Annotation of texi2mdoc/util.c, Revision 1.12
1.12 ! kristaps 1: /* $Id: util.c,v 1.11 2015/02/23 17:24:51 kristaps Exp $ */
1.1 kristaps 2: /*
3: * Copyright (c) 2015 Kristaps Dzonsons <kristaps@bsd.lv>
4: *
5: * Permission to use, copy, modify, and distribute this software for any
6: * purpose with or without fee is hereby granted, provided that the above
7: * copyright notice and this permission notice appear in all copies.
8: *
9: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
10: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
11: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
12: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
13: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
14: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
15: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
16: */
17: #include <sys/mman.h>
18: #include <sys/stat.h>
19:
20: #include <assert.h>
21: #include <ctype.h>
22: #include <fcntl.h>
23: #include <getopt.h>
24: #include <libgen.h>
25: #include <limits.h>
26: #include <stdarg.h>
27: #include <stdio.h>
28: #include <stdlib.h>
29: #include <string.h>
30: #include <time.h>
31: #include <unistd.h>
32:
33: #include "extern.h"
34:
35: /*
36: * Unmap the top-most file in the stack of files currently opened (that
37: * is, nested calls to parsefile()).
38: */
39: void
40: texifilepop(struct texi *p)
41: {
42: struct texifile *f;
43:
44: assert(p->filepos > 0);
45: f = &p->files[--p->filepos];
1.12 ! kristaps 46: if (TEXISRC_FILE == f->type)
! 47: munmap(f->map, f->mapsz);
! 48: else
! 49: free(f->map);
1.1 kristaps 50: }
51:
1.7 kristaps 52: static void
53: teximacrofree(struct teximacro *p)
54: {
55: size_t i;
56:
57: for (i = 0; i < p->argsz; i++)
58: free(p->args[i]);
59:
60: free(p->args);
61: free(p->key);
62: free(p->value);
63: }
64:
65: static void
66: texivaluefree(struct texivalue *p)
67: {
68:
69: free(p->key);
70: free(p->value);
71: }
72:
1.1 kristaps 73: /*
74: * Unmap all files that we're currently using and free all resources
75: * that we've allocated during the parse.
76: * The utility should exit(...) after this is called.
77: */
78: void
79: texiexit(struct texi *p)
80: {
81: size_t i;
82:
83: /* Make sure we're newline-terminated. */
84: if (p->outcol)
85: putchar('\n');
86:
87: /* Unmap all files. */
88: while (p->filepos > 0)
89: texifilepop(p);
90:
1.7 kristaps 91: for (i = 0; i < p->macrosz; i++)
92: teximacrofree(&p->macros[i]);
1.1 kristaps 93: for (i = 0; i < p->dirsz; i++)
94: free(p->dirs[i]);
1.4 kristaps 95: for (i = 0; i < p->indexsz; i++)
96: free(p->indexs[i]);
1.7 kristaps 97: for (i = 0; i < p->valsz; i++)
98: texivaluefree(&p->vals[i]);
1.4 kristaps 99:
1.7 kristaps 100: free(p->macros);
1.1 kristaps 101: free(p->vals);
1.4 kristaps 102: free(p->indexs);
1.1 kristaps 103: free(p->dirs);
104: free(p->subtitle);
105: free(p->title);
106: }
107:
108: /*
109: * Fatal error: unmap all files and exit.
110: * The "errstring" is passed to perror(3).
111: */
112: void
113: texiabort(struct texi *p, const char *errstring)
114: {
115:
116: perror(errstring);
117: texiexit(p);
118: exit(EXIT_FAILURE);
119: }
120:
121: /*
122: * Print a generic warning message (to stderr) tied to our current
123: * location in the parse sequence.
124: */
125: void
126: texiwarn(const struct texi *p, const char *fmt, ...)
127: {
128: va_list ap;
129:
130: fprintf(stderr, "%s:%zu:%zu: warning: ",
131: p->files[p->filepos - 1].name,
132: p->files[p->filepos - 1].line + 1,
133: p->files[p->filepos - 1].col + 1);
134: va_start(ap, fmt);
135: vfprintf(stderr, fmt, ap);
136: va_end(ap);
137: fputc('\n', stderr);
138: }
139:
140: /*
141: * Print an error message (to stderr) tied to our current location in
142: * the parse sequence, invoke texiexit(), then die.
143: */
144: void
145: texierr(struct texi *p, const char *fmt, ...)
146: {
147: va_list ap;
148:
149: fprintf(stderr, "%s:%zu:%zu: error: ",
150: p->files[p->filepos - 1].name,
151: p->files[p->filepos - 1].line + 1,
152: p->files[p->filepos - 1].col + 1);
153: va_start(ap, fmt);
154: vfprintf(stderr, fmt, ap);
155: va_end(ap);
156: fputc('\n', stderr);
157: texiexit(p);
158: exit(EXIT_FAILURE);
159: }
160:
161: /*
162: * Put a single data character to the output if we're not ignoring.
163: * Makes sure we don't spurriously start a macro.
164: * Adjusts our output status.
165: * This shouldn't be called for macros: just for ordinary text.
166: */
167: void
168: texiputchar(struct texi *p, char c)
169: {
170:
171: if (p->ign)
172: return;
173:
174: if ('.' == c && 0 == p->outcol)
175: fputs("\\&", stdout);
1.10 kristaps 176: if ('\'' == c && 0 == p->outcol)
177: fputs("\\&", stdout);
1.1 kristaps 178:
179: putchar(c);
180: p->seenvs = 0;
181: if ('\n' == c) {
182: p->outcol = 0;
183: p->seenws = 0;
184: } else
185: p->outcol++;
186: }
187:
188: /*
189: * Put multiple characters (see texiputchar()).
190: * This shouldn't be called for macros: just for ordinary text.
191: */
192: void
193: texiputchars(struct texi *p, const char *s)
194: {
195:
196: while ('\0' != *s)
197: texiputchar(p, *s++);
1.9 kristaps 198: }
199:
200: /*
201: * This puts all characters onto the output stream but makes sure to
202: * escape mdoc(7) slashes.
203: */
204: void
205: texiputbuf(struct texi *p, const char *buf, size_t start, size_t end)
206: {
207:
208: for ( ; start < end; start++) {
209: texiputchar(p, buf[start]);
210: if ('\\' == buf[start])
211: texiputchar(p, 'e');
212: }
1.1 kristaps 213: }
214:
215: /*
216: * Close an mdoc(7) macro opened with teximacroopen().
217: * If there are no more macros on the line, prints a newline.
218: */
219: void
220: teximacroclose(struct texi *p)
221: {
222:
223: if (p->ign)
224: return;
225:
226: if (0 == --p->outmacro) {
227: putchar('\n');
228: p->outcol = p->seenws = 0;
229: }
230: }
231:
232: /*
233: * Open a mdoc(7) macro.
234: * This is used for line macros, e.g., Qq [foo bar baz].
235: * It can be invoked for nested macros, e.g., Qq Li foo .
236: * TODO: flush-right punctuation (e.g., parenthesis).
237: */
238: void
239: teximacroopen(struct texi *p, const char *s)
240: {
241: int rc;
242:
243: if (p->ign)
244: return;
245:
246: if (p->outcol && 0 == p->outmacro) {
247: putchar('\n');
248: p->outcol = 0;
249: }
250:
251: if (0 == p->outmacro)
252: putchar('.');
253: else
254: putchar(' ');
255:
256: if (EOF != (rc = fputs(s, stdout)))
257: p->outcol += rc;
258:
259: putchar(' ');
260: p->outcol++;
261: p->outmacro++;
262: p->seenws = 0;
263: }
264:
265: /*
266: * Put a stadnalone mdoc(7) command with the trailing newline.
267: */
268: void
269: teximacro(struct texi *p, const char *s)
270: {
271:
272: if (p->ign)
273: return;
274:
275: if (p->outmacro)
276: texierr(p, "\"%s\" in open line scope!?", s);
277: if (p->literal)
278: texierr(p, "\"%s\" in a literal scope!?", s);
279:
280: if (p->outcol)
281: putchar('\n');
282:
283: putchar('.');
284: puts(s);
285: p->outcol = p->seenws = 0;
286: }
287:
288: /*
289: * Introduce vertical space during normal (non-macro) input.
290: */
291: void
292: texivspace(struct texi *p)
293: {
294:
1.5 kristaps 295: if (p->seenvs || TEXILIST_TABLE == p->list)
1.1 kristaps 296: return;
297: teximacro(p, "Pp");
298: p->seenvs = 1;
299: }
300:
301: /*
302: * Advance by a single byte in the input stream, adjusting our location
303: * in the current input file.
304: */
305: void
306: advance(struct texi *p, const char *buf, size_t *pos)
307: {
308:
309: if ('\n' == buf[*pos]) {
310: p->files[p->filepos - 1].line++;
311: p->files[p->filepos - 1].col = 0;
312: } else
313: p->files[p->filepos - 1].col++;
314:
315: (*pos)++;
316: }
317:
318: /*
319: * It's common to wait punctuation to float on the right side of macro
320: * lines in mdoc(7), e.g., ".Em hello ) ."
321: * This function does so, and should be called before teximacroclose().
322: * It will detect that it's the last in the nested macros and
323: * appropriately flush-left punctuation alongside the macro.
324: */
325: void
326: texipunctuate(struct texi *p, const char *buf, size_t sz, size_t *pos)
327: {
328: size_t start, end;
329:
330: if (1 != p->outmacro)
331: return;
332:
333: for (start = end = *pos; end < sz; end++) {
334: switch (buf[end]) {
335: case (','):
336: case (')'):
337: case ('.'):
338: case ('"'):
339: case (':'):
340: case ('!'):
341: case ('?'):
342: continue;
343: default:
344: break;
345: }
346: break;
347: }
348: if (end == *pos)
349: return;
350: if (end + 1 == sz || ' ' == buf[end] || '\n' == buf[end]) {
351: for ( ; start < end; start++) {
352: texiputchar(p, ' ');
353: texiputchar(p, buf[start]);
354: advance(p, buf, pos);
355: }
356: }
357: }
358:
359: /*
360: * Advance to the next non-whitespace word in the input stream.
361: * If we're in literal mode, then print all of the whitespace as we're
362: * doing so.
363: */
364: static size_t
365: advancenext(struct texi *p, const char *buf, size_t sz, size_t *pos)
366: {
367:
368: if (p->literal) {
369: while (*pos < sz && ismspace(buf[*pos])) {
370: if (*pos && '\n' == buf[*pos] &&
371: '\\' == buf[*pos - 1])
372: texiputchar(p, 'e');
373: texiputchar(p, buf[*pos]);
374: advance(p, buf, pos);
375: }
376: return(*pos);
377: }
378:
379: while (*pos < sz && ismspace(buf[*pos])) {
380: p->seenws = 1;
381: /*
382: * If it looks like we've printed a double-line, then
383: * output a paragraph.
384: * FIXME: this is stupid.
385: */
386: if (*pos && '\n' == buf[*pos] && '\n' == buf[*pos - 1])
387: texivspace(p);
388: advance(p, buf, pos);
389: }
390: return(*pos);
391: }
392:
393: /*
394: * Advance to the EOLN in the input stream.
395: * NOTE: THIS SHOULD NOT BE CALLED ON BLANK TEXT, as it will read up to
396: * the @\n.
397: */
398: size_t
399: advanceeoln(struct texi *p, const char *buf,
400: size_t sz, size_t *pos, int consumenl)
401: {
402:
403: while (*pos < sz && '\n' != buf[*pos])
404: advance(p, buf, pos);
405: if (*pos < sz && consumenl)
406: advance(p, buf, pos);
407: return(*pos);
408: }
409:
410: /*
411: * Advance to position "end", which is an absolute position in the
412: * current buffer greater than or equal to the current position.
413: */
414: void
415: advanceto(struct texi *p, const char *buf, size_t *pos, size_t end)
416: {
417:
418: assert(*pos <= end);
419: while (*pos < end)
420: advance(p, buf, pos);
421: }
422:
1.7 kristaps 423: static void
424: texiexecmacro(struct texi *p, struct teximacro *m,
425: const char *buf, size_t sz, size_t *pos)
426: {
1.11 kristaps 427: size_t valsz, realsz, aasz, asz,
428: ssz, i, j, k, start, end;
429: char *val;
430: char **args;
431: const char *cp;
1.7 kristaps 432:
1.8 kristaps 433: args = argparse(p, buf, sz, pos, &asz, m->argsz);
1.7 kristaps 434: if (asz != m->argsz)
435: texiwarn(p, "invalid macro argument length");
436: aasz = asz < m->argsz ? asz : m->argsz;
437:
438: if (0 == aasz) {
1.8 kristaps 439: parsemembuf(p, m->value, strlen(m->value));
1.7 kristaps 440: return;
441: }
442:
443: valsz = realsz = strlen(m->value);
444: val = strdup(m->value);
445:
446: for (i = j = 0; i < realsz; i++) {
447: /* Parse blindly til the backslash delimiter. */
448: if ('\\' != m->value[i]) {
449: val[j++] = m->value[i];
450: val[j] = '\0';
451: continue;
452: } else if (i == realsz - 1)
453: texierr(p, "trailing argument name delimiter");
454:
455: /* Double-backslash is escaped. */
456: if ('\\' == m->value[i + 1]) {
457: val[j++] = m->value[i++];
458: val[j] = '\0';
459: continue;
460: }
461:
462: assert('\\' == m->value[i] && i < realsz - 1);
463:
464: /* Parse to terminating delimiter. */
465: /* FIXME: embedded, escaped delimiters? */
466: for (start = end = i + 1; end < realsz; end++)
467: if ('\\' == m->value[end])
468: break;
469: if (end == realsz)
470: texierr(p, "unterminated argument name");
471:
472: for (k = 0; k < aasz; k++) {
473: if ((ssz = strlen(m->args[k])) != (end - start))
474: continue;
475: if (strncmp(&m->value[start], m->args[k], ssz))
476: continue;
477: break;
478: }
479:
480: /*
481: * Argument didn't exist in argument table.
482: * No need to reallocate here: we just copy the text
483: * directly from the macro value into the buffer.
484: */
485: if (k == aasz) {
486: for ( ; i < end; i++)
487: val[j++] = m->value[i];
488: assert('\\' == m->value[i]);
489: val[j++] = m->value[i];
490: val[j] = '\0';
491: continue;
492: }
493:
494: if (strlen(args[k]) > ssz) {
495: valsz += strlen(args[k]);
496: val = realloc(val, valsz + 1);
497: if (NULL == val)
498: texiabort(p, NULL);
499: }
500:
1.11 kristaps 501: for (cp = args[k]; '\0' != *cp; cp++)
502: val[j++] = *cp;
503:
504: val[j] = '\0';
1.7 kristaps 505: i = end;
506: }
507:
1.8 kristaps 508: parsemembuf(p, val, strlen(val));
1.7 kristaps 509:
510: for (i = 0; i < asz; i++)
511: free(args[i]);
512: free(args);
513: free(val);
514: }
515:
1.1 kristaps 516: /*
517: * Output a free-form word in the input stream, progressing to the next
518: * command or white-space.
519: * This also will advance the input stream.
520: */
521: static void
522: texiword(struct texi *p, const char *buf,
523: size_t sz, size_t *pos, char extra)
524: {
525:
526: if (p->seenws && 0 == p->outmacro &&
527: p->outcol > 72 && 0 == p->literal)
528: texiputchar(p, '\n');
529: /* FIXME: abstract this: we use it elsewhere. */
530: if (p->seenws && p->outcol && 0 == p->literal)
531: texiputchar(p, ' ');
532:
533: p->seenws = 0;
534:
535: while (*pos < sz && ! ismspace(buf[*pos])) {
536: switch (buf[*pos]) {
537: case ('@'):
538: case ('}'):
539: case ('{'):
540: return;
541: }
542: if ('\0' != extra && buf[*pos] == extra)
543: return;
544: if (*pos < sz - 1 &&
545: '`' == buf[*pos] &&
546: '`' == buf[*pos + 1]) {
547: texiputchars(p, "\\(lq");
548: advance(p, buf, pos);
549: } else if (*pos < sz - 1 &&
550: '\'' == buf[*pos] &&
551: '\'' == buf[*pos + 1]) {
552: texiputchars(p, "\\(rq");
553: advance(p, buf, pos);
1.10 kristaps 554: } else if ('\\' == buf[*pos]) {
555: texiputchar(p, buf[*pos]);
556: texiputchar(p, 'e');
1.1 kristaps 557: } else
558: texiputchar(p, buf[*pos]);
559: advance(p, buf, pos);
560: }
561: }
562:
563: /*
564: * Look up the command at position "pos" in the buffer, returning it (or
565: * TEXICMD__MAX if none found) and setting "end" to be the absolute
566: * index after the command name.
567: */
568: enum texicmd
1.7 kristaps 569: texicmd(struct texi *p, const char *buf, size_t pos,
570: size_t sz, size_t *end, struct teximacro **macro)
1.1 kristaps 571: {
1.4 kristaps 572: size_t i, len, toksz;
1.1 kristaps 573:
574: assert('@' == buf[pos]);
575:
1.7 kristaps 576: if (NULL != macro)
577: *macro = NULL;
578:
1.1 kristaps 579: if ((*end = pos) == sz)
580: return(TEXICMD__MAX);
581: else if ((*end = ++pos) == sz)
582: return(TEXICMD__MAX);
583:
584: /* Alphabetic commands are special. */
585: if ( ! isalpha(buf[pos])) {
586: if ((*end = pos + 1) == sz)
587: return(TEXICMD__MAX);
588: for (i = 0; i < TEXICMD__MAX; i++) {
589: if (1 != texitoks[i].len)
590: continue;
591: if (0 == strncmp(texitoks[i].tok, &buf[pos], 1))
592: return(i);
593: }
594: texiwarn(p, "bad command: @%c", buf[pos]);
595: return(TEXICMD__MAX);
596: }
597:
1.4 kristaps 598: /* Scan to the end of the possible command name. */
1.1 kristaps 599: for (*end = pos; *end < sz && ! ismspace(buf[*end]); (*end)++)
600: if ((*end > pos && ('@' == buf[*end] ||
601: '{' == buf[*end] || '}' == buf[*end])))
602: break;
603:
1.4 kristaps 604: /* Look for the command. */
1.1 kristaps 605: len = *end - pos;
606: for (i = 0; i < TEXICMD__MAX; i++) {
607: if (len != texitoks[i].len)
608: continue;
609: if (0 == strncmp(texitoks[i].tok, &buf[pos], len))
610: return(i);
611: }
612:
1.4 kristaps 613: /* Look for it in our indices. */
614: for (i = 0; i < p->indexsz; i++) {
615: toksz = strlen(p->indexs[i]);
616: if (len != 5 + toksz)
617: continue;
618: if (strncmp(&buf[pos], p->indexs[i], toksz))
619: continue;
620: if (0 == strncmp(&buf[pos + toksz], "index", 5))
1.7 kristaps 621: return(TEXICMD_USER_INDEX);
622: }
623:
624: for (i = 0; i < p->macrosz; i++) {
625: if (len != strlen(p->macros[i].key))
626: continue;
627: if (strncmp(&buf[pos], p->macros[i].key, len))
628: continue;
629: if (NULL != macro)
630: *macro = &p->macros[i];
631: return(TEXICMD__MAX);
1.4 kristaps 632: }
633:
1.1 kristaps 634: texiwarn(p, "bad command: @%.*s", (int)len, &buf[pos]);
635: return(TEXICMD__MAX);
636: }
637:
638: /*
639: * Parse an argument from a bracketed command, e.g., @url{foo, baz}.
640: * Num should be set to the argument we're currently parsing, although
641: * it suffixes for it to be zero or non-zero.
642: * This will return 1 if there are more arguments, 0 otherwise.
643: * This will stop (returning 0) in the event of EOF or if we're not at a
644: * bracket for the zeroth parse.
645: */
646: int
647: parsearg(struct texi *p, const char *buf,
648: size_t sz, size_t *pos, size_t num)
649: {
1.7 kristaps 650: size_t end;
651: enum texicmd cmd;
652: struct teximacro *macro;
1.1 kristaps 653:
654: while (*pos < sz && ismspace(buf[*pos]))
655: advance(p, buf, pos);
656: if (*pos == sz || (0 == num && '{' != buf[*pos]))
657: return(0);
658: if (0 == num)
659: advance(p, buf, pos);
660:
661: while ((*pos = advancenext(p, buf, sz, pos)) < sz) {
662: switch (buf[*pos]) {
663: case (','):
664: advance(p, buf, pos);
665: return(1);
666: case ('}'):
667: advance(p, buf, pos);
668: return(0);
669: case ('{'):
670: if (0 == p->ign)
671: texiwarn(p, "unexpected \"{\"");
672: advance(p, buf, pos);
673: continue;
674: case ('@'):
675: break;
676: default:
677: texiword(p, buf, sz, pos, ',');
678: continue;
679: }
680:
1.7 kristaps 681: cmd = texicmd(p, buf, *pos, sz, &end, ¯o);
1.1 kristaps 682: advanceto(p, buf, pos, end);
1.7 kristaps 683: if (NULL != macro)
684: texiexecmacro(p, macro, buf, sz, pos);
1.1 kristaps 685: if (TEXICMD__MAX == cmd)
686: continue;
687: if (NULL != texitoks[cmd].fp)
688: (*texitoks[cmd].fp)(p, cmd, buf, sz, pos);
689: }
690: return(0);
691: }
692:
693: /*
694: * Parse until the end of a bracketed statement, e.g., @foo{bar baz}.
695: * This will stop in the event of EOF or if we're not at a bracket.
696: */
697: void
698: parsebracket(struct texi *p, const char *buf, size_t sz, size_t *pos)
699: {
1.7 kristaps 700: size_t end;
701: enum texicmd cmd;
702: struct teximacro *macro;
1.1 kristaps 703:
704: while (*pos < sz && ismspace(buf[*pos]))
705: advance(p, buf, pos);
706:
707: if (*pos == sz || '{' != buf[*pos])
708: return;
709: advance(p, buf, pos);
710:
711: while ((*pos = advancenext(p, buf, sz, pos)) < sz) {
712: switch (buf[*pos]) {
713: case ('}'):
714: advance(p, buf, pos);
715: return;
716: case ('{'):
717: if (0 == p->ign)
718: texiwarn(p, "unexpected \"{\"");
719: advance(p, buf, pos);
720: continue;
721: case ('@'):
722: break;
723: default:
724: texiword(p, buf, sz, pos, '\0');
725: continue;
726: }
727:
1.7 kristaps 728: cmd = texicmd(p, buf, *pos, sz, &end, ¯o);
1.1 kristaps 729: advanceto(p, buf, pos, end);
1.7 kristaps 730: if (NULL != macro)
731: texiexecmacro(p, macro, buf, sz, pos);
1.1 kristaps 732: if (TEXICMD__MAX == cmd)
733: continue;
734: if (NULL != texitoks[cmd].fp)
735: (*texitoks[cmd].fp)(p, cmd, buf, sz, pos);
736: }
737: }
738:
739: /*
740: * This should be invoked when we're on a macro line and want to process
741: * to the end of the current input line, doing all of our macros along
742: * the way.
743: */
744: void
745: parseeoln(struct texi *p, const char *buf, size_t sz, size_t *pos)
746: {
1.7 kristaps 747: size_t end;
748: enum texicmd cmd;
749: struct teximacro *macro;
1.1 kristaps 750:
751: while (*pos < sz && '\n' != buf[*pos]) {
752: while (*pos < sz && isws(buf[*pos])) {
753: p->seenws = 1;
754: if (p->literal)
755: texiputchar(p, buf[*pos]);
756: advance(p, buf, pos);
757: }
758: switch (buf[*pos]) {
759: case ('}'):
760: if (0 == p->ign)
761: texiwarn(p, "unexpected \"}\"");
762: advance(p, buf, pos);
763: continue;
764: case ('{'):
765: if (0 == p->ign)
766: texiwarn(p, "unexpected \"{\"");
767: advance(p, buf, pos);
768: continue;
769: case ('@'):
770: break;
771: default:
772: texiword(p, buf, sz, pos, '\0');
773: continue;
774: }
775:
1.7 kristaps 776: cmd = texicmd(p, buf, *pos, sz, &end, ¯o);
1.1 kristaps 777: advanceto(p, buf, pos, end);
1.7 kristaps 778: if (NULL != macro)
779: texiexecmacro(p, macro, buf, sz, pos);
1.1 kristaps 780: if (TEXICMD__MAX == cmd)
781: continue;
782: if (NULL != texitoks[cmd].fp)
783: (*texitoks[cmd].fp)(p, cmd, buf, sz, pos);
784: }
785: }
786:
787: /*
788: * Parse a single word or command.
789: * This will return immediately at the EOF.
790: */
791: void
792: parsesingle(struct texi *p, const char *buf, size_t sz, size_t *pos)
793: {
1.7 kristaps 794: size_t end;
795: enum texicmd cmd;
796: struct teximacro *macro;
1.1 kristaps 797:
798: if ((*pos = advancenext(p, buf, sz, pos)) >= sz)
799: return;
800:
801: switch (buf[*pos]) {
802: case ('}'):
803: if (0 == p->ign)
804: texiwarn(p, "unexpected \"}\"");
805: advance(p, buf, pos);
806: return;
807: case ('{'):
808: if (0 == p->ign)
809: texiwarn(p, "unexpected \"{\"");
810: advance(p, buf, pos);
811: return;
812: case ('@'):
813: break;
814: default:
815: texiword(p, buf, sz, pos, '\0');
816: return;
817: }
818:
1.7 kristaps 819: cmd = texicmd(p, buf, *pos, sz, &end, ¯o);
1.1 kristaps 820: advanceto(p, buf, pos, end);
1.7 kristaps 821: if (NULL != macro)
822: texiexecmacro(p, macro, buf, sz, pos);
1.1 kristaps 823: if (TEXICMD__MAX == cmd)
824: return;
825: if (NULL != texitoks[cmd].fp)
826: (*texitoks[cmd].fp)(p, cmd, buf, sz, pos);
827: }
828:
829: /*
830: * This is used in the @deffn type of command.
831: * These have an arbitrary number of line arguments; however, these
832: * arguments may or may not be surrounded by brackets.
833: * In this function, we parse each one as either a bracketed or
834: * non-bracketed argument, returning 0 when we've reached the end of
835: * line or 1 otherwise.
836: */
837: int
838: parselinearg(struct texi *p, const char *buf, size_t sz, size_t *pos)
839: {
840:
841: while (*pos < sz && isws(buf[*pos])) {
842: p->seenws = 1;
843: advance(p, buf, pos);
844: }
845:
846: if (*pos < sz && '{' == buf[*pos])
847: parsebracket(p, buf, sz, pos);
1.3 kristaps 848: else if (*pos < sz && '\n' != buf[*pos])
1.1 kristaps 849: parsesingle(p, buf, sz, pos);
850: else
851: return(0);
852:
853: return(1);
854: }
855:
856: /*
857: * Parse til the end of the buffer.
858: */
859: void
860: parseeof(struct texi *p, const char *buf, size_t sz)
861: {
862: size_t pos;
863:
864: for (pos = 0; pos < sz; )
865: parsesingle(p, buf, sz, &pos);
866: }
867:
868: /*
1.8 kristaps 869: * This is like parseeof() except that it's to be invoked on memory
870: * buffers while parsing a larger scope.
871: * This is useful for parsing macro sequences.
872: * The line, column, and name of the calling file context are saved, the
873: * column and line reset, then all of these restored after parse.
874: */
875: void
876: parsemembuf(struct texi *p, const char *buf, size_t sz)
877: {
878: size_t svln, svcol;
879: const char *svname;
880:
881: svln = p->files[p->filepos - 1].line;
882: svcol = p->files[p->filepos - 1].col;
883: svname = p->files[p->filepos - 1].name;
884:
885: p->files[p->filepos - 1].line = 0;
886: p->files[p->filepos - 1].col = 0;
887: p->files[p->filepos - 1].name = "<macro buffer>";
888:
889: parseeof(p, buf, sz);
890:
891: p->files[p->filepos - 1].line = svln;
892: p->files[p->filepos - 1].col = svcol;
893: p->files[p->filepos - 1].name = svname;
894: }
895:
896: /*
1.1 kristaps 897: * Parse a block sequence until we have the "@end endtoken" command
898: * invocation.
899: * This will return immediately at EOF.
900: */
901: void
902: parseto(struct texi *p, const char *buf,
903: size_t sz, size_t *pos, const char *endtoken)
904: {
1.7 kristaps 905: size_t end;
906: enum texicmd cmd;
907: size_t endtoksz;
908: struct teximacro *macro;
1.1 kristaps 909:
910: endtoksz = strlen(endtoken);
911: assert(endtoksz > 0);
912:
913: while ((*pos = advancenext(p, buf, sz, pos)) < sz) {
914: switch (buf[*pos]) {
915: case ('}'):
916: if (0 == p->ign)
917: texiwarn(p, "unexpected \"}\"");
918: advance(p, buf, pos);
919: continue;
920: case ('{'):
921: if (0 == p->ign)
922: texiwarn(p, "unexpected \"{\"");
923: advance(p, buf, pos);
924: continue;
925: case ('@'):
926: break;
927: default:
928: texiword(p, buf, sz, pos, '\0');
929: continue;
930: }
931:
1.7 kristaps 932: cmd = texicmd(p, buf, *pos, sz, &end, ¯o);
1.1 kristaps 933: advanceto(p, buf, pos, end);
934: if (TEXICMD_END == cmd) {
935: while (*pos < sz && isws(buf[*pos]))
936: advance(p, buf, pos);
937: /*
938: * FIXME: check the full word, not just its
939: * initial substring!
940: */
941: if (sz - *pos >= endtoksz && 0 == strncmp
942: (&buf[*pos], endtoken, endtoksz)) {
943: advanceeoln(p, buf, sz, pos, 0);
944: break;
945: }
946: if (0 == p->ign)
947: texiwarn(p, "unexpected \"end\"");
948: advanceeoln(p, buf, sz, pos, 0);
949: continue;
1.7 kristaps 950: }
951: if (NULL != macro)
952: texiexecmacro(p, macro, buf, sz, pos);
953: if (TEXICMD__MAX == cmd)
954: continue;
955: if (NULL != texitoks[cmd].fp)
956: (*texitoks[cmd].fp)(p, cmd, buf, sz, pos);
1.1 kristaps 957: }
958: }
959:
960: /*
1.12 ! kristaps 961: * Like parsefile() but used for reading from stdandard input.
! 962: * This can only be called for the first file!
! 963: */
! 964: void
! 965: parsestdin(struct texi *p)
! 966: {
! 967: struct texifile *f;
! 968: size_t off;
! 969: ssize_t ssz;
! 970:
! 971: assert(0 == p->filepos);
! 972: f = &p->files[p->filepos];
! 973: memset(f, 0, sizeof(struct texifile));
! 974:
! 975: f->type = TEXISRC_STDIN;
! 976: f->name = "<stdin>";
! 977:
! 978: for (off = 0; ; off += (size_t)ssz) {
! 979: if (off == f->mapsz) {
! 980: if (f->mapsz == (1U << 31))
! 981: texierr(p, "stdin buffer too long");
! 982: f->mapsz = f->mapsz > 65536 / 2 ?
! 983: 2 * f->mapsz : 65536;
! 984: f->map = realloc(f->map, f->mapsz);
! 985: if (NULL == f->map)
! 986: texiabort(p, NULL);
! 987: }
! 988: ssz = read(STDIN_FILENO,
! 989: f->map + (int)off, f->mapsz - off);
! 990: if (0 == ssz)
! 991: break;
! 992: else if (-1 == ssz)
! 993: texiabort(p, NULL);
! 994: }
! 995:
! 996: p->filepos++;
! 997: parseeof(p, f->map, off);
! 998: texifilepop(p);
! 999: }
! 1000:
! 1001: /*
1.1 kristaps 1002: * Memory-map the file "fname" and begin parsing it unless "parse" is
1003: * zero, in which case we just dump the file to stdout (making sure it
1004: * doesn't trip up mdoc(7) along the way).
1005: * This can be called in a nested context.
1006: */
1007: void
1008: parsefile(struct texi *p, const char *fname, int parse)
1009: {
1010: struct texifile *f;
1011: int fd;
1012: struct stat st;
1013: size_t i;
1014:
1.5 kristaps 1015: if (64 == p->filepos)
1.6 kristaps 1016: texierr(p, "too many open files");
1.1 kristaps 1017: f = &p->files[p->filepos];
1018: memset(f, 0, sizeof(struct texifile));
1019:
1.12 ! kristaps 1020: f->type = TEXISRC_FILE;
1.1 kristaps 1021: f->name = fname;
1022: if (-1 == (fd = open(fname, O_RDONLY, 0))) {
1023: texiabort(p, fname);
1024: } else if (-1 == fstat(fd, &st)) {
1025: close(fd);
1026: texiabort(p, fname);
1027: }
1028:
1029: f->mapsz = st.st_size;
1030: f->map = mmap(NULL, f->mapsz,
1031: PROT_READ, MAP_SHARED, fd, 0);
1032: close(fd);
1033:
1034: if (MAP_FAILED == f->map)
1035: texiabort(p, fname);
1036:
1037: p->filepos++;
1038: if ( ! parse) {
1039: /*
1040: * We're printing verbatim output.
1041: * Make sure it doesn't get interpreted as mdoc by
1042: * escaping escapes and making sure leading dots don't
1043: * trigger mdoc(7) expansion.
1044: */
1045: for (i = 0; i < f->mapsz; i++) {
1046: if (i > 0 && '.' == f->map[i])
1047: if ('\n' == f->map[i - 1])
1048: fputs("\\&", stdout);
1049: putchar(f->map[i]);
1050: if ('\\' == f->map[i])
1051: putchar('e');
1052: }
1053: } else
1054: parseeof(p, f->map, f->mapsz);
1055: texifilepop(p);
1056: }
1057:
1.2 kristaps 1058: /*
1059: * Look up the value to a stored pair's value starting in "buf" from
1060: * start to end.
1061: * Return the pointer to the value memory, which can be NULL if the
1062: * pointer key does not exist.
1063: * The pointer can point to NULL if the value has been unset.
1064: */
1065: static char **
1066: valuequery(const struct texi *p,
1067: const char *buf, size_t start, size_t end)
1068: {
1069: size_t i, sz, len;
1070:
1071: assert(end >= start);
1072: /* Ignore zero-length. */
1073: if (0 == (len = (end - start)))
1074: return(NULL);
1075: for (i = 0; i < p->valsz; i++) {
1076: sz = strlen(p->vals[i].key);
1077: if (sz != len)
1078: continue;
1079: if (0 == strncmp(p->vals[i].key, &buf[start], len))
1080: return(&p->vals[i].value);
1081: }
1082: return(NULL);
1083: }
1084:
1085: /*
1086: * Parse a key until the end of line, e.g., @clear foo\n, and return the
1087: * pointer to its value via valuequery().
1088: */
1089: static char **
1090: valuelquery(struct texi *p, const char *buf, size_t sz, size_t *pos)
1091: {
1092: size_t start, end;
1093: char **ret;
1094:
1095: while (*pos < sz && isws(buf[*pos]))
1096: advance(p, buf, pos);
1097: if (*pos == sz)
1098: return(NULL);
1099: for (start = end = *pos; end < sz; end++)
1100: if ('\n' == buf[end])
1101: break;
1102: advanceto(p, buf, pos, end);
1103: if (*pos < sz) {
1104: assert('\n' == buf[*pos]);
1105: advance(p, buf, pos);
1106: }
1107: if (NULL == (ret = valuequery(p, buf, start, end)))
1108: return(NULL);
1109: return(ret);
1110: }
1111:
1112: void
1113: valuelclear(struct texi *p, const char *buf, size_t sz, size_t *pos)
1114: {
1115: char **ret;
1116:
1117: if (NULL == (ret = valuelquery(p, buf, sz, pos)))
1118: return;
1119: free(*ret);
1120: *ret = NULL;
1121: }
1122:
1123: const char *
1124: valuellookup(struct texi *p, const char *buf, size_t sz, size_t *pos)
1125: {
1126: char **ret;
1127:
1128: if (NULL == (ret = valuelquery(p, buf, sz, pos)))
1129: return(NULL);
1130: return(*ret);
1131: }
1132:
1133: /*
1134: * Parse a key from a bracketed string, e.g., @value{foo}, and return
1135: * the pointer to its value.
1136: * If the returned pointer is NULL, either there was no string within
1137: * the brackets (or no brackets), or the value was not found, or the
1138: * value had previously been unset.
1139: */
1140: const char *
1141: valueblookup(struct texi *p, const char *buf, size_t sz, size_t *pos)
1142: {
1143: size_t start, end;
1144: char **ret;
1145:
1146: while (*pos < sz && isws(buf[*pos]))
1147: advance(p, buf, pos);
1148: if (*pos == sz || '{' != buf[*pos])
1149: return(NULL);
1150: advance(p, buf, pos);
1151: for (start = end = *pos; end < sz; end++)
1152: if ('}' == buf[end])
1153: break;
1154: advanceto(p, buf, pos, end);
1155: if (*pos < sz) {
1156: assert('}' == buf[*pos]);
1157: advance(p, buf, pos);
1158: }
1159: if (NULL == (ret = valuequery(p, buf, start, end)))
1160: return(NULL);
1161: return(*ret);
1162: }
1163:
1164: void
1165: valueadd(struct texi *p, char *key, char *val)
1166: {
1167: size_t i;
1168:
1169: assert(NULL != key);
1170: assert(NULL != val);
1171:
1172: for (i = 0; i < p->valsz; i++)
1173: if (0 == strcmp(p->vals[i].key, key))
1174: break;
1175:
1176: if (i < p->valsz) {
1177: free(key);
1178: free(p->vals[i].value);
1179: p->vals[i].value = val;
1180: } else {
1.4 kristaps 1181: /* FIXME: reallocarray() */
1.2 kristaps 1182: p->vals = realloc(p->vals,
1183: (p->valsz + 1) *
1184: sizeof(struct texivalue));
1.4 kristaps 1185: if (NULL == p->vals)
1186: texiabort(p, NULL);
1.2 kristaps 1187: p->vals[p->valsz].key = key;
1188: p->vals[p->valsz].value = val;
1189: p->valsz++;
1190: }
1.7 kristaps 1191: }
1192:
1193: /*
1194: * Take the arguments to a macro, e.g., @foo{bar, baz, xyzzy} (or the
1195: * declaration form, @macro foo {arg1, ...}) and textually convert it to
1196: * an array of arguments of size "argsz".
1197: * These need to be freed individually and as a whole.
1198: * NOTE: this will puke on @, or @} macros, which can trick it into
1199: * stopping argument parsing earlier.
1200: * Ergo, textual: this doesn't interpret the arguments in any way.
1201: */
1202: char **
1203: argparse(struct texi *p, const char *buf,
1.8 kristaps 1204: size_t sz, size_t *pos, size_t *argsz, size_t hint)
1.7 kristaps 1205: {
1206: char **args;
1207: size_t start, end, stack;
1208:
1209: while (*pos < sz && isws(buf[*pos]))
1210: advance(p, buf, pos);
1211:
1212: args = NULL;
1213: *argsz = 0;
1214:
1.10 kristaps 1215: if ('{' != buf[*pos] && hint) {
1216: /*
1217: * Special case: if we encounter an unbracketed argument
1218: * and we're being invoked with non-zero arguments
1219: * (versus being set, i.e., hint>0), then parse until
1220: * the end of line.
1221: */
1222: *argsz = 1;
1223: args = calloc(1, sizeof(char *));
1224: if (NULL == args)
1225: texiabort(p, NULL);
1226: start = *pos;
1227: while (*pos < sz) {
1228: if ('\n' == buf[*pos])
1229: break;
1230: advance(p, buf, pos);
1231: }
1232: args[0] = malloc(*pos - start + 1);
1233: memcpy(args[0], &buf[start], *pos - start);
1234: args[0][*pos - start] = '\0';
1235: if (*pos < sz && '\n' == buf[*pos])
1236: advance(p, buf, pos);
1237: return(args);
1238: } else if ('{' != buf[*pos])
1.7 kristaps 1239: return(args);
1240:
1241: /* Parse til the closing '}', putting into the array. */
1242: advance(p, buf, pos);
1243: while (*pos < sz) {
1244: while (*pos < sz && isws(buf[*pos]))
1245: advance(p, buf, pos);
1246: start = *pos;
1247: stack = 0;
1248: while (*pos < sz) {
1249: /*
1250: * According to the manual, commas within
1251: * embedded commands are escaped.
1252: * We keep track of embedded-ness in the "stack"
1253: * state anyway, so this is free.
1254: */
1.8 kristaps 1255: if (',' == buf[*pos] && 0 == stack && 1 != hint)
1.7 kristaps 1256: break;
1257: else if (0 == stack && '}' == buf[*pos])
1258: break;
1259: else if (0 != stack && '}' == buf[*pos])
1260: stack--;
1261: else if ('{' == buf[*pos])
1262: stack++;
1263: advance(p, buf, pos);
1264: }
1265: if (stack)
1266: texiwarn(p, "unterminated macro "
1267: "in macro arguments");
1268: if ((end = *pos) == sz)
1269: break;
1270: /* Test for zero-length '{ }'. */
1271: if (start == end && '}' == buf[*pos] && 0 == *argsz)
1272: break;
1273: /* FIXME: use reallocarray. */
1274: args = realloc
1275: (args, sizeof(char *) *
1276: (*argsz + 1));
1277: if (NULL == args)
1278: texiabort(p, NULL);
1279: args[*argsz] = malloc(end - start + 1);
1280: if (NULL == args[*argsz])
1281: texiabort(p, NULL);
1282: memcpy(args[*argsz],
1283: &buf[start], end - start);
1284: args[*argsz][end - start] = '\0';
1285: (*argsz)++;
1286: if ('}' == buf[*pos])
1287: break;
1288: advance(p, buf, pos);
1289: }
1290:
1291: if (*pos == sz)
1292: texierr(p, "unterminated arguments");
1293: assert('}' == buf[*pos]);
1294: advance(p, buf, pos);
1295: return(args);
1.2 kristaps 1296: }
CVSweb