Annotation of texi2mdoc/util.c, Revision 1.13
1.13 ! kristaps 1: /* $Id: util.c,v 1.12 2015/02/23 22:50:11 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.
1.13 ! kristaps 163: * Escape starting a line with a control character and slashes.
1.1 kristaps 164: */
165: void
166: texiputchar(struct texi *p, char c)
167: {
168:
169: if (p->ign)
170: return;
171: if ('.' == c && 0 == p->outcol)
172: fputs("\\&", stdout);
1.10 kristaps 173: if ('\'' == c && 0 == p->outcol)
174: fputs("\\&", stdout);
1.1 kristaps 175:
176: putchar(c);
1.13 ! kristaps 177: if ('\\' == c)
! 178: putchar('e');
1.1 kristaps 179: p->seenvs = 0;
180: if ('\n' == c) {
181: p->outcol = 0;
182: p->seenws = 0;
183: } else
184: p->outcol++;
185: }
186:
187: /*
1.13 ! kristaps 188: * Put an opaque series of characters.
! 189: * Characters starting a line with a control character are escaped, but
! 190: * that's it, so don't use this for non-controlled sequences of text.
1.1 kristaps 191: */
192: void
193: texiputchars(struct texi *p, const char *s)
194: {
195:
1.13 ! kristaps 196: if (p->ign)
! 197: return;
! 198: if ('.' == *s && 0 == p->outcol)
! 199: fputs("\\&", stdout);
! 200: if ('\'' == *s && 0 == p->outcol)
! 201: fputs("\\&", stdout);
! 202: p->outcol += fputs(s, stdout);
! 203: p->seenvs = 0;
1.9 kristaps 204: }
205:
206: /*
207: * This puts all characters onto the output stream but makes sure to
208: * escape mdoc(7) slashes.
209: */
210: void
211: texiputbuf(struct texi *p, const char *buf, size_t start, size_t end)
212: {
213:
214: for ( ; start < end; start++) {
215: texiputchar(p, buf[start]);
216: if ('\\' == buf[start])
217: texiputchar(p, 'e');
218: }
1.1 kristaps 219: }
220:
221: /*
222: * Close an mdoc(7) macro opened with teximacroopen().
223: * If there are no more macros on the line, prints a newline.
224: */
225: void
226: teximacroclose(struct texi *p)
227: {
228:
229: if (p->ign)
230: return;
231:
232: if (0 == --p->outmacro) {
233: putchar('\n');
234: p->outcol = p->seenws = 0;
235: }
236: }
237:
238: /*
239: * Open a mdoc(7) macro.
240: * This is used for line macros, e.g., Qq [foo bar baz].
241: * It can be invoked for nested macros, e.g., Qq Li foo .
242: * TODO: flush-right punctuation (e.g., parenthesis).
243: */
244: void
245: teximacroopen(struct texi *p, const char *s)
246: {
247: int rc;
248:
249: if (p->ign)
250: return;
251:
252: if (p->outcol && 0 == p->outmacro) {
253: putchar('\n');
254: p->outcol = 0;
255: }
256:
257: if (0 == p->outmacro)
258: putchar('.');
259: else
260: putchar(' ');
261:
262: if (EOF != (rc = fputs(s, stdout)))
263: p->outcol += rc;
264:
265: putchar(' ');
266: p->outcol++;
267: p->outmacro++;
268: p->seenws = 0;
269: }
270:
271: /*
272: * Put a stadnalone mdoc(7) command with the trailing newline.
273: */
274: void
275: teximacro(struct texi *p, const char *s)
276: {
277:
278: if (p->ign)
279: return;
280:
281: if (p->outmacro)
282: texierr(p, "\"%s\" in open line scope!?", s);
283: if (p->literal)
284: texierr(p, "\"%s\" in a literal scope!?", s);
285:
286: if (p->outcol)
287: putchar('\n');
288:
289: putchar('.');
290: puts(s);
291: p->outcol = p->seenws = 0;
292: }
293:
294: /*
295: * Introduce vertical space during normal (non-macro) input.
296: */
297: void
298: texivspace(struct texi *p)
299: {
300:
1.5 kristaps 301: if (p->seenvs || TEXILIST_TABLE == p->list)
1.1 kristaps 302: return;
303: teximacro(p, "Pp");
304: p->seenvs = 1;
305: }
306:
307: /*
308: * Advance by a single byte in the input stream, adjusting our location
309: * in the current input file.
310: */
311: void
312: advance(struct texi *p, const char *buf, size_t *pos)
313: {
314:
315: if ('\n' == buf[*pos]) {
316: p->files[p->filepos - 1].line++;
317: p->files[p->filepos - 1].col = 0;
318: } else
319: p->files[p->filepos - 1].col++;
320:
321: (*pos)++;
322: }
323:
324: /*
325: * It's common to wait punctuation to float on the right side of macro
326: * lines in mdoc(7), e.g., ".Em hello ) ."
327: * This function does so, and should be called before teximacroclose().
328: * It will detect that it's the last in the nested macros and
329: * appropriately flush-left punctuation alongside the macro.
330: */
331: void
332: texipunctuate(struct texi *p, const char *buf, size_t sz, size_t *pos)
333: {
334: size_t start, end;
335:
336: if (1 != p->outmacro)
337: return;
338:
339: for (start = end = *pos; end < sz; end++) {
340: switch (buf[end]) {
341: case (','):
342: case (')'):
343: case ('.'):
344: case ('"'):
345: case (':'):
346: case ('!'):
347: case ('?'):
348: continue;
349: default:
350: break;
351: }
352: break;
353: }
354: if (end == *pos)
355: return;
356: if (end + 1 == sz || ' ' == buf[end] || '\n' == buf[end]) {
357: for ( ; start < end; start++) {
358: texiputchar(p, ' ');
359: texiputchar(p, buf[start]);
360: advance(p, buf, pos);
361: }
362: }
363: }
364:
365: /*
366: * Advance to the next non-whitespace word in the input stream.
367: * If we're in literal mode, then print all of the whitespace as we're
368: * doing so.
369: */
370: static size_t
371: advancenext(struct texi *p, const char *buf, size_t sz, size_t *pos)
372: {
373:
374: if (p->literal) {
375: while (*pos < sz && ismspace(buf[*pos])) {
376: texiputchar(p, buf[*pos]);
377: advance(p, buf, pos);
378: }
379: return(*pos);
380: }
381:
382: while (*pos < sz && ismspace(buf[*pos])) {
383: p->seenws = 1;
384: /*
385: * If it looks like we've printed a double-line, then
386: * output a paragraph.
387: * FIXME: this is stupid.
388: */
389: if (*pos && '\n' == buf[*pos] && '\n' == buf[*pos - 1])
390: texivspace(p);
391: advance(p, buf, pos);
392: }
393: return(*pos);
394: }
395:
396: /*
397: * Advance to the EOLN in the input stream.
398: * NOTE: THIS SHOULD NOT BE CALLED ON BLANK TEXT, as it will read up to
399: * the @\n.
400: */
401: size_t
402: advanceeoln(struct texi *p, const char *buf,
403: size_t sz, size_t *pos, int consumenl)
404: {
405:
406: while (*pos < sz && '\n' != buf[*pos])
407: advance(p, buf, pos);
408: if (*pos < sz && consumenl)
409: advance(p, buf, pos);
410: return(*pos);
411: }
412:
413: /*
414: * Advance to position "end", which is an absolute position in the
415: * current buffer greater than or equal to the current position.
416: */
417: void
418: advanceto(struct texi *p, const char *buf, size_t *pos, size_t end)
419: {
420:
421: assert(*pos <= end);
422: while (*pos < end)
423: advance(p, buf, pos);
424: }
425:
1.7 kristaps 426: static void
427: texiexecmacro(struct texi *p, struct teximacro *m,
428: const char *buf, size_t sz, size_t *pos)
429: {
1.11 kristaps 430: size_t valsz, realsz, aasz, asz,
431: ssz, i, j, k, start, end;
432: char *val;
433: char **args;
434: const char *cp;
1.7 kristaps 435:
1.8 kristaps 436: args = argparse(p, buf, sz, pos, &asz, m->argsz);
1.7 kristaps 437: if (asz != m->argsz)
438: texiwarn(p, "invalid macro argument length");
439: aasz = asz < m->argsz ? asz : m->argsz;
440:
441: if (0 == aasz) {
1.8 kristaps 442: parsemembuf(p, m->value, strlen(m->value));
1.7 kristaps 443: return;
444: }
445:
446: valsz = realsz = strlen(m->value);
447: val = strdup(m->value);
448:
449: for (i = j = 0; i < realsz; i++) {
450: /* Parse blindly til the backslash delimiter. */
451: if ('\\' != m->value[i]) {
452: val[j++] = m->value[i];
453: val[j] = '\0';
454: continue;
455: } else if (i == realsz - 1)
456: texierr(p, "trailing argument name delimiter");
457:
458: /* Double-backslash is escaped. */
459: if ('\\' == m->value[i + 1]) {
460: val[j++] = m->value[i++];
461: val[j] = '\0';
462: continue;
463: }
464:
465: assert('\\' == m->value[i] && i < realsz - 1);
466:
467: /* Parse to terminating delimiter. */
468: /* FIXME: embedded, escaped delimiters? */
469: for (start = end = i + 1; end < realsz; end++)
470: if ('\\' == m->value[end])
471: break;
472: if (end == realsz)
473: texierr(p, "unterminated argument name");
474:
475: for (k = 0; k < aasz; k++) {
476: if ((ssz = strlen(m->args[k])) != (end - start))
477: continue;
478: if (strncmp(&m->value[start], m->args[k], ssz))
479: continue;
480: break;
481: }
482:
483: /*
484: * Argument didn't exist in argument table.
485: * No need to reallocate here: we just copy the text
486: * directly from the macro value into the buffer.
487: */
488: if (k == aasz) {
489: for ( ; i < end; i++)
490: val[j++] = m->value[i];
491: assert('\\' == m->value[i]);
492: val[j++] = m->value[i];
493: val[j] = '\0';
494: continue;
495: }
496:
497: if (strlen(args[k]) > ssz) {
498: valsz += strlen(args[k]);
499: val = realloc(val, valsz + 1);
500: if (NULL == val)
501: texiabort(p, NULL);
502: }
503:
1.11 kristaps 504: for (cp = args[k]; '\0' != *cp; cp++)
505: val[j++] = *cp;
506:
507: val[j] = '\0';
1.7 kristaps 508: i = end;
509: }
510:
1.8 kristaps 511: parsemembuf(p, val, strlen(val));
1.7 kristaps 512:
513: for (i = 0; i < asz; i++)
514: free(args[i]);
515: free(args);
516: free(val);
517: }
518:
1.1 kristaps 519: /*
520: * Output a free-form word in the input stream, progressing to the next
521: * command or white-space.
522: * This also will advance the input stream.
523: */
524: static void
525: texiword(struct texi *p, const char *buf,
526: size_t sz, size_t *pos, char extra)
527: {
528:
529: if (p->seenws && 0 == p->outmacro &&
530: p->outcol > 72 && 0 == p->literal)
531: texiputchar(p, '\n');
532: /* FIXME: abstract this: we use it elsewhere. */
533: if (p->seenws && p->outcol && 0 == p->literal)
534: texiputchar(p, ' ');
535:
536: p->seenws = 0;
537:
538: while (*pos < sz && ! ismspace(buf[*pos])) {
539: switch (buf[*pos]) {
540: case ('@'):
541: case ('}'):
542: case ('{'):
543: return;
544: }
545: if ('\0' != extra && buf[*pos] == extra)
546: return;
547: if (*pos < sz - 1 &&
548: '`' == buf[*pos] &&
549: '`' == buf[*pos + 1]) {
550: texiputchars(p, "\\(lq");
551: advance(p, buf, pos);
552: } else if (*pos < sz - 1 &&
553: '\'' == buf[*pos] &&
554: '\'' == buf[*pos + 1]) {
555: texiputchars(p, "\\(rq");
556: advance(p, buf, pos);
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) {
1.13 ! kristaps 1039: for (i = 0; i < f->mapsz; i++)
! 1040: texiputchar(p, f->map[i]);
! 1041: if (p->outcol)
! 1042: texiputchar(p, '\n');
1.1 kristaps 1043: } else
1044: parseeof(p, f->map, f->mapsz);
1045: texifilepop(p);
1046: }
1047:
1.2 kristaps 1048: /*
1049: * Look up the value to a stored pair's value starting in "buf" from
1050: * start to end.
1051: * Return the pointer to the value memory, which can be NULL if the
1052: * pointer key does not exist.
1053: * The pointer can point to NULL if the value has been unset.
1054: */
1055: static char **
1056: valuequery(const struct texi *p,
1057: const char *buf, size_t start, size_t end)
1058: {
1059: size_t i, sz, len;
1060:
1061: assert(end >= start);
1062: /* Ignore zero-length. */
1063: if (0 == (len = (end - start)))
1064: return(NULL);
1065: for (i = 0; i < p->valsz; i++) {
1066: sz = strlen(p->vals[i].key);
1067: if (sz != len)
1068: continue;
1069: if (0 == strncmp(p->vals[i].key, &buf[start], len))
1070: return(&p->vals[i].value);
1071: }
1072: return(NULL);
1073: }
1074:
1075: /*
1076: * Parse a key until the end of line, e.g., @clear foo\n, and return the
1077: * pointer to its value via valuequery().
1078: */
1079: static char **
1080: valuelquery(struct texi *p, const char *buf, size_t sz, size_t *pos)
1081: {
1082: size_t start, end;
1083: char **ret;
1084:
1085: while (*pos < sz && isws(buf[*pos]))
1086: advance(p, buf, pos);
1087: if (*pos == sz)
1088: return(NULL);
1089: for (start = end = *pos; end < sz; end++)
1090: if ('\n' == buf[end])
1091: break;
1092: advanceto(p, buf, pos, end);
1093: if (*pos < sz) {
1094: assert('\n' == buf[*pos]);
1095: advance(p, buf, pos);
1096: }
1097: if (NULL == (ret = valuequery(p, buf, start, end)))
1098: return(NULL);
1099: return(ret);
1100: }
1101:
1102: void
1103: valuelclear(struct texi *p, const char *buf, size_t sz, size_t *pos)
1104: {
1105: char **ret;
1106:
1107: if (NULL == (ret = valuelquery(p, buf, sz, pos)))
1108: return;
1109: free(*ret);
1110: *ret = NULL;
1111: }
1112:
1113: const char *
1114: valuellookup(struct texi *p, const char *buf, size_t sz, size_t *pos)
1115: {
1116: char **ret;
1117:
1118: if (NULL == (ret = valuelquery(p, buf, sz, pos)))
1119: return(NULL);
1120: return(*ret);
1121: }
1122:
1123: /*
1124: * Parse a key from a bracketed string, e.g., @value{foo}, and return
1125: * the pointer to its value.
1126: * If the returned pointer is NULL, either there was no string within
1127: * the brackets (or no brackets), or the value was not found, or the
1128: * value had previously been unset.
1129: */
1130: const char *
1131: valueblookup(struct texi *p, const char *buf, size_t sz, size_t *pos)
1132: {
1133: size_t start, end;
1134: char **ret;
1135:
1136: while (*pos < sz && isws(buf[*pos]))
1137: advance(p, buf, pos);
1138: if (*pos == sz || '{' != buf[*pos])
1139: return(NULL);
1140: advance(p, buf, pos);
1141: for (start = end = *pos; end < sz; end++)
1142: if ('}' == buf[end])
1143: break;
1144: advanceto(p, buf, pos, end);
1145: if (*pos < sz) {
1146: assert('}' == buf[*pos]);
1147: advance(p, buf, pos);
1148: }
1149: if (NULL == (ret = valuequery(p, buf, start, end)))
1150: return(NULL);
1151: return(*ret);
1152: }
1153:
1154: void
1155: valueadd(struct texi *p, char *key, char *val)
1156: {
1157: size_t i;
1158:
1159: assert(NULL != key);
1160: assert(NULL != val);
1161:
1162: for (i = 0; i < p->valsz; i++)
1163: if (0 == strcmp(p->vals[i].key, key))
1164: break;
1165:
1166: if (i < p->valsz) {
1167: free(key);
1168: free(p->vals[i].value);
1169: p->vals[i].value = val;
1170: } else {
1.4 kristaps 1171: /* FIXME: reallocarray() */
1.2 kristaps 1172: p->vals = realloc(p->vals,
1173: (p->valsz + 1) *
1174: sizeof(struct texivalue));
1.4 kristaps 1175: if (NULL == p->vals)
1176: texiabort(p, NULL);
1.2 kristaps 1177: p->vals[p->valsz].key = key;
1178: p->vals[p->valsz].value = val;
1179: p->valsz++;
1180: }
1.7 kristaps 1181: }
1182:
1183: /*
1184: * Take the arguments to a macro, e.g., @foo{bar, baz, xyzzy} (or the
1185: * declaration form, @macro foo {arg1, ...}) and textually convert it to
1186: * an array of arguments of size "argsz".
1187: * These need to be freed individually and as a whole.
1188: * NOTE: this will puke on @, or @} macros, which can trick it into
1189: * stopping argument parsing earlier.
1190: * Ergo, textual: this doesn't interpret the arguments in any way.
1191: */
1192: char **
1193: argparse(struct texi *p, const char *buf,
1.8 kristaps 1194: size_t sz, size_t *pos, size_t *argsz, size_t hint)
1.7 kristaps 1195: {
1196: char **args;
1197: size_t start, end, stack;
1198:
1199: while (*pos < sz && isws(buf[*pos]))
1200: advance(p, buf, pos);
1201:
1202: args = NULL;
1203: *argsz = 0;
1204:
1.10 kristaps 1205: if ('{' != buf[*pos] && hint) {
1206: /*
1207: * Special case: if we encounter an unbracketed argument
1208: * and we're being invoked with non-zero arguments
1209: * (versus being set, i.e., hint>0), then parse until
1210: * the end of line.
1211: */
1212: *argsz = 1;
1213: args = calloc(1, sizeof(char *));
1214: if (NULL == args)
1215: texiabort(p, NULL);
1216: start = *pos;
1217: while (*pos < sz) {
1218: if ('\n' == buf[*pos])
1219: break;
1220: advance(p, buf, pos);
1221: }
1222: args[0] = malloc(*pos - start + 1);
1223: memcpy(args[0], &buf[start], *pos - start);
1224: args[0][*pos - start] = '\0';
1225: if (*pos < sz && '\n' == buf[*pos])
1226: advance(p, buf, pos);
1227: return(args);
1228: } else if ('{' != buf[*pos])
1.7 kristaps 1229: return(args);
1230:
1231: /* Parse til the closing '}', putting into the array. */
1232: advance(p, buf, pos);
1233: while (*pos < sz) {
1234: while (*pos < sz && isws(buf[*pos]))
1235: advance(p, buf, pos);
1236: start = *pos;
1237: stack = 0;
1238: while (*pos < sz) {
1239: /*
1240: * According to the manual, commas within
1241: * embedded commands are escaped.
1242: * We keep track of embedded-ness in the "stack"
1243: * state anyway, so this is free.
1244: */
1.8 kristaps 1245: if (',' == buf[*pos] && 0 == stack && 1 != hint)
1.7 kristaps 1246: break;
1247: else if (0 == stack && '}' == buf[*pos])
1248: break;
1249: else if (0 != stack && '}' == buf[*pos])
1250: stack--;
1251: else if ('{' == buf[*pos])
1252: stack++;
1253: advance(p, buf, pos);
1254: }
1255: if (stack)
1256: texiwarn(p, "unterminated macro "
1257: "in macro arguments");
1258: if ((end = *pos) == sz)
1259: break;
1260: /* Test for zero-length '{ }'. */
1261: if (start == end && '}' == buf[*pos] && 0 == *argsz)
1262: break;
1263: /* FIXME: use reallocarray. */
1264: args = realloc
1265: (args, sizeof(char *) *
1266: (*argsz + 1));
1267: if (NULL == args)
1268: texiabort(p, NULL);
1269: args[*argsz] = malloc(end - start + 1);
1270: if (NULL == args[*argsz])
1271: texiabort(p, NULL);
1272: memcpy(args[*argsz],
1273: &buf[start], end - start);
1274: args[*argsz][end - start] = '\0';
1275: (*argsz)++;
1276: if ('}' == buf[*pos])
1277: break;
1278: advance(p, buf, pos);
1279: }
1280:
1281: if (*pos == sz)
1282: texierr(p, "unterminated arguments");
1283: assert('}' == buf[*pos]);
1284: advance(p, buf, pos);
1285: return(args);
1.2 kristaps 1286: }
CVSweb