Add "core.eol" config variable
[git/dscho.git] / convert.c
blob061fb23d6a196ce96d1d5b3e514d173a58ffe425
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 fd, 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[] = { params->cmd, NULL };
332 memset(&child_process, 0, sizeof(child_process));
333 child_process.argv = argv;
334 child_process.use_shell = 1;
335 child_process.in = -1;
336 child_process.out = fd;
338 if (start_command(&child_process))
339 return error("cannot fork to run external filter %s", params->cmd);
341 write_err = (write_in_full(child_process.in, params->src, params->size) < 0);
342 if (close(child_process.in))
343 write_err = 1;
344 if (write_err)
345 error("cannot feed the input to external filter %s", params->cmd);
347 status = finish_command(&child_process);
348 if (status)
349 error("external filter %s failed %d", params->cmd, status);
350 return (write_err || status);
353 static int apply_filter(const char *path, const char *src, size_t len,
354 struct strbuf *dst, const char *cmd)
357 * Create a pipeline to have the command filter the buffer's
358 * contents.
360 * (child --> cmd) --> us
362 int ret = 1;
363 struct strbuf nbuf = STRBUF_INIT;
364 struct async async;
365 struct filter_params params;
367 if (!cmd)
368 return 0;
370 memset(&async, 0, sizeof(async));
371 async.proc = filter_buffer;
372 async.data = &params;
373 params.src = src;
374 params.size = len;
375 params.cmd = cmd;
377 fflush(NULL);
378 if (start_async(&async))
379 return 0; /* error was already reported */
381 if (strbuf_read(&nbuf, async.out, len) < 0) {
382 error("read from external filter %s failed", cmd);
383 ret = 0;
385 if (close(async.out)) {
386 error("read from external filter %s failed", cmd);
387 ret = 0;
389 if (finish_async(&async)) {
390 error("external filter %s failed", cmd);
391 ret = 0;
394 if (ret) {
395 strbuf_swap(dst, &nbuf);
397 strbuf_release(&nbuf);
398 return ret;
401 static struct convert_driver {
402 const char *name;
403 struct convert_driver *next;
404 const char *smudge;
405 const char *clean;
406 } *user_convert, **user_convert_tail;
408 static int read_convert_config(const char *var, const char *value, void *cb)
410 const char *ep, *name;
411 int namelen;
412 struct convert_driver *drv;
415 * External conversion drivers are configured using
416 * "filter.<name>.variable".
418 if (prefixcmp(var, "filter.") || (ep = strrchr(var, '.')) == var + 6)
419 return 0;
420 name = var + 7;
421 namelen = ep - name;
422 for (drv = user_convert; drv; drv = drv->next)
423 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
424 break;
425 if (!drv) {
426 drv = xcalloc(1, sizeof(struct convert_driver));
427 drv->name = xmemdupz(name, namelen);
428 *user_convert_tail = drv;
429 user_convert_tail = &(drv->next);
432 ep++;
435 * filter.<name>.smudge and filter.<name>.clean specifies
436 * the command line:
438 * command-line
440 * The command-line will not be interpolated in any way.
443 if (!strcmp("smudge", ep))
444 return git_config_string(&drv->smudge, var, value);
446 if (!strcmp("clean", ep))
447 return git_config_string(&drv->clean, var, value);
449 return 0;
452 static void setup_convert_check(struct git_attr_check *check)
454 static struct git_attr *attr_text;
455 static struct git_attr *attr_crlf;
456 static struct git_attr *attr_eol;
457 static struct git_attr *attr_ident;
458 static struct git_attr *attr_filter;
460 if (!attr_text) {
461 attr_text = git_attr("text");
462 attr_crlf = git_attr("crlf");
463 attr_eol = git_attr("eol");
464 attr_ident = git_attr("ident");
465 attr_filter = git_attr("filter");
466 user_convert_tail = &user_convert;
467 git_config(read_convert_config, NULL);
469 check[0].attr = attr_crlf;
470 check[1].attr = attr_ident;
471 check[2].attr = attr_filter;
472 check[3].attr = attr_eol;
473 check[4].attr = attr_text;
476 static int count_ident(const char *cp, unsigned long size)
479 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
481 int cnt = 0;
482 char ch;
484 while (size) {
485 ch = *cp++;
486 size--;
487 if (ch != '$')
488 continue;
489 if (size < 3)
490 break;
491 if (memcmp("Id", cp, 2))
492 continue;
493 ch = cp[2];
494 cp += 3;
495 size -= 3;
496 if (ch == '$')
497 cnt++; /* $Id$ */
498 if (ch != ':')
499 continue;
502 * "$Id: ... "; scan up to the closing dollar sign and discard.
504 while (size) {
505 ch = *cp++;
506 size--;
507 if (ch == '$') {
508 cnt++;
509 break;
513 return cnt;
516 static int ident_to_git(const char *path, const char *src, size_t len,
517 struct strbuf *buf, int ident)
519 char *dst, *dollar;
521 if (!ident || !count_ident(src, len))
522 return 0;
524 /* only grow if not in place */
525 if (strbuf_avail(buf) + buf->len < len)
526 strbuf_grow(buf, len - buf->len);
527 dst = buf->buf;
528 for (;;) {
529 dollar = memchr(src, '$', len);
530 if (!dollar)
531 break;
532 memcpy(dst, src, dollar + 1 - src);
533 dst += dollar + 1 - src;
534 len -= dollar + 1 - src;
535 src = dollar + 1;
537 if (len > 3 && !memcmp(src, "Id:", 3)) {
538 dollar = memchr(src + 3, '$', len - 3);
539 if (!dollar)
540 break;
541 memcpy(dst, "Id$", 3);
542 dst += 3;
543 len -= dollar + 1 - src;
544 src = dollar + 1;
547 memcpy(dst, src, len);
548 strbuf_setlen(buf, dst + len - buf->buf);
549 return 1;
552 static int ident_to_worktree(const char *path, const char *src, size_t len,
553 struct strbuf *buf, int ident)
555 unsigned char sha1[20];
556 char *to_free = NULL, *dollar;
557 int cnt;
559 if (!ident)
560 return 0;
562 cnt = count_ident(src, len);
563 if (!cnt)
564 return 0;
566 /* are we "faking" in place editing ? */
567 if (src == buf->buf)
568 to_free = strbuf_detach(buf, NULL);
569 hash_sha1_file(src, len, "blob", sha1);
571 strbuf_grow(buf, len + cnt * 43);
572 for (;;) {
573 /* step 1: run to the next '$' */
574 dollar = memchr(src, '$', len);
575 if (!dollar)
576 break;
577 strbuf_add(buf, src, dollar + 1 - src);
578 len -= dollar + 1 - src;
579 src = dollar + 1;
581 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
582 if (len < 3 || memcmp("Id", src, 2))
583 continue;
585 /* step 3: skip over Id$ or Id:xxxxx$ */
586 if (src[2] == '$') {
587 src += 3;
588 len -= 3;
589 } else if (src[2] == ':') {
591 * It's possible that an expanded Id has crept its way into the
592 * repository, we cope with that by stripping the expansion out
594 dollar = memchr(src + 3, '$', len - 3);
595 if (!dollar) {
596 /* incomplete keyword, no more '$', so just quit the loop */
597 break;
600 len -= dollar + 1 - src;
601 src = dollar + 1;
602 } else {
603 /* it wasn't a "Id$" or "Id:xxxx$" */
604 continue;
607 /* step 4: substitute */
608 strbuf_addstr(buf, "Id: ");
609 strbuf_add(buf, sha1_to_hex(sha1), 40);
610 strbuf_addstr(buf, " $");
612 strbuf_add(buf, src, len);
614 free(to_free);
615 return 1;
618 static int git_path_check_crlf(const char *path, struct git_attr_check *check)
620 const char *value = check->value;
622 if (ATTR_TRUE(value))
623 return CRLF_TEXT;
624 else if (ATTR_FALSE(value))
625 return CRLF_BINARY;
626 else if (ATTR_UNSET(value))
628 else if (!strcmp(value, "input"))
629 return CRLF_INPUT;
630 else if (!strcmp(value, "auto"))
631 return CRLF_AUTO;
632 return CRLF_GUESS;
635 static int git_path_check_eol(const char *path, struct git_attr_check *check)
637 const char *value = check->value;
639 if (ATTR_UNSET(value))
641 else if (!strcmp(value, "lf"))
642 return EOL_LF;
643 else if (!strcmp(value, "crlf"))
644 return EOL_CRLF;
645 return EOL_UNSET;
648 static struct convert_driver *git_path_check_convert(const char *path,
649 struct git_attr_check *check)
651 const char *value = check->value;
652 struct convert_driver *drv;
654 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
655 return NULL;
656 for (drv = user_convert; drv; drv = drv->next)
657 if (!strcmp(value, drv->name))
658 return drv;
659 return NULL;
662 static int git_path_check_ident(const char *path, struct git_attr_check *check)
664 const char *value = check->value;
666 return !!ATTR_TRUE(value);
669 enum action determine_action(enum action text_attr, enum eol eol_attr) {
670 if (text_attr == CRLF_BINARY)
671 return CRLF_BINARY;
672 if (eol_attr == EOL_LF)
673 return CRLF_INPUT;
674 if (eol_attr == EOL_CRLF)
675 return CRLF_CRLF;
676 return text_attr;
679 int convert_to_git(const char *path, const char *src, size_t len,
680 struct strbuf *dst, enum safe_crlf checksafe)
682 struct git_attr_check check[5];
683 enum action action = CRLF_GUESS;
684 enum eol eol_attr = EOL_UNSET;
685 int ident = 0, ret = 0;
686 const char *filter = NULL;
688 setup_convert_check(check);
689 if (!git_checkattr(path, ARRAY_SIZE(check), check)) {
690 struct convert_driver *drv;
691 action = git_path_check_crlf(path, check + 4);
692 if (action == CRLF_GUESS)
693 action = git_path_check_crlf(path, check + 0);
694 ident = git_path_check_ident(path, check + 1);
695 drv = git_path_check_convert(path, check + 2);
696 eol_attr = git_path_check_eol(path, check + 3);
697 if (drv && drv->clean)
698 filter = drv->clean;
701 ret |= apply_filter(path, src, len, dst, filter);
702 if (ret) {
703 src = dst->buf;
704 len = dst->len;
706 action = determine_action(action, eol_attr);
707 ret |= crlf_to_git(path, src, len, dst, action, checksafe);
708 if (ret) {
709 src = dst->buf;
710 len = dst->len;
712 return ret | ident_to_git(path, src, len, dst, ident);
715 int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
717 struct git_attr_check check[5];
718 enum action action = CRLF_GUESS;
719 enum eol eol_attr = EOL_UNSET;
720 int ident = 0, ret = 0;
721 const char *filter = NULL;
723 setup_convert_check(check);
724 if (!git_checkattr(path, ARRAY_SIZE(check), check)) {
725 struct convert_driver *drv;
726 action = git_path_check_crlf(path, check + 4);
727 if (action == CRLF_GUESS)
728 action = git_path_check_crlf(path, check + 0);
729 ident = git_path_check_ident(path, check + 1);
730 drv = git_path_check_convert(path, check + 2);
731 eol_attr = git_path_check_eol(path, check + 3);
732 if (drv && drv->smudge)
733 filter = drv->smudge;
736 ret |= ident_to_worktree(path, src, len, dst, ident);
737 if (ret) {
738 src = dst->buf;
739 len = dst->len;
741 action = determine_action(action, eol_attr);
742 ret |= crlf_to_worktree(path, src, len, dst, action);
743 if (ret) {
744 src = dst->buf;
745 len = dst->len;
747 return ret | apply_filter(path, src, len, dst, filter);