Merge branch 'ab/blame-textconv'
[git/raj.git] / convert.c
blobe41a31e4807e92e210854214d46767b9752c5181
1 #include "cache.h"
2 #include "attr.h"
3 #include "run-command.h"
5 /*
6 * convert.c - convert a file when checking it out and checking it in.
8 * This should use the pathname to decide on whether it wants to do some
9 * more interesting conversions (automatic gzip/unzip, general format
10 * conversions etc etc), but by default it just does automatic CRLF<->LF
11 * translation when the "text" attribute or "auto_crlf" option is set.
14 enum action {
15 CRLF_GUESS = -1,
16 CRLF_BINARY = 0,
17 CRLF_TEXT,
18 CRLF_INPUT,
19 CRLF_CRLF,
20 CRLF_AUTO,
23 struct text_stat {
24 /* NUL, CR, LF and CRLF counts */
25 unsigned nul, cr, lf, crlf;
27 /* These are just approximations! */
28 unsigned printable, nonprintable;
31 static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
33 unsigned long i;
35 memset(stats, 0, sizeof(*stats));
37 for (i = 0; i < size; i++) {
38 unsigned char c = buf[i];
39 if (c == '\r') {
40 stats->cr++;
41 if (i+1 < size && buf[i+1] == '\n')
42 stats->crlf++;
43 continue;
45 if (c == '\n') {
46 stats->lf++;
47 continue;
49 if (c == 127)
50 /* DEL */
51 stats->nonprintable++;
52 else if (c < 32) {
53 switch (c) {
54 /* BS, HT, ESC and FF */
55 case '\b': case '\t': case '\033': case '\014':
56 stats->printable++;
57 break;
58 case 0:
59 stats->nul++;
60 /* fall through */
61 default:
62 stats->nonprintable++;
65 else
66 stats->printable++;
69 /* If file ends with EOF then don't count this EOF as non-printable. */
70 if (size >= 1 && buf[size-1] == '\032')
71 stats->nonprintable--;
75 * The same heuristics as diff.c::mmfile_is_binary()
77 static int is_binary(unsigned long size, struct text_stat *stats)
80 if (stats->nul)
81 return 1;
82 if ((stats->printable >> 7) < stats->nonprintable)
83 return 1;
85 * Other heuristics? Average line length might be relevant,
86 * as might LF vs CR vs CRLF counts..
88 * NOTE! It might be normal to have a low ratio of CRLF to LF
89 * (somebody starts with a LF-only file and edits it with an editor
90 * that adds CRLF only to lines that are added..). But do we
91 * want to support CR-only? Probably not.
93 return 0;
96 static enum eol determine_output_conversion(enum action action) {
97 switch (action) {
98 case CRLF_BINARY:
99 return EOL_UNSET;
100 case CRLF_CRLF:
101 return EOL_CRLF;
102 case CRLF_INPUT:
103 return EOL_LF;
104 case CRLF_GUESS:
105 if (!auto_crlf)
106 return EOL_UNSET;
107 /* fall through */
108 case CRLF_TEXT:
109 case CRLF_AUTO:
110 if (auto_crlf == AUTO_CRLF_TRUE)
111 return EOL_CRLF;
112 else if (auto_crlf == AUTO_CRLF_INPUT)
113 return EOL_LF;
114 else if (eol == EOL_UNSET)
115 return EOL_NATIVE;
117 return eol;
120 static void check_safe_crlf(const char *path, enum action action,
121 struct text_stat *stats, enum safe_crlf checksafe)
123 if (!checksafe)
124 return;
126 if (determine_output_conversion(action) == EOL_LF) {
128 * CRLFs would not be restored by checkout:
129 * check if we'd remove CRLFs
131 if (stats->crlf) {
132 if (checksafe == SAFE_CRLF_WARN)
133 warning("CRLF will be replaced by LF in %s.\nThe file will have its original line endings in your working directory.", path);
134 else /* i.e. SAFE_CRLF_FAIL */
135 die("CRLF would be replaced by LF in %s.", path);
137 } else if (determine_output_conversion(action) == EOL_CRLF) {
139 * CRLFs would be added by checkout:
140 * check if we have "naked" LFs
142 if (stats->lf != stats->crlf) {
143 if (checksafe == SAFE_CRLF_WARN)
144 warning("LF will be replaced by CRLF in %s.\nThe file will have its original line endings in your working directory.", path);
145 else /* i.e. SAFE_CRLF_FAIL */
146 die("LF would be replaced by CRLF in %s", path);
151 static int has_cr_in_index(const char *path)
153 int pos, len;
154 unsigned long sz;
155 enum object_type type;
156 void *data;
157 int has_cr;
158 struct index_state *istate = &the_index;
160 len = strlen(path);
161 pos = index_name_pos(istate, path, len);
162 if (pos < 0) {
164 * We might be in the middle of a merge, in which
165 * case we would read stage #2 (ours).
167 int i;
168 for (i = -pos - 1;
169 (pos < 0 && i < istate->cache_nr &&
170 !strcmp(istate->cache[i]->name, path));
171 i++)
172 if (ce_stage(istate->cache[i]) == 2)
173 pos = i;
175 if (pos < 0)
176 return 0;
177 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
178 if (!data || type != OBJ_BLOB) {
179 free(data);
180 return 0;
183 has_cr = memchr(data, '\r', sz) != NULL;
184 free(data);
185 return has_cr;
188 static int crlf_to_git(const char *path, const char *src, size_t len,
189 struct strbuf *buf, enum action action, enum safe_crlf checksafe)
191 struct text_stat stats;
192 char *dst;
194 if (action == CRLF_BINARY ||
195 (action == CRLF_GUESS && auto_crlf == AUTO_CRLF_FALSE) || !len)
196 return 0;
198 gather_stats(src, len, &stats);
200 if (action == CRLF_AUTO || action == CRLF_GUESS) {
202 * We're currently not going to even try to convert stuff
203 * that has bare CR characters. Does anybody do that crazy
204 * stuff?
206 if (stats.cr != stats.crlf)
207 return 0;
210 * And add some heuristics for binary vs text, of course...
212 if (is_binary(len, &stats))
213 return 0;
215 if (action == CRLF_GUESS) {
217 * If the file in the index has any CR in it, do not convert.
218 * This is the new safer autocrlf handling.
220 if (has_cr_in_index(path))
221 return 0;
225 check_safe_crlf(path, action, &stats, checksafe);
227 /* Optimization: No CR? Nothing to convert, regardless. */
228 if (!stats.cr)
229 return 0;
231 /* only grow if not in place */
232 if (strbuf_avail(buf) + buf->len < len)
233 strbuf_grow(buf, len - buf->len);
234 dst = buf->buf;
235 if (action == CRLF_AUTO || action == CRLF_GUESS) {
237 * If we guessed, we already know we rejected a file with
238 * lone CR, and we can strip a CR without looking at what
239 * follow it.
241 do {
242 unsigned char c = *src++;
243 if (c != '\r')
244 *dst++ = c;
245 } while (--len);
246 } else {
247 do {
248 unsigned char c = *src++;
249 if (! (c == '\r' && (1 < len && *src == '\n')))
250 *dst++ = c;
251 } while (--len);
253 strbuf_setlen(buf, dst - buf->buf);
254 return 1;
257 static int crlf_to_worktree(const char *path, const char *src, size_t len,
258 struct strbuf *buf, enum action action)
260 char *to_free = NULL;
261 struct text_stat stats;
263 if (!len || determine_output_conversion(action) != EOL_CRLF)
264 return 0;
266 gather_stats(src, len, &stats);
268 /* No LF? Nothing to convert, regardless. */
269 if (!stats.lf)
270 return 0;
272 /* Was it already in CRLF format? */
273 if (stats.lf == stats.crlf)
274 return 0;
276 if (action == CRLF_AUTO || action == CRLF_GUESS) {
277 if (action == CRLF_GUESS) {
278 /* If we have any CR or CRLF line endings, we do not touch it */
279 /* This is the new safer autocrlf-handling */
280 if (stats.cr > 0 || stats.crlf > 0)
281 return 0;
284 /* If we have any bare CR characters, we're not going to touch it */
285 if (stats.cr != stats.crlf)
286 return 0;
288 if (is_binary(len, &stats))
289 return 0;
292 /* are we "faking" in place editing ? */
293 if (src == buf->buf)
294 to_free = strbuf_detach(buf, NULL);
296 strbuf_grow(buf, len + stats.lf - stats.crlf);
297 for (;;) {
298 const char *nl = memchr(src, '\n', len);
299 if (!nl)
300 break;
301 if (nl > src && nl[-1] == '\r') {
302 strbuf_add(buf, src, nl + 1 - src);
303 } else {
304 strbuf_add(buf, src, nl - src);
305 strbuf_addstr(buf, "\r\n");
307 len -= nl + 1 - src;
308 src = nl + 1;
310 strbuf_add(buf, src, len);
312 free(to_free);
313 return 1;
316 struct filter_params {
317 const char *src;
318 unsigned long size;
319 const char *cmd;
322 static int filter_buffer(int in, int out, void *data)
325 * Spawn cmd and feed the buffer contents through its stdin.
327 struct child_process child_process;
328 struct filter_params *params = (struct filter_params *)data;
329 int write_err, status;
330 const char *argv[] = { NULL, NULL };
332 argv[0] = params->cmd;
334 memset(&child_process, 0, sizeof(child_process));
335 child_process.argv = argv;
336 child_process.use_shell = 1;
337 child_process.in = -1;
338 child_process.out = out;
340 if (start_command(&child_process))
341 return error("cannot fork to run external filter %s", params->cmd);
343 write_err = (write_in_full(child_process.in, params->src, params->size) < 0);
344 if (close(child_process.in))
345 write_err = 1;
346 if (write_err)
347 error("cannot feed the input to external filter %s", params->cmd);
349 status = finish_command(&child_process);
350 if (status)
351 error("external filter %s failed %d", params->cmd, status);
352 return (write_err || status);
355 static int apply_filter(const char *path, const char *src, size_t len,
356 struct strbuf *dst, const char *cmd)
359 * Create a pipeline to have the command filter the buffer's
360 * contents.
362 * (child --> cmd) --> us
364 int ret = 1;
365 struct strbuf nbuf = STRBUF_INIT;
366 struct async async;
367 struct filter_params params;
369 if (!cmd)
370 return 0;
372 memset(&async, 0, sizeof(async));
373 async.proc = filter_buffer;
374 async.data = &params;
375 async.out = -1;
376 params.src = src;
377 params.size = len;
378 params.cmd = cmd;
380 fflush(NULL);
381 if (start_async(&async))
382 return 0; /* error was already reported */
384 if (strbuf_read(&nbuf, async.out, len) < 0) {
385 error("read from external filter %s failed", cmd);
386 ret = 0;
388 if (close(async.out)) {
389 error("read from external filter %s failed", cmd);
390 ret = 0;
392 if (finish_async(&async)) {
393 error("external filter %s failed", cmd);
394 ret = 0;
397 if (ret) {
398 strbuf_swap(dst, &nbuf);
400 strbuf_release(&nbuf);
401 return ret;
404 static struct convert_driver {
405 const char *name;
406 struct convert_driver *next;
407 const char *smudge;
408 const char *clean;
409 } *user_convert, **user_convert_tail;
411 static int read_convert_config(const char *var, const char *value, void *cb)
413 const char *ep, *name;
414 int namelen;
415 struct convert_driver *drv;
418 * External conversion drivers are configured using
419 * "filter.<name>.variable".
421 if (prefixcmp(var, "filter.") || (ep = strrchr(var, '.')) == var + 6)
422 return 0;
423 name = var + 7;
424 namelen = ep - name;
425 for (drv = user_convert; drv; drv = drv->next)
426 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
427 break;
428 if (!drv) {
429 drv = xcalloc(1, sizeof(struct convert_driver));
430 drv->name = xmemdupz(name, namelen);
431 *user_convert_tail = drv;
432 user_convert_tail = &(drv->next);
435 ep++;
438 * filter.<name>.smudge and filter.<name>.clean specifies
439 * the command line:
441 * command-line
443 * The command-line will not be interpolated in any way.
446 if (!strcmp("smudge", ep))
447 return git_config_string(&drv->smudge, var, value);
449 if (!strcmp("clean", ep))
450 return git_config_string(&drv->clean, var, value);
452 return 0;
455 static void setup_convert_check(struct git_attr_check *check)
457 static struct git_attr *attr_text;
458 static struct git_attr *attr_crlf;
459 static struct git_attr *attr_eol;
460 static struct git_attr *attr_ident;
461 static struct git_attr *attr_filter;
463 if (!attr_text) {
464 attr_text = git_attr("text");
465 attr_crlf = git_attr("crlf");
466 attr_eol = git_attr("eol");
467 attr_ident = git_attr("ident");
468 attr_filter = git_attr("filter");
469 user_convert_tail = &user_convert;
470 git_config(read_convert_config, NULL);
472 check[0].attr = attr_crlf;
473 check[1].attr = attr_ident;
474 check[2].attr = attr_filter;
475 check[3].attr = attr_eol;
476 check[4].attr = attr_text;
479 static int count_ident(const char *cp, unsigned long size)
482 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
484 int cnt = 0;
485 char ch;
487 while (size) {
488 ch = *cp++;
489 size--;
490 if (ch != '$')
491 continue;
492 if (size < 3)
493 break;
494 if (memcmp("Id", cp, 2))
495 continue;
496 ch = cp[2];
497 cp += 3;
498 size -= 3;
499 if (ch == '$')
500 cnt++; /* $Id$ */
501 if (ch != ':')
502 continue;
505 * "$Id: ... "; scan up to the closing dollar sign and discard.
507 while (size) {
508 ch = *cp++;
509 size--;
510 if (ch == '$') {
511 cnt++;
512 break;
514 if (ch == '\n')
515 break;
518 return cnt;
521 static int ident_to_git(const char *path, const char *src, size_t len,
522 struct strbuf *buf, int ident)
524 char *dst, *dollar;
526 if (!ident || !count_ident(src, len))
527 return 0;
529 /* only grow if not in place */
530 if (strbuf_avail(buf) + buf->len < len)
531 strbuf_grow(buf, len - buf->len);
532 dst = buf->buf;
533 for (;;) {
534 dollar = memchr(src, '$', len);
535 if (!dollar)
536 break;
537 memcpy(dst, src, dollar + 1 - src);
538 dst += dollar + 1 - src;
539 len -= dollar + 1 - src;
540 src = dollar + 1;
542 if (len > 3 && !memcmp(src, "Id:", 3)) {
543 dollar = memchr(src + 3, '$', len - 3);
544 if (!dollar)
545 break;
546 if (memchr(src + 3, '\n', dollar - src - 3)) {
547 /* Line break before the next dollar. */
548 continue;
551 memcpy(dst, "Id$", 3);
552 dst += 3;
553 len -= dollar + 1 - src;
554 src = dollar + 1;
557 memcpy(dst, src, len);
558 strbuf_setlen(buf, dst + len - buf->buf);
559 return 1;
562 static int ident_to_worktree(const char *path, const char *src, size_t len,
563 struct strbuf *buf, int ident)
565 unsigned char sha1[20];
566 char *to_free = NULL, *dollar, *spc;
567 int cnt;
569 if (!ident)
570 return 0;
572 cnt = count_ident(src, len);
573 if (!cnt)
574 return 0;
576 /* are we "faking" in place editing ? */
577 if (src == buf->buf)
578 to_free = strbuf_detach(buf, NULL);
579 hash_sha1_file(src, len, "blob", sha1);
581 strbuf_grow(buf, len + cnt * 43);
582 for (;;) {
583 /* step 1: run to the next '$' */
584 dollar = memchr(src, '$', len);
585 if (!dollar)
586 break;
587 strbuf_add(buf, src, dollar + 1 - src);
588 len -= dollar + 1 - src;
589 src = dollar + 1;
591 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
592 if (len < 3 || memcmp("Id", src, 2))
593 continue;
595 /* step 3: skip over Id$ or Id:xxxxx$ */
596 if (src[2] == '$') {
597 src += 3;
598 len -= 3;
599 } else if (src[2] == ':') {
601 * It's possible that an expanded Id has crept its way into the
602 * repository, we cope with that by stripping the expansion out.
603 * This is probably not a good idea, since it will cause changes
604 * on checkout, which won't go away by stash, but let's keep it
605 * for git-style ids.
607 dollar = memchr(src + 3, '$', len - 3);
608 if (!dollar) {
609 /* incomplete keyword, no more '$', so just quit the loop */
610 break;
613 if (memchr(src + 3, '\n', dollar - src - 3)) {
614 /* Line break before the next dollar. */
615 continue;
618 spc = memchr(src + 4, ' ', dollar - src - 4);
619 if (spc && spc < dollar-1) {
620 /* There are spaces in unexpected places.
621 * This is probably an id from some other
622 * versioning system. Keep it for now.
624 continue;
627 len -= dollar + 1 - src;
628 src = dollar + 1;
629 } else {
630 /* it wasn't a "Id$" or "Id:xxxx$" */
631 continue;
634 /* step 4: substitute */
635 strbuf_addstr(buf, "Id: ");
636 strbuf_add(buf, sha1_to_hex(sha1), 40);
637 strbuf_addstr(buf, " $");
639 strbuf_add(buf, src, len);
641 free(to_free);
642 return 1;
645 static int git_path_check_crlf(const char *path, struct git_attr_check *check)
647 const char *value = check->value;
649 if (ATTR_TRUE(value))
650 return CRLF_TEXT;
651 else if (ATTR_FALSE(value))
652 return CRLF_BINARY;
653 else if (ATTR_UNSET(value))
655 else if (!strcmp(value, "input"))
656 return CRLF_INPUT;
657 else if (!strcmp(value, "auto"))
658 return CRLF_AUTO;
659 return CRLF_GUESS;
662 static int git_path_check_eol(const char *path, struct git_attr_check *check)
664 const char *value = check->value;
666 if (ATTR_UNSET(value))
668 else if (!strcmp(value, "lf"))
669 return EOL_LF;
670 else if (!strcmp(value, "crlf"))
671 return EOL_CRLF;
672 return EOL_UNSET;
675 static struct convert_driver *git_path_check_convert(const char *path,
676 struct git_attr_check *check)
678 const char *value = check->value;
679 struct convert_driver *drv;
681 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
682 return NULL;
683 for (drv = user_convert; drv; drv = drv->next)
684 if (!strcmp(value, drv->name))
685 return drv;
686 return NULL;
689 static int git_path_check_ident(const char *path, struct git_attr_check *check)
691 const char *value = check->value;
693 return !!ATTR_TRUE(value);
696 enum action determine_action(enum action text_attr, enum eol eol_attr) {
697 if (text_attr == CRLF_BINARY)
698 return CRLF_BINARY;
699 if (eol_attr == EOL_LF)
700 return CRLF_INPUT;
701 if (eol_attr == EOL_CRLF)
702 return CRLF_CRLF;
703 return text_attr;
706 int convert_to_git(const char *path, const char *src, size_t len,
707 struct strbuf *dst, enum safe_crlf checksafe)
709 struct git_attr_check check[5];
710 enum action action = CRLF_GUESS;
711 enum eol eol_attr = EOL_UNSET;
712 int ident = 0, ret = 0;
713 const char *filter = NULL;
715 setup_convert_check(check);
716 if (!git_checkattr(path, ARRAY_SIZE(check), check)) {
717 struct convert_driver *drv;
718 action = git_path_check_crlf(path, check + 4);
719 if (action == CRLF_GUESS)
720 action = git_path_check_crlf(path, check + 0);
721 ident = git_path_check_ident(path, check + 1);
722 drv = git_path_check_convert(path, check + 2);
723 eol_attr = git_path_check_eol(path, check + 3);
724 if (drv && drv->clean)
725 filter = drv->clean;
728 ret |= apply_filter(path, src, len, dst, filter);
729 if (ret) {
730 src = dst->buf;
731 len = dst->len;
733 action = determine_action(action, eol_attr);
734 ret |= crlf_to_git(path, src, len, dst, action, checksafe);
735 if (ret) {
736 src = dst->buf;
737 len = dst->len;
739 return ret | ident_to_git(path, src, len, dst, ident);
742 int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
744 struct git_attr_check check[5];
745 enum action action = CRLF_GUESS;
746 enum eol eol_attr = EOL_UNSET;
747 int ident = 0, ret = 0;
748 const char *filter = NULL;
750 setup_convert_check(check);
751 if (!git_checkattr(path, ARRAY_SIZE(check), check)) {
752 struct convert_driver *drv;
753 action = git_path_check_crlf(path, check + 4);
754 if (action == CRLF_GUESS)
755 action = git_path_check_crlf(path, check + 0);
756 ident = git_path_check_ident(path, check + 1);
757 drv = git_path_check_convert(path, check + 2);
758 eol_attr = git_path_check_eol(path, check + 3);
759 if (drv && drv->smudge)
760 filter = drv->smudge;
763 ret |= ident_to_worktree(path, src, len, dst, ident);
764 if (ret) {
765 src = dst->buf;
766 len = dst->len;
768 action = determine_action(action, eol_attr);
769 ret |= crlf_to_worktree(path, src, len, dst, action);
770 if (ret) {
771 src = dst->buf;
772 len = dst->len;
774 return ret | apply_filter(path, src, len, dst, filter);