]>
Commit | Line | Data |
---|---|---|
1 | /* braces.c -- code for doing word expansion in curly braces. */ | |
2 | ||
3 | /* Copyright (C) 1987-2012 Free Software Foundation, Inc. | |
4 | ||
5 | This file is part of GNU Bash, the Bourne Again SHell. | |
6 | ||
7 | Bash is free software: you can redistribute it and/or modify | |
8 | it under the terms of the GNU General Public License as published by | |
9 | the Free Software Foundation, either version 3 of the License, or | |
10 | (at your option) any later version. | |
11 | ||
12 | Bash is distributed in the hope that it will be useful, | |
13 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | GNU General Public License for more details. | |
16 | ||
17 | You should have received a copy of the GNU General Public License | |
18 | along with Bash. If not, see <http://www.gnu.org/licenses/>. | |
19 | */ | |
20 | ||
21 | /* Stuff in curly braces gets expanded before all other shell expansions. */ | |
22 | ||
23 | #include "config.h" | |
24 | ||
25 | #if defined (BRACE_EXPANSION) | |
26 | ||
27 | #if defined (HAVE_UNISTD_H) | |
28 | # ifdef _MINIX | |
29 | # include <sys/types.h> | |
30 | # endif | |
31 | # include <unistd.h> | |
32 | #endif | |
33 | ||
34 | #include <errno.h> | |
35 | ||
36 | #include "bashansi.h" | |
37 | #include "bashintl.h" | |
38 | ||
39 | #if defined (SHELL) | |
40 | # include "shell.h" | |
41 | #else | |
42 | # if defined (TEST) | |
43 | typedef char *WORD_DESC; | |
44 | typedef char **WORD_LIST; | |
45 | #define _(X) X | |
46 | # endif /* TEST */ | |
47 | #endif /* SHELL */ | |
48 | ||
49 | #include "typemax.h" /* INTMAX_MIN, INTMAX_MAX */ | |
50 | #include "general.h" | |
51 | #include "shmbutil.h" | |
52 | #include "chartypes.h" | |
53 | ||
54 | #ifndef errno | |
55 | extern int errno; | |
56 | #endif | |
57 | ||
58 | #define brace_whitespace(c) (!(c) || (c) == ' ' || (c) == '\t' || (c) == '\n') | |
59 | ||
60 | #define BRACE_SEQ_SPECIFIER ".." | |
61 | ||
62 | extern int asprintf __P((char **, const char *, ...)) __attribute__((__format__ (printf, 2, 3))); | |
63 | ||
64 | extern int last_command_exit_value; | |
65 | ||
66 | /* Basic idea: | |
67 | ||
68 | Segregate the text into 3 sections: preamble (stuff before an open brace), | |
69 | postamble (stuff after the matching close brace) and amble (stuff after | |
70 | preamble, and before postamble). Expand amble, and then tack on the | |
71 | expansions to preamble. Expand postamble, and tack on the expansions to | |
72 | the result so far. | |
73 | */ | |
74 | ||
75 | /* The character which is used to separate arguments. */ | |
76 | static const int brace_arg_separator = ','; | |
77 | ||
78 | #if defined (__P) | |
79 | static int brace_gobbler __P((char *, size_t, int *, int)); | |
80 | static char **expand_amble __P((char *, size_t, int)); | |
81 | static char **expand_seqterm __P((char *, size_t)); | |
82 | static char **mkseq __P((intmax_t, intmax_t, intmax_t, int, int)); | |
83 | static char **array_concat __P((char **, char **)); | |
84 | #else | |
85 | static int brace_gobbler (); | |
86 | static char **expand_amble (); | |
87 | static char **expand_seqterm (); | |
88 | static char **mkseq(); | |
89 | static char **array_concat (); | |
90 | #endif | |
91 | ||
92 | #if 0 | |
93 | static void | |
94 | dump_result (a) | |
95 | char **a; | |
96 | { | |
97 | int i; | |
98 | ||
99 | for (i = 0; a[i]; i++) | |
100 | printf ("dump_result: a[%d] = -%s-\n", i, a[i]); | |
101 | } | |
102 | #endif | |
103 | ||
104 | /* Return an array of strings; the brace expansion of TEXT. */ | |
105 | char ** | |
106 | brace_expand (text) | |
107 | char *text; | |
108 | { | |
109 | register int start; | |
110 | size_t tlen; | |
111 | char *preamble, *postamble, *amble; | |
112 | size_t alen; | |
113 | char **tack, **result; | |
114 | int i, j, c, c1; | |
115 | ||
116 | DECLARE_MBSTATE; | |
117 | ||
118 | /* Find the text of the preamble. */ | |
119 | tlen = strlen (text); | |
120 | i = 0; | |
121 | #if defined (CSH_BRACE_COMPAT) | |
122 | c = brace_gobbler (text, tlen, &i, '{'); /* } */ | |
123 | #else | |
124 | /* Make sure that when we exit this loop, c == 0 or text[i] begins a | |
125 | valid brace expansion sequence. */ | |
126 | do | |
127 | { | |
128 | c = brace_gobbler (text, tlen, &i, '{'); /* } */ | |
129 | c1 = c; | |
130 | /* Verify that c begins a valid brace expansion word. If it doesn't, we | |
131 | go on. Loop stops when there are no more open braces in the word. */ | |
132 | if (c) | |
133 | { | |
134 | start = j = i + 1; /* { */ | |
135 | c = brace_gobbler (text, tlen, &j, '}'); | |
136 | if (c == 0) /* it's not */ | |
137 | { | |
138 | i++; | |
139 | c = c1; | |
140 | continue; | |
141 | } | |
142 | else /* it is */ | |
143 | { | |
144 | c = c1; | |
145 | break; | |
146 | } | |
147 | } | |
148 | else | |
149 | break; | |
150 | } | |
151 | while (c); | |
152 | #endif /* !CSH_BRACE_COMPAT */ | |
153 | ||
154 | preamble = (char *)xmalloc (i + 1); | |
155 | if (i > 0) | |
156 | strncpy (preamble, text, i); | |
157 | preamble[i] = '\0'; | |
158 | ||
159 | result = (char **)xmalloc (2 * sizeof (char *)); | |
160 | result[0] = preamble; | |
161 | result[1] = (char *)NULL; | |
162 | ||
163 | /* Special case. If we never found an exciting character, then | |
164 | the preamble is all of the text, so just return that. */ | |
165 | if (c != '{') | |
166 | return (result); | |
167 | ||
168 | /* Find the amble. This is the stuff inside this set of braces. */ | |
169 | start = ++i; | |
170 | c = brace_gobbler (text, tlen, &i, '}'); | |
171 | ||
172 | /* What if there isn't a matching close brace? */ | |
173 | if (c == 0) | |
174 | { | |
175 | #if defined (NOTDEF) | |
176 | /* Well, if we found an unquoted BRACE_ARG_SEPARATOR between START | |
177 | and I, then this should be an error. Otherwise, it isn't. */ | |
178 | j = start; | |
179 | while (j < i) | |
180 | { | |
181 | if (text[j] == '\\') | |
182 | { | |
183 | j++; | |
184 | ADVANCE_CHAR (text, tlen, j); | |
185 | continue; | |
186 | } | |
187 | ||
188 | if (text[j] == brace_arg_separator) | |
189 | { /* { */ | |
190 | strvec_dispose (result); | |
191 | last_command_exit_value = 1; | |
192 | report_error ("no closing `%c' in %s", '}', text); | |
193 | throw_to_top_level (); | |
194 | } | |
195 | ADVANCE_CHAR (text, tlen, j); | |
196 | } | |
197 | #endif | |
198 | free (preamble); /* Same as result[0]; see initialization. */ | |
199 | result[0] = savestring (text); | |
200 | return (result); | |
201 | } | |
202 | ||
203 | #if defined (SHELL) | |
204 | amble = substring (text, start, i); | |
205 | alen = i - start; | |
206 | #else | |
207 | amble = (char *)xmalloc (1 + (i - start)); | |
208 | strncpy (amble, &text[start], (i - start)); | |
209 | alen = i - start; | |
210 | amble[alen] = '\0'; | |
211 | #endif | |
212 | ||
213 | #if defined (SHELL) | |
214 | INITIALIZE_MBSTATE; | |
215 | ||
216 | /* If the amble does not contain an unquoted BRACE_ARG_SEPARATOR, then | |
217 | just return without doing any expansion. */ | |
218 | j = 0; | |
219 | while (amble[j]) | |
220 | { | |
221 | if (amble[j] == '\\') | |
222 | { | |
223 | j++; | |
224 | ADVANCE_CHAR (amble, alen, j); | |
225 | continue; | |
226 | } | |
227 | ||
228 | if (amble[j] == brace_arg_separator) | |
229 | break; | |
230 | ||
231 | ADVANCE_CHAR (amble, alen, j); | |
232 | } | |
233 | ||
234 | if (amble[j] == 0) | |
235 | { | |
236 | tack = expand_seqterm (amble, alen); | |
237 | if (tack) | |
238 | goto add_tack; | |
239 | else if (text[i + 1]) | |
240 | { | |
241 | /* If the sequence expansion fails (e.g., because the integers | |
242 | overflow), but there is more in the string, try and process | |
243 | the rest of the string, which may contain additional brace | |
244 | expansions. Treat the unexpanded sequence term as a simple | |
245 | string (including the braces). */ | |
246 | tack = strvec_create (2); | |
247 | tack[0] = savestring (text+start-1); | |
248 | tack[0][i-start+2] = '\0'; | |
249 | tack[1] = (char *)0; | |
250 | goto add_tack; | |
251 | } | |
252 | else | |
253 | { | |
254 | free (amble); | |
255 | free (preamble); | |
256 | result[0] = savestring (text); | |
257 | return (result); | |
258 | } | |
259 | } | |
260 | #endif /* SHELL */ | |
261 | ||
262 | tack = expand_amble (amble, alen, 0); | |
263 | add_tack: | |
264 | result = array_concat (result, tack); | |
265 | free (amble); | |
266 | if (tack != result) | |
267 | strvec_dispose (tack); | |
268 | ||
269 | postamble = text + i + 1; | |
270 | ||
271 | if (postamble && *postamble) | |
272 | { | |
273 | tack = brace_expand (postamble); | |
274 | result = array_concat (result, tack); | |
275 | if (tack != result) | |
276 | strvec_dispose (tack); | |
277 | } | |
278 | ||
279 | return (result); | |
280 | } | |
281 | ||
282 | /* Expand the text found inside of braces. We simply try to split the | |
283 | text at BRACE_ARG_SEPARATORs into separate strings. We then brace | |
284 | expand each slot which needs it, until there are no more slots which | |
285 | need it. */ | |
286 | static char ** | |
287 | expand_amble (text, tlen, flags) | |
288 | char *text; | |
289 | size_t tlen; | |
290 | int flags; | |
291 | { | |
292 | char **result, **partial, **tresult; | |
293 | char *tem; | |
294 | int start, i, c; | |
295 | ||
296 | #if defined (SHELL) | |
297 | DECLARE_MBSTATE; | |
298 | #endif | |
299 | ||
300 | result = (char **)NULL; | |
301 | ||
302 | start = i = 0; | |
303 | c = 1; | |
304 | while (c) | |
305 | { | |
306 | c = brace_gobbler (text, tlen, &i, brace_arg_separator); | |
307 | #if defined (SHELL) | |
308 | tem = substring (text, start, i); | |
309 | #else | |
310 | tem = (char *)xmalloc (1 + (i - start)); | |
311 | strncpy (tem, &text[start], (i - start)); | |
312 | tem[i - start] = '\0'; | |
313 | #endif | |
314 | ||
315 | partial = brace_expand (tem); | |
316 | ||
317 | if (!result) | |
318 | result = partial; | |
319 | else | |
320 | { | |
321 | register int lr, lp, j; | |
322 | ||
323 | lr = strvec_len (result); | |
324 | lp = strvec_len (partial); | |
325 | ||
326 | tresult = strvec_mresize (result, lp + lr + 1); | |
327 | if (tresult == 0) | |
328 | { | |
329 | internal_error (_("brace expansion: cannot allocate memory for %s"), tem); | |
330 | free (tem); | |
331 | strvec_dispose (partial); | |
332 | strvec_dispose (result); | |
333 | result = (char **)NULL; | |
334 | return result; | |
335 | } | |
336 | else | |
337 | result = tresult; | |
338 | ||
339 | for (j = 0; j < lp; j++) | |
340 | result[lr + j] = partial[j]; | |
341 | ||
342 | result[lr + j] = (char *)NULL; | |
343 | free (partial); | |
344 | } | |
345 | free (tem); | |
346 | #if defined (SHELL) | |
347 | ADVANCE_CHAR (text, tlen, i); | |
348 | #else | |
349 | i++; | |
350 | #endif | |
351 | start = i; | |
352 | } | |
353 | return (result); | |
354 | } | |
355 | ||
356 | #define ST_BAD 0 | |
357 | #define ST_INT 1 | |
358 | #define ST_CHAR 2 | |
359 | #define ST_ZINT 3 | |
360 | ||
361 | #ifndef sh_imaxabs | |
362 | # define sh_imaxabs(x) (((x) >= 0) ? (x) : -(x)) | |
363 | #endif | |
364 | ||
365 | /* Handle signed arithmetic overflow and underflow. Have to do it this way | |
366 | to avoid compilers optimizing out simpler overflow checks. */ | |
367 | ||
368 | /* Make sure that a+b does not exceed MAXV or is smaller than MINV (if b < 0). | |
369 | Assumes that b > 0 if a > 0 and b < 0 if a < 0 */ | |
370 | #define ADDOVERFLOW(a,b,minv,maxv) \ | |
371 | ((((a) > 0) && ((b) > ((maxv) - (a)))) || \ | |
372 | (((a) < 0) && ((b) < ((minv) - (a))))) | |
373 | ||
374 | /* Make sure that a-b is not smaller than MINV or exceeds MAXV (if b < 0). | |
375 | Assumes that b > 0 if a > 0 and b < 0 if a < 0 */ | |
376 | #define SUBOVERFLOW(a,b,minv,maxv) \ | |
377 | ((((b) > 0) && ((a) < ((minv) + (b)))) || \ | |
378 | (((b) < 0) && ((a) > ((maxv) + (b))))) | |
379 | ||
380 | static char ** | |
381 | mkseq (start, end, incr, type, width) | |
382 | intmax_t start, end, incr; | |
383 | int type, width; | |
384 | { | |
385 | intmax_t n, prevn; | |
386 | int i, j, nelem; | |
387 | char **result, *t; | |
388 | ||
389 | if (incr == 0) | |
390 | incr = 1; | |
391 | ||
392 | if (start > end && incr > 0) | |
393 | incr = -incr; | |
394 | else if (start < end && incr < 0) | |
395 | { | |
396 | if (incr == INTMAX_MIN) /* Don't use -INTMAX_MIN */ | |
397 | return ((char **)NULL); | |
398 | incr = -incr; | |
399 | } | |
400 | ||
401 | /* Check that end-start will not overflow INTMAX_MIN, INTMAX_MAX. The +3 | |
402 | and -2, not strictly necessary, are there because of the way the number | |
403 | of elements and value passed to strvec_create() are calculated below. */ | |
404 | if (SUBOVERFLOW (end, start, INTMAX_MIN+3, INTMAX_MAX-2)) | |
405 | return ((char **)NULL); | |
406 | ||
407 | prevn = sh_imaxabs (end - start); | |
408 | /* Need to check this way in case INT_MAX == INTMAX_MAX */ | |
409 | if (INT_MAX == INTMAX_MAX && (ADDOVERFLOW (prevn, 2, INT_MIN, INT_MAX))) | |
410 | return ((char **)NULL); | |
411 | /* Make sure the assignment to nelem below doesn't end up <= 0 due to | |
412 | intmax_t overflow */ | |
413 | else if (ADDOVERFLOW ((prevn/sh_imaxabs(incr)), 1, INTMAX_MIN, INTMAX_MAX)) | |
414 | return ((char **)NULL); | |
415 | ||
416 | /* XXX - TOFIX: potentially allocating a lot of extra memory if | |
417 | imaxabs(incr) != 1 */ | |
418 | /* Instead of a simple nelem = prevn + 1, something like: | |
419 | nelem = (prevn / imaxabs(incr)) + 1; | |
420 | would work */ | |
421 | nelem = (prevn / sh_imaxabs(incr)) + 1; | |
422 | if (nelem > INT_MAX - 2) /* Don't overflow int */ | |
423 | return ((char **)NULL); | |
424 | result = strvec_mcreate (nelem + 1); | |
425 | if (result == 0) | |
426 | { | |
427 | internal_error (_("brace expansion: failed to allocate memory for %d elements"), nelem); | |
428 | return ((char **)NULL); | |
429 | } | |
430 | ||
431 | /* Make sure we go through the loop at least once, so {3..3} prints `3' */ | |
432 | i = 0; | |
433 | n = start; | |
434 | do | |
435 | { | |
436 | #if defined (SHELL) | |
437 | if (ISINTERRUPT) | |
438 | { | |
439 | strvec_dispose (result); | |
440 | result = (char **)NULL; | |
441 | } | |
442 | QUIT; | |
443 | #endif | |
444 | if (type == ST_INT) | |
445 | result[i++] = t = itos (n); | |
446 | else if (type == ST_ZINT) | |
447 | { | |
448 | int len, arg; | |
449 | arg = n; | |
450 | len = asprintf (&t, "%0*d", width, arg); | |
451 | result[i++] = t; | |
452 | } | |
453 | else | |
454 | { | |
455 | if (t = (char *)malloc (2)) | |
456 | { | |
457 | t[0] = n; | |
458 | t[1] = '\0'; | |
459 | } | |
460 | result[i++] = t; | |
461 | } | |
462 | ||
463 | /* We failed to allocate memory for this number, so we bail. */ | |
464 | if (t == 0) | |
465 | { | |
466 | char *p, lbuf[INT_STRLEN_BOUND(intmax_t) + 1]; | |
467 | ||
468 | /* Easier to do this than mess around with various intmax_t printf | |
469 | formats (%ld? %lld? %jd?) and PRIdMAX. */ | |
470 | p = inttostr (n, lbuf, sizeof (lbuf)); | |
471 | internal_error (_("brace expansion: failed to allocate memory for `%s'"), p); | |
472 | strvec_dispose (result); | |
473 | return ((char **)NULL); | |
474 | } | |
475 | ||
476 | /* Handle overflow and underflow of n+incr */ | |
477 | if (ADDOVERFLOW (n, incr, INTMAX_MIN, INTMAX_MAX)) | |
478 | break; | |
479 | ||
480 | n += incr; | |
481 | ||
482 | if ((incr < 0 && n < end) || (incr > 0 && n > end)) | |
483 | break; | |
484 | } | |
485 | while (1); | |
486 | ||
487 | result[i] = (char *)0; | |
488 | return (result); | |
489 | } | |
490 | ||
491 | static char ** | |
492 | expand_seqterm (text, tlen) | |
493 | char *text; | |
494 | size_t tlen; | |
495 | { | |
496 | char *t, *lhs, *rhs; | |
497 | int i, lhs_t, rhs_t, lhs_l, rhs_l, width; | |
498 | intmax_t lhs_v, rhs_v, incr; | |
499 | intmax_t tl, tr; | |
500 | char **result, *ep, *oep; | |
501 | ||
502 | t = strstr (text, BRACE_SEQ_SPECIFIER); | |
503 | if (t == 0) | |
504 | return ((char **)NULL); | |
505 | ||
506 | lhs_l = t - text; /* index of start of BRACE_SEQ_SPECIFIER */ | |
507 | lhs = substring (text, 0, lhs_l); | |
508 | rhs = substring (text, lhs_l + sizeof(BRACE_SEQ_SPECIFIER) - 1, tlen); | |
509 | ||
510 | if (lhs[0] == 0 || rhs[0] == 0) | |
511 | { | |
512 | free (lhs); | |
513 | free (rhs); | |
514 | return ((char **)NULL); | |
515 | } | |
516 | ||
517 | /* Now figure out whether LHS and RHS are integers or letters. Both | |
518 | sides have to match. */ | |
519 | lhs_t = (legal_number (lhs, &tl)) ? ST_INT : | |
520 | ((ISALPHA (lhs[0]) && lhs[1] == 0) ? ST_CHAR : ST_BAD); | |
521 | ||
522 | /* Decide on rhs and whether or not it looks like the user specified | |
523 | an increment */ | |
524 | ep = 0; | |
525 | if (ISDIGIT (rhs[0]) || ((rhs[0] == '+' || rhs[0] == '-') && ISDIGIT (rhs[1]))) | |
526 | { | |
527 | rhs_t = ST_INT; | |
528 | errno = 0; | |
529 | tr = strtoimax (rhs, &ep, 10); | |
530 | if (errno == ERANGE || (ep && *ep != 0 && *ep != '.')) | |
531 | rhs_t = ST_BAD; /* invalid */ | |
532 | } | |
533 | else if (ISALPHA (rhs[0]) && (rhs[1] == 0 || rhs[1] == '.')) | |
534 | { | |
535 | rhs_t = ST_CHAR; | |
536 | ep = rhs + 1; | |
537 | } | |
538 | else | |
539 | { | |
540 | rhs_t = ST_BAD; | |
541 | ep = 0; | |
542 | } | |
543 | ||
544 | incr = 1; | |
545 | if (rhs_t != ST_BAD) | |
546 | { | |
547 | oep = ep; | |
548 | errno = 0; | |
549 | if (ep && *ep == '.' && ep[1] == '.' && ep[2]) | |
550 | incr = strtoimax (ep + 2, &ep, 10); | |
551 | if (*ep != 0 || errno == ERANGE) | |
552 | rhs_t = ST_BAD; /* invalid incr or overflow */ | |
553 | tlen -= ep - oep; | |
554 | } | |
555 | ||
556 | if (lhs_t != rhs_t || lhs_t == ST_BAD || rhs_t == ST_BAD) | |
557 | { | |
558 | free (lhs); | |
559 | free (rhs); | |
560 | return ((char **)NULL); | |
561 | } | |
562 | ||
563 | /* OK, we have something. It's either a sequence of integers, ascending | |
564 | or descending, or a sequence or letters, ditto. Generate the sequence, | |
565 | put it into a string vector, and return it. */ | |
566 | ||
567 | if (lhs_t == ST_CHAR) | |
568 | { | |
569 | lhs_v = (unsigned char)lhs[0]; | |
570 | rhs_v = (unsigned char)rhs[0]; | |
571 | width = 1; | |
572 | } | |
573 | else | |
574 | { | |
575 | lhs_v = tl; /* integer truncation */ | |
576 | rhs_v = tr; | |
577 | ||
578 | /* Decide whether or not the terms need zero-padding */ | |
579 | rhs_l = tlen - lhs_l - sizeof (BRACE_SEQ_SPECIFIER) + 1; | |
580 | width = 0; | |
581 | if (lhs_l > 1 && lhs[0] == '0') | |
582 | width = lhs_l, lhs_t = ST_ZINT; | |
583 | if (lhs_l > 2 && lhs[0] == '-' && lhs[1] == '0') | |
584 | width = lhs_l, lhs_t = ST_ZINT; | |
585 | if (rhs_l > 1 && rhs[0] == '0' && width < rhs_l) | |
586 | width = rhs_l, lhs_t = ST_ZINT; | |
587 | if (rhs_l > 2 && rhs[0] == '-' && rhs[1] == '0' && width < rhs_l) | |
588 | width = rhs_l, lhs_t = ST_ZINT; | |
589 | ||
590 | if (width < lhs_l && lhs_t == ST_ZINT) | |
591 | width = lhs_l; | |
592 | if (width < rhs_l && lhs_t == ST_ZINT) | |
593 | width = rhs_l; | |
594 | } | |
595 | ||
596 | result = mkseq (lhs_v, rhs_v, incr, lhs_t, width); | |
597 | ||
598 | free (lhs); | |
599 | free (rhs); | |
600 | ||
601 | return (result); | |
602 | } | |
603 | ||
604 | /* Start at INDEX, and skip characters in TEXT. Set INDEX to the | |
605 | index of the character matching SATISFY. This understands about | |
606 | quoting. Return the character that caused us to stop searching; | |
607 | this is either the same as SATISFY, or 0. */ | |
608 | /* If SATISFY is `}', we are looking for a brace expression, so we | |
609 | should enforce the rules that govern valid brace expansions: | |
610 | 1) to count as an arg separator, a comma or `..' has to be outside | |
611 | an inner set of braces. | |
612 | */ | |
613 | static int | |
614 | brace_gobbler (text, tlen, indx, satisfy) | |
615 | char *text; | |
616 | size_t tlen; | |
617 | int *indx; | |
618 | int satisfy; | |
619 | { | |
620 | register int i, c, quoted, level, commas, pass_next; | |
621 | #if defined (SHELL) | |
622 | int si; | |
623 | char *t; | |
624 | #endif | |
625 | DECLARE_MBSTATE; | |
626 | ||
627 | level = quoted = pass_next = 0; | |
628 | #if defined (CSH_BRACE_COMPAT) | |
629 | commas = 1; | |
630 | #else | |
631 | commas = (satisfy == '}') ? 0 : 1; | |
632 | #endif | |
633 | ||
634 | i = *indx; | |
635 | while (c = text[i]) | |
636 | { | |
637 | if (pass_next) | |
638 | { | |
639 | pass_next = 0; | |
640 | #if defined (SHELL) | |
641 | ADVANCE_CHAR (text, tlen, i); | |
642 | #else | |
643 | i++; | |
644 | #endif | |
645 | continue; | |
646 | } | |
647 | ||
648 | /* A backslash escapes the next character. This allows backslash to | |
649 | escape the quote character in a double-quoted string. */ | |
650 | if (c == '\\' && (quoted == 0 || quoted == '"' || quoted == '`')) | |
651 | { | |
652 | pass_next = 1; | |
653 | i++; | |
654 | continue; | |
655 | } | |
656 | ||
657 | #if defined (SHELL) | |
658 | /* If compiling for the shell, treat ${...} like \{...} */ | |
659 | if (c == '$' && text[i+1] == '{' && quoted != '\'') /* } */ | |
660 | { | |
661 | pass_next = 1; | |
662 | i++; | |
663 | if (quoted == 0) | |
664 | level++; | |
665 | continue; | |
666 | } | |
667 | #endif | |
668 | ||
669 | if (quoted) | |
670 | { | |
671 | if (c == quoted) | |
672 | quoted = 0; | |
673 | #if defined (SHELL) | |
674 | /* The shell allows quoted command substitutions */ | |
675 | if (quoted == '"' && c == '$' && text[i+1] == '(') /*)*/ | |
676 | goto comsub; | |
677 | #endif | |
678 | #if defined (SHELL) | |
679 | ADVANCE_CHAR (text, tlen, i); | |
680 | #else | |
681 | i++; | |
682 | #endif | |
683 | continue; | |
684 | } | |
685 | ||
686 | if (c == '"' || c == '\'' || c == '`') | |
687 | { | |
688 | quoted = c; | |
689 | i++; | |
690 | continue; | |
691 | } | |
692 | ||
693 | #if defined (SHELL) | |
694 | /* Pass new-style command and process substitutions through unchanged. */ | |
695 | if ((c == '$' || c == '<' || c == '>') && text[i+1] == '(') /* ) */ | |
696 | { | |
697 | comsub: | |
698 | si = i + 2; | |
699 | t = extract_command_subst (text, &si, 0); | |
700 | i = si; | |
701 | free (t); | |
702 | i++; | |
703 | continue; | |
704 | } | |
705 | #endif | |
706 | ||
707 | if (c == satisfy && level == 0 && quoted == 0 && commas > 0) | |
708 | { | |
709 | /* We ignore an open brace surrounded by whitespace, and also | |
710 | an open brace followed immediately by a close brace preceded | |
711 | by whitespace. */ | |
712 | if (c == '{' && | |
713 | ((!i || brace_whitespace (text[i - 1])) && | |
714 | (brace_whitespace (text[i + 1]) || text[i + 1] == '}'))) | |
715 | { | |
716 | i++; | |
717 | continue; | |
718 | } | |
719 | ||
720 | break; | |
721 | } | |
722 | ||
723 | if (c == '{') | |
724 | level++; | |
725 | else if (c == '}' && level) | |
726 | level--; | |
727 | #if !defined (CSH_BRACE_COMPAT) | |
728 | else if (satisfy == '}' && c == brace_arg_separator && level == 0) | |
729 | commas++; | |
730 | else if (satisfy == '}' && STREQN (text+i, BRACE_SEQ_SPECIFIER, 2) && | |
731 | text[i+2] != satisfy && level == 0) | |
732 | commas++; | |
733 | #endif | |
734 | ||
735 | #if defined (SHELL) | |
736 | ADVANCE_CHAR (text, tlen, i); | |
737 | #else | |
738 | i++; | |
739 | #endif | |
740 | } | |
741 | ||
742 | *indx = i; | |
743 | return (c); | |
744 | } | |
745 | ||
746 | /* Return 1 if ARR has any non-empty-string members. Used to short-circuit | |
747 | in array_concat() below. */ | |
748 | static int | |
749 | degenerate_array (arr) | |
750 | char **arr; | |
751 | { | |
752 | register int i; | |
753 | ||
754 | for (i = 0; arr[i]; i++) | |
755 | if (arr[i][0] != '\0') | |
756 | return 0; | |
757 | return 1; | |
758 | } | |
759 | ||
760 | /* Return a new array of strings which is the result of appending each | |
761 | string in ARR2 to each string in ARR1. The resultant array is | |
762 | len (arr1) * len (arr2) long. For convenience, ARR1 (and its contents) | |
763 | are free ()'ed. ARR1 can be NULL, in that case, a new version of ARR2 | |
764 | is returned. */ | |
765 | static char ** | |
766 | array_concat (arr1, arr2) | |
767 | char **arr1, **arr2; | |
768 | { | |
769 | register int i, j, len, len1, len2; | |
770 | register char **result; | |
771 | ||
772 | if (arr1 == 0) | |
773 | return (arr2); /* XXX - see if we can get away without copying? */ | |
774 | ||
775 | if (arr2 == 0) | |
776 | return (arr1); /* XXX - caller expects us to free arr1 */ | |
777 | ||
778 | /* We can only short-circuit if the array consists of a single null element; | |
779 | otherwise we need to replicate the contents of the other array and | |
780 | prefix (or append, below) an empty element to each one. */ | |
781 | if (arr1[0] && arr1[0][0] == 0 && arr1[1] == 0) | |
782 | { | |
783 | strvec_dispose (arr1); | |
784 | return (arr2); /* XXX - use flags to see if we can avoid copying here */ | |
785 | } | |
786 | ||
787 | if (arr2[0] && arr2[0][0] == 0 && arr2[1] == 0) | |
788 | return (arr1); /* XXX - rather than copying and freeing it */ | |
789 | ||
790 | len1 = strvec_len (arr1); | |
791 | len2 = strvec_len (arr2); | |
792 | ||
793 | result = (char **)xmalloc ((1 + (len1 * len2)) * sizeof (char *)); | |
794 | ||
795 | len = 0; | |
796 | for (i = 0; i < len1; i++) | |
797 | { | |
798 | int strlen_1 = strlen (arr1[i]); | |
799 | ||
800 | for (j = 0; j < len2; j++) | |
801 | { | |
802 | result[len] = (char *)xmalloc (1 + strlen_1 + strlen (arr2[j])); | |
803 | strcpy (result[len], arr1[i]); | |
804 | strcpy (result[len] + strlen_1, arr2[j]); | |
805 | len++; | |
806 | } | |
807 | free (arr1[i]); | |
808 | } | |
809 | free (arr1); | |
810 | ||
811 | result[len] = (char *)NULL; | |
812 | return (result); | |
813 | } | |
814 | ||
815 | #if defined (TEST) | |
816 | #include <stdio.h> | |
817 | ||
818 | void * | |
819 | xmalloc(n) | |
820 | size_t n; | |
821 | { | |
822 | return (malloc (n)); | |
823 | } | |
824 | ||
825 | void * | |
826 | xrealloc(p, n) | |
827 | void *p; | |
828 | size_t n; | |
829 | { | |
830 | return (realloc (p, n)); | |
831 | } | |
832 | ||
833 | int | |
834 | internal_error (format, arg1, arg2) | |
835 | char *format, *arg1, *arg2; | |
836 | { | |
837 | fprintf (stderr, format, arg1, arg2); | |
838 | fprintf (stderr, "\n"); | |
839 | } | |
840 | ||
841 | main () | |
842 | { | |
843 | char example[256]; | |
844 | ||
845 | for (;;) | |
846 | { | |
847 | char **result; | |
848 | int i; | |
849 | ||
850 | fprintf (stderr, "brace_expand> "); | |
851 | ||
852 | if ((!fgets (example, 256, stdin)) || | |
853 | (strncmp (example, "quit", 4) == 0)) | |
854 | break; | |
855 | ||
856 | if (strlen (example)) | |
857 | example[strlen (example) - 1] = '\0'; | |
858 | ||
859 | result = brace_expand (example); | |
860 | ||
861 | for (i = 0; result[i]; i++) | |
862 | printf ("%s\n", result[i]); | |
863 | ||
864 | strvec_dispose (result); | |
865 | } | |
866 | } | |
867 | \f | |
868 | /* | |
869 | * Local variables: | |
870 | * compile-command: "gcc -g -Bstatic -DTEST -o brace_expand braces.c general.o" | |
871 | * end: | |
872 | */ | |
873 | ||
874 | #endif /* TEST */ | |
875 | #endif /* BRACE_EXPANSION */ |