Annotation of pod2mdoc/pod2mdoc.c, Revision 1.23
1.23 ! kristaps 1: /* $Id: pod2mdoc.c,v 1.22 2014/04/03 16:50:32 kristaps Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2014 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 AUTHORS DISCLAIM ALL WARRANTIES
10: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
11: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS 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/stat.h>
18: #include <sys/time.h>
19:
20: #include <assert.h>
21: #include <ctype.h>
22: #include <fcntl.h>
23: #include <getopt.h>
24: #include <stdio.h>
25: #include <stdlib.h>
26: #include <string.h>
27: #include <unistd.h>
28:
1.10 kristaps 29: /*
1.19 kristaps 30: * In what section can we find Perl module manuals?
31: * Sometimes (Mac OS X) it's 3pm, sometimes (OpenBSD, etc.) 3p.
32: * XXX IF YOU CHANGE THIS, CHANGE POD2MDOC.1 AS WELL.
1.10 kristaps 33: */
34: #define PERL_SECTION "3p"
35:
1.1 schwarze 36: struct args {
37: const char *title; /* override "Dt" title */
38: const char *date; /* override "Dd" date */
39: const char *section; /* override "Dt" section */
40: };
41:
1.4 schwarze 42: enum list {
43: LIST_BULLET = 0,
44: LIST_ENUM,
45: LIST_TAG,
46: LIST__MAX
47: };
48:
1.11 kristaps 49: enum sect {
50: SECT_NONE = 0,
51: SECT_NAME, /* NAME section */
52: SECT_SYNOPSIS, /* SYNOPSIS section */
53: };
54:
1.1 schwarze 55: struct state {
56: int parsing; /* after =cut of before command */
57: int paused; /* in =begin and before =end */
58: int haspar; /* in paragraph: do we need Pp? */
1.11 kristaps 59: enum sect sect; /* which section are we in? */
1.1 schwarze 60: const char *fname; /* file being parsed */
1.4 schwarze 61: #define LIST_STACKSZ 128
62: enum list lstack[LIST_STACKSZ]; /* open lists */
63: size_t lpos; /* where in list stack */
1.1 schwarze 64: };
65:
66: enum fmt {
67: FMT_ITALIC,
68: FMT_BOLD,
69: FMT_CODE,
70: FMT_LINK,
71: FMT_ESCAPE,
72: FMT_FILE,
73: FMT_NBSP,
74: FMT_INDEX,
75: FMT_NULL,
76: FMT__MAX
77: };
78:
79: enum cmd {
80: CMD_POD = 0,
81: CMD_HEAD1,
82: CMD_HEAD2,
83: CMD_HEAD3,
84: CMD_HEAD4,
85: CMD_OVER,
86: CMD_ITEM,
87: CMD_BACK,
88: CMD_BEGIN,
89: CMD_END,
90: CMD_FOR,
91: CMD_ENCODING,
92: CMD_CUT,
93: CMD__MAX
94: };
95:
96: static const char *const cmds[CMD__MAX] = {
97: "pod", /* CMD_POD */
98: "head1", /* CMD_HEAD1 */
99: "head2", /* CMD_HEAD2 */
100: "head3", /* CMD_HEAD3 */
101: "head4", /* CMD_HEAD4 */
102: "over", /* CMD_OVER */
103: "item", /* CMD_ITEM */
104: "back", /* CMD_BACK */
105: "begin", /* CMD_BEGIN */
106: "end", /* CMD_END */
107: "for", /* CMD_FOR */
108: "encoding", /* CMD_ENCODING */
109: "cut" /* CMD_CUT */
110: };
111:
112: static const char fmts[FMT__MAX] = {
113: 'I', /* FMT_ITALIC */
114: 'B', /* FMT_BOLD */
115: 'C', /* FMT_CODE */
116: 'L', /* FMT_LINK */
117: 'E', /* FMT_ESCAPE */
118: 'F', /* FMT_FILE */
119: 'S', /* FMT_NBSP */
120: 'X', /* FMT_INDEX */
121: 'Z' /* FMT_NULL */
122: };
123:
1.6 kristaps 124: static int last;
125:
1.1 schwarze 126: /*
127: * Given buf[*start] is at the start of an escape name, read til the end
128: * of the escape ('>') then try to do something with it.
129: * Sets start to be one after the '>'.
130: */
131: static void
132: formatescape(const char *buf, size_t *start, size_t end)
133: {
134: char esc[16]; /* no more needed */
135: size_t i, max;
136:
137: max = sizeof(esc) - 1;
138: i = 0;
139: /* Read til our buffer is full. */
140: while (*start < end && '>' != buf[*start] && i < max)
141: esc[i++] = buf[(*start)++];
142: esc[i] = '\0';
143:
144: if (i == max) {
145: /* Too long... skip til we end. */
146: while (*start < end && '>' != buf[*start])
147: (*start)++;
148: return;
149: } else if (*start >= end)
150: return;
151:
152: assert('>' == buf[*start]);
153: (*start)++;
154:
155: /*
156: * TODO: right now, we only recognise the named escapes.
157: * Just let the rest of them go.
158: */
1.6 kristaps 159: if (0 == strcmp(esc, "lt"))
1.1 schwarze 160: printf("\\(la");
161: else if (0 == strcmp(esc, "gt"))
162: printf("\\(ra");
163: else if (0 == strcmp(esc, "vb"))
164: printf("\\(ba");
165: else if (0 == strcmp(esc, "sol"))
166: printf("\\(sl");
1.6 kristaps 167: else
168: return;
169:
170: last = 'a';
1.1 schwarze 171: }
172:
173: /*
1.9 kristaps 174: * Run some heuristics to intuit a link format.
1.19 kristaps 175: * I set "start" to be the end of the sequence (last right-carrot) so
1.9 kristaps 176: * that the caller can safely just continue processing.
1.19 kristaps 177: * If this is just an empty tag, I'll return 0.
1.9 kristaps 178: */
179: static int
180: trylink(const char *buf, size_t *start, size_t end, size_t dsz)
181: {
1.21 kristaps 182: size_t linkstart, realend, linkend,
183: i, j, textsz, stack;
1.18 kristaps 184: const char *text;
1.9 kristaps 185:
186: /*
187: * Scan to the start of the terminus.
188: * This function is more or less replicated in the formatcode()
189: * for null or index formatting codes.
1.23 ! kristaps 190: * However, we're slightly different because we might have
! 191: * nested escapes we need to ignore.
1.9 kristaps 192: */
1.21 kristaps 193: stack = 0;
1.19 kristaps 194: for (linkstart = realend = *start; realend < end; realend++) {
1.23 ! kristaps 195: if ('<' == buf[realend])
! 196: stack++;
1.19 kristaps 197: if ('>' != buf[realend])
1.9 kristaps 198: continue;
1.23 ! kristaps 199: else if (stack-- > 0)
! 200: continue;
! 201: if (dsz == 1)
1.9 kristaps 202: break;
1.19 kristaps 203: assert(realend > 0);
204: if (' ' != buf[realend - 1])
1.9 kristaps 205: continue;
1.19 kristaps 206: for (i = realend, j = 0; i < end && j < dsz; j++)
1.9 kristaps 207: if ('>' != buf[i++])
208: break;
209: if (dsz == j)
210: break;
211: }
1.19 kristaps 212:
213: /* Ignore stubs. */
214: if (realend == end || realend == *start)
1.9 kristaps 215: return(0);
216:
1.19 kristaps 217: /* Set linkend to the end of content. */
218: linkend = dsz > 1 ? realend - 1 : realend;
1.18 kristaps 219:
1.19 kristaps 220: /* Re-scan to see if we have a title or section. */
1.18 kristaps 221: text = &buf[*start];
1.19 kristaps 222: for (textsz = *start; textsz < linkend; textsz++)
223: if ('|' == buf[textsz] || '/' == buf[textsz])
1.18 kristaps 224: break;
225:
1.19 kristaps 226: if (textsz < linkend && '|' == buf[textsz]) {
1.20 kristaps 227: /* With title: set start, then end at section. */
1.19 kristaps 228: linkstart = textsz + 1;
1.18 kristaps 229: textsz = textsz - *start;
1.19 kristaps 230: for (i = linkstart; i < linkend; i++)
231: if ('/' == buf[i])
232: break;
233: if (i < linkend)
234: linkend = i;
1.20 kristaps 235: } else if (textsz < linkend && '/' == buf[textsz]) {
236: /* With section: set end at section. */
237: linkend = textsz;
238: textsz = 0;
239: } else
240: /* No title, no section. */
1.18 kristaps 241: textsz = 0;
1.19 kristaps 242:
243: *start = realend;
244: j = linkend - linkstart;
245:
1.20 kristaps 246: /* Do we have only subsection material? */
247: if (0 == j && '/' == buf[linkend]) {
248: linkstart = linkend + 1;
249: linkend = dsz > 1 ? realend - 1 : realend;
250: if (0 == (j = linkend - linkstart))
251: return(0);
252: printf("Sx %.*s", (int)j, &buf[linkstart]);
253: return(1);
254: } else if (0 == j)
1.19 kristaps 255: return(0);
256:
257: /* See if we qualify as being a link or not. */
1.20 kristaps 258: if ((j > 4 && 0 == memcmp("http:", &buf[linkstart], j)) ||
259: (j > 5 && 0 == memcmp("https:", &buf[linkstart], j)) ||
260: (j > 3 && 0 == memcmp("ftp:", &buf[linkstart], j)) ||
261: (j > 4 && 0 == memcmp("sftp:", &buf[linkstart], j)) ||
262: (j > 3 && 0 == memcmp("smb:", &buf[linkstart], j)) ||
263: (j > 3 && 0 == memcmp("afs:", &buf[linkstart], j))) {
264: /* Gross. */
265: printf("Lk %.*s", (int)((dsz > 1 ? realend - 1 :
266: realend) - linkstart), &buf[linkstart]);
1.19 kristaps 267: return(1);
268: }
269:
270: /* See if we qualify as a mailto. */
1.20 kristaps 271: if (j > 6 && 0 == memcmp("mailto:", &buf[linkstart], j)) {
1.19 kristaps 272: printf("Mt %.*s", (int)j, &buf[linkstart]);
273: return(1);
274: }
275:
276: /* See if we're a foo(5), foo(5x), or foo(5xx) manpage. */
277: if ((j > 3 && ')' == buf[linkend - 1]) &&
278: ('(' == buf[linkend - 3])) {
279: printf("Xr %.*s %c", (int)(j - 3),
280: &buf[linkstart], buf[linkend - 2]);
281: return(1);
282: } else if ((j > 4 && ')' == buf[linkend - 1]) &&
283: ('(' == buf[linkend - 4])) {
284: printf("Xr %.*s %.*s", (int)(j - 4),
285: &buf[linkstart], 2, &buf[linkend - 3]);
286: return(1);
287: } else if ((j > 5 && ')' == buf[linkend - 1]) &&
288: ('(' == buf[linkend - 5])) {
289: printf("Xr %.*s %.*s", (int)(j - 5),
290: &buf[linkstart], 3, &buf[linkend - 4]);
291: return(1);
292: }
293:
294: /* Last try: do we have a double-colon? */
295: for (i = linkstart + 1; i < linkend; i++)
296: if (':' == buf[i] && ':' == buf[i - 1])
1.18 kristaps 297: break;
1.9 kristaps 298:
1.19 kristaps 299: if (i < linkend)
1.10 kristaps 300: printf("Xr %.*s " PERL_SECTION,
1.19 kristaps 301: (int)j, &buf[linkstart]);
1.9 kristaps 302: else
1.19 kristaps 303: printf("Xr %.*s 1", (int)j, &buf[linkstart]);
1.9 kristaps 304:
305: return(1);
306: }
307:
1.13 kristaps 308: /*
309: * Doclifting: if we're a bold "-xx" and we're in the SYNOPSIS section,
310: * then it's likely that we're a flag.
311: * Our flag might be followed by an argument, so make sure that we're
312: * accounting for that, too.
313: * If we don't have a flag at all, however, then assume we're an "Ar".
314: */
315: static void
316: dosynopsisfl(const char *buf, size_t *start, size_t end)
317: {
318: size_t i;
319: again:
1.14 kristaps 320: assert(*start + 1 < end);
321: assert('-' == buf[*start]);
322:
323: if ( ! isalnum((int)buf[*start + 1]) &&
324: '?' != buf[*start + 1] &&
325: '-' != buf[*start + 1]) {
326: (*start)--;
327: fputs("Ar ", stdout);
328: return;
329: }
330:
1.13 kristaps 331: (*start)++;
332: for (i = *start; i < end; i++)
333: if (isalnum((int)buf[i]))
334: continue;
1.14 kristaps 335: else if ('?' == buf[i])
336: continue;
1.13 kristaps 337: else if ('-' == buf[i])
338: continue;
339: else if ('_' == buf[i])
340: continue;
341: else
342: break;
343:
344: assert(i < end);
345:
346: if ( ! (' ' == buf[i] || '>' == buf[i])) {
347: printf("Ar ");
348: return;
349: }
350:
351: printf("Fl ");
352: if (end - *start > 1 &&
353: isupper((int)buf[*start]) &&
354: islower((int)buf[*start + 1]) &&
355: (end - *start == 2 ||
356: ' ' == buf[*start + 2]))
357: printf("\\&");
358: printf("%.*s ", (int)(i - *start), &buf[*start]);
359: *start = i;
360:
361: if (' ' == buf[i]) {
362: while (i < end && ' ' == buf[i])
363: i++;
364: assert(i < end);
365: if ('-' == buf[i]) {
366: *start = i;
367: goto again;
368: }
369: printf("Ar ");
370: *start = i;
371: }
372: }
373:
1.9 kristaps 374: /*
1.1 schwarze 375: * We're at the character in front of a format code, which is structured
376: * like X<...> and can contain nested format codes.
377: * This consumes the whole format code, and any nested format codes, til
378: * the end of matched production.
379: * If "reentrant", then we're being called after a macro has already
380: * been printed to the current line.
1.6 kristaps 381: * If "nomacro", then we don't print any macros, just contained data
382: * (e.g., following "Sh" or "Nm").
1.15 kristaps 383: * "pos" is only significant in SYNOPSIS, and should be 0 when invoked
384: * as the first format code on a line (for decoration as an "Nm"),
385: * non-zero otherwise.
1.6 kristaps 386: * Return whether we've printed a macro or not--in other words, whether
387: * this should trigger a subsequent newline (this should be ignored when
388: * reentrant).
1.1 schwarze 389: */
390: static int
1.15 kristaps 391: formatcode(struct state *st, const char *buf, size_t *start,
392: size_t end, int reentrant, int nomacro, int pos)
1.1 schwarze 393: {
394: enum fmt fmt;
1.5 kristaps 395: size_t i, j, dsz;
1.1 schwarze 396:
397: assert(*start + 1 < end);
398: assert('<' == buf[*start + 1]);
399:
1.6 kristaps 400: /*
401: * First, look up the format code.
402: * If it's not valid, then exit immediately.
403: */
404: for (fmt = 0; fmt < FMT__MAX; fmt++)
405: if (buf[*start] == fmts[fmt])
406: break;
407:
408: if (FMT__MAX == fmt) {
409: putchar(last = buf[(*start)++]);
1.8 kristaps 410: if ('\\' == last)
411: putchar('e');
1.6 kristaps 412: return(0);
413: }
414:
1.5 kristaps 415: /*
416: * Determine whether we're overriding our delimiter.
417: * According to POD, if we have more than one '<' followed by a
418: * space, then we need a space followed by matching '>' to close
419: * the expression.
420: * Otherwise we use the usual '<' and '>' matched pair.
421: */
422: i = *start + 1;
423: while (i < end && '<' == buf[i])
424: i++;
425: assert(i > *start + 1);
426: dsz = i - (*start + 1);
427: if (dsz > 1 && (i >= end || ' ' != buf[i]))
428: dsz = 1;
429:
430: /* Remember, if dsz>1, to jump the trailing space. */
431: *start += dsz + 1 + (dsz > 1 ? 1 : 0);
1.1 schwarze 432:
433: /*
1.6 kristaps 434: * Escapes and ignored codes (NULL and INDEX) don't print macro
435: * sequences, so just output them like normal text before
436: * processing for real macros.
1.1 schwarze 437: */
438: if (FMT_ESCAPE == fmt) {
439: formatescape(buf, start, end);
440: return(0);
441: } else if (FMT_NULL == fmt || FMT_INDEX == fmt) {
1.5 kristaps 442: /*
1.6 kristaps 443: * Just consume til the end delimiter, accounting for
444: * whether it's a custom one.
1.5 kristaps 445: */
446: for ( ; *start < end; (*start)++) {
447: if ('>' != buf[*start])
448: continue;
449: else if (dsz == 1)
450: break;
451: assert(*start > 0);
452: if (' ' != buf[*start - 1])
453: continue;
454: i = *start;
455: for (j = 0; i < end && j < dsz; j++)
456: if ('>' != buf[i++])
457: break;
458: if (dsz != j)
459: continue;
460: (*start) += dsz;
461: break;
462: }
1.1 schwarze 463: return(0);
464: }
465:
1.6 kristaps 466: /*
467: * Check whether we're supposed to print macro stuff (this is
468: * suppressed in, e.g., "Nm" and "Sh" macros).
469: */
1.1 schwarze 470: if ( ! nomacro) {
471: /*
472: * Print out the macro describing this format code.
473: * If we're not "reentrant" (not yet on a macro line)
474: * then print a newline, if necessary, and the macro
475: * indicator.
476: * Otherwise, offset us with a space.
477: */
1.6 kristaps 478: if ( ! reentrant) {
479: if (last != '\n')
480: putchar('\n');
1.1 schwarze 481: putchar('.');
1.6 kristaps 482: } else
1.1 schwarze 483: putchar(' ');
484:
485: /*
1.6 kristaps 486: * If we don't have whitespace before us (and none after
487: * the opening delimiter), then suppress macro
488: * whitespace with Pf.
1.1 schwarze 489: */
1.6 kristaps 490: if (' ' != last && '\n' != last && ' ' != buf[*start])
491: printf("Pf ");
492:
1.1 schwarze 493: switch (fmt) {
494: case (FMT_ITALIC):
495: printf("Em ");
496: break;
497: case (FMT_BOLD):
1.14 kristaps 498: if (SECT_SYNOPSIS == st->sect) {
499: if (1 == dsz && '-' == buf[*start])
500: dosynopsisfl(buf, start, end);
1.15 kristaps 501: else if (0 == pos)
502: printf("Nm ");
1.14 kristaps 503: else
504: printf("Ar ");
505: break;
506: }
507: printf("Sy ");
1.1 schwarze 508: break;
509: case (FMT_CODE):
1.2 schwarze 510: printf("Qo Li ");
1.1 schwarze 511: break;
512: case (FMT_LINK):
1.19 kristaps 513: /* Try to link; use "No" if it's empty. */
1.9 kristaps 514: if ( ! trylink(buf, start, end, dsz))
515: printf("No ");
1.1 schwarze 516: break;
517: case (FMT_FILE):
518: printf("Pa ");
519: break;
520: case (FMT_NBSP):
521: printf("No ");
522: break;
523: default:
524: abort();
525: }
526: }
527:
528: /*
1.6 kristaps 529: * Process until we reach the end marker (e.g., '>') or until we
1.5 kristaps 530: * find a nested format code.
1.1 schwarze 531: * Don't emit any newlines: since we're on a macro line, we
532: * don't want to break the line.
533: */
534: while (*start < end) {
1.5 kristaps 535: if ('>' == buf[*start] && 1 == dsz) {
1.1 schwarze 536: (*start)++;
537: break;
1.5 kristaps 538: } else if ('>' == buf[*start] &&
539: ' ' == buf[*start - 1]) {
540: /*
541: * Handle custom delimiters.
542: * These require a certain number of
543: * space-preceded carrots before we're really at
544: * the end.
545: */
546: i = *start;
547: for (j = 0; i < end && j < dsz; j++)
548: if ('>' != buf[i++])
549: break;
550: if (dsz == j) {
551: *start += dsz;
552: break;
553: }
1.1 schwarze 554: }
555: if (*start + 1 < end && '<' == buf[*start + 1]) {
1.15 kristaps 556: formatcode(st, buf, start, end, 1, nomacro, 1);
1.1 schwarze 557: continue;
558: }
1.3 schwarze 559:
1.4 schwarze 560: /*
561: * Make sure that any macro-like words (or
562: * really any word starting with a capital
563: * letter) is assumed to be a macro that must be
564: * escaped.
565: * This matches "Xx " and "XxEOLN".
566: */
567: if ((' ' == last || '\n' == last) &&
568: end - *start > 1 &&
569: isupper((int)buf[*start]) &&
570: islower((int)buf[*start + 1]) &&
571: (end - *start == 2 ||
572: ' ' == buf[*start + 2]))
573: printf("\\&");
1.3 schwarze 574:
1.4 schwarze 575: /* Suppress newline. */
1.6 kristaps 576: if ('\n' == buf[*start])
577: putchar(last = ' ');
578: else
579: putchar(last = buf[*start]);
1.4 schwarze 580:
1.8 kristaps 581: /* Protect against character escapes. */
582: if ('\\' == last)
583: putchar('e');
584:
1.6 kristaps 585: (*start)++;
586:
587: if (' ' == last)
588: while (*start < end && ' ' == buf[*start])
589: (*start)++;
1.1 schwarze 590: }
1.2 schwarze 591:
592: if ( ! nomacro && FMT_CODE == fmt)
593: printf(" Qc ");
1.1 schwarze 594:
595: /*
1.6 kristaps 596: * We're now subsequent the format code.
597: * If there isn't a space (or newline) here, and we haven't just
598: * printed a space, then suppress space.
1.1 schwarze 599: */
1.6 kristaps 600: if ( ! nomacro && ' ' != last)
601: if (' ' != buf[*start] && '\n' != buf[*start])
602: printf(" Ns ");
1.5 kristaps 603:
1.1 schwarze 604: return(1);
605: }
606:
607: /*
608: * Calls formatcode() til the end of a paragraph.
609: */
610: static void
1.11 kristaps 611: formatcodeln(struct state *st, const char *buf,
612: size_t *start, size_t end, int nomacro)
1.1 schwarze 613: {
614:
1.4 schwarze 615: last = ' ';
1.1 schwarze 616: while (*start < end) {
617: if (*start + 1 < end && '<' == buf[*start + 1]) {
1.15 kristaps 618: formatcode(st, buf, start, end, 1, nomacro, 1);
1.1 schwarze 619: continue;
620: }
1.4 schwarze 621: /*
622: * Since we're already on a macro line, we want to make
623: * sure that we don't inadvertently invoke a macro.
624: * We need to do this carefully because section names
625: * are used in troff and we don't want to escape
626: * something that needn't be escaped.
627: */
628: if (' ' == last && end - *start > 1 &&
629: isupper((int)buf[*start]) &&
630: islower((int)buf[*start + 1]) &&
631: (end - *start == 2 ||
632: ' ' == buf[*start + 2]))
633: printf("\\&");
634:
1.8 kristaps 635: if ('\n' == buf[*start])
636: putchar(last = ' ');
637: else
1.1 schwarze 638: putchar(last = buf[*start]);
1.8 kristaps 639:
640: /* Protect against character escapes. */
641: if ('\\' == last)
642: putchar('e');
643:
1.1 schwarze 644: (*start)++;
645: }
646: }
647:
648: /*
1.4 schwarze 649: * Guess at what kind of list we are.
650: * These are taken straight from the POD manual.
651: * I don't know what people do in real life.
652: */
653: static enum list
654: listguess(const char *buf, size_t start, size_t end)
655: {
656: size_t len = end - start;
657:
658: assert(end >= start);
659:
660: if (len == 1 && '*' == buf[start])
661: return(LIST_BULLET);
662: if (len == 2 && '1' == buf[start] && '.' == buf[start + 1])
663: return(LIST_ENUM);
664: else if (len == 1 && '1' == buf[start])
665: return(LIST_ENUM);
666: else
667: return(LIST_TAG);
668: }
669:
670: /*
1.1 schwarze 671: * A command paragraph, as noted in the perlpod manual, just indicates
672: * that we should do something, optionally with some text to print as
673: * well.
674: */
675: static void
676: command(struct state *st, const char *buf, size_t start, size_t end)
677: {
678: size_t len, csz;
679: enum cmd cmd;
680:
681: assert('=' == buf[start]);
682: start++;
683: len = end - start;
684:
685: for (cmd = 0; cmd < CMD__MAX; cmd++) {
686: csz = strlen(cmds[cmd]);
687: if (len < csz)
688: continue;
689: if (0 == memcmp(&buf[start], cmd[cmds], csz))
690: break;
691: }
692:
693: /* Ignore bogus commands. */
694:
695: if (CMD__MAX == cmd)
696: return;
697:
698: start += csz;
1.8 kristaps 699: while (start < end && ' ' == buf[start])
700: start++;
701:
1.1 schwarze 702: len = end - start;
703:
704: if (st->paused) {
705: st->paused = CMD_END != cmd;
706: return;
707: }
708:
709: switch (cmd) {
710: case (CMD_POD):
711: break;
712: case (CMD_HEAD1):
713: /*
714: * The behaviour of head= follows from a quick glance at
715: * how pod2man handles it.
716: */
717: printf(".Sh ");
1.11 kristaps 718: st->sect = SECT_NONE;
719: if (end - start == 4) {
1.1 schwarze 720: if (0 == memcmp(&buf[start], "NAME", 4))
1.11 kristaps 721: st->sect = SECT_NAME;
722: } else if (end - start == 8) {
723: if (0 == memcmp(&buf[start], "SYNOPSIS", 8))
724: st->sect = SECT_SYNOPSIS;
725: }
726: formatcodeln(st, buf, &start, end, 1);
1.1 schwarze 727: putchar('\n');
728: st->haspar = 1;
729: break;
730: case (CMD_HEAD2):
731: printf(".Ss ");
1.11 kristaps 732: formatcodeln(st, buf, &start, end, 1);
1.1 schwarze 733: putchar('\n');
734: st->haspar = 1;
735: break;
736: case (CMD_HEAD3):
737: puts(".Pp");
738: printf(".Em ");
1.11 kristaps 739: formatcodeln(st, buf, &start, end, 0);
1.1 schwarze 740: putchar('\n');
741: puts(".Pp");
742: st->haspar = 1;
743: break;
744: case (CMD_HEAD4):
745: puts(".Pp");
746: printf(".No ");
1.11 kristaps 747: formatcodeln(st, buf, &start, end, 0);
1.1 schwarze 748: putchar('\n');
749: puts(".Pp");
750: st->haspar = 1;
751: break;
752: case (CMD_OVER):
1.4 schwarze 753: /*
754: * If we have an existing list that hasn't had an =item
755: * yet, then make sure that we open it now.
756: * We use the default list type, but that can't be
757: * helped (we haven't seen any items yet).
1.1 schwarze 758: */
1.4 schwarze 759: if (st->lpos > 0)
760: if (LIST__MAX == st->lstack[st->lpos - 1]) {
761: st->lstack[st->lpos - 1] = LIST_TAG;
762: puts(".Bl -tag -width Ds");
763: }
764: st->lpos++;
765: assert(st->lpos < LIST_STACKSZ);
766: st->lstack[st->lpos - 1] = LIST__MAX;
1.1 schwarze 767: break;
768: case (CMD_ITEM):
1.6 kristaps 769: if (0 == st->lpos) {
770: /*
771: * Bad markup.
772: * Try to compensate.
773: */
774: st->lstack[st->lpos] = LIST__MAX;
775: st->lpos++;
776: }
1.4 schwarze 777: assert(st->lpos > 0);
778: /*
779: * If we're the first =item, guess at what our content
780: * will be: "*" is a bullet list, "1." is a numbered
781: * list, and everything is tagged.
782: */
783: if (LIST__MAX == st->lstack[st->lpos - 1]) {
784: st->lstack[st->lpos - 1] =
785: listguess(buf, start, end);
786: switch (st->lstack[st->lpos - 1]) {
787: case (LIST_BULLET):
788: puts(".Bl -bullet");
789: break;
790: case (LIST_ENUM):
791: puts(".Bl -enum");
792: break;
793: default:
794: puts(".Bl -tag -width Ds");
795: break;
796: }
797: }
798: switch (st->lstack[st->lpos - 1]) {
799: case (LIST_TAG):
800: printf(".It ");
1.11 kristaps 801: formatcodeln(st, buf, &start, end, 0);
1.4 schwarze 802: putchar('\n');
803: break;
804: case (LIST_ENUM):
805: /* FALLTHROUGH */
806: case (LIST_BULLET):
807: /*
808: * Abandon the remainder of the paragraph
809: * because we're going to be a bulletted or
810: * numbered list.
811: */
812: puts(".It");
813: break;
814: default:
815: abort();
816: }
1.1 schwarze 817: st->haspar = 1;
818: break;
819: case (CMD_BACK):
1.4 schwarze 820: /* Make sure we don't back over the stack. */
821: if (st->lpos > 0) {
822: st->lpos--;
823: puts(".El");
824: }
1.1 schwarze 825: break;
826: case (CMD_BEGIN):
827: /*
828: * We disregard all types for now.
829: * TODO: process at least "text" in a -literal block.
830: */
831: st->paused = 1;
832: break;
833: case (CMD_FOR):
834: /*
835: * We ignore all types of encodings and formats
836: * unilaterally.
837: */
838: break;
839: case (CMD_ENCODING):
840: break;
841: case (CMD_CUT):
842: st->parsing = 0;
843: return;
844: default:
845: abort();
846: }
847:
848: /* Any command (but =cut) makes us start parsing. */
849: st->parsing = 1;
850: }
851:
852: /*
853: * Just pump out the line in a verbatim block.
854: */
855: static void
856: verbatim(struct state *st, const char *buf, size_t start, size_t end)
857: {
1.8 kristaps 858: int last;
1.22 kristaps 859: size_t i;
1.1 schwarze 860:
861: if ( ! st->parsing || st->paused)
862: return;
1.22 kristaps 863: again:
864: /*
865: * If we're in the SYNOPSIS, see if we're an #include block.
866: * If we are, then print the "In" macro and re-loop.
867: * This handles any number of inclusions, but only when they
868: * come before the remaining parts...
869: */
870: if (SECT_SYNOPSIS == st->sect) {
871: i = start;
872: for (i = start; i < end && ' ' == buf[i]; i++)
873: /* Spin. */ ;
874: if (i == end)
875: return;
876: /* We're an include block! */
877: if (end - i > 10 &&
878: 0 == memcmp(&buf[i], "#include <", 10)) {
879: start = i + 10;
880: while (start < end && ' ' == buf[start])
881: start++;
882: fputs(".In ", stdout);
883: /* Stop til the '>' marker or we hit eoln. */
884: while (start < end &&
885: '>' != buf[start] && '\n' != buf[start])
886: putchar(buf[start++]);
887: putchar('\n');
888: if (start < end && '>' == buf[start])
889: start++;
890: if (start < end && '\n' == buf[start])
891: start++;
892: if (start < end)
893: goto again;
894: return;
895: }
896: }
897:
898: if (start == end)
899: return;
1.1 schwarze 900: puts(".Bd -literal");
1.8 kristaps 901: for (last = ' '; start < end; start++) {
902: /*
903: * Handle accidental macros (newline starting with
904: * control character) and escapes.
905: */
906: if ('\n' == last)
1.7 kristaps 907: if ('.' == buf[start] || '\'' == buf[start])
908: printf("\\&");
1.8 kristaps 909: putchar(last = buf[start]);
910: if ('\\' == buf[start])
911: printf("e");
1.7 kristaps 912: }
913: putchar('\n');
1.1 schwarze 914: puts(".Ed");
915: }
916:
917: /*
1.13 kristaps 918: * See dosynopsisop().
919: */
920: static int
921: hasmatch(const char *buf, size_t start, size_t end)
922: {
923: size_t stack;
924:
925: for (stack = 0; start < end; start++)
926: if (buf[start] == '[')
927: stack++;
928: else if (buf[start] == ']' && 0 == stack)
929: return(1);
930: else if (buf[start] == ']')
931: stack--;
932: return(0);
933: }
934:
935: /*
936: * If we're in the SYNOPSIS section and we've encounter braces in an
937: * ordinary paragraph, then try to see whether we're an [-option].
938: * Do this, if we're an opening bracket, by first seeing if we have a
939: * matching end via hasmatch().
940: * If we're an ending bracket, see if we have a stack already.
941: */
942: static int
943: dosynopsisop(const char *buf, int *last,
944: size_t *start, size_t end, size_t *opstack)
945: {
946:
947: assert('[' == buf[*start] || ']' == buf[*start]);
948:
949: if ('[' == buf[*start] && hasmatch(buf, *start + 1, end)) {
950: if ('\n' != *last)
951: putchar('\n');
952: puts(".Oo");
953: (*opstack)++;
954: } else if ('[' == buf[*start])
955: return(0);
956:
957: if (']' == buf[*start] && *opstack > 0) {
958: if ('\n' != *last)
959: putchar('\n');
960: puts(".Oc");
961: (*opstack)--;
962: } else if (']' == buf[*start])
963: return(0);
964:
965: (*start)++;
966: *last = '\n';
967: while (' ' == buf[*start])
968: (*start)++;
969: return(1);
970: }
971:
972: /*
1.17 kristaps 973: * Format multiple "Nm" manpage names in the NAME section.
974: */
975: static void
976: donamenm(struct state *st, const char *buf, size_t *start, size_t end)
977: {
978: size_t word;
979:
980: while (*start < end && ' ' == buf[*start])
981: (*start)++;
982:
983: if (end == *start) {
984: puts(".Nm unknown");
985: return;
986: }
987:
988: while (*start < end) {
989: fputs(".Nm ", stdout);
990: for (word = *start; word < end; word++)
991: if (',' == buf[word])
992: break;
993: formatcodeln(st, buf, start, word, 1);
994: if (*start == end) {
995: putchar('\n');
996: continue;
997: }
998: assert(',' == buf[*start]);
999: puts(" ,");
1000: (*start)++;
1001: while (*start < end && ' ' == buf[*start])
1002: (*start)++;
1003: }
1004: }
1005:
1006: /*
1.1 schwarze 1007: * Ordinary paragraph.
1008: * Well, this is really the hardest--POD seems to assume that, for
1009: * example, a leading space implies a newline, and so on.
1010: * Lots of other snakes in the grass: escaping a newline followed by a
1011: * period (accidental mdoc(7) control), double-newlines after macro
1012: * passages, etc.
1013: */
1014: static void
1015: ordinary(struct state *st, const char *buf, size_t start, size_t end)
1016: {
1.13 kristaps 1017: size_t i, j, opstack;
1.15 kristaps 1018: int seq;
1.1 schwarze 1019:
1020: if ( ! st->parsing || st->paused)
1021: return;
1022:
1023: /*
1024: * Special-case: the NAME section.
1025: * If we find a "-" when searching from the end, assume that
1026: * we're in "name - description" format.
1027: * To wit, print out a "Nm" and "Nd" in that format.
1028: */
1.11 kristaps 1029: if (SECT_NAME == st->sect) {
1.15 kristaps 1030: for (i = end - 2; i > start; i--)
1031: if ('-' == buf[i] && ' ' == buf[i + 1])
1.1 schwarze 1032: break;
1033: if ('-' == buf[i]) {
1034: j = i;
1035: /* Roll over multiple "-". */
1036: for ( ; i > start; i--)
1037: if ('-' != buf[i])
1038: break;
1.17 kristaps 1039: donamenm(st, buf, &start, i + 1);
1.5 kristaps 1040: start = j + 1;
1.17 kristaps 1041: while (start < end && ' ' == buf[start])
1042: start++;
1.15 kristaps 1043: fputs(".Nd ", stdout);
1.11 kristaps 1044: formatcodeln(st, buf, &start, end, 1);
1.5 kristaps 1045: putchar('\n');
1.1 schwarze 1046: return;
1047: }
1048: }
1049:
1050: if ( ! st->haspar)
1051: puts(".Pp");
1052:
1053: st->haspar = 0;
1054: last = '\n';
1.13 kristaps 1055: opstack = 0;
1.1 schwarze 1056:
1.15 kristaps 1057: for (seq = 0; start < end; seq++) {
1.1 schwarze 1058: /*
1059: * Loop til we get either to a newline or escape.
1060: * Escape initial control characters.
1061: */
1062: while (start < end) {
1063: if (start < end - 1 && '<' == buf[start + 1])
1064: break;
1065: else if ('\n' == buf[start])
1066: break;
1067: else if ('\n' == last && '.' == buf[start])
1068: printf("\\&");
1069: else if ('\n' == last && '\'' == buf[start])
1070: printf("\\&");
1.12 kristaps 1071: /*
1072: * If we're in the SYNOPSIS, have square
1073: * brackets indicate that we're opening and
1074: * closing an optional context.
1075: */
1.13 kristaps 1076: if (SECT_SYNOPSIS == st->sect &&
1077: ('[' == buf[start] ||
1078: ']' == buf[start]) &&
1079: dosynopsisop(buf, &last,
1080: &start, end, &opstack))
1081: continue;
1.1 schwarze 1082: putchar(last = buf[start++]);
1.8 kristaps 1083: if ('\\' == last)
1084: putchar('e');
1.1 schwarze 1085: }
1086:
1087: if (start < end - 1 && '<' == buf[start + 1]) {
1088: /*
1089: * We've encountered a format code.
1090: * This is going to trigger a macro no matter
1091: * what, so print a newline now.
1092: * Then print the (possibly nested) macros and
1093: * following that, a newline.
1.8 kristaps 1094: * Consume all whitespace so we don't
1095: * accidentally start an implicit literal line.
1.16 kristaps 1096: * If the macro ends with a flush comma or
1097: * period, let mdoc(7) handle it for us.
1.1 schwarze 1098: */
1.15 kristaps 1099: if (formatcode(st, buf, &start, end, 0, 0, seq)) {
1.16 kristaps 1100: if ((start == end - 1 ||
1101: (start < end - 1 &&
1102: (' ' == buf[start + 1] ||
1103: '\n' == buf[start + 1]))) &&
1104: ('.' == buf[start] ||
1105: ',' == buf[start])) {
1106: putchar(' ');
1107: putchar(buf[start++]);
1108: }
1.1 schwarze 1109: putchar(last = '\n');
1.6 kristaps 1110: while (start < end && ' ' == buf[start])
1111: start++;
1112: }
1.1 schwarze 1113: } else if (start < end && '\n' == buf[start]) {
1114: /*
1115: * Print the newline only if we haven't already
1116: * printed a newline.
1117: */
1118: if (last != '\n')
1119: putchar(last = buf[start]);
1120: if (++start >= end)
1121: continue;
1122: /*
1123: * If we have whitespace next, eat it to prevent
1124: * mdoc(7) from thinking that it's meant for
1125: * verbatim text.
1126: * It is--but if we start with that, we can't
1127: * have a macro subsequent it, which may be
1128: * possible if we have an escape next.
1129: */
1130: if (' ' == buf[start] || '\t' == buf[start]) {
1131: puts(".br");
1132: last = '\n';
1133: }
1134: for ( ; start < end; start++)
1135: if (' ' != buf[start] && '\t' != buf[start])
1136: break;
1.12 kristaps 1137: }
1.1 schwarze 1138: }
1139:
1140: if (last != '\n')
1141: putchar('\n');
1142: }
1143:
1144: /*
1145: * There are three kinds of paragraphs: verbatim (starts with whitespace
1146: * of some sort), ordinary (starts without "=" marker), or a command
1147: * (default: starts with "=").
1148: */
1149: static void
1150: dopar(struct state *st, const char *buf, size_t start, size_t end)
1151: {
1152:
1153: if (end == start)
1154: return;
1155: if (' ' == buf[start] || '\t' == buf[start])
1156: verbatim(st, buf, start, end);
1157: else if ('=' != buf[start])
1158: ordinary(st, buf, start, end);
1159: else
1160: command(st, buf, start, end);
1161: }
1162:
1163: /*
1164: * Loop around paragraphs within a document, processing each one in the
1165: * POD way.
1166: */
1167: static void
1168: dofile(const struct args *args, const char *fname,
1169: const struct tm *tm, const char *buf, size_t sz)
1170: {
1171: size_t sup, end, i, cur = 0;
1172: struct state st;
1173: const char *section, *date;
1174: char datebuf[64];
1175: char *title, *cp;
1176:
1177: if (0 == sz)
1178: return;
1179:
1180: /* Title is last path component of the filename. */
1181:
1182: if (NULL != args->title)
1183: title = strdup(args->title);
1184: else if (NULL != (cp = strrchr(fname, '/')))
1185: title = strdup(cp + 1);
1186: else
1187: title = strdup(fname);
1188:
1189: if (NULL == title) {
1190: perror(NULL);
1191: exit(EXIT_FAILURE);
1192: }
1193:
1194: /* Section is 1 unless suffix is "pm". */
1195:
1196: if (NULL == (section = args->section)) {
1197: section = "1";
1198: if (NULL != (cp = strrchr(title, '.'))) {
1199: *cp++ = '\0';
1200: if (0 == strcmp(cp, "pm"))
1.10 kristaps 1201: section = PERL_SECTION;
1.1 schwarze 1202: }
1203: }
1204:
1205: /* Date. Or the given "tm" if not supplied. */
1206:
1207: if (NULL == (date = args->date)) {
1208: strftime(datebuf, sizeof(datebuf), "%B %d, %Y", tm);
1209: date = datebuf;
1210: }
1211:
1212: for (cp = title; '\0' != *cp; cp++)
1213: *cp = toupper((int)*cp);
1214:
1215: /* The usual mdoc(7) preamble. */
1216:
1217: printf(".Dd %s\n", date);
1218: printf(".Dt %s %s\n", title, section);
1219: puts(".Os");
1220:
1221: free(title);
1222:
1223: memset(&st, 0, sizeof(struct state));
1224: assert(sz > 0);
1225:
1226: /* Main loop over file contents. */
1227:
1228: while (cur < sz) {
1229: /* Read until next paragraph. */
1230: for (i = cur + 1; i < sz; i++)
1231: if ('\n' == buf[i] && '\n' == buf[i - 1]) {
1232: /* Consume blank paragraphs. */
1233: while (i + 1 < sz && '\n' == buf[i + 1])
1234: i++;
1235: break;
1236: }
1237:
1238: /* Adjust end marker for EOF. */
1239: end = i < sz ? i - 1 :
1240: ('\n' == buf[sz - 1] ? sz - 1 : sz);
1241: sup = i < sz ? end + 2 : sz;
1242:
1243: /* Process paragraph and adjust start. */
1244: dopar(&st, buf, cur, end);
1245: cur = sup;
1246: }
1247: }
1248:
1249: /*
1250: * Read a single file fully into memory.
1251: * If the file is "-", do it from stdin.
1252: * If successfully read, send the input buffer to dofile() for further
1253: * processing.
1254: */
1255: static int
1256: readfile(const struct args *args, const char *fname)
1257: {
1258: int fd;
1259: char *buf;
1260: size_t bufsz, cur;
1261: ssize_t ssz;
1262: struct tm *tm;
1263: time_t ttm;
1264: struct stat st;
1265:
1266: assert(NULL != fname);
1267:
1268: fd = 0 != strcmp("-", fname) ?
1269: open(fname, O_RDONLY, 0) : STDIN_FILENO;
1270:
1271: if (-1 == fd) {
1272: perror(fname);
1273: return(0);
1274: }
1275:
1276: if (STDIN_FILENO == fd || -1 == fstat(fd, &st)) {
1277: ttm = time(NULL);
1278: tm = localtime(&ttm);
1279: } else
1280: tm = localtime(&st.st_mtime);
1281:
1282: /*
1283: * Arbitrarily-sized initial buffer.
1284: * Should be big enough for most files...
1285: */
1286: cur = 0;
1287: bufsz = 1 << 14;
1288: if (NULL == (buf = malloc(bufsz))) {
1289: perror(NULL);
1290: exit(EXIT_FAILURE);
1291: }
1292:
1293: while ((ssz = read(fd, buf + cur, bufsz - cur)) > 0) {
1294: /* Double buffer size on fill. */
1295: if ((size_t)ssz == bufsz - cur) {
1296: bufsz *= 2;
1297: if (NULL == (buf = realloc(buf, bufsz))) {
1298: perror(NULL);
1299: exit(EXIT_FAILURE);
1300: }
1301: }
1302: cur += (size_t)ssz;
1303: }
1304: if (ssz < 0) {
1305: perror(fname);
1306: free(buf);
1307: return(0);
1308: }
1309:
1310: dofile(args, STDIN_FILENO == fd ?
1311: "STDIN" : fname, tm, buf, cur);
1312: free(buf);
1313: if (STDIN_FILENO != fd)
1314: close(fd);
1315: return(1);
1316: }
1317:
1318: int
1319: main(int argc, char *argv[])
1320: {
1321: const char *fname, *name;
1322: struct args args;
1323: int c;
1324:
1325: name = strrchr(argv[0], '/');
1326: if (name == NULL)
1327: name = argv[0];
1328: else
1329: ++name;
1330:
1331: memset(&args, 0, sizeof(struct args));
1332: fname = "-";
1333:
1334: /* Accept no arguments for now. */
1335:
1336: while (-1 != (c = getopt(argc, argv, "c:d:hln:oq:rs:uv")))
1337: switch (c) {
1338: case ('h'):
1339: /* FALLTHROUGH */
1340: case ('l'):
1341: /* FALLTHROUGH */
1342: case ('c'):
1343: /* FALLTHROUGH */
1344: case ('o'):
1345: /* FALLTHROUGH */
1346: case ('q'):
1347: /* FALLTHROUGH */
1348: case ('r'):
1349: /* FALLTHROUGH */
1350: case ('u'):
1351: /* FALLTHROUGH */
1352: case ('v'):
1353: /* Ignore these. */
1354: break;
1355: case ('d'):
1356: args.date = optarg;
1357: break;
1358: case ('n'):
1359: args.title = optarg;
1360: break;
1361: case ('s'):
1362: args.section = optarg;
1363: break;
1364: default:
1365: goto usage;
1366: }
1367:
1368: argc -= optind;
1369: argv += optind;
1370:
1371: /* Accept only a single input file. */
1372:
1373: if (argc > 2)
1374: return(EXIT_FAILURE);
1375: else if (1 == argc)
1376: fname = *argv;
1377:
1378: return(readfile(&args, fname) ?
1379: EXIT_SUCCESS : EXIT_FAILURE);
1380:
1381: usage:
1382: fprintf(stderr, "usage: %s [-d date] "
1383: "[-n title] [-s section]\n", name);
1384:
1385: return(EXIT_FAILURE);
1386: }
CVSweb