. By storing an offset into the file instead of the literal line,
670 * we can decrease the memory footprint for the buffered output. At first we
671 * may want to only have indirection for the content lines, but we could also
672 * enhance the state for emitting prefabricated lines, e.g. the similarity
673 * score line or hunk/file headers would only need to store a number or path
674 * and then the output can be constructed later on depending on state.
676 struct emitted_diff_symbol {
677 const char *line;
678 int len;
679 int flags;
680 enum diff_symbol s;
682 #define EMITTED_DIFF_SYMBOL_INIT {NULL}
684 struct emitted_diff_symbols {
685 struct emitted_diff_symbol *buf;
686 int nr, alloc;
688 #define EMITTED_DIFF_SYMBOLS_INIT {NULL, 0, 0}
690 static void append_emitted_diff_symbol(struct diff_options *o,
691 struct emitted_diff_symbol *e)
693 struct emitted_diff_symbol *f;
695 ALLOC_GROW(o->emitted_symbols->buf,
696 o->emitted_symbols->nr + 1,
697 o->emitted_symbols->alloc);
698 f = &o->emitted_symbols->buf[o->emitted_symbols->nr++];
700 memcpy(f, e, sizeof(struct emitted_diff_symbol));
701 f->line = e->line ? xmemdupz(e->line, e->len) : NULL;
704 struct moved_entry {
705 struct hashmap_entry ent;
706 const struct emitted_diff_symbol *es;
707 struct moved_entry *next_line;
710 static int next_byte(const char **cp, const char **endp,
711 const struct diff_options *diffopt)
713 int retval;
715 if (*cp > *endp)
716 return -1;
718 if (isspace(**cp)) {
719 if (DIFF_XDL_TST(diffopt, IGNORE_WHITESPACE_CHANGE)) {
720 while (*cp < *endp && isspace(**cp))
721 (*cp)++;
723 * After skipping a couple of whitespaces,
724 * we still have to account for one space.
726 return (int)' ';
729 if (DIFF_XDL_TST(diffopt, IGNORE_WHITESPACE)) {
730 while (*cp < *endp && isspace(**cp))
731 (*cp)++;
732 /* return the first non-ws character via the usual below */
736 retval = (unsigned char)(**cp);
737 (*cp)++;
738 return retval;
741 static int moved_entry_cmp(const struct diff_options *diffopt,
742 const struct moved_entry *a,
743 const struct moved_entry *b,
744 const void *keydata)
746 const char *ap = a->es->line, *ae = a->es->line + a->es->len;
747 const char *bp = b->es->line, *be = b->es->line + b->es->len;
749 if (!(diffopt->xdl_opts & XDF_WHITESPACE_FLAGS))
750 return a->es->len != b->es->len || memcmp(ap, bp, a->es->len);
752 if (DIFF_XDL_TST(diffopt, IGNORE_WHITESPACE_AT_EOL)) {
753 while (ae > ap && isspace(*ae))
754 ae--;
755 while (be > bp && isspace(*be))
756 be--;
759 while (1) {
760 int ca, cb;
761 ca = next_byte(&ap, &ae, diffopt);
762 cb = next_byte(&bp, &be, diffopt);
763 if (ca != cb)
764 return 1;
765 if (ca < 0)
766 return 0;
770 static unsigned get_string_hash(struct emitted_diff_symbol *es, struct diff_options *o)
772 if (o->xdl_opts & XDF_WHITESPACE_FLAGS) {
773 static struct strbuf sb = STRBUF_INIT;
774 const char *ap = es->line, *ae = es->line + es->len;
775 int c;
777 strbuf_reset(&sb);
778 while (ae > ap && isspace(*ae))
779 ae--;
780 while ((c = next_byte(&ap, &ae, o)) > 0)
781 strbuf_addch(&sb, c);
783 return memhash(sb.buf, sb.len);
784 } else {
785 return memhash(es->line, es->len);
789 static struct moved_entry *prepare_entry(struct diff_options *o,
790 int line_no)
792 struct moved_entry *ret = xmalloc(sizeof(*ret));
793 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[line_no];
795 ret->ent.hash = get_string_hash(l, o);
796 ret->es = l;
797 ret->next_line = NULL;
799 return ret;
802 static void add_lines_to_move_detection(struct diff_options *o,
803 struct hashmap *add_lines,
804 struct hashmap *del_lines)
806 struct moved_entry *prev_line = NULL;
808 int n;
809 for (n = 0; n < o->emitted_symbols->nr; n++) {
810 struct hashmap *hm;
811 struct moved_entry *key;
813 switch (o->emitted_symbols->buf[n].s) {
814 case DIFF_SYMBOL_PLUS:
815 hm = add_lines;
816 break;
817 case DIFF_SYMBOL_MINUS:
818 hm = del_lines;
819 break;
820 default:
821 prev_line = NULL;
822 continue;
825 key = prepare_entry(o, n);
826 if (prev_line && prev_line->es->s == o->emitted_symbols->buf[n].s)
827 prev_line->next_line = key;
829 hashmap_add(hm, key);
830 prev_line = key;
834 static int shrink_potential_moved_blocks(struct moved_entry **pmb,
835 int pmb_nr)
837 int lp, rp;
839 /* Shrink the set of potential block to the remaining running */
840 for (lp = 0, rp = pmb_nr - 1; lp <= rp;) {
841 while (lp < pmb_nr && pmb[lp])
842 lp++;
843 /* lp points at the first NULL now */
845 while (rp > -1 && !pmb[rp])
846 rp--;
847 /* rp points at the last non-NULL */
849 if (lp < pmb_nr && rp > -1 && lp < rp) {
850 pmb[lp] = pmb[rp];
851 pmb[rp] = NULL;
852 rp--;
853 lp++;
857 /* Remember the number of running sets */
858 return rp + 1;
862 * If o->color_moved is COLOR_MOVED_PLAIN, this function does nothing.
864 * Otherwise, if the last block has fewer alphanumeric characters than
865 * COLOR_MOVED_MIN_ALNUM_COUNT, unset DIFF_SYMBOL_MOVED_LINE on all lines in
866 * that block.
868 * The last block consists of the (n - block_length)'th line up to but not
869 * including the nth line.
871 * NEEDSWORK: This uses the same heuristic as blame_entry_score() in blame.c.
872 * Think of a way to unify them.
874 static void adjust_last_block(struct diff_options *o, int n, int block_length)
876 int i, alnum_count = 0;
877 if (o->color_moved == COLOR_MOVED_PLAIN)
878 return;
879 for (i = 1; i < block_length + 1; i++) {
880 const char *c = o->emitted_symbols->buf[n - i].line;
881 for (; *c; c++) {
882 if (!isalnum(*c))
883 continue;
884 alnum_count++;
885 if (alnum_count >= COLOR_MOVED_MIN_ALNUM_COUNT)
886 return;
889 for (i = 1; i < block_length + 1; i++)
890 o->emitted_symbols->buf[n - i].flags &= ~DIFF_SYMBOL_MOVED_LINE;
893 /* Find blocks of moved code, delegate actual coloring decision to helper */
894 static void mark_color_as_moved(struct diff_options *o,
895 struct hashmap *add_lines,
896 struct hashmap *del_lines)
898 struct moved_entry **pmb = NULL; /* potentially moved blocks */
899 int pmb_nr = 0, pmb_alloc = 0;
900 int n, flipped_block = 1, block_length = 0;
903 for (n = 0; n < o->emitted_symbols->nr; n++) {
904 struct hashmap *hm = NULL;
905 struct moved_entry *key;
906 struct moved_entry *match = NULL;
907 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
908 int i;
910 switch (l->s) {
911 case DIFF_SYMBOL_PLUS:
912 hm = del_lines;
913 key = prepare_entry(o, n);
914 match = hashmap_get(hm, key, o);
915 free(key);
916 break;
917 case DIFF_SYMBOL_MINUS:
918 hm = add_lines;
919 key = prepare_entry(o, n);
920 match = hashmap_get(hm, key, o);
921 free(key);
922 break;
923 default:
924 flipped_block = 1;
927 if (!match) {
928 adjust_last_block(o, n, block_length);
929 pmb_nr = 0;
930 block_length = 0;
931 continue;
934 l->flags |= DIFF_SYMBOL_MOVED_LINE;
936 if (o->color_moved == COLOR_MOVED_PLAIN)
937 continue;
939 /* Check any potential block runs, advance each or nullify */
940 for (i = 0; i < pmb_nr; i++) {
941 struct moved_entry *p = pmb[i];
942 struct moved_entry *pnext = (p && p->next_line) ?
943 p->next_line : NULL;
944 if (pnext && !hm->cmpfn(o, pnext, match, NULL)) {
945 pmb[i] = p->next_line;
946 } else {
947 pmb[i] = NULL;
951 pmb_nr = shrink_potential_moved_blocks(pmb, pmb_nr);
953 if (pmb_nr == 0) {
955 * The current line is the start of a new block.
956 * Setup the set of potential blocks.
958 for (; match; match = hashmap_get_next(hm, match)) {
959 ALLOC_GROW(pmb, pmb_nr + 1, pmb_alloc);
960 pmb[pmb_nr++] = match;
963 flipped_block = (flipped_block + 1) % 2;
965 adjust_last_block(o, n, block_length);
966 block_length = 0;
969 block_length++;
971 if (flipped_block)
972 l->flags |= DIFF_SYMBOL_MOVED_LINE_ALT;
974 adjust_last_block(o, n, block_length);
976 free(pmb);
979 #define DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK \
980 (DIFF_SYMBOL_MOVED_LINE | DIFF_SYMBOL_MOVED_LINE_ALT)
981 static void dim_moved_lines(struct diff_options *o)
983 int n;
984 for (n = 0; n < o->emitted_symbols->nr; n++) {
985 struct emitted_diff_symbol *prev = (n != 0) ?
986 &o->emitted_symbols->buf[n - 1] : NULL;
987 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
988 struct emitted_diff_symbol *next =
989 (n < o->emitted_symbols->nr - 1) ?
990 &o->emitted_symbols->buf[n + 1] : NULL;
992 /* Not a plus or minus line? */
993 if (l->s != DIFF_SYMBOL_PLUS && l->s != DIFF_SYMBOL_MINUS)
994 continue;
996 /* Not a moved line? */
997 if (!(l->flags & DIFF_SYMBOL_MOVED_LINE))
998 continue;
1001 * If prev or next are not a plus or minus line,
1002 * pretend they don't exist
1004 if (prev && prev->s != DIFF_SYMBOL_PLUS &&
1005 prev->s != DIFF_SYMBOL_MINUS)
1006 prev = NULL;
1007 if (next && next->s != DIFF_SYMBOL_PLUS &&
1008 next->s != DIFF_SYMBOL_MINUS)
1009 next = NULL;
1011 /* Inside a block? */
1012 if ((prev &&
1013 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
1014 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK)) &&
1015 (next &&
1016 (next->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
1017 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK))) {
1018 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1019 continue;
1022 /* Check if we are at an interesting bound: */
1023 if (prev && (prev->flags & DIFF_SYMBOL_MOVED_LINE) &&
1024 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1025 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1026 continue;
1027 if (next && (next->flags & DIFF_SYMBOL_MOVED_LINE) &&
1028 (next->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1029 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1030 continue;
1033 * The boundary to prev and next are not interesting,
1034 * so this line is not interesting as a whole
1036 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1040 static void emit_line_ws_markup(struct diff_options *o,
1041 const char *set, const char *reset,
1042 const char *line, int len, char sign,
1043 unsigned ws_rule, int blank_at_eof)
1045 const char *ws = NULL;
1047 if (o->ws_error_highlight & ws_rule) {
1048 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
1049 if (!*ws)
1050 ws = NULL;
1053 if (!ws)
1054 emit_line_0(o, set, reset, sign, line, len);
1055 else if (blank_at_eof)
1056 /* Blank line at EOF - paint '+' as well */
1057 emit_line_0(o, ws, reset, sign, line, len);
1058 else {
1059 /* Emit just the prefix, then the rest. */
1060 emit_line_0(o, set, reset, sign, "", 0);
1061 ws_check_emit(line, len, ws_rule,
1062 o->file, set, reset, ws);
1066 static void emit_diff_symbol_from_struct(struct diff_options *o,
1067 struct emitted_diff_symbol *eds)
1069 static const char *nneof = " No newline at end of file\n";
1070 const char *context, *reset, *set, *meta, *fraginfo;
1071 struct strbuf sb = STRBUF_INIT;
1073 enum diff_symbol s = eds->s;
1074 const char *line = eds->line;
1075 int len = eds->len;
1076 unsigned flags = eds->flags;
1078 switch (s) {
1079 case DIFF_SYMBOL_NO_LF_EOF:
1080 context = diff_get_color_opt(o, DIFF_CONTEXT);
1081 reset = diff_get_color_opt(o, DIFF_RESET);
1082 putc('\n', o->file);
1083 emit_line_0(o, context, reset, '\\',
1084 nneof, strlen(nneof));
1085 break;
1086 case DIFF_SYMBOL_SUBMODULE_HEADER:
1087 case DIFF_SYMBOL_SUBMODULE_ERROR:
1088 case DIFF_SYMBOL_SUBMODULE_PIPETHROUGH:
1089 case DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES:
1090 case DIFF_SYMBOL_SUMMARY:
1091 case DIFF_SYMBOL_STATS_LINE:
1092 case DIFF_SYMBOL_BINARY_DIFF_BODY:
1093 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
1094 emit_line(o, "", "", line, len);
1095 break;
1096 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
1097 case DIFF_SYMBOL_CONTEXT_MARKER:
1098 context = diff_get_color_opt(o, DIFF_CONTEXT);
1099 reset = diff_get_color_opt(o, DIFF_RESET);
1100 emit_line(o, context, reset, line, len);
1101 break;
1102 case DIFF_SYMBOL_SEPARATOR:
1103 fprintf(o->file, "%s%c",
1104 diff_line_prefix(o),
1105 o->line_termination);
1106 break;
1107 case DIFF_SYMBOL_CONTEXT:
1108 set = diff_get_color_opt(o, DIFF_CONTEXT);
1109 reset = diff_get_color_opt(o, DIFF_RESET);
1110 emit_line_ws_markup(o, set, reset, line, len, ' ',
1111 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
1112 break;
1113 case DIFF_SYMBOL_PLUS:
1114 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1115 DIFF_SYMBOL_MOVED_LINE_ALT |
1116 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1117 case DIFF_SYMBOL_MOVED_LINE |
1118 DIFF_SYMBOL_MOVED_LINE_ALT |
1119 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1120 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT_DIM);
1121 break;
1122 case DIFF_SYMBOL_MOVED_LINE |
1123 DIFF_SYMBOL_MOVED_LINE_ALT:
1124 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT);
1125 break;
1126 case DIFF_SYMBOL_MOVED_LINE |
1127 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1128 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_DIM);
1129 break;
1130 case DIFF_SYMBOL_MOVED_LINE:
1131 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED);
1132 break;
1133 default:
1134 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1136 reset = diff_get_color_opt(o, DIFF_RESET);
1137 emit_line_ws_markup(o, set, reset, line, len, '+',
1138 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
1139 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
1140 break;
1141 case DIFF_SYMBOL_MINUS:
1142 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1143 DIFF_SYMBOL_MOVED_LINE_ALT |
1144 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1145 case DIFF_SYMBOL_MOVED_LINE |
1146 DIFF_SYMBOL_MOVED_LINE_ALT |
1147 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1148 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT_DIM);
1149 break;
1150 case DIFF_SYMBOL_MOVED_LINE |
1151 DIFF_SYMBOL_MOVED_LINE_ALT:
1152 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT);
1153 break;
1154 case DIFF_SYMBOL_MOVED_LINE |
1155 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1156 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_DIM);
1157 break;
1158 case DIFF_SYMBOL_MOVED_LINE:
1159 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED);
1160 break;
1161 default:
1162 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1164 reset = diff_get_color_opt(o, DIFF_RESET);
1165 emit_line_ws_markup(o, set, reset, line, len, '-',
1166 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
1167 break;
1168 case DIFF_SYMBOL_WORDS_PORCELAIN:
1169 context = diff_get_color_opt(o, DIFF_CONTEXT);
1170 reset = diff_get_color_opt(o, DIFF_RESET);
1171 emit_line(o, context, reset, line, len);
1172 fputs("~\n", o->file);
1173 break;
1174 case DIFF_SYMBOL_WORDS:
1175 context = diff_get_color_opt(o, DIFF_CONTEXT);
1176 reset = diff_get_color_opt(o, DIFF_RESET);
1178 * Skip the prefix character, if any. With
1179 * diff_suppress_blank_empty, there may be
1180 * none.
1182 if (line[0] != '\n') {
1183 line++;
1184 len--;
1186 emit_line(o, context, reset, line, len);
1187 break;
1188 case DIFF_SYMBOL_FILEPAIR_PLUS:
1189 meta = diff_get_color_opt(o, DIFF_METAINFO);
1190 reset = diff_get_color_opt(o, DIFF_RESET);
1191 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
1192 line, reset,
1193 strchr(line, ' ') ? "\t" : "");
1194 break;
1195 case DIFF_SYMBOL_FILEPAIR_MINUS:
1196 meta = diff_get_color_opt(o, DIFF_METAINFO);
1197 reset = diff_get_color_opt(o, DIFF_RESET);
1198 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
1199 line, reset,
1200 strchr(line, ' ') ? "\t" : "");
1201 break;
1202 case DIFF_SYMBOL_BINARY_FILES:
1203 case DIFF_SYMBOL_HEADER:
1204 fprintf(o->file, "%s", line);
1205 break;
1206 case DIFF_SYMBOL_BINARY_DIFF_HEADER:
1207 fprintf(o->file, "%sGIT binary patch\n", diff_line_prefix(o));
1208 break;
1209 case DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA:
1210 fprintf(o->file, "%sdelta %s\n", diff_line_prefix(o), line);
1211 break;
1212 case DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL:
1213 fprintf(o->file, "%sliteral %s\n", diff_line_prefix(o), line);
1214 break;
1215 case DIFF_SYMBOL_BINARY_DIFF_FOOTER:
1216 fputs(diff_line_prefix(o), o->file);
1217 fputc('\n', o->file);
1218 break;
1219 case DIFF_SYMBOL_REWRITE_DIFF:
1220 fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
1221 reset = diff_get_color_opt(o, DIFF_RESET);
1222 emit_line(o, fraginfo, reset, line, len);
1223 break;
1224 case DIFF_SYMBOL_SUBMODULE_ADD:
1225 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1226 reset = diff_get_color_opt(o, DIFF_RESET);
1227 emit_line(o, set, reset, line, len);
1228 break;
1229 case DIFF_SYMBOL_SUBMODULE_DEL:
1230 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1231 reset = diff_get_color_opt(o, DIFF_RESET);
1232 emit_line(o, set, reset, line, len);
1233 break;
1234 case DIFF_SYMBOL_SUBMODULE_UNTRACKED:
1235 fprintf(o->file, "%sSubmodule %s contains untracked content\n",
1236 diff_line_prefix(o), line);
1237 break;
1238 case DIFF_SYMBOL_SUBMODULE_MODIFIED:
1239 fprintf(o->file, "%sSubmodule %s contains modified content\n",
1240 diff_line_prefix(o), line);
1241 break;
1242 case DIFF_SYMBOL_STATS_SUMMARY_NO_FILES:
1243 emit_line(o, "", "", " 0 files changed\n",
1244 strlen(" 0 files changed\n"));
1245 break;
1246 case DIFF_SYMBOL_STATS_SUMMARY_ABBREV:
1247 emit_line(o, "", "", " ...\n", strlen(" ...\n"));
1248 break;
1249 case DIFF_SYMBOL_WORD_DIFF:
1250 fprintf(o->file, "%.*s", len, line);
1251 break;
1252 case DIFF_SYMBOL_STAT_SEP:
1253 fputs(o->stat_sep, o->file);
1254 break;
1255 default:
1256 die("BUG: unknown diff symbol");
1258 strbuf_release(&sb);
1261 static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
1262 const char *line, int len, unsigned flags)
1264 struct emitted_diff_symbol e = {line, len, flags, s};
1266 if (o->emitted_symbols)
1267 append_emitted_diff_symbol(o, &e);
1268 else
1269 emit_diff_symbol_from_struct(o, &e);
1272 void diff_emit_submodule_del(struct diff_options *o, const char *line)
1274 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_DEL, line, strlen(line), 0);
1277 void diff_emit_submodule_add(struct diff_options *o, const char *line)
1279 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ADD, line, strlen(line), 0);
1282 void diff_emit_submodule_untracked(struct diff_options *o, const char *path)
1284 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_UNTRACKED,
1285 path, strlen(path), 0);
1288 void diff_emit_submodule_modified(struct diff_options *o, const char *path)
1290 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_MODIFIED,
1291 path, strlen(path), 0);
1294 void diff_emit_submodule_header(struct diff_options *o, const char *header)
1296 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_HEADER,
1297 header, strlen(header), 0);
1300 void diff_emit_submodule_error(struct diff_options *o, const char *err)
1302 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ERROR, err, strlen(err), 0);
1305 void diff_emit_submodule_pipethrough(struct diff_options *o,
1306 const char *line, int len)
1308 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_PIPETHROUGH, line, len, 0);
1311 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
1313 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
1314 ecbdata->blank_at_eof_in_preimage &&
1315 ecbdata->blank_at_eof_in_postimage &&
1316 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
1317 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
1318 return 0;
1319 return ws_blank_line(line, len, ecbdata->ws_rule);
1322 static void emit_add_line(const char *reset,
1323 struct emit_callback *ecbdata,
1324 const char *line, int len)
1326 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
1327 if (new_blank_line_at_eof(ecbdata, line, len))
1328 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
1330 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
1333 static void emit_del_line(const char *reset,
1334 struct emit_callback *ecbdata,
1335 const char *line, int len)
1337 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
1338 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
1341 static void emit_context_line(const char *reset,
1342 struct emit_callback *ecbdata,
1343 const char *line, int len)
1345 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
1346 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
1349 static void emit_hunk_header(struct emit_callback *ecbdata,
1350 const char *line, int len)
1352 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1353 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
1354 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
1355 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1356 static const char atat[2] = { '@', '@' };
1357 const char *cp, *ep;
1358 struct strbuf msgbuf = STRBUF_INIT;
1359 int org_len = len;
1360 int i = 1;
1363 * As a hunk header must begin with "@@ -<old>, +<new> @@",
1364 * it always is at least 10 bytes long.
1366 if (len < 10 ||
1367 memcmp(line, atat, 2) ||
1368 !(ep = memmem(line + 2, len - 2, atat, 2))) {
1369 emit_diff_symbol(ecbdata->opt,
1370 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
1371 return;
1373 ep += 2; /* skip over @@ */
1375 /* The hunk header in fraginfo color */
1376 strbuf_addstr(&msgbuf, frag);
1377 strbuf_add(&msgbuf, line, ep - line);
1378 strbuf_addstr(&msgbuf, reset);
1381 * trailing "\r\n"
1383 for ( ; i < 3; i++)
1384 if (line[len - i] == '\r' || line[len - i] == '\n')
1385 len--;
1387 /* blank before the func header */
1388 for (cp = ep; ep - line < len; ep++)
1389 if (*ep != ' ' && *ep != '\t')
1390 break;
1391 if (ep != cp) {
1392 strbuf_addstr(&msgbuf, context);
1393 strbuf_add(&msgbuf, cp, ep - cp);
1394 strbuf_addstr(&msgbuf, reset);
1397 if (ep < line + len) {
1398 strbuf_addstr(&msgbuf, func);
1399 strbuf_add(&msgbuf, ep, line + len - ep);
1400 strbuf_addstr(&msgbuf, reset);
1403 strbuf_add(&msgbuf, line + len, org_len - len);
1404 strbuf_complete_line(&msgbuf);
1405 emit_diff_symbol(ecbdata->opt,
1406 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
1407 strbuf_release(&msgbuf);
1410 static struct diff_tempfile *claim_diff_tempfile(void) {
1411 int i;
1412 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
1413 if (!diff_temp[i].name)
1414 return diff_temp + i;
1415 die("BUG: diff is failing to clean up its tempfiles");
1418 static void remove_tempfile(void)
1420 int i;
1421 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
1422 if (is_tempfile_active(diff_temp[i].tempfile))
1423 delete_tempfile(&diff_temp[i].tempfile);
1424 diff_temp[i].name = NULL;
1428 static void add_line_count(struct strbuf *out, int count)
1430 switch (count) {
1431 case 0:
1432 strbuf_addstr(out, "0,0");
1433 break;
1434 case 1:
1435 strbuf_addstr(out, "1");
1436 break;
1437 default:
1438 strbuf_addf(out, "1,%d", count);
1439 break;
1443 static void emit_rewrite_lines(struct emit_callback *ecb,
1444 int prefix, const char *data, int size)
1446 const char *endp = NULL;
1447 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
1449 while (0 < size) {
1450 int len;
1452 endp = memchr(data, '\n', size);
1453 len = endp ? (endp - data + 1) : size;
1454 if (prefix != '+') {
1455 ecb->lno_in_preimage++;
1456 emit_del_line(reset, ecb, data, len);
1457 } else {
1458 ecb->lno_in_postimage++;
1459 emit_add_line(reset, ecb, data, len);
1461 size -= len;
1462 data += len;
1464 if (!endp)
1465 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
1468 static void emit_rewrite_diff(const char *name_a,
1469 const char *name_b,
1470 struct diff_filespec *one,
1471 struct diff_filespec *two,
1472 struct userdiff_driver *textconv_one,
1473 struct userdiff_driver *textconv_two,
1474 struct diff_options *o)
1476 int lc_a, lc_b;
1477 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
1478 const char *a_prefix, *b_prefix;
1479 char *data_one, *data_two;
1480 size_t size_one, size_two;
1481 struct emit_callback ecbdata;
1482 struct strbuf out = STRBUF_INIT;
1484 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
1485 a_prefix = o->b_prefix;
1486 b_prefix = o->a_prefix;
1487 } else {
1488 a_prefix = o->a_prefix;
1489 b_prefix = o->b_prefix;
1492 name_a += (*name_a == '/');
1493 name_b += (*name_b == '/');
1495 strbuf_reset(&a_name);
1496 strbuf_reset(&b_name);
1497 quote_two_c_style(&a_name, a_prefix, name_a, 0);
1498 quote_two_c_style(&b_name, b_prefix, name_b, 0);
1500 size_one = fill_textconv(textconv_one, one, &data_one);
1501 size_two = fill_textconv(textconv_two, two, &data_two);
1503 memset(&ecbdata, 0, sizeof(ecbdata));
1504 ecbdata.color_diff = want_color(o->use_color);
1505 ecbdata.ws_rule = whitespace_rule(name_b);
1506 ecbdata.opt = o;
1507 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
1508 mmfile_t mf1, mf2;
1509 mf1.ptr = (char *)data_one;
1510 mf2.ptr = (char *)data_two;
1511 mf1.size = size_one;
1512 mf2.size = size_two;
1513 check_blank_at_eof(&mf1, &mf2, &ecbdata);
1515 ecbdata.lno_in_preimage = 1;
1516 ecbdata.lno_in_postimage = 1;
1518 lc_a = count_lines(data_one, size_one);
1519 lc_b = count_lines(data_two, size_two);
1521 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1522 a_name.buf, a_name.len, 0);
1523 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1524 b_name.buf, b_name.len, 0);
1526 strbuf_addstr(&out, "@@ -");
1527 if (!o->irreversible_delete)
1528 add_line_count(&out, lc_a);
1529 else
1530 strbuf_addstr(&out, "?,?");
1531 strbuf_addstr(&out, " +");
1532 add_line_count(&out, lc_b);
1533 strbuf_addstr(&out, " @@\n");
1534 emit_diff_symbol(o, DIFF_SYMBOL_REWRITE_DIFF, out.buf, out.len, 0);
1535 strbuf_release(&out);
1537 if (lc_a && !o->irreversible_delete)
1538 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
1539 if (lc_b)
1540 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
1541 if (textconv_one)
1542 free((char *)data_one);
1543 if (textconv_two)
1544 free((char *)data_two);
1547 struct diff_words_buffer {
1548 mmfile_t text;
1549 unsigned long alloc;
1550 struct diff_words_orig {
1551 const char *begin, *end;
1552 } *orig;
1553 int orig_nr, orig_alloc;
1556 static void diff_words_append(char *line, unsigned long len,
1557 struct diff_words_buffer *buffer)
1559 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
1560 line++;
1561 len--;
1562 memcpy(buffer->text.ptr + buffer->text.size, line, len);
1563 buffer->text.size += len;
1564 buffer->text.ptr[buffer->text.size] = '\0';
1567 struct diff_words_style_elem {
1568 const char *prefix;
1569 const char *suffix;
1570 const char *color; /* NULL; filled in by the setup code if
1571 * color is enabled */
1574 struct diff_words_style {
1575 enum diff_words_type type;
1576 struct diff_words_style_elem new, old, ctx;
1577 const char *newline;
1580 static struct diff_words_style diff_words_styles[] = {
1581 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
1582 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
1583 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
1586 struct diff_words_data {
1587 struct diff_words_buffer minus, plus;
1588 const char *current_plus;
1589 int last_minus;
1590 struct diff_options *opt;
1591 regex_t *word_regex;
1592 enum diff_words_type type;
1593 struct diff_words_style *style;
1596 static int fn_out_diff_words_write_helper(struct diff_options *o,
1597 struct diff_words_style_elem *st_el,
1598 const char *newline,
1599 size_t count, const char *buf)
1601 int print = 0;
1602 struct strbuf sb = STRBUF_INIT;
1604 while (count) {
1605 char *p = memchr(buf, '\n', count);
1606 if (print)
1607 strbuf_addstr(&sb, diff_line_prefix(o));
1609 if (p != buf) {
1610 const char *reset = st_el->color && *st_el->color ?
1611 GIT_COLOR_RESET : NULL;
1612 if (st_el->color && *st_el->color)
1613 strbuf_addstr(&sb, st_el->color);
1614 strbuf_addstr(&sb, st_el->prefix);
1615 strbuf_add(&sb, buf, p ? p - buf : count);
1616 strbuf_addstr(&sb, st_el->suffix);
1617 if (reset)
1618 strbuf_addstr(&sb, reset);
1620 if (!p)
1621 goto out;
1623 strbuf_addstr(&sb, newline);
1624 count -= p + 1 - buf;
1625 buf = p + 1;
1626 print = 1;
1627 if (count) {
1628 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1629 sb.buf, sb.len, 0);
1630 strbuf_reset(&sb);
1634 out:
1635 if (sb.len)
1636 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1637 sb.buf, sb.len, 0);
1638 strbuf_release(&sb);
1639 return 0;
1643 * '--color-words' algorithm can be described as:
1645 * 1. collect the minus/plus lines of a diff hunk, divided into
1646 * minus-lines and plus-lines;
1648 * 2. break both minus-lines and plus-lines into words and
1649 * place them into two mmfile_t with one word for each line;
1651 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1653 * And for the common parts of the both file, we output the plus side text.
1654 * diff_words->current_plus is used to trace the current position of the plus file
1655 * which printed. diff_words->last_minus is used to trace the last minus word
1656 * printed.
1658 * For '--graph' to work with '--color-words', we need to output the graph prefix
1659 * on each line of color words output. Generally, there are two conditions on
1660 * which we should output the prefix.
1662 * 1. diff_words->last_minus == 0 &&
1663 * diff_words->current_plus == diff_words->plus.text.ptr
1665 * that is: the plus text must start as a new line, and if there is no minus
1666 * word printed, a graph prefix must be printed.
1668 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1669 * *(diff_words->current_plus - 1) == '\n'
1671 * that is: a graph prefix must be printed following a '\n'
1673 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1675 if ((diff_words->last_minus == 0 &&
1676 diff_words->current_plus == diff_words->plus.text.ptr) ||
1677 (diff_words->current_plus > diff_words->plus.text.ptr &&
1678 *(diff_words->current_plus - 1) == '\n')) {
1679 return 1;
1680 } else {
1681 return 0;
1685 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1687 struct diff_words_data *diff_words = priv;
1688 struct diff_words_style *style = diff_words->style;
1689 int minus_first, minus_len, plus_first, plus_len;
1690 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1691 struct diff_options *opt = diff_words->opt;
1692 const char *line_prefix;
1694 if (line[0] != '@' || parse_hunk_header(line, len,
1695 &minus_first, &minus_len, &plus_first, &plus_len))
1696 return;
1698 assert(opt);
1699 line_prefix = diff_line_prefix(opt);
1701 /* POSIX requires that first be decremented by one if len == 0... */
1702 if (minus_len) {
1703 minus_begin = diff_words->minus.orig[minus_first].begin;
1704 minus_end =
1705 diff_words->minus.orig[minus_first + minus_len - 1].end;
1706 } else
1707 minus_begin = minus_end =
1708 diff_words->minus.orig[minus_first].end;
1710 if (plus_len) {
1711 plus_begin = diff_words->plus.orig[plus_first].begin;
1712 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1713 } else
1714 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1716 if (color_words_output_graph_prefix(diff_words)) {
1717 fputs(line_prefix, diff_words->opt->file);
1719 if (diff_words->current_plus != plus_begin) {
1720 fn_out_diff_words_write_helper(diff_words->opt,
1721 &style->ctx, style->newline,
1722 plus_begin - diff_words->current_plus,
1723 diff_words->current_plus);
1725 if (minus_begin != minus_end) {
1726 fn_out_diff_words_write_helper(diff_words->opt,
1727 &style->old, style->newline,
1728 minus_end - minus_begin, minus_begin);
1730 if (plus_begin != plus_end) {
1731 fn_out_diff_words_write_helper(diff_words->opt,
1732 &style->new, style->newline,
1733 plus_end - plus_begin, plus_begin);
1736 diff_words->current_plus = plus_end;
1737 diff_words->last_minus = minus_first;
1740 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1741 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1742 int *begin, int *end)
1744 if (word_regex && *begin < buffer->size) {
1745 regmatch_t match[1];
1746 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1747 buffer->size - *begin, 1, match, 0)) {
1748 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1749 '\n', match[0].rm_eo - match[0].rm_so);
1750 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1751 *begin += match[0].rm_so;
1752 return *begin >= *end;
1754 return -1;
1757 /* find the next word */
1758 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1759 (*begin)++;
1760 if (*begin >= buffer->size)
1761 return -1;
1763 /* find the end of the word */
1764 *end = *begin + 1;
1765 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1766 (*end)++;
1768 return 0;
1772 * This function splits the words in buffer->text, stores the list with
1773 * newline separator into out, and saves the offsets of the original words
1774 * in buffer->orig.
1776 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1777 regex_t *word_regex)
1779 int i, j;
1780 long alloc = 0;
1782 out->size = 0;
1783 out->ptr = NULL;
1785 /* fake an empty "0th" word */
1786 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1787 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1788 buffer->orig_nr = 1;
1790 for (i = 0; i < buffer->text.size; i++) {
1791 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1792 return;
1794 /* store original boundaries */
1795 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1796 buffer->orig_alloc);
1797 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1798 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1799 buffer->orig_nr++;
1801 /* store one word */
1802 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1803 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1804 out->ptr[out->size + j - i] = '\n';
1805 out->size += j - i + 1;
1807 i = j - 1;
1811 /* this executes the word diff on the accumulated buffers */
1812 static void diff_words_show(struct diff_words_data *diff_words)
1814 xpparam_t xpp;
1815 xdemitconf_t xecfg;
1816 mmfile_t minus, plus;
1817 struct diff_words_style *style = diff_words->style;
1819 struct diff_options *opt = diff_words->opt;
1820 const char *line_prefix;
1822 assert(opt);
1823 line_prefix = diff_line_prefix(opt);
1825 /* special case: only removal */
1826 if (!diff_words->plus.text.size) {
1827 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
1828 line_prefix, strlen(line_prefix), 0);
1829 fn_out_diff_words_write_helper(diff_words->opt,
1830 &style->old, style->newline,
1831 diff_words->minus.text.size,
1832 diff_words->minus.text.ptr);
1833 diff_words->minus.text.size = 0;
1834 return;
1837 diff_words->current_plus = diff_words->plus.text.ptr;
1838 diff_words->last_minus = 0;
1840 memset(&xpp, 0, sizeof(xpp));
1841 memset(&xecfg, 0, sizeof(xecfg));
1842 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1843 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1844 xpp.flags = 0;
1845 /* as only the hunk header will be parsed, we need a 0-context */
1846 xecfg.ctxlen = 0;
1847 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1848 &xpp, &xecfg))
1849 die("unable to generate word diff");
1850 free(minus.ptr);
1851 free(plus.ptr);
1852 if (diff_words->current_plus != diff_words->plus.text.ptr +
1853 diff_words->plus.text.size) {
1854 if (color_words_output_graph_prefix(diff_words))
1855 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
1856 line_prefix, strlen(line_prefix), 0);
1857 fn_out_diff_words_write_helper(diff_words->opt,
1858 &style->ctx, style->newline,
1859 diff_words->plus.text.ptr + diff_words->plus.text.size
1860 - diff_words->current_plus, diff_words->current_plus);
1862 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1865 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1866 static void diff_words_flush(struct emit_callback *ecbdata)
1868 struct diff_options *wo = ecbdata->diff_words->opt;
1870 if (ecbdata->diff_words->minus.text.size ||
1871 ecbdata->diff_words->plus.text.size)
1872 diff_words_show(ecbdata->diff_words);
1874 if (wo->emitted_symbols) {
1875 struct diff_options *o = ecbdata->opt;
1876 struct emitted_diff_symbols *wol = wo->emitted_symbols;
1877 int i;
1880 * NEEDSWORK:
1881 * Instead of appending each, concat all words to a line?
1883 for (i = 0; i < wol->nr; i++)
1884 append_emitted_diff_symbol(o, &wol->buf[i]);
1886 for (i = 0; i < wol->nr; i++)
1887 free((void *)wol->buf[i].line);
1889 wol->nr = 0;
1893 static void diff_filespec_load_driver(struct diff_filespec *one)
1895 /* Use already-loaded driver */
1896 if (one->driver)
1897 return;
1899 if (S_ISREG(one->mode))
1900 one->driver = userdiff_find_by_path(one->path);
1902 /* Fallback to default settings */
1903 if (!one->driver)
1904 one->driver = userdiff_find_by_name("default");
1907 static const char *userdiff_word_regex(struct diff_filespec *one)
1909 diff_filespec_load_driver(one);
1910 return one->driver->word_regex;
1913 static void init_diff_words_data(struct emit_callback *ecbdata,
1914 struct diff_options *orig_opts,
1915 struct diff_filespec *one,
1916 struct diff_filespec *two)
1918 int i;
1919 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1920 memcpy(o, orig_opts, sizeof(struct diff_options));
1922 ecbdata->diff_words =
1923 xcalloc(1, sizeof(struct diff_words_data));
1924 ecbdata->diff_words->type = o->word_diff;
1925 ecbdata->diff_words->opt = o;
1927 if (orig_opts->emitted_symbols)
1928 o->emitted_symbols =
1929 xcalloc(1, sizeof(struct emitted_diff_symbols));
1931 if (!o->word_regex)
1932 o->word_regex = userdiff_word_regex(one);
1933 if (!o->word_regex)
1934 o->word_regex = userdiff_word_regex(two);
1935 if (!o->word_regex)
1936 o->word_regex = diff_word_regex_cfg;
1937 if (o->word_regex) {
1938 ecbdata->diff_words->word_regex = (regex_t *)
1939 xmalloc(sizeof(regex_t));
1940 if (regcomp(ecbdata->diff_words->word_regex,
1941 o->word_regex,
1942 REG_EXTENDED | REG_NEWLINE))
1943 die ("Invalid regular expression: %s",
1944 o->word_regex);
1946 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1947 if (o->word_diff == diff_words_styles[i].type) {
1948 ecbdata->diff_words->style =
1949 &diff_words_styles[i];
1950 break;
1953 if (want_color(o->use_color)) {
1954 struct diff_words_style *st = ecbdata->diff_words->style;
1955 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1956 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1957 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1961 static void free_diff_words_data(struct emit_callback *ecbdata)
1963 if (ecbdata->diff_words) {
1964 diff_words_flush(ecbdata);
1965 free (ecbdata->diff_words->opt->emitted_symbols);
1966 free (ecbdata->diff_words->opt);
1967 free (ecbdata->diff_words->minus.text.ptr);
1968 free (ecbdata->diff_words->minus.orig);
1969 free (ecbdata->diff_words->plus.text.ptr);
1970 free (ecbdata->diff_words->plus.orig);
1971 if (ecbdata->diff_words->word_regex) {
1972 regfree(ecbdata->diff_words->word_regex);
1973 free(ecbdata->diff_words->word_regex);
1975 FREE_AND_NULL(ecbdata->diff_words);
1979 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1981 if (want_color(diff_use_color))
1982 return diff_colors[ix];
1983 return "";
1986 const char *diff_line_prefix(struct diff_options *opt)
1988 struct strbuf *msgbuf;
1989 if (!opt->output_prefix)
1990 return "";
1992 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1993 return msgbuf->buf;
1996 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1998 const char *cp;
1999 unsigned long allot;
2000 size_t l = len;
2002 cp = line;
2003 allot = l;
2004 while (0 < l) {
2005 (void) utf8_width(&cp, &l);
2006 if (!cp)
2007 break; /* truncated in the middle? */
2009 return allot - l;
2012 static void find_lno(const char *line, struct emit_callback *ecbdata)
2014 const char *p;
2015 ecbdata->lno_in_preimage = 0;
2016 ecbdata->lno_in_postimage = 0;
2017 p = strchr(line, '-');
2018 if (!p)
2019 return; /* cannot happen */
2020 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
2021 p = strchr(p, '+');
2022 if (!p)
2023 return; /* cannot happen */
2024 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
2027 static void fn_out_consume(void *priv, char *line, unsigned long len)
2029 struct emit_callback *ecbdata = priv;
2030 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
2031 struct diff_options *o = ecbdata->opt;
2033 o->found_changes = 1;
2035 if (ecbdata->header) {
2036 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2037 ecbdata->header->buf, ecbdata->header->len, 0);
2038 strbuf_reset(ecbdata->header);
2039 ecbdata->header = NULL;
2042 if (ecbdata->label_path[0]) {
2043 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
2044 ecbdata->label_path[0],
2045 strlen(ecbdata->label_path[0]), 0);
2046 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
2047 ecbdata->label_path[1],
2048 strlen(ecbdata->label_path[1]), 0);
2049 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
2052 if (diff_suppress_blank_empty
2053 && len == 2 && line[0] == ' ' && line[1] == '\n') {
2054 line[0] = '\n';
2055 len = 1;
2058 if (line[0] == '@') {
2059 if (ecbdata->diff_words)
2060 diff_words_flush(ecbdata);
2061 len = sane_truncate_line(ecbdata, line, len);
2062 find_lno(line, ecbdata);
2063 emit_hunk_header(ecbdata, line, len);
2064 return;
2067 if (ecbdata->diff_words) {
2068 enum diff_symbol s =
2069 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
2070 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
2071 if (line[0] == '-') {
2072 diff_words_append(line, len,
2073 &ecbdata->diff_words->minus);
2074 return;
2075 } else if (line[0] == '+') {
2076 diff_words_append(line, len,
2077 &ecbdata->diff_words->plus);
2078 return;
2079 } else if (starts_with(line, "\\ ")) {
2081 * Eat the "no newline at eof" marker as if we
2082 * saw a "+" or "-" line with nothing on it,
2083 * and return without diff_words_flush() to
2084 * defer processing. If this is the end of
2085 * preimage, more "+" lines may come after it.
2087 return;
2089 diff_words_flush(ecbdata);
2090 emit_diff_symbol(o, s, line, len, 0);
2091 return;
2094 switch (line[0]) {
2095 case '+':
2096 ecbdata->lno_in_postimage++;
2097 emit_add_line(reset, ecbdata, line + 1, len - 1);
2098 break;
2099 case '-':
2100 ecbdata->lno_in_preimage++;
2101 emit_del_line(reset, ecbdata, line + 1, len - 1);
2102 break;
2103 case ' ':
2104 ecbdata->lno_in_postimage++;
2105 ecbdata->lno_in_preimage++;
2106 emit_context_line(reset, ecbdata, line + 1, len - 1);
2107 break;
2108 default:
2109 /* incomplete line at the end */
2110 ecbdata->lno_in_preimage++;
2111 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
2112 line, len, 0);
2113 break;
2117 static char *pprint_rename(const char *a, const char *b)
2119 const char *old = a;
2120 const char *new = b;
2121 struct strbuf name = STRBUF_INIT;
2122 int pfx_length, sfx_length;
2123 int pfx_adjust_for_slash;
2124 int len_a = strlen(a);
2125 int len_b = strlen(b);
2126 int a_midlen, b_midlen;
2127 int qlen_a = quote_c_style(a, NULL, NULL, 0);
2128 int qlen_b = quote_c_style(b, NULL, NULL, 0);
2130 if (qlen_a || qlen_b) {
2131 quote_c_style(a, &name, NULL, 0);
2132 strbuf_addstr(&name, " => ");
2133 quote_c_style(b, &name, NULL, 0);
2134 return strbuf_detach(&name, NULL);
2137 /* Find common prefix */
2138 pfx_length = 0;
2139 while (*old && *new && *old == *new) {
2140 if (*old == '/')
2141 pfx_length = old - a + 1;
2142 old++;
2143 new++;
2146 /* Find common suffix */
2147 old = a + len_a;
2148 new = b + len_b;
2149 sfx_length = 0;
2151 * If there is a common prefix, it must end in a slash. In
2152 * that case we let this loop run 1 into the prefix to see the
2153 * same slash.
2155 * If there is no common prefix, we cannot do this as it would
2156 * underrun the input strings.
2158 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
2159 while (a + pfx_length - pfx_adjust_for_slash <= old &&
2160 b + pfx_length - pfx_adjust_for_slash <= new &&
2161 *old == *new) {
2162 if (*old == '/')
2163 sfx_length = len_a - (old - a);
2164 old--;
2165 new--;
2169 * pfx{mid-a => mid-b}sfx
2170 * {pfx-a => pfx-b}sfx
2171 * pfx{sfx-a => sfx-b}
2172 * name-a => name-b
2174 a_midlen = len_a - pfx_length - sfx_length;
2175 b_midlen = len_b - pfx_length - sfx_length;
2176 if (a_midlen < 0)
2177 a_midlen = 0;
2178 if (b_midlen < 0)
2179 b_midlen = 0;
2181 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
2182 if (pfx_length + sfx_length) {
2183 strbuf_add(&name, a, pfx_length);
2184 strbuf_addch(&name, '{');
2186 strbuf_add(&name, a + pfx_length, a_midlen);
2187 strbuf_addstr(&name, " => ");
2188 strbuf_add(&name, b + pfx_length, b_midlen);
2189 if (pfx_length + sfx_length) {
2190 strbuf_addch(&name, '}');
2191 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
2193 return strbuf_detach(&name, NULL);
2196 struct diffstat_t {
2197 int nr;
2198 int alloc;
2199 struct diffstat_file {
2200 char *from_name;
2201 char *name;
2202 char *print_name;
2203 unsigned is_unmerged:1;
2204 unsigned is_binary:1;
2205 unsigned is_renamed:1;
2206 unsigned is_interesting:1;
2207 uintmax_t added, deleted;
2208 } **files;
2211 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
2212 const char *name_a,
2213 const char *name_b)
2215 struct diffstat_file *x;
2216 x = xcalloc(1, sizeof(*x));
2217 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
2218 diffstat->files[diffstat->nr++] = x;
2219 if (name_b) {
2220 x->from_name = xstrdup(name_a);
2221 x->name = xstrdup(name_b);
2222 x->is_renamed = 1;
2224 else {
2225 x->from_name = NULL;
2226 x->name = xstrdup(name_a);
2228 return x;
2231 static void diffstat_consume(void *priv, char *line, unsigned long len)
2233 struct diffstat_t *diffstat = priv;
2234 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
2236 if (line[0] == '+')
2237 x->added++;
2238 else if (line[0] == '-')
2239 x->deleted++;
2242 const char mime_boundary_leader[] = "------------";
2244 static int scale_linear(int it, int width, int max_change)
2246 if (!it)
2247 return 0;
2249 * make sure that at least one '-' or '+' is printed if
2250 * there is any change to this path. The easiest way is to
2251 * scale linearly as if the alloted width is one column shorter
2252 * than it is, and then add 1 to the result.
2254 return 1 + (it * (width - 1) / max_change);
2257 static void show_graph(struct strbuf *out, char ch, int cnt,
2258 const char *set, const char *reset)
2260 if (cnt <= 0)
2261 return;
2262 strbuf_addstr(out, set);
2263 strbuf_addchars(out, ch, cnt);
2264 strbuf_addstr(out, reset);
2267 static void fill_print_name(struct diffstat_file *file)
2269 char *pname;
2271 if (file->print_name)
2272 return;
2274 if (!file->is_renamed) {
2275 struct strbuf buf = STRBUF_INIT;
2276 if (quote_c_style(file->name, &buf, NULL, 0)) {
2277 pname = strbuf_detach(&buf, NULL);
2278 } else {
2279 pname = file->name;
2280 strbuf_release(&buf);
2282 } else {
2283 pname = pprint_rename(file->from_name, file->name);
2285 file->print_name = pname;
2288 static void print_stat_summary_inserts_deletes(struct diff_options *options,
2289 int files, int insertions, int deletions)
2291 struct strbuf sb = STRBUF_INIT;
2293 if (!files) {
2294 assert(insertions == 0 && deletions == 0);
2295 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
2296 NULL, 0, 0);
2297 return;
2300 strbuf_addf(&sb,
2301 (files == 1) ? " %d file changed" : " %d files changed",
2302 files);
2305 * For binary diff, the caller may want to print "x files
2306 * changed" with insertions == 0 && deletions == 0.
2308 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
2309 * is probably less confusing (i.e skip over "2 files changed
2310 * but nothing about added/removed lines? Is this a bug in Git?").
2312 if (insertions || deletions == 0) {
2313 strbuf_addf(&sb,
2314 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
2315 insertions);
2318 if (deletions || insertions == 0) {
2319 strbuf_addf(&sb,
2320 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
2321 deletions);
2323 strbuf_addch(&sb, '\n');
2324 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
2325 sb.buf, sb.len, 0);
2326 strbuf_release(&sb);
2329 void print_stat_summary(FILE *fp, int files,
2330 int insertions, int deletions)
2332 struct diff_options o;
2333 memset(&o, 0, sizeof(o));
2334 o.file = fp;
2336 print_stat_summary_inserts_deletes(&o, files, insertions, deletions);
2339 static void show_stats(struct diffstat_t *data, struct diff_options *options)
2341 int i, len, add, del, adds = 0, dels = 0;
2342 uintmax_t max_change = 0, max_len = 0;
2343 int total_files = data->nr, count;
2344 int width, name_width, graph_width, number_width = 0, bin_width = 0;
2345 const char *reset, *add_c, *del_c;
2346 int extra_shown = 0;
2347 const char *line_prefix = diff_line_prefix(options);
2348 struct strbuf out = STRBUF_INIT;
2350 if (data->nr == 0)
2351 return;
2353 count = options->stat_count ? options->stat_count : data->nr;
2355 reset = diff_get_color_opt(options, DIFF_RESET);
2356 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
2357 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
2360 * Find the longest filename and max number of changes
2362 for (i = 0; (i < count) && (i < data->nr); i++) {
2363 struct diffstat_file *file = data->files[i];
2364 uintmax_t change = file->added + file->deleted;
2366 if (!file->is_interesting && (change == 0)) {
2367 count++; /* not shown == room for one more */
2368 continue;
2370 fill_print_name(file);
2371 len = strlen(file->print_name);
2372 if (max_len < len)
2373 max_len = len;
2375 if (file->is_unmerged) {
2376 /* "Unmerged" is 8 characters */
2377 bin_width = bin_width < 8 ? 8 : bin_width;
2378 continue;
2380 if (file->is_binary) {
2381 /* "Bin XXX -> YYY bytes" */
2382 int w = 14 + decimal_width(file->added)
2383 + decimal_width(file->deleted);
2384 bin_width = bin_width < w ? w : bin_width;
2385 /* Display change counts aligned with "Bin" */
2386 number_width = 3;
2387 continue;
2390 if (max_change < change)
2391 max_change = change;
2393 count = i; /* where we can stop scanning in data->files[] */
2396 * We have width = stat_width or term_columns() columns total.
2397 * We want a maximum of min(max_len, stat_name_width) for the name part.
2398 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
2399 * We also need 1 for " " and 4 + decimal_width(max_change)
2400 * for " | NNNN " and one the empty column at the end, altogether
2401 * 6 + decimal_width(max_change).
2403 * If there's not enough space, we will use the smaller of
2404 * stat_name_width (if set) and 5/8*width for the filename,
2405 * and the rest for constant elements + graph part, but no more
2406 * than stat_graph_width for the graph part.
2407 * (5/8 gives 50 for filename and 30 for the constant parts + graph
2408 * for the standard terminal size).
2410 * In other words: stat_width limits the maximum width, and
2411 * stat_name_width fixes the maximum width of the filename,
2412 * and is also used to divide available columns if there
2413 * aren't enough.
2415 * Binary files are displayed with "Bin XXX -> YYY bytes"
2416 * instead of the change count and graph. This part is treated
2417 * similarly to the graph part, except that it is not
2418 * "scaled". If total width is too small to accommodate the
2419 * guaranteed minimum width of the filename part and the
2420 * separators and this message, this message will "overflow"
2421 * making the line longer than the maximum width.
2424 if (options->stat_width == -1)
2425 width = term_columns() - strlen(line_prefix);
2426 else
2427 width = options->stat_width ? options->stat_width : 80;
2428 number_width = decimal_width(max_change) > number_width ?
2429 decimal_width(max_change) : number_width;
2431 if (options->stat_graph_width == -1)
2432 options->stat_graph_width = diff_stat_graph_width;
2435 * Guarantee 3/8*16==6 for the graph part
2436 * and 5/8*16==10 for the filename part
2438 if (width < 16 + 6 + number_width)
2439 width = 16 + 6 + number_width;
2442 * First assign sizes that are wanted, ignoring available width.
2443 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
2444 * starting from "XXX" should fit in graph_width.
2446 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
2447 if (options->stat_graph_width &&
2448 options->stat_graph_width < graph_width)
2449 graph_width = options->stat_graph_width;
2451 name_width = (options->stat_name_width > 0 &&
2452 options->stat_name_width < max_len) ?
2453 options->stat_name_width : max_len;
2456 * Adjust adjustable widths not to exceed maximum width
2458 if (name_width + number_width + 6 + graph_width > width) {
2459 if (graph_width > width * 3/8 - number_width - 6) {
2460 graph_width = width * 3/8 - number_width - 6;
2461 if (graph_width < 6)
2462 graph_width = 6;
2465 if (options->stat_graph_width &&
2466 graph_width > options->stat_graph_width)
2467 graph_width = options->stat_graph_width;
2468 if (name_width > width - number_width - 6 - graph_width)
2469 name_width = width - number_width - 6 - graph_width;
2470 else
2471 graph_width = width - number_width - 6 - name_width;
2475 * From here name_width is the width of the name area,
2476 * and graph_width is the width of the graph area.
2477 * max_change is used to scale graph properly.
2479 for (i = 0; i < count; i++) {
2480 const char *prefix = "";
2481 struct diffstat_file *file = data->files[i];
2482 char *name = file->print_name;
2483 uintmax_t added = file->added;
2484 uintmax_t deleted = file->deleted;
2485 int name_len;
2487 if (!file->is_interesting && (added + deleted == 0))
2488 continue;
2491 * "scale" the filename
2493 len = name_width;
2494 name_len = strlen(name);
2495 if (name_width < name_len) {
2496 char *slash;
2497 prefix = "...";
2498 len -= 3;
2499 name += name_len - len;
2500 slash = strchr(name, '/');
2501 if (slash)
2502 name = slash;
2505 if (file->is_binary) {
2506 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2507 strbuf_addf(&out, " %*s", number_width, "Bin");
2508 if (!added && !deleted) {
2509 strbuf_addch(&out, '\n');
2510 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2511 out.buf, out.len, 0);
2512 strbuf_reset(&out);
2513 continue;
2515 strbuf_addf(&out, " %s%"PRIuMAX"%s",
2516 del_c, deleted, reset);
2517 strbuf_addstr(&out, " -> ");
2518 strbuf_addf(&out, "%s%"PRIuMAX"%s",
2519 add_c, added, reset);
2520 strbuf_addstr(&out, " bytes\n");
2521 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2522 out.buf, out.len, 0);
2523 strbuf_reset(&out);
2524 continue;
2526 else if (file->is_unmerged) {
2527 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2528 strbuf_addstr(&out, " Unmerged\n");
2529 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2530 out.buf, out.len, 0);
2531 strbuf_reset(&out);
2532 continue;
2536 * scale the add/delete
2538 add = added;
2539 del = deleted;
2541 if (graph_width <= max_change) {
2542 int total = scale_linear(add + del, graph_width, max_change);
2543 if (total < 2 && add && del)
2544 /* width >= 2 due to the sanity check */
2545 total = 2;
2546 if (add < del) {
2547 add = scale_linear(add, graph_width, max_change);
2548 del = total - add;
2549 } else {
2550 del = scale_linear(del, graph_width, max_change);
2551 add = total - del;
2554 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2555 strbuf_addf(&out, " %*"PRIuMAX"%s",
2556 number_width, added + deleted,
2557 added + deleted ? " " : "");
2558 show_graph(&out, '+', add, add_c, reset);
2559 show_graph(&out, '-', del, del_c, reset);
2560 strbuf_addch(&out, '\n');
2561 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2562 out.buf, out.len, 0);
2563 strbuf_reset(&out);
2566 for (i = 0; i < data->nr; i++) {
2567 struct diffstat_file *file = data->files[i];
2568 uintmax_t added = file->added;
2569 uintmax_t deleted = file->deleted;
2571 if (file->is_unmerged ||
2572 (!file->is_interesting && (added + deleted == 0))) {
2573 total_files--;
2574 continue;
2577 if (!file->is_binary) {
2578 adds += added;
2579 dels += deleted;
2581 if (i < count)
2582 continue;
2583 if (!extra_shown)
2584 emit_diff_symbol(options,
2585 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
2586 NULL, 0, 0);
2587 extra_shown = 1;
2590 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2591 strbuf_release(&out);
2594 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
2596 int i, adds = 0, dels = 0, total_files = data->nr;
2598 if (data->nr == 0)
2599 return;
2601 for (i = 0; i < data->nr; i++) {
2602 int added = data->files[i]->added;
2603 int deleted = data->files[i]->deleted;
2605 if (data->files[i]->is_unmerged ||
2606 (!data->files[i]->is_interesting && (added + deleted == 0))) {
2607 total_files--;
2608 } else if (!data->files[i]->is_binary) { /* don't count bytes */
2609 adds += added;
2610 dels += deleted;
2613 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2616 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
2618 int i;
2620 if (data->nr == 0)
2621 return;
2623 for (i = 0; i < data->nr; i++) {
2624 struct diffstat_file *file = data->files[i];
2626 fprintf(options->file, "%s", diff_line_prefix(options));
2628 if (file->is_binary)
2629 fprintf(options->file, "-\t-\t");
2630 else
2631 fprintf(options->file,
2632 "%"PRIuMAX"\t%"PRIuMAX"\t",
2633 file->added, file->deleted);
2634 if (options->line_termination) {
2635 fill_print_name(file);
2636 if (!file->is_renamed)
2637 write_name_quoted(file->name, options->file,
2638 options->line_termination);
2639 else {
2640 fputs(file->print_name, options->file);
2641 putc(options->line_termination, options->file);
2643 } else {
2644 if (file->is_renamed) {
2645 putc('\0', options->file);
2646 write_name_quoted(file->from_name, options->file, '\0');
2648 write_name_quoted(file->name, options->file, '\0');
2653 struct dirstat_file {
2654 const char *name;
2655 unsigned long changed;
2658 struct dirstat_dir {
2659 struct dirstat_file *files;
2660 int alloc, nr, permille, cumulative;
2663 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2664 unsigned long changed, const char *base, int baselen)
2666 unsigned long this_dir = 0;
2667 unsigned int sources = 0;
2668 const char *line_prefix = diff_line_prefix(opt);
2670 while (dir->nr) {
2671 struct dirstat_file *f = dir->files;
2672 int namelen = strlen(f->name);
2673 unsigned long this;
2674 char *slash;
2676 if (namelen < baselen)
2677 break;
2678 if (memcmp(f->name, base, baselen))
2679 break;
2680 slash = strchr(f->name + baselen, '/');
2681 if (slash) {
2682 int newbaselen = slash + 1 - f->name;
2683 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2684 sources++;
2685 } else {
2686 this = f->changed;
2687 dir->files++;
2688 dir->nr--;
2689 sources += 2;
2691 this_dir += this;
2695 * We don't report dirstat's for
2696 * - the top level
2697 * - or cases where everything came from a single directory
2698 * under this directory (sources == 1).
2700 if (baselen && sources != 1) {
2701 if (this_dir) {
2702 int permille = this_dir * 1000 / changed;
2703 if (permille >= dir->permille) {
2704 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2705 permille / 10, permille % 10, baselen, base);
2706 if (!dir->cumulative)
2707 return 0;
2711 return this_dir;
2714 static int dirstat_compare(const void *_a, const void *_b)
2716 const struct dirstat_file *a = _a;
2717 const struct dirstat_file *b = _b;
2718 return strcmp(a->name, b->name);
2721 static void show_dirstat(struct diff_options *options)
2723 int i;
2724 unsigned long changed;
2725 struct dirstat_dir dir;
2726 struct diff_queue_struct *q = &diff_queued_diff;
2728 dir.files = NULL;
2729 dir.alloc = 0;
2730 dir.nr = 0;
2731 dir.permille = options->dirstat_permille;
2732 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2734 changed = 0;
2735 for (i = 0; i < q->nr; i++) {
2736 struct diff_filepair *p = q->queue[i];
2737 const char *name;
2738 unsigned long copied, added, damage;
2739 int content_changed;
2741 name = p->two->path ? p->two->path : p->one->path;
2743 if (p->one->oid_valid && p->two->oid_valid)
2744 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2745 else
2746 content_changed = 1;
2748 if (!content_changed) {
2750 * The SHA1 has not changed, so pre-/post-content is
2751 * identical. We can therefore skip looking at the
2752 * file contents altogether.
2754 damage = 0;
2755 goto found_damage;
2758 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2760 * In --dirstat-by-file mode, we don't really need to
2761 * look at the actual file contents at all.
2762 * The fact that the SHA1 changed is enough for us to
2763 * add this file to the list of results
2764 * (with each file contributing equal damage).
2766 damage = 1;
2767 goto found_damage;
2770 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2771 diff_populate_filespec(p->one, 0);
2772 diff_populate_filespec(p->two, 0);
2773 diffcore_count_changes(p->one, p->two, NULL, NULL,
2774 &copied, &added);
2775 diff_free_filespec_data(p->one);
2776 diff_free_filespec_data(p->two);
2777 } else if (DIFF_FILE_VALID(p->one)) {
2778 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2779 copied = added = 0;
2780 diff_free_filespec_data(p->one);
2781 } else if (DIFF_FILE_VALID(p->two)) {
2782 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2783 copied = 0;
2784 added = p->two->size;
2785 diff_free_filespec_data(p->two);
2786 } else
2787 continue;
2790 * Original minus copied is the removed material,
2791 * added is the new material. They are both damages
2792 * made to the preimage.
2793 * If the resulting damage is zero, we know that
2794 * diffcore_count_changes() considers the two entries to
2795 * be identical, but since content_changed is true, we
2796 * know that there must have been _some_ kind of change,
2797 * so we force all entries to have damage > 0.
2799 damage = (p->one->size - copied) + added;
2800 if (!damage)
2801 damage = 1;
2803 found_damage:
2804 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2805 dir.files[dir.nr].name = name;
2806 dir.files[dir.nr].changed = damage;
2807 changed += damage;
2808 dir.nr++;
2811 /* This can happen even with many files, if everything was renames */
2812 if (!changed)
2813 return;
2815 /* Show all directories with more than x% of the changes */
2816 QSORT(dir.files, dir.nr, dirstat_compare);
2817 gather_dirstat(options, &dir, changed, "", 0);
2820 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2822 int i;
2823 unsigned long changed;
2824 struct dirstat_dir dir;
2826 if (data->nr == 0)
2827 return;
2829 dir.files = NULL;
2830 dir.alloc = 0;
2831 dir.nr = 0;
2832 dir.permille = options->dirstat_permille;
2833 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2835 changed = 0;
2836 for (i = 0; i < data->nr; i++) {
2837 struct diffstat_file *file = data->files[i];
2838 unsigned long damage = file->added + file->deleted;
2839 if (file->is_binary)
2841 * binary files counts bytes, not lines. Must find some
2842 * way to normalize binary bytes vs. textual lines.
2843 * The following heuristic assumes that there are 64
2844 * bytes per "line".
2845 * This is stupid and ugly, but very cheap...
2847 damage = DIV_ROUND_UP(damage, 64);
2848 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2849 dir.files[dir.nr].name = file->name;
2850 dir.files[dir.nr].changed = damage;
2851 changed += damage;
2852 dir.nr++;
2855 /* This can happen even with many files, if everything was renames */
2856 if (!changed)
2857 return;
2859 /* Show all directories with more than x% of the changes */
2860 QSORT(dir.files, dir.nr, dirstat_compare);
2861 gather_dirstat(options, &dir, changed, "", 0);
2864 static void free_diffstat_info(struct diffstat_t *diffstat)
2866 int i;
2867 for (i = 0; i < diffstat->nr; i++) {
2868 struct diffstat_file *f = diffstat->files[i];
2869 if (f->name != f->print_name)
2870 free(f->print_name);
2871 free(f->name);
2872 free(f->from_name);
2873 free(f);
2875 free(diffstat->files);
2878 struct checkdiff_t {
2879 const char *filename;
2880 int lineno;
2881 int conflict_marker_size;
2882 struct diff_options *o;
2883 unsigned ws_rule;
2884 unsigned status;
2887 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2889 char firstchar;
2890 int cnt;
2892 if (len < marker_size + 1)
2893 return 0;
2894 firstchar = line[0];
2895 switch (firstchar) {
2896 case '=': case '>': case '<': case '|':
2897 break;
2898 default:
2899 return 0;
2901 for (cnt = 1; cnt < marker_size; cnt++)
2902 if (line[cnt] != firstchar)
2903 return 0;
2904 /* line[1] thru line[marker_size-1] are same as firstchar */
2905 if (len < marker_size + 1 || !isspace(line[marker_size]))
2906 return 0;
2907 return 1;
2910 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2912 struct checkdiff_t *data = priv;
2913 int marker_size = data->conflict_marker_size;
2914 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2915 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2916 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2917 char *err;
2918 const char *line_prefix;
2920 assert(data->o);
2921 line_prefix = diff_line_prefix(data->o);
2923 if (line[0] == '+') {
2924 unsigned bad;
2925 data->lineno++;
2926 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2927 data->status |= 1;
2928 fprintf(data->o->file,
2929 "%s%s:%d: leftover conflict marker\n",
2930 line_prefix, data->filename, data->lineno);
2932 bad = ws_check(line + 1, len - 1, data->ws_rule);
2933 if (!bad)
2934 return;
2935 data->status |= bad;
2936 err = whitespace_error_string(bad);
2937 fprintf(data->o->file, "%s%s:%d: %s.\n",
2938 line_prefix, data->filename, data->lineno, err);
2939 free(err);
2940 emit_line(data->o, set, reset, line, 1);
2941 ws_check_emit(line + 1, len - 1, data->ws_rule,
2942 data->o->file, set, reset, ws);
2943 } else if (line[0] == ' ') {
2944 data->lineno++;
2945 } else if (line[0] == '@') {
2946 char *plus = strchr(line, '+');
2947 if (plus)
2948 data->lineno = strtol(plus, NULL, 10) - 1;
2949 else
2950 die("invalid diff");
2954 static unsigned char *deflate_it(char *data,
2955 unsigned long size,
2956 unsigned long *result_size)
2958 int bound;
2959 unsigned char *deflated;
2960 git_zstream stream;
2962 git_deflate_init(&stream, zlib_compression_level);
2963 bound = git_deflate_bound(&stream, size);
2964 deflated = xmalloc(bound);
2965 stream.next_out = deflated;
2966 stream.avail_out = bound;
2968 stream.next_in = (unsigned char *)data;
2969 stream.avail_in = size;
2970 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2971 ; /* nothing */
2972 git_deflate_end(&stream);
2973 *result_size = stream.total_out;
2974 return deflated;
2977 static void emit_binary_diff_body(struct diff_options *o,
2978 mmfile_t *one, mmfile_t *two)
2980 void *cp;
2981 void *delta;
2982 void *deflated;
2983 void *data;
2984 unsigned long orig_size;
2985 unsigned long delta_size;
2986 unsigned long deflate_size;
2987 unsigned long data_size;
2989 /* We could do deflated delta, or we could do just deflated two,
2990 * whichever is smaller.
2992 delta = NULL;
2993 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2994 if (one->size && two->size) {
2995 delta = diff_delta(one->ptr, one->size,
2996 two->ptr, two->size,
2997 &delta_size, deflate_size);
2998 if (delta) {
2999 void *to_free = delta;
3000 orig_size = delta_size;
3001 delta = deflate_it(delta, delta_size, &delta_size);
3002 free(to_free);
3006 if (delta && delta_size < deflate_size) {
3007 char *s = xstrfmt("%lu", orig_size);
3008 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
3009 s, strlen(s), 0);
3010 free(s);
3011 free(deflated);
3012 data = delta;
3013 data_size = delta_size;
3014 } else {
3015 char *s = xstrfmt("%lu", two->size);
3016 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
3017 s, strlen(s), 0);
3018 free(s);
3019 free(delta);
3020 data = deflated;
3021 data_size = deflate_size;
3024 /* emit data encoded in base85 */
3025 cp = data;
3026 while (data_size) {
3027 int len;
3028 int bytes = (52 < data_size) ? 52 : data_size;
3029 char line[71];
3030 data_size -= bytes;
3031 if (bytes <= 26)
3032 line[0] = bytes + 'A' - 1;
3033 else
3034 line[0] = bytes - 26 + 'a' - 1;
3035 encode_85(line + 1, cp, bytes);
3036 cp = (char *) cp + bytes;
3038 len = strlen(line);
3039 line[len++] = '\n';
3040 line[len] = '\0';
3042 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_BODY,
3043 line, len, 0);
3045 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_FOOTER, NULL, 0, 0);
3046 free(data);
3049 static void emit_binary_diff(struct diff_options *o,
3050 mmfile_t *one, mmfile_t *two)
3052 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER, NULL, 0, 0);
3053 emit_binary_diff_body(o, one, two);
3054 emit_binary_diff_body(o, two, one);
3057 int diff_filespec_is_binary(struct diff_filespec *one)
3059 if (one->is_binary == -1) {
3060 diff_filespec_load_driver(one);
3061 if (one->driver->binary != -1)
3062 one->is_binary = one->driver->binary;
3063 else {
3064 if (!one->data && DIFF_FILE_VALID(one))
3065 diff_populate_filespec(one, CHECK_BINARY);
3066 if (one->is_binary == -1 && one->data)
3067 one->is_binary = buffer_is_binary(one->data,
3068 one->size);
3069 if (one->is_binary == -1)
3070 one->is_binary = 0;
3073 return one->is_binary;
3076 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
3078 diff_filespec_load_driver(one);
3079 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
3082 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
3084 if (!options->a_prefix)
3085 options->a_prefix = a;
3086 if (!options->b_prefix)
3087 options->b_prefix = b;
3090 struct userdiff_driver *get_textconv(struct diff_filespec *one)
3092 if (!DIFF_FILE_VALID(one))
3093 return NULL;
3095 diff_filespec_load_driver(one);
3096 return userdiff_get_textconv(one->driver);
3099 static void builtin_diff(const char *name_a,
3100 const char *name_b,
3101 struct diff_filespec *one,
3102 struct diff_filespec *two,
3103 const char *xfrm_msg,
3104 int must_show_header,
3105 struct diff_options *o,
3106 int complete_rewrite)
3108 mmfile_t mf1, mf2;
3109 const char *lbl[2];
3110 char *a_one, *b_two;
3111 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
3112 const char *reset = diff_get_color_opt(o, DIFF_RESET);
3113 const char *a_prefix, *b_prefix;
3114 struct userdiff_driver *textconv_one = NULL;
3115 struct userdiff_driver *textconv_two = NULL;
3116 struct strbuf header = STRBUF_INIT;
3117 const char *line_prefix = diff_line_prefix(o);
3119 diff_set_mnemonic_prefix(o, "a/", "b/");
3120 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
3121 a_prefix = o->b_prefix;
3122 b_prefix = o->a_prefix;
3123 } else {
3124 a_prefix = o->a_prefix;
3125 b_prefix = o->b_prefix;
3128 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
3129 (!one->mode || S_ISGITLINK(one->mode)) &&
3130 (!two->mode || S_ISGITLINK(two->mode))) {
3131 show_submodule_summary(o, one->path ? one->path : two->path,
3132 &one->oid, &two->oid,
3133 two->dirty_submodule);
3134 return;
3135 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
3136 (!one->mode || S_ISGITLINK(one->mode)) &&
3137 (!two->mode || S_ISGITLINK(two->mode))) {
3138 show_submodule_inline_diff(o, one->path ? one->path : two->path,
3139 &one->oid, &two->oid,
3140 two->dirty_submodule);
3141 return;
3144 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
3145 textconv_one = get_textconv(one);
3146 textconv_two = get_textconv(two);
3149 /* Never use a non-valid filename anywhere if at all possible */
3150 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
3151 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
3153 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
3154 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
3155 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
3156 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
3157 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
3158 if (lbl[0][0] == '/') {
3159 /* /dev/null */
3160 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
3161 if (xfrm_msg)
3162 strbuf_addstr(&header, xfrm_msg);
3163 must_show_header = 1;
3165 else if (lbl[1][0] == '/') {
3166 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
3167 if (xfrm_msg)
3168 strbuf_addstr(&header, xfrm_msg);
3169 must_show_header = 1;
3171 else {
3172 if (one->mode != two->mode) {
3173 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
3174 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
3175 must_show_header = 1;
3177 if (xfrm_msg)
3178 strbuf_addstr(&header, xfrm_msg);
3181 * we do not run diff between different kind
3182 * of objects.
3184 if ((one->mode ^ two->mode) & S_IFMT)
3185 goto free_ab_and_return;
3186 if (complete_rewrite &&
3187 (textconv_one || !diff_filespec_is_binary(one)) &&
3188 (textconv_two || !diff_filespec_is_binary(two))) {
3189 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3190 header.buf, header.len, 0);
3191 strbuf_reset(&header);
3192 emit_rewrite_diff(name_a, name_b, one, two,
3193 textconv_one, textconv_two, o);
3194 o->found_changes = 1;
3195 goto free_ab_and_return;
3199 if (o->irreversible_delete && lbl[1][0] == '/') {
3200 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
3201 header.len, 0);
3202 strbuf_reset(&header);
3203 goto free_ab_and_return;
3204 } else if (!DIFF_OPT_TST(o, TEXT) &&
3205 ( (!textconv_one && diff_filespec_is_binary(one)) ||
3206 (!textconv_two && diff_filespec_is_binary(two)) )) {
3207 struct strbuf sb = STRBUF_INIT;
3208 if (!one->data && !two->data &&
3209 S_ISREG(one->mode) && S_ISREG(two->mode) &&
3210 !DIFF_OPT_TST(o, BINARY)) {
3211 if (!oidcmp(&one->oid, &two->oid)) {
3212 if (must_show_header)
3213 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3214 header.buf, header.len,
3216 goto free_ab_and_return;
3218 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3219 header.buf, header.len, 0);
3220 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3221 diff_line_prefix(o), lbl[0], lbl[1]);
3222 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3223 sb.buf, sb.len, 0);
3224 strbuf_release(&sb);
3225 goto free_ab_and_return;
3227 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3228 die("unable to read files to diff");
3229 /* Quite common confusing case */
3230 if (mf1.size == mf2.size &&
3231 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
3232 if (must_show_header)
3233 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3234 header.buf, header.len, 0);
3235 goto free_ab_and_return;
3237 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
3238 strbuf_reset(&header);
3239 if (DIFF_OPT_TST(o, BINARY))
3240 emit_binary_diff(o, &mf1, &mf2);
3241 else {
3242 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3243 diff_line_prefix(o), lbl[0], lbl[1]);
3244 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3245 sb.buf, sb.len, 0);
3246 strbuf_release(&sb);
3248 o->found_changes = 1;
3249 } else {
3250 /* Crazy xdl interfaces.. */
3251 const char *diffopts = getenv("GIT_DIFF_OPTS");
3252 const char *v;
3253 xpparam_t xpp;
3254 xdemitconf_t xecfg;
3255 struct emit_callback ecbdata;
3256 const struct userdiff_funcname *pe;
3258 if (must_show_header) {
3259 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3260 header.buf, header.len, 0);
3261 strbuf_reset(&header);
3264 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
3265 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
3267 pe = diff_funcname_pattern(one);
3268 if (!pe)
3269 pe = diff_funcname_pattern(two);
3271 memset(&xpp, 0, sizeof(xpp));
3272 memset(&xecfg, 0, sizeof(xecfg));
3273 memset(&ecbdata, 0, sizeof(ecbdata));
3274 ecbdata.label_path = lbl;
3275 ecbdata.color_diff = want_color(o->use_color);
3276 ecbdata.ws_rule = whitespace_rule(name_b);
3277 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
3278 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3279 ecbdata.opt = o;
3280 ecbdata.header = header.len ? &header : NULL;
3281 xpp.flags = o->xdl_opts;
3282 xecfg.ctxlen = o->context;
3283 xecfg.interhunkctxlen = o->interhunkcontext;
3284 xecfg.flags = XDL_EMIT_FUNCNAMES;
3285 if (DIFF_OPT_TST(o, FUNCCONTEXT))
3286 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
3287 if (pe)
3288 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
3289 if (!diffopts)
3291 else if (skip_prefix(diffopts, "--unified=", &v))
3292 xecfg.ctxlen = strtoul(v, NULL, 10);
3293 else if (skip_prefix(diffopts, "-u", &v))
3294 xecfg.ctxlen = strtoul(v, NULL, 10);
3295 if (o->word_diff)
3296 init_diff_words_data(&ecbdata, o, one, two);
3297 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
3298 &xpp, &xecfg))
3299 die("unable to generate diff for %s", one->path);
3300 if (o->word_diff)
3301 free_diff_words_data(&ecbdata);
3302 if (textconv_one)
3303 free(mf1.ptr);
3304 if (textconv_two)
3305 free(mf2.ptr);
3306 xdiff_clear_find_func(&xecfg);
3309 free_ab_and_return:
3310 strbuf_release(&header);
3311 diff_free_filespec_data(one);
3312 diff_free_filespec_data(two);
3313 free(a_one);
3314 free(b_two);
3315 return;
3318 static void builtin_diffstat(const char *name_a, const char *name_b,
3319 struct diff_filespec *one,
3320 struct diff_filespec *two,
3321 struct diffstat_t *diffstat,
3322 struct diff_options *o,
3323 struct diff_filepair *p)
3325 mmfile_t mf1, mf2;
3326 struct diffstat_file *data;
3327 int same_contents;
3328 int complete_rewrite = 0;
3330 if (!DIFF_PAIR_UNMERGED(p)) {
3331 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3332 complete_rewrite = 1;
3335 data = diffstat_add(diffstat, name_a, name_b);
3336 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
3338 if (!one || !two) {
3339 data->is_unmerged = 1;
3340 return;
3343 same_contents = !oidcmp(&one->oid, &two->oid);
3345 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
3346 data->is_binary = 1;
3347 if (same_contents) {
3348 data->added = 0;
3349 data->deleted = 0;
3350 } else {
3351 data->added = diff_filespec_size(two);
3352 data->deleted = diff_filespec_size(one);
3356 else if (complete_rewrite) {
3357 diff_populate_filespec(one, 0);
3358 diff_populate_filespec(two, 0);
3359 data->deleted = count_lines(one->data, one->size);
3360 data->added = count_lines(two->data, two->size);
3363 else if (!same_contents) {
3364 /* Crazy xdl interfaces.. */
3365 xpparam_t xpp;
3366 xdemitconf_t xecfg;
3368 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3369 die("unable to read files to diff");
3371 memset(&xpp, 0, sizeof(xpp));
3372 memset(&xecfg, 0, sizeof(xecfg));
3373 xpp.flags = o->xdl_opts;
3374 xecfg.ctxlen = o->context;
3375 xecfg.interhunkctxlen = o->interhunkcontext;
3376 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
3377 &xpp, &xecfg))
3378 die("unable to generate diffstat for %s", one->path);
3381 diff_free_filespec_data(one);
3382 diff_free_filespec_data(two);
3385 static void builtin_checkdiff(const char *name_a, const char *name_b,
3386 const char *attr_path,
3387 struct diff_filespec *one,
3388 struct diff_filespec *two,
3389 struct diff_options *o)
3391 mmfile_t mf1, mf2;
3392 struct checkdiff_t data;
3394 if (!two)
3395 return;
3397 memset(&data, 0, sizeof(data));
3398 data.filename = name_b ? name_b : name_a;
3399 data.lineno = 0;
3400 data.o = o;
3401 data.ws_rule = whitespace_rule(attr_path);
3402 data.conflict_marker_size = ll_merge_marker_size(attr_path);
3404 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3405 die("unable to read files to diff");
3408 * All the other codepaths check both sides, but not checking
3409 * the "old" side here is deliberate. We are checking the newly
3410 * introduced changes, and as long as the "new" side is text, we
3411 * can and should check what it introduces.
3413 if (diff_filespec_is_binary(two))
3414 goto free_and_return;
3415 else {
3416 /* Crazy xdl interfaces.. */
3417 xpparam_t xpp;
3418 xdemitconf_t xecfg;
3420 memset(&xpp, 0, sizeof(xpp));
3421 memset(&xecfg, 0, sizeof(xecfg));
3422 xecfg.ctxlen = 1; /* at least one context line */
3423 xpp.flags = 0;
3424 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
3425 &xpp, &xecfg))
3426 die("unable to generate checkdiff for %s", one->path);
3428 if (data.ws_rule & WS_BLANK_AT_EOF) {
3429 struct emit_callback ecbdata;
3430 int blank_at_eof;
3432 ecbdata.ws_rule = data.ws_rule;
3433 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3434 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
3436 if (blank_at_eof) {
3437 static char *err;
3438 if (!err)
3439 err = whitespace_error_string(WS_BLANK_AT_EOF);
3440 fprintf(o->file, "%s:%d: %s.\n",
3441 data.filename, blank_at_eof, err);
3442 data.status = 1; /* report errors */
3446 free_and_return:
3447 diff_free_filespec_data(one);
3448 diff_free_filespec_data(two);
3449 if (data.status)
3450 DIFF_OPT_SET(o, CHECK_FAILED);
3453 struct diff_filespec *alloc_filespec(const char *path)
3455 struct diff_filespec *spec;
3457 FLEXPTR_ALLOC_STR(spec, path, path);
3458 spec->count = 1;
3459 spec->is_binary = -1;
3460 return spec;
3463 void free_filespec(struct diff_filespec *spec)
3465 if (!--spec->count) {
3466 diff_free_filespec_data(spec);
3467 free(spec);
3471 void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
3472 int oid_valid, unsigned short mode)
3474 if (mode) {
3475 spec->mode = canon_mode(mode);
3476 oidcpy(&spec->oid, oid);
3477 spec->oid_valid = oid_valid;
3482 * Given a name and sha1 pair, if the index tells us the file in
3483 * the work tree has that object contents, return true, so that
3484 * prepare_temp_file() does not have to inflate and extract.
3486 static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
3488 const struct cache_entry *ce;
3489 struct stat st;
3490 int pos, len;
3493 * We do not read the cache ourselves here, because the
3494 * benchmark with my previous version that always reads cache
3495 * shows that it makes things worse for diff-tree comparing
3496 * two linux-2.6 kernel trees in an already checked out work
3497 * tree. This is because most diff-tree comparisons deal with
3498 * only a small number of files, while reading the cache is
3499 * expensive for a large project, and its cost outweighs the
3500 * savings we get by not inflating the object to a temporary
3501 * file. Practically, this code only helps when we are used
3502 * by diff-cache --cached, which does read the cache before
3503 * calling us.
3505 if (!active_cache)
3506 return 0;
3508 /* We want to avoid the working directory if our caller
3509 * doesn't need the data in a normal file, this system
3510 * is rather slow with its stat/open/mmap/close syscalls,
3511 * and the object is contained in a pack file. The pack
3512 * is probably already open and will be faster to obtain
3513 * the data through than the working directory. Loose
3514 * objects however would tend to be slower as they need
3515 * to be individually opened and inflated.
3517 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
3518 return 0;
3521 * Similarly, if we'd have to convert the file contents anyway, that
3522 * makes the optimization not worthwhile.
3524 if (!want_file && would_convert_to_git(&the_index, name))
3525 return 0;
3527 len = strlen(name);
3528 pos = cache_name_pos(name, len);
3529 if (pos < 0)
3530 return 0;
3531 ce = active_cache[pos];
3534 * This is not the sha1 we are looking for, or
3535 * unreusable because it is not a regular file.
3537 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
3538 return 0;
3541 * If ce is marked as "assume unchanged", there is no
3542 * guarantee that work tree matches what we are looking for.
3544 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
3545 return 0;
3548 * If ce matches the file in the work tree, we can reuse it.
3550 if (ce_uptodate(ce) ||
3551 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
3552 return 1;
3554 return 0;
3557 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
3559 struct strbuf buf = STRBUF_INIT;
3560 char *dirty = "";
3562 /* Are we looking at the work tree? */
3563 if (s->dirty_submodule)
3564 dirty = "-dirty";
3566 strbuf_addf(&buf, "Subproject commit %s%s\n",
3567 oid_to_hex(&s->oid), dirty);
3568 s->size = buf.len;
3569 if (size_only) {
3570 s->data = NULL;
3571 strbuf_release(&buf);
3572 } else {
3573 s->data = strbuf_detach(&buf, NULL);
3574 s->should_free = 1;
3576 return 0;
3580 * While doing rename detection and pickaxe operation, we may need to
3581 * grab the data for the blob (or file) for our own in-core comparison.
3582 * diff_filespec has data and size fields for this purpose.
3584 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
3586 int size_only = flags & CHECK_SIZE_ONLY;
3587 int err = 0;
3589 * demote FAIL to WARN to allow inspecting the situation
3590 * instead of refusing.
3592 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
3593 ? SAFE_CRLF_WARN
3594 : safe_crlf);
3596 if (!DIFF_FILE_VALID(s))
3597 die("internal error: asking to populate invalid file.");
3598 if (S_ISDIR(s->mode))
3599 return -1;
3601 if (s->data)
3602 return 0;
3604 if (size_only && 0 < s->size)
3605 return 0;
3607 if (S_ISGITLINK(s->mode))
3608 return diff_populate_gitlink(s, size_only);
3610 if (!s->oid_valid ||
3611 reuse_worktree_file(s->path, &s->oid, 0)) {
3612 struct strbuf buf = STRBUF_INIT;
3613 struct stat st;
3614 int fd;
3616 if (lstat(s->path, &st) < 0) {
3617 if (errno == ENOENT) {
3618 err_empty:
3619 err = -1;
3620 empty:
3621 s->data = (char *)"";
3622 s->size = 0;
3623 return err;
3626 s->size = xsize_t(st.st_size);
3627 if (!s->size)
3628 goto empty;
3629 if (S_ISLNK(st.st_mode)) {
3630 struct strbuf sb = STRBUF_INIT;
3632 if (strbuf_readlink(&sb, s->path, s->size))
3633 goto err_empty;
3634 s->size = sb.len;
3635 s->data = strbuf_detach(&sb, NULL);
3636 s->should_free = 1;
3637 return 0;
3641 * Even if the caller would be happy with getting
3642 * only the size, we cannot return early at this
3643 * point if the path requires us to run the content
3644 * conversion.
3646 if (size_only && !would_convert_to_git(&the_index, s->path))
3647 return 0;
3650 * Note: this check uses xsize_t(st.st_size) that may
3651 * not be the true size of the blob after it goes
3652 * through convert_to_git(). This may not strictly be
3653 * correct, but the whole point of big_file_threshold
3654 * and is_binary check being that we want to avoid
3655 * opening the file and inspecting the contents, this
3656 * is probably fine.
3658 if ((flags & CHECK_BINARY) &&
3659 s->size > big_file_threshold && s->is_binary == -1) {
3660 s->is_binary = 1;
3661 return 0;
3663 fd = open(s->path, O_RDONLY);
3664 if (fd < 0)
3665 goto err_empty;
3666 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
3667 close(fd);
3668 s->should_munmap = 1;
3671 * Convert from working tree format to canonical git format
3673 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
3674 size_t size = 0;
3675 munmap(s->data, s->size);
3676 s->should_munmap = 0;
3677 s->data = strbuf_detach(&buf, &size);
3678 s->size = size;
3679 s->should_free = 1;
3682 else {
3683 enum object_type type;
3684 if (size_only || (flags & CHECK_BINARY)) {
3685 type = sha1_object_info(s->oid.hash, &s->size);
3686 if (type < 0)
3687 die("unable to read %s",
3688 oid_to_hex(&s->oid));
3689 if (size_only)
3690 return 0;
3691 if (s->size > big_file_threshold && s->is_binary == -1) {
3692 s->is_binary = 1;
3693 return 0;
3696 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
3697 if (!s->data)
3698 die("unable to read %s", oid_to_hex(&s->oid));
3699 s->should_free = 1;
3701 return 0;
3704 void diff_free_filespec_blob(struct diff_filespec *s)
3706 if (s->should_free)
3707 free(s->data);
3708 else if (s->should_munmap)
3709 munmap(s->data, s->size);
3711 if (s->should_free || s->should_munmap) {
3712 s->should_free = s->should_munmap = 0;
3713 s->data = NULL;
3717 void diff_free_filespec_data(struct diff_filespec *s)
3719 diff_free_filespec_blob(s);
3720 FREE_AND_NULL(s->cnt_data);
3723 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
3724 void *blob,
3725 unsigned long size,
3726 const struct object_id *oid,
3727 int mode)
3729 struct strbuf buf = STRBUF_INIT;
3730 struct strbuf template = STRBUF_INIT;
3731 char *path_dup = xstrdup(path);
3732 const char *base = basename(path_dup);
3734 /* Generate "XXXXXX_basename.ext" */
3735 strbuf_addstr(&template, "XXXXXX_");
3736 strbuf_addstr(&template, base);
3738 temp->tempfile = mks_tempfile_ts(template.buf, strlen(base) + 1);
3739 if (!temp->tempfile)
3740 die_errno("unable to create temp-file");
3741 if (convert_to_working_tree(path,
3742 (const char *)blob, (size_t)size, &buf)) {
3743 blob = buf.buf;
3744 size = buf.len;
3746 if (write_in_full(temp->tempfile->fd, blob, size) < 0 ||
3747 close_tempfile_gently(temp->tempfile))
3748 die_errno("unable to write temp-file");
3749 temp->name = get_tempfile_path(temp->tempfile);
3750 oid_to_hex_r(temp->hex, oid);
3751 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3752 strbuf_release(&buf);
3753 strbuf_release(&template);
3754 free(path_dup);
3757 static struct diff_tempfile *prepare_temp_file(const char *name,
3758 struct diff_filespec *one)
3760 struct diff_tempfile *temp = claim_diff_tempfile();
3762 if (!DIFF_FILE_VALID(one)) {
3763 not_a_valid_file:
3764 /* A '-' entry produces this for file-2, and
3765 * a '+' entry produces this for file-1.
3767 temp->name = "/dev/null";
3768 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3769 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3770 return temp;
3773 if (!S_ISGITLINK(one->mode) &&
3774 (!one->oid_valid ||
3775 reuse_worktree_file(name, &one->oid, 1))) {
3776 struct stat st;
3777 if (lstat(name, &st) < 0) {
3778 if (errno == ENOENT)
3779 goto not_a_valid_file;
3780 die_errno("stat(%s)", name);
3782 if (S_ISLNK(st.st_mode)) {
3783 struct strbuf sb = STRBUF_INIT;
3784 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3785 die_errno("readlink(%s)", name);
3786 prep_temp_blob(name, temp, sb.buf, sb.len,
3787 (one->oid_valid ?
3788 &one->oid : &null_oid),
3789 (one->oid_valid ?
3790 one->mode : S_IFLNK));
3791 strbuf_release(&sb);
3793 else {
3794 /* we can borrow from the file in the work tree */
3795 temp->name = name;
3796 if (!one->oid_valid)
3797 oid_to_hex_r(temp->hex, &null_oid);
3798 else
3799 oid_to_hex_r(temp->hex, &one->oid);
3800 /* Even though we may sometimes borrow the
3801 * contents from the work tree, we always want
3802 * one->mode. mode is trustworthy even when
3803 * !(one->oid_valid), as long as
3804 * DIFF_FILE_VALID(one).
3806 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3808 return temp;
3810 else {
3811 if (diff_populate_filespec(one, 0))
3812 die("cannot read data blob for %s", one->path);
3813 prep_temp_blob(name, temp, one->data, one->size,
3814 &one->oid, one->mode);
3816 return temp;
3819 static void add_external_diff_name(struct argv_array *argv,
3820 const char *name,
3821 struct diff_filespec *df)
3823 struct diff_tempfile *temp = prepare_temp_file(name, df);
3824 argv_array_push(argv, temp->name);
3825 argv_array_push(argv, temp->hex);
3826 argv_array_push(argv, temp->mode);
3829 /* An external diff command takes:
3831 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3832 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3835 static void run_external_diff(const char *pgm,
3836 const char *name,
3837 const char *other,
3838 struct diff_filespec *one,
3839 struct diff_filespec *two,
3840 const char *xfrm_msg,
3841 int complete_rewrite,
3842 struct diff_options *o)
3844 struct argv_array argv = ARGV_ARRAY_INIT;
3845 struct argv_array env = ARGV_ARRAY_INIT;
3846 struct diff_queue_struct *q = &diff_queued_diff;
3848 argv_array_push(&argv, pgm);
3849 argv_array_push(&argv, name);
3851 if (one && two) {
3852 add_external_diff_name(&argv, name, one);
3853 if (!other)
3854 add_external_diff_name(&argv, name, two);
3855 else {
3856 add_external_diff_name(&argv, other, two);
3857 argv_array_push(&argv, other);
3858 argv_array_push(&argv, xfrm_msg);
3862 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3863 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3865 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3866 die(_("external diff died, stopping at %s"), name);
3868 remove_tempfile();
3869 argv_array_clear(&argv);
3870 argv_array_clear(&env);