Annotation of pod2mdoc/pod2mdoc.c, Revision 1.55
1.55 ! schwarze 1: /* $Id: pod2mdoc.c,v 1.54 2015/02/19 15:26:45 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2014 Kristaps Dzonsons <kristaps@bsd.lv>
1.37 schwarze 4: * Copyright (c) 2014, 2015 Ingo Schwarze <schwarze@openbsd.org>
1.1 schwarze 5: *
6: * Permission to use, copy, modify, and distribute this software for any
7: * purpose with or without fee is hereby granted, provided that the above
8: * copyright notice and this permission notice appear in all copies.
9: *
10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
13: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17: */
18: #include <sys/stat.h>
19: #include <sys/time.h>
20:
21: #include <assert.h>
22: #include <ctype.h>
23: #include <fcntl.h>
24: #include <getopt.h>
25: #include <stdio.h>
26: #include <stdlib.h>
27: #include <string.h>
28: #include <unistd.h>
29:
1.37 schwarze 30: #include "dict.h"
31:
1.10 kristaps 32: /*
1.19 kristaps 33: * In what section can we find Perl module manuals?
34: * Sometimes (Mac OS X) it's 3pm, sometimes (OpenBSD, etc.) 3p.
35: * XXX IF YOU CHANGE THIS, CHANGE POD2MDOC.1 AS WELL.
1.10 kristaps 36: */
37: #define PERL_SECTION "3p"
38:
1.1 schwarze 39: struct args {
40: const char *title; /* override "Dt" title */
41: const char *date; /* override "Dd" date */
42: const char *section; /* override "Dt" section */
43: };
44:
1.4 schwarze 45: enum list {
46: LIST_BULLET = 0,
47: LIST_ENUM,
48: LIST_TAG,
49: LIST__MAX
50: };
51:
1.11 kristaps 52: enum sect {
53: SECT_NONE = 0,
54: SECT_NAME, /* NAME section */
55: SECT_SYNOPSIS, /* SYNOPSIS section */
56: };
57:
1.32 schwarze 58: enum outstate {
59: OUST_NL = 0, /* just started a new output line */
60: OUST_TXT, /* text line output in progress */
61: OUST_MAC /* macro line output in progress */
62: };
63:
1.1 schwarze 64: struct state {
1.31 schwarze 65: const char *fname; /* file being parsed */
1.1 schwarze 66: int parsing; /* after =cut of before command */
67: int paused; /* in =begin and before =end */
1.11 kristaps 68: enum sect sect; /* which section are we in? */
1.4 schwarze 69: #define LIST_STACKSZ 128
70: enum list lstack[LIST_STACKSZ]; /* open lists */
71: size_t lpos; /* where in list stack */
1.31 schwarze 72: int haspar; /* in paragraph: do we need Pp? */
1.32 schwarze 73: enum outstate oust; /* state of the mdoc output stream */
74: int wantws; /* let mdoc(7) output whitespace here */
1.31 schwarze 75: char *outbuf; /* text buffered for output */
76: size_t outbufsz; /* allocated size of outbuf */
77: size_t outbuflen; /* current length of outbuf */
1.1 schwarze 78: };
79:
80: enum fmt {
81: FMT_ITALIC,
82: FMT_BOLD,
83: FMT_CODE,
84: FMT_LINK,
85: FMT_ESCAPE,
86: FMT_FILE,
87: FMT_NBSP,
88: FMT_INDEX,
89: FMT_NULL,
90: FMT__MAX
91: };
92:
93: enum cmd {
94: CMD_POD = 0,
95: CMD_HEAD1,
96: CMD_HEAD2,
97: CMD_HEAD3,
98: CMD_HEAD4,
99: CMD_OVER,
100: CMD_ITEM,
101: CMD_BACK,
102: CMD_BEGIN,
103: CMD_END,
104: CMD_FOR,
105: CMD_ENCODING,
106: CMD_CUT,
107: CMD__MAX
108: };
1.55 ! schwarze 109:
! 110: static void command(struct state *, const char *, size_t, size_t);
! 111: static void dofile(const struct args *, const char *,
! 112: const struct tm *, char *, size_t);
! 113: static void donamenm(struct state *, const char *, size_t *, size_t);
! 114: static void dopar(struct state *, char *, size_t, size_t);
! 115: static void dosynopsisfl(const char *, size_t *, size_t);
! 116: static int dosynopsisop(struct state *, const char *, size_t *,
! 117: size_t, size_t *);
! 118: static int formatcode(struct state *, const char *, size_t *,
! 119: size_t, int, int);
! 120: static void formatcodeln(struct state *, const char *, const char *,
! 121: size_t *, size_t, int);
! 122: static void formatescape(struct state *, const char *, size_t *, size_t);
! 123: static int hasmatch(const char *, size_t, size_t);
! 124: static void ordinary(struct state *, const char *, size_t, size_t);
! 125: static void outbuf_addchar(struct state *);
! 126: static void outbuf_addstr(struct state *, const char *);
! 127: static void outbuf_flush(struct state *);
! 128: static void outbuf_grow(struct state *, size_t);
! 129: static enum list listguess(const char *, size_t, size_t);
! 130: static void mdoc_newln(struct state *);
! 131: static int readfile(const struct args *, const char *);
! 132: static void register_type(const char *);
! 133: static int trylink(const char *, size_t *, size_t, size_t);
! 134: static void verbatim(struct state *, char *, size_t, size_t);
1.1 schwarze 135:
136: static const char *const cmds[CMD__MAX] = {
137: "pod", /* CMD_POD */
138: "head1", /* CMD_HEAD1 */
139: "head2", /* CMD_HEAD2 */
140: "head3", /* CMD_HEAD3 */
141: "head4", /* CMD_HEAD4 */
142: "over", /* CMD_OVER */
143: "item", /* CMD_ITEM */
144: "back", /* CMD_BACK */
145: "begin", /* CMD_BEGIN */
146: "end", /* CMD_END */
147: "for", /* CMD_FOR */
148: "encoding", /* CMD_ENCODING */
149: "cut" /* CMD_CUT */
150: };
151:
152: static const char fmts[FMT__MAX] = {
153: 'I', /* FMT_ITALIC */
154: 'B', /* FMT_BOLD */
155: 'C', /* FMT_CODE */
156: 'L', /* FMT_LINK */
157: 'E', /* FMT_ESCAPE */
158: 'F', /* FMT_FILE */
159: 'S', /* FMT_NBSP */
160: 'X', /* FMT_INDEX */
161: 'Z' /* FMT_NULL */
162: };
163:
1.42 schwarze 164: static unsigned char last;
1.6 kristaps 165:
1.31 schwarze 166:
167: static void
168: outbuf_grow(struct state *st, size_t by)
169: {
170:
171: st->outbufsz += (by / 128 + 1) * 128;
172: st->outbuf = realloc(st->outbuf, st->outbufsz);
173: if (NULL == st->outbuf) {
174: perror(NULL);
175: exit(EXIT_FAILURE);
176: }
177: }
178:
179: static void
180: outbuf_addchar(struct state *st)
181: {
182:
183: if (st->outbuflen + 2 >= st->outbufsz)
184: outbuf_grow(st, 1);
185: st->outbuf[st->outbuflen++] = last;
186: if ('\\' == last)
187: st->outbuf[st->outbuflen++] = 'e';
188: st->outbuf[st->outbuflen] = '\0';
189: }
190:
191: static void
192: outbuf_addstr(struct state *st, const char *str)
193: {
194: size_t slen;
195:
196: slen = strlen(str);
197: if (st->outbuflen + slen >= st->outbufsz)
198: outbuf_grow(st, slen);
199: memcpy(st->outbuf + st->outbuflen, str, slen+1);
1.33 schwarze 200: st->outbuflen += slen;
1.31 schwarze 201: last = str[slen - 1];
202: }
203:
204: static void
205: outbuf_flush(struct state *st)
206: {
207:
208: if (0 == st->outbuflen)
209: return;
210:
1.40 schwarze 211: if (OUST_TXT == st->oust && st->wantws)
212: putchar(' ');
213:
1.54 schwarze 214: if (OUST_MAC == st->oust && '"' == *st->outbuf)
215: printf("\\(dq%s", st->outbuf + 1);
216: else
217: fputs(st->outbuf, stdout);
218:
1.31 schwarze 219: *st->outbuf = '\0';
220: st->outbuflen = 0;
1.32 schwarze 221:
222: if (OUST_NL == st->oust)
223: st->oust = OUST_TXT;
1.31 schwarze 224: }
225:
226: static void
1.32 schwarze 227: mdoc_newln(struct state *st)
1.31 schwarze 228: {
229:
1.32 schwarze 230: if (OUST_NL == st->oust)
1.31 schwarze 231: return;
1.32 schwarze 232:
1.31 schwarze 233: putchar('\n');
234: last = '\n';
1.32 schwarze 235: st->oust = OUST_NL;
236: st->wantws = 1;
1.31 schwarze 237: }
238:
1.1 schwarze 239: /*
240: * Given buf[*start] is at the start of an escape name, read til the end
241: * of the escape ('>') then try to do something with it.
242: * Sets start to be one after the '>'.
1.32 schwarze 243: *
244: * This function does not care about output modes,
245: * it merely appends text to the output buffer,
246: * which can then be used in any mode.
1.1 schwarze 247: */
248: static void
1.31 schwarze 249: formatescape(struct state *st, const char *buf, size_t *start, size_t end)
1.1 schwarze 250: {
251: char esc[16]; /* no more needed */
252: size_t i, max;
253:
254: max = sizeof(esc) - 1;
255: i = 0;
256: /* Read til our buffer is full. */
257: while (*start < end && '>' != buf[*start] && i < max)
258: esc[i++] = buf[(*start)++];
259: esc[i] = '\0';
260:
261: if (i == max) {
262: /* Too long... skip til we end. */
263: while (*start < end && '>' != buf[*start])
264: (*start)++;
265: return;
266: } else if (*start >= end)
267: return;
268:
269: assert('>' == buf[*start]);
270: (*start)++;
271:
272: /*
273: * TODO: right now, we only recognise the named escapes.
274: * Just let the rest of them go.
275: */
1.6 kristaps 276: if (0 == strcmp(esc, "lt"))
1.31 schwarze 277: outbuf_addstr(st, "\\(la");
1.1 schwarze 278: else if (0 == strcmp(esc, "gt"))
1.31 schwarze 279: outbuf_addstr(st, "\\(ra");
1.33 schwarze 280: else if (0 == strcmp(esc, "verbar"))
1.31 schwarze 281: outbuf_addstr(st, "\\(ba");
1.1 schwarze 282: else if (0 == strcmp(esc, "sol"))
1.31 schwarze 283: outbuf_addstr(st, "\\(sl");
1.1 schwarze 284: }
285:
286: /*
1.9 kristaps 287: * Run some heuristics to intuit a link format.
1.19 kristaps 288: * I set "start" to be the end of the sequence (last right-carrot) so
1.9 kristaps 289: * that the caller can safely just continue processing.
1.19 kristaps 290: * If this is just an empty tag, I'll return 0.
1.32 schwarze 291: *
292: * Always operates in OUST_MAC mode.
293: * Mode handling is done by the caller.
1.9 kristaps 294: */
295: static int
296: trylink(const char *buf, size_t *start, size_t end, size_t dsz)
297: {
1.21 kristaps 298: size_t linkstart, realend, linkend,
299: i, j, textsz, stack;
1.9 kristaps 300:
301: /*
302: * Scan to the start of the terminus.
303: * This function is more or less replicated in the formatcode()
304: * for null or index formatting codes.
1.23 kristaps 305: * However, we're slightly different because we might have
306: * nested escapes we need to ignore.
1.9 kristaps 307: */
1.21 kristaps 308: stack = 0;
1.19 kristaps 309: for (linkstart = realend = *start; realend < end; realend++) {
1.23 kristaps 310: if ('<' == buf[realend])
311: stack++;
1.19 kristaps 312: if ('>' != buf[realend])
1.9 kristaps 313: continue;
1.23 kristaps 314: else if (stack-- > 0)
315: continue;
316: if (dsz == 1)
1.9 kristaps 317: break;
1.19 kristaps 318: assert(realend > 0);
319: if (' ' != buf[realend - 1])
1.9 kristaps 320: continue;
1.19 kristaps 321: for (i = realend, j = 0; i < end && j < dsz; j++)
1.9 kristaps 322: if ('>' != buf[i++])
323: break;
324: if (dsz == j)
325: break;
326: }
1.19 kristaps 327:
328: /* Ignore stubs. */
329: if (realend == end || realend == *start)
1.9 kristaps 330: return(0);
331:
1.19 kristaps 332: /* Set linkend to the end of content. */
333: linkend = dsz > 1 ? realend - 1 : realend;
1.18 kristaps 334:
1.19 kristaps 335: /* Re-scan to see if we have a title or section. */
336: for (textsz = *start; textsz < linkend; textsz++)
337: if ('|' == buf[textsz] || '/' == buf[textsz])
1.18 kristaps 338: break;
339:
1.19 kristaps 340: if (textsz < linkend && '|' == buf[textsz]) {
1.20 kristaps 341: /* With title: set start, then end at section. */
1.19 kristaps 342: linkstart = textsz + 1;
1.18 kristaps 343: textsz = textsz - *start;
1.19 kristaps 344: for (i = linkstart; i < linkend; i++)
345: if ('/' == buf[i])
346: break;
347: if (i < linkend)
348: linkend = i;
1.20 kristaps 349: } else if (textsz < linkend && '/' == buf[textsz]) {
350: /* With section: set end at section. */
351: linkend = textsz;
352: textsz = 0;
353: } else
354: /* No title, no section. */
1.18 kristaps 355: textsz = 0;
1.19 kristaps 356:
357: *start = realend;
358: j = linkend - linkstart;
359:
1.20 kristaps 360: /* Do we have only subsection material? */
361: if (0 == j && '/' == buf[linkend]) {
362: linkstart = linkend + 1;
363: linkend = dsz > 1 ? realend - 1 : realend;
364: if (0 == (j = linkend - linkstart))
365: return(0);
366: printf("Sx %.*s", (int)j, &buf[linkstart]);
367: return(1);
368: } else if (0 == j)
1.19 kristaps 369: return(0);
370:
371: /* See if we qualify as being a link or not. */
1.20 kristaps 372: if ((j > 4 && 0 == memcmp("http:", &buf[linkstart], j)) ||
373: (j > 5 && 0 == memcmp("https:", &buf[linkstart], j)) ||
374: (j > 3 && 0 == memcmp("ftp:", &buf[linkstart], j)) ||
375: (j > 4 && 0 == memcmp("sftp:", &buf[linkstart], j)) ||
376: (j > 3 && 0 == memcmp("smb:", &buf[linkstart], j)) ||
377: (j > 3 && 0 == memcmp("afs:", &buf[linkstart], j))) {
378: /* Gross. */
379: printf("Lk %.*s", (int)((dsz > 1 ? realend - 1 :
380: realend) - linkstart), &buf[linkstart]);
1.19 kristaps 381: return(1);
382: }
383:
384: /* See if we qualify as a mailto. */
1.20 kristaps 385: if (j > 6 && 0 == memcmp("mailto:", &buf[linkstart], j)) {
1.19 kristaps 386: printf("Mt %.*s", (int)j, &buf[linkstart]);
387: return(1);
388: }
389:
390: /* See if we're a foo(5), foo(5x), or foo(5xx) manpage. */
391: if ((j > 3 && ')' == buf[linkend - 1]) &&
392: ('(' == buf[linkend - 3])) {
393: printf("Xr %.*s %c", (int)(j - 3),
394: &buf[linkstart], buf[linkend - 2]);
395: return(1);
396: } else if ((j > 4 && ')' == buf[linkend - 1]) &&
397: ('(' == buf[linkend - 4])) {
398: printf("Xr %.*s %.*s", (int)(j - 4),
399: &buf[linkstart], 2, &buf[linkend - 3]);
400: return(1);
401: } else if ((j > 5 && ')' == buf[linkend - 1]) &&
402: ('(' == buf[linkend - 5])) {
403: printf("Xr %.*s %.*s", (int)(j - 5),
404: &buf[linkstart], 3, &buf[linkend - 4]);
405: return(1);
406: }
407:
408: /* Last try: do we have a double-colon? */
409: for (i = linkstart + 1; i < linkend; i++)
410: if (':' == buf[i] && ':' == buf[i - 1])
1.18 kristaps 411: break;
1.9 kristaps 412:
1.19 kristaps 413: if (i < linkend)
1.10 kristaps 414: printf("Xr %.*s " PERL_SECTION,
1.19 kristaps 415: (int)j, &buf[linkstart]);
1.9 kristaps 416: else
1.19 kristaps 417: printf("Xr %.*s 1", (int)j, &buf[linkstart]);
1.9 kristaps 418:
419: return(1);
420: }
421:
1.13 kristaps 422: /*
423: * Doclifting: if we're a bold "-xx" and we're in the SYNOPSIS section,
424: * then it's likely that we're a flag.
425: * Our flag might be followed by an argument, so make sure that we're
426: * accounting for that, too.
427: * If we don't have a flag at all, however, then assume we're an "Ar".
1.32 schwarze 428: *
429: * Always operates in OUST_MAC mode.
430: * Mode handlinf is done by the caller.
1.13 kristaps 431: */
432: static void
433: dosynopsisfl(const char *buf, size_t *start, size_t end)
434: {
435: size_t i;
436: again:
1.14 kristaps 437: assert(*start + 1 < end);
438: assert('-' == buf[*start]);
439:
440: if ( ! isalnum((int)buf[*start + 1]) &&
441: '?' != buf[*start + 1] &&
442: '-' != buf[*start + 1]) {
443: (*start)--;
444: fputs("Ar ", stdout);
445: return;
446: }
447:
1.13 kristaps 448: (*start)++;
449: for (i = *start; i < end; i++)
450: if (isalnum((int)buf[i]))
451: continue;
1.14 kristaps 452: else if ('?' == buf[i])
453: continue;
1.13 kristaps 454: else if ('-' == buf[i])
455: continue;
456: else if ('_' == buf[i])
457: continue;
458: else
459: break;
460:
461: assert(i < end);
462:
463: if ( ! (' ' == buf[i] || '>' == buf[i])) {
464: printf("Ar ");
465: return;
466: }
467:
468: printf("Fl ");
469: if (end - *start > 1 &&
470: isupper((int)buf[*start]) &&
471: islower((int)buf[*start + 1]) &&
472: (end - *start == 2 ||
473: ' ' == buf[*start + 2]))
474: printf("\\&");
475: printf("%.*s ", (int)(i - *start), &buf[*start]);
476: *start = i;
477:
478: if (' ' == buf[i]) {
479: while (i < end && ' ' == buf[i])
480: i++;
481: assert(i < end);
482: if ('-' == buf[i]) {
483: *start = i;
484: goto again;
485: }
486: printf("Ar ");
487: *start = i;
488: }
489: }
490:
1.9 kristaps 491: /*
1.1 schwarze 492: * We're at the character in front of a format code, which is structured
493: * like X<...> and can contain nested format codes.
494: * This consumes the whole format code, and any nested format codes, til
495: * the end of matched production.
1.6 kristaps 496: * If "nomacro", then we don't print any macros, just contained data
497: * (e.g., following "Sh" or "Nm").
1.15 kristaps 498: * "pos" is only significant in SYNOPSIS, and should be 0 when invoked
499: * as the first format code on a line (for decoration as an "Nm"),
500: * non-zero otherwise.
1.32 schwarze 501: *
502: * Output mode handling is most complicated here.
503: * We may enter in any mode.
504: * We usually exit in OUST_MAC mode, except when
505: * entering without OUST_MAC and the code is invalid.
1.1 schwarze 506: */
1.33 schwarze 507: static int
1.15 kristaps 508: formatcode(struct state *st, const char *buf, size_t *start,
1.32 schwarze 509: size_t end, int nomacro, int pos)
1.1 schwarze 510: {
1.40 schwarze 511: size_t i, j, dsz;
1.1 schwarze 512: enum fmt fmt;
1.39 schwarze 513: unsigned char uc;
1.1 schwarze 514:
515: assert(*start + 1 < end);
516: assert('<' == buf[*start + 1]);
517:
1.6 kristaps 518: /*
519: * First, look up the format code.
1.30 schwarze 520: * If it's not valid, treat it as a NOOP.
1.6 kristaps 521: */
522: for (fmt = 0; fmt < FMT__MAX; fmt++)
523: if (buf[*start] == fmts[fmt])
524: break;
525:
1.5 kristaps 526: /*
527: * Determine whether we're overriding our delimiter.
528: * According to POD, if we have more than one '<' followed by a
529: * space, then we need a space followed by matching '>' to close
530: * the expression.
531: * Otherwise we use the usual '<' and '>' matched pair.
532: */
533: i = *start + 1;
534: while (i < end && '<' == buf[i])
535: i++;
536: assert(i > *start + 1);
537: dsz = i - (*start + 1);
538: if (dsz > 1 && (i >= end || ' ' != buf[i]))
539: dsz = 1;
540:
541: /* Remember, if dsz>1, to jump the trailing space. */
542: *start += dsz + 1 + (dsz > 1 ? 1 : 0);
1.1 schwarze 543:
544: /*
1.6 kristaps 545: * Escapes and ignored codes (NULL and INDEX) don't print macro
546: * sequences, so just output them like normal text before
547: * processing for real macros.
1.1 schwarze 548: */
549: if (FMT_ESCAPE == fmt) {
1.31 schwarze 550: formatescape(st, buf, start, end);
1.33 schwarze 551: return(0);
1.1 schwarze 552: } else if (FMT_NULL == fmt || FMT_INDEX == fmt) {
1.5 kristaps 553: /*
1.6 kristaps 554: * Just consume til the end delimiter, accounting for
555: * whether it's a custom one.
1.5 kristaps 556: */
557: for ( ; *start < end; (*start)++) {
558: if ('>' != buf[*start])
559: continue;
560: else if (dsz == 1)
561: break;
562: assert(*start > 0);
563: if (' ' != buf[*start - 1])
564: continue;
565: i = *start;
566: for (j = 0; i < end && j < dsz; j++)
567: if ('>' != buf[i++])
568: break;
569: if (dsz != j)
570: continue;
571: (*start) += dsz;
572: break;
573: }
1.24 kristaps 574: if (*start < end) {
575: assert('>' == buf[*start]);
576: (*start)++;
577: }
578: if (isspace(last))
579: while (*start < end && isspace((int)buf[*start]))
580: (*start)++;
1.33 schwarze 581: return(0);
1.1 schwarze 582: }
583:
1.6 kristaps 584: /*
585: * Check whether we're supposed to print macro stuff (this is
586: * suppressed in, e.g., "Nm" and "Sh" macros).
587: */
1.30 schwarze 588: if (FMT__MAX != fmt && !nomacro) {
1.32 schwarze 589:
590: /*
1.31 schwarze 591: * If we are on a text line and there is no
592: * whitespace before our content, we have to make
593: * the previous word a prefix to the macro line.
1.1 schwarze 594: */
1.31 schwarze 595:
1.54 schwarze 596: if (OUST_MAC != st->oust && ' ' != buf[*start] &&
597: st->outbuflen) {
1.32 schwarze 598: if (OUST_NL != st->oust)
1.54 schwarze 599: mdoc_newln(st);
1.31 schwarze 600: printf(".Pf ");
1.54 schwarze 601: st->oust = OUST_MAC;
602: st->wantws = 1;
1.31 schwarze 603: }
604:
605: outbuf_flush(st);
606:
1.54 schwarze 607: /*
608: * Whitespace is easier to suppress on macro lines.
609: * We may already have wantws if there was whitespace
610: * before the code ("text B<text"), or there may be
611: * whitespace inside our scope ("textB< text").
612: */
1.31 schwarze 613:
1.54 schwarze 614: if (OUST_MAC == st->oust && ' ' != buf[*start] &&
615: ! st->wantws)
616: printf(" Ns");
1.31 schwarze 617:
618: /* Unless we are on a macro line, start one. */
619:
1.54 schwarze 620: if (OUST_MAC != st->oust) {
1.32 schwarze 621: if (OUST_NL != st->oust)
1.54 schwarze 622: mdoc_newln(st);
1.1 schwarze 623: putchar('.');
1.54 schwarze 624: st->oust = OUST_MAC;
1.31 schwarze 625: } else
1.1 schwarze 626: putchar(' ');
1.54 schwarze 627: st->wantws = 1;
1.31 schwarze 628:
1.32 schwarze 629: /*
630: * Print the macro corresponding to this format code,
631: * and update the output state afterwards.
632: */
1.6 kristaps 633:
1.1 schwarze 634: switch (fmt) {
635: case (FMT_ITALIC):
636: printf("Em ");
637: break;
638: case (FMT_BOLD):
1.14 kristaps 639: if (SECT_SYNOPSIS == st->sect) {
640: if (1 == dsz && '-' == buf[*start])
641: dosynopsisfl(buf, start, end);
1.15 kristaps 642: else if (0 == pos)
643: printf("Nm ");
1.14 kristaps 644: else
645: printf("Ar ");
646: break;
1.39 schwarze 647: }
648: i = 0;
649: uc = buf[*start];
650: while (isalnum(uc) || '_' == uc || ' ' == uc)
651: uc = buf[*start + ++i];
652: if ('=' != uc && '>' != uc)
653: i = 0;
654: if (4 == i && ! strncmp(buf + *start, "NULL", 4)) {
1.27 schwarze 655: printf("Dv ");
1.38 schwarze 656: break;
657: }
1.39 schwarze 658: switch (i ? dict_get(buf + *start, i) : MDOC_MAX) {
659: case MDOC_Fa:
1.38 schwarze 660: printf("Fa ");
1.39 schwarze 661: break;
662: case MDOC_Vt:
663: printf("Vt ");
664: break;
665: default:
1.27 schwarze 666: printf("Sy ");
1.39 schwarze 667: break;
668: }
1.1 schwarze 669: break;
670: case (FMT_CODE):
1.2 schwarze 671: printf("Qo Li ");
1.1 schwarze 672: break;
673: case (FMT_LINK):
1.19 kristaps 674: /* Try to link; use "No" if it's empty. */
1.9 kristaps 675: if ( ! trylink(buf, start, end, dsz))
676: printf("No ");
1.1 schwarze 677: break;
678: case (FMT_FILE):
679: printf("Pa ");
680: break;
681: case (FMT_NBSP):
682: printf("No ");
683: break;
684: default:
685: abort();
686: }
1.31 schwarze 687: } else
688: outbuf_flush(st);
1.1 schwarze 689:
690: /*
1.6 kristaps 691: * Process until we reach the end marker (e.g., '>') or until we
1.5 kristaps 692: * find a nested format code.
1.1 schwarze 693: * Don't emit any newlines: since we're on a macro line, we
694: * don't want to break the line.
695: */
696: while (*start < end) {
1.5 kristaps 697: if ('>' == buf[*start] && 1 == dsz) {
1.1 schwarze 698: (*start)++;
699: break;
1.5 kristaps 700: } else if ('>' == buf[*start] &&
701: ' ' == buf[*start - 1]) {
702: /*
703: * Handle custom delimiters.
704: * These require a certain number of
705: * space-preceded carrots before we're really at
706: * the end.
707: */
708: i = *start;
709: for (j = 0; i < end && j < dsz; j++)
710: if ('>' != buf[i++])
711: break;
712: if (dsz == j) {
713: *start += dsz;
714: break;
715: }
1.1 schwarze 716: }
1.34 schwarze 717: if (*start + 1 < end && '<' == buf[*start + 1] &&
718: 'A' <= buf[*start] && 'Z' >= buf[*start]) {
1.40 schwarze 719: if ( ! formatcode(st, buf, start, end, nomacro, 1))
720: st->wantws = 1;
1.1 schwarze 721: continue;
722: }
1.3 schwarze 723:
1.32 schwarze 724: /* Suppress newlines and multiple spaces. */
725:
726: last = buf[(*start)++];
727: if (' ' == last || '\n' == last) {
728: putchar(' ');
729: while (*start < end && ' ' == buf[*start])
730: (*start)++;
731: continue;
732: }
733:
1.33 schwarze 734: if (OUST_MAC == st->oust && FMT__MAX != fmt) {
1.32 schwarze 735: if ( ! st->wantws) {
736: printf(" Ns ");
737: st->wantws = 1;
738: }
739:
740: /*
741: * Escape macro-like words.
742: * This matches "Xx " and "XxEOLN".
743: */
744:
745: if (end - *start > 0 &&
746: isupper((unsigned char)last) &&
747: islower((unsigned char)buf[*start]) &&
748: (end - *start == 1 ||
749: ' ' == buf[*start + 1] ||
750: '>' == buf[*start + 1]))
751: printf("\\&");
752: }
1.3 schwarze 753:
1.32 schwarze 754: putchar(last);
1.4 schwarze 755:
1.8 kristaps 756: /* Protect against character escapes. */
1.32 schwarze 757:
1.8 kristaps 758: if ('\\' == last)
759: putchar('e');
1.1 schwarze 760: }
1.2 schwarze 761:
762: if ( ! nomacro && FMT_CODE == fmt)
763: printf(" Qc ");
1.1 schwarze 764:
1.33 schwarze 765: st->wantws = ' ' == last;
1.40 schwarze 766: return(FMT__MAX != fmt);
1.1 schwarze 767: }
768:
769: /*
770: * Calls formatcode() til the end of a paragraph.
1.32 schwarze 771: * Goes to OUST_MAC mode and stays there when returning,
772: * such that the caller can add arguments to the macro line
773: * before closing it out.
1.1 schwarze 774: */
775: static void
1.32 schwarze 776: formatcodeln(struct state *st, const char *linemac,
777: const char *buf, size_t *start, size_t end, int nomacro)
1.1 schwarze 778: {
1.33 schwarze 779: int gotmacro, wantws;
1.1 schwarze 780:
1.32 schwarze 781: assert(OUST_NL == st->oust);
782: assert(st->wantws);
783: printf(".%s ", linemac);
784: st->oust = OUST_MAC;
785:
1.33 schwarze 786: gotmacro = 0;
1.1 schwarze 787: while (*start < end) {
1.33 schwarze 788: wantws = ' ' == buf[*start] || '\n' == buf[*start];
789: if (wantws) {
790: last = ' ';
791: do {
792: (*start)++;
793: } while (*start < end && ' ' == buf[*start]);
794: }
795:
1.34 schwarze 796: if (*start + 1 < end && '<' == buf[*start + 1] &&
797: 'A' <= buf[*start] && 'Z' >= buf[*start]) {
1.33 schwarze 798: st->wantws |= wantws;
799: gotmacro = formatcode(st, buf,
800: start, end, nomacro, 1);
1.1 schwarze 801: continue;
802: }
1.32 schwarze 803:
1.33 schwarze 804: if (gotmacro) {
805: if (*start < end || st->outbuflen) {
806: if (st->wantws ||
807: (wantws && !st->outbuflen))
808: printf(" No ");
809: else
810: printf(" Ns ");
811: }
812: gotmacro = 0;
813: }
814: outbuf_flush(st);
815: st->wantws = wantws;
816:
817: if (*start >= end)
818: break;
819:
820: if (st->wantws) {
821: putchar(' ');
822: st->wantws = 0;
1.32 schwarze 823: }
824:
1.4 schwarze 825: /*
826: * Since we're already on a macro line, we want to make
827: * sure that we don't inadvertently invoke a macro.
828: * We need to do this carefully because section names
829: * are used in troff and we don't want to escape
830: * something that needn't be escaped.
831: */
832: if (' ' == last && end - *start > 1 &&
1.33 schwarze 833: isupper((unsigned char)buf[*start]) &&
834: islower((unsigned char)buf[*start + 1]) &&
835: (end - *start == 2 || ' ' == buf[*start + 2]))
1.4 schwarze 836: printf("\\&");
837:
1.33 schwarze 838: putchar(last = buf[*start]);
1.8 kristaps 839:
840: /* Protect against character escapes. */
1.33 schwarze 841:
1.8 kristaps 842: if ('\\' == last)
843: putchar('e');
844:
1.1 schwarze 845: (*start)++;
846: }
847: }
848:
849: /*
1.4 schwarze 850: * Guess at what kind of list we are.
851: * These are taken straight from the POD manual.
852: * I don't know what people do in real life.
853: */
854: static enum list
855: listguess(const char *buf, size_t start, size_t end)
856: {
857: size_t len = end - start;
858:
859: assert(end >= start);
860:
861: if (len == 1 && '*' == buf[start])
862: return(LIST_BULLET);
863: if (len == 2 && '1' == buf[start] && '.' == buf[start + 1])
864: return(LIST_ENUM);
865: else if (len == 1 && '1' == buf[start])
866: return(LIST_ENUM);
867: else
868: return(LIST_TAG);
869: }
870:
871: /*
1.1 schwarze 872: * A command paragraph, as noted in the perlpod manual, just indicates
873: * that we should do something, optionally with some text to print as
874: * well.
1.32 schwarze 875: * From the perspective of external callers,
876: * always stays in OUST_NL/wantws mode,
877: * but its children do use OUST_MAC.
1.1 schwarze 878: */
879: static void
880: command(struct state *st, const char *buf, size_t start, size_t end)
881: {
882: size_t len, csz;
883: enum cmd cmd;
884:
885: assert('=' == buf[start]);
886: start++;
887: len = end - start;
888:
889: for (cmd = 0; cmd < CMD__MAX; cmd++) {
890: csz = strlen(cmds[cmd]);
891: if (len < csz)
892: continue;
893: if (0 == memcmp(&buf[start], cmd[cmds], csz))
894: break;
895: }
896:
897: /* Ignore bogus commands. */
898:
899: if (CMD__MAX == cmd)
900: return;
901:
902: start += csz;
1.8 kristaps 903: while (start < end && ' ' == buf[start])
904: start++;
905:
1.1 schwarze 906: len = end - start;
907:
908: if (st->paused) {
909: st->paused = CMD_END != cmd;
910: return;
911: }
912:
913: switch (cmd) {
914: case (CMD_POD):
915: break;
916: case (CMD_HEAD1):
917: /*
918: * The behaviour of head= follows from a quick glance at
919: * how pod2man handles it.
920: */
1.11 kristaps 921: st->sect = SECT_NONE;
922: if (end - start == 4) {
1.1 schwarze 923: if (0 == memcmp(&buf[start], "NAME", 4))
1.11 kristaps 924: st->sect = SECT_NAME;
925: } else if (end - start == 8) {
926: if (0 == memcmp(&buf[start], "SYNOPSIS", 8))
927: st->sect = SECT_SYNOPSIS;
928: }
1.32 schwarze 929: formatcodeln(st, "Sh", buf, &start, end, 1);
930: mdoc_newln(st);
1.1 schwarze 931: st->haspar = 1;
932: break;
933: case (CMD_HEAD2):
1.32 schwarze 934: formatcodeln(st, "Ss", buf, &start, end, 1);
935: mdoc_newln(st);
1.1 schwarze 936: st->haspar = 1;
937: break;
938: case (CMD_HEAD3):
939: puts(".Pp");
1.32 schwarze 940: formatcodeln(st, "Em", buf, &start, end, 0);
941: mdoc_newln(st);
1.1 schwarze 942: puts(".Pp");
943: st->haspar = 1;
944: break;
945: case (CMD_HEAD4):
946: puts(".Pp");
1.32 schwarze 947: formatcodeln(st, "No", buf, &start, end, 0);
948: mdoc_newln(st);
1.1 schwarze 949: puts(".Pp");
950: st->haspar = 1;
951: break;
952: case (CMD_OVER):
1.4 schwarze 953: /*
954: * If we have an existing list that hasn't had an =item
955: * yet, then make sure that we open it now.
956: * We use the default list type, but that can't be
957: * helped (we haven't seen any items yet).
1.1 schwarze 958: */
1.4 schwarze 959: if (st->lpos > 0)
960: if (LIST__MAX == st->lstack[st->lpos - 1]) {
961: st->lstack[st->lpos - 1] = LIST_TAG;
962: puts(".Bl -tag -width Ds");
963: }
964: st->lpos++;
965: assert(st->lpos < LIST_STACKSZ);
966: st->lstack[st->lpos - 1] = LIST__MAX;
1.1 schwarze 967: break;
968: case (CMD_ITEM):
1.6 kristaps 969: if (0 == st->lpos) {
970: /*
971: * Bad markup.
972: * Try to compensate.
973: */
974: st->lstack[st->lpos] = LIST__MAX;
975: st->lpos++;
976: }
1.4 schwarze 977: assert(st->lpos > 0);
978: /*
979: * If we're the first =item, guess at what our content
980: * will be: "*" is a bullet list, "1." is a numbered
981: * list, and everything is tagged.
982: */
983: if (LIST__MAX == st->lstack[st->lpos - 1]) {
984: st->lstack[st->lpos - 1] =
985: listguess(buf, start, end);
986: switch (st->lstack[st->lpos - 1]) {
987: case (LIST_BULLET):
988: puts(".Bl -bullet");
989: break;
990: case (LIST_ENUM):
991: puts(".Bl -enum");
992: break;
993: default:
994: puts(".Bl -tag -width Ds");
995: break;
996: }
997: }
998: switch (st->lstack[st->lpos - 1]) {
999: case (LIST_TAG):
1.32 schwarze 1000: formatcodeln(st, "It", buf, &start, end, 0);
1001: mdoc_newln(st);
1.4 schwarze 1002: break;
1003: case (LIST_ENUM):
1004: /* FALLTHROUGH */
1005: case (LIST_BULLET):
1006: /*
1007: * Abandon the remainder of the paragraph
1008: * because we're going to be a bulletted or
1009: * numbered list.
1010: */
1011: puts(".It");
1012: break;
1013: default:
1014: abort();
1015: }
1.1 schwarze 1016: st->haspar = 1;
1017: break;
1018: case (CMD_BACK):
1.4 schwarze 1019: /* Make sure we don't back over the stack. */
1020: if (st->lpos > 0) {
1021: st->lpos--;
1022: puts(".El");
1023: }
1.1 schwarze 1024: break;
1025: case (CMD_BEGIN):
1026: /*
1027: * We disregard all types for now.
1028: * TODO: process at least "text" in a -literal block.
1029: */
1030: st->paused = 1;
1031: break;
1032: case (CMD_FOR):
1033: /*
1034: * We ignore all types of encodings and formats
1035: * unilaterally.
1036: */
1037: break;
1038: case (CMD_ENCODING):
1039: break;
1040: case (CMD_CUT):
1041: st->parsing = 0;
1042: return;
1043: default:
1044: abort();
1045: }
1046:
1047: /* Any command (but =cut) makes us start parsing. */
1048: st->parsing = 1;
1049: }
1050:
1051: /*
1.39 schwarze 1052: * Put the type provided as an argument into the dictionary.
1053: */
1054: static void
1055: register_type(const char *ptype)
1056: {
1057: const char *pname, *pend;
1058:
1059: pname = ptype;
1060: while (isalnum((unsigned char)*pname) || '_' == *pname)
1061: pname++;
1062: if ((pname - ptype == 6 && ! strncmp(ptype, "struct", 6)) ||
1063: (pname - ptype == 4 && ! strncmp(ptype, "enum", 4))) {
1064: while (' ' == *pname)
1065: pname++;
1066: pend = pname;
1067: while (isalnum((unsigned char)*pend) || '_' == *pend)
1068: pend++;
1069: if (pend > pname)
1070: dict_put(pname, pend - pname, MDOC_Vt);
1071: } else
1072: pend = pname;
1073: if (pend > ptype)
1074: dict_put(ptype, pend - ptype, MDOC_Vt);
1075: }
1076:
1077: /*
1.1 schwarze 1078: * Just pump out the line in a verbatim block.
1.32 schwarze 1079: * From the perspective of external callers,
1080: * always stays in OUST_NL/wantws mode.
1.1 schwarze 1081: */
1082: static void
1.35 schwarze 1083: verbatim(struct state *st, char *buf, size_t start, size_t end)
1.1 schwarze 1084: {
1.36 schwarze 1085: size_t i, ift, ifo, ifa, ifc, inl;
1.38 schwarze 1086: char *cp, *cp2;
1.53 schwarze 1087: int indisplay, nopen, wantsp;
1.1 schwarze 1088:
1.53 schwarze 1089: if (st->paused || ! st->parsing)
1.1 schwarze 1090: return;
1.53 schwarze 1091:
1092: indisplay = wantsp = 0;
1093:
1.22 kristaps 1094: again:
1.53 schwarze 1095: if (start == end) {
1096: if (indisplay)
1097: puts(".Ed");
1098: return;
1099: }
1100:
1101: if ('\n' == buf[start]) {
1102: wantsp = 1;
1103: start++;
1104: goto again;
1105: }
1106:
1.22 kristaps 1107: /*
1108: * If we're in the SYNOPSIS, see if we're an #include block.
1109: * If we are, then print the "In" macro and re-loop.
1110: * This handles any number of inclusions, but only when they
1111: * come before the remaining parts...
1112: */
1113: if (SECT_SYNOPSIS == st->sect) {
1114: i = start;
1.35 schwarze 1115: while (i < end && buf[i] == ' ')
1116: i++;
1.22 kristaps 1117: if (i == end)
1.53 schwarze 1118: goto again;
1.35 schwarze 1119:
1.22 kristaps 1120: /* We're an include block! */
1121: if (end - i > 10 &&
1122: 0 == memcmp(&buf[i], "#include <", 10)) {
1123: start = i + 10;
1124: while (start < end && ' ' == buf[start])
1125: start++;
1.53 schwarze 1126: if (indisplay)
1127: puts(".Ed");
1128: indisplay = wantsp = 0;
1.22 kristaps 1129: fputs(".In ", stdout);
1130: /* Stop til the '>' marker or we hit eoln. */
1131: while (start < end &&
1132: '>' != buf[start] && '\n' != buf[start])
1133: putchar(buf[start++]);
1134: putchar('\n');
1135: if (start < end && '>' == buf[start])
1136: start++;
1137: if (start < end && '\n' == buf[start])
1138: start++;
1.41 schwarze 1139: goto again;
1140: }
1141:
1142: /* Other preprocessor directives. */
1143: if ('#' == buf[i]) {
1.53 schwarze 1144: if (indisplay)
1145: puts(".Ed");
1146: indisplay = wantsp = 0;
1.41 schwarze 1147: fputs(".Fd ", stdout);
1148: start = i;
1149: while(start < end && '\n' != buf[start])
1150: putchar(buf[start++]);
1151: putchar('\n');
1152: if (start < end && '\n' == buf[start])
1153: start++;
1.49 schwarze 1154:
1155: /* Remember #define for Dv or Fn. */
1156:
1157: if (strncmp(buf + i + 1, "define", 6) ||
1158: ! isspace((unsigned char)buf[i + 7]))
1159: goto again;
1160:
1161: ifo = i + 7;
1162: while (ifo < start &&
1163: isspace((unsigned char)buf[ifo]))
1164: ifo++;
1165: ifa = ifo;
1166: while ('_' == buf[ifa] ||
1167: isalnum((unsigned char)buf[ifa]))
1168: ifa++;
1169: dict_put(buf + ifo, ifa - ifo,
1170: '(' == buf[ifa] ? MDOC_Fo : MDOC_Dv);
1171:
1.41 schwarze 1172: goto again;
1.22 kristaps 1173: }
1.35 schwarze 1174:
1175: /* Parse function declaration. */
1176: ifo = ifa = ifc = 0;
1.36 schwarze 1177: inl = end;
1178: nopen = 0;
1179: for (ift = i; i < end; i++) {
1180: if (ifc) {
1181: if (buf[i] != '\n')
1182: continue;
1183: inl = i;
1184: break;
1185: }
1186: switch (buf[i]) {
1.45 schwarze 1187: case '\t':
1188: /* FALLTHROUGH */
1.36 schwarze 1189: case ' ':
1190: if ( ! ifa)
1191: ifo = i;
1192: break;
1193: case '(':
1194: if (ifo) {
1195: nopen++;
1196: if ( ! ifa)
1197: ifa = i;
1198: } else
1199: i = end;
1200: break;
1201: case ')':
1202: switch (nopen) {
1203: case 0:
1204: i = end;
1205: break;
1206: case 1:
1.35 schwarze 1207: ifc = i;
1.36 schwarze 1208: break;
1209: default:
1210: nopen--;
1211: break;
1212: }
1213: break;
1214: default:
1215: break;
1216: }
1.35 schwarze 1217: }
1218:
1219: /* Encode function declaration. */
1220: if (ifc) {
1.36 schwarze 1221: for (i = ifa; i < ifc; i++)
1222: if (buf[i] == '\n')
1223: buf[i] = ' ';
1.35 schwarze 1224: buf[ifo++] = '\0';
1.39 schwarze 1225: register_type(buf + ift);
1.53 schwarze 1226: if (indisplay)
1227: puts(".Ed");
1228: indisplay = wantsp = 0;
1.35 schwarze 1229: printf(".Ft %s", buf + ift);
1230: if (buf[ifo] == '*') {
1231: fputs(" *", stdout);
1232: ifo++;
1233: }
1234: putchar('\n');
1235: buf[ifa++] = '\0';
1236: printf(".Fo %s\n", buf + ifo);
1.39 schwarze 1237: dict_put(buf + ifo, 0, MDOC_Fo);
1.35 schwarze 1238: buf[ifc++] = '\0';
1239: for (;;) {
1240: cp = strchr(buf + ifa, ',');
1.38 schwarze 1241: if (cp != NULL) {
1242: cp2 = cp;
1.36 schwarze 1243: *cp++ = '\0';
1.38 schwarze 1244: } else
1245: cp2 = strchr(buf + ifa, '\0');
1246: while (isalnum((unsigned char)cp2[-1]) ||
1247: '_' == cp2[-1])
1248: cp2--;
1249: if ('\0' != *cp2)
1.39 schwarze 1250: dict_put(cp2, 0, MDOC_Fa);
1251: register_type(buf + ifa);
1.50 schwarze 1252: if (strchr(buf + ifa, ' ') == NULL)
1253: printf(".Fa %s\n", buf + ifa);
1254: else
1255: printf(".Fa \"%s\"\n", buf + ifa);
1.35 schwarze 1256: if (cp == NULL)
1257: break;
1.45 schwarze 1258: while (*cp == ' ' || *cp == '\t')
1.36 schwarze 1259: cp++;
1260: ifa = cp - buf;
1.35 schwarze 1261: }
1262: puts(".Fc");
1263: if (buf[ifc] == ';')
1264: ifc++;
1.36 schwarze 1265: if (ifc < inl) {
1266: buf[inl] = '\0';
1.35 schwarze 1267: puts(buf + ifc);
1268: }
1.53 schwarze 1269: start = inl < end ? inl + 1 : end;
1270: goto again;
1.35 schwarze 1271: }
1.22 kristaps 1272: }
1.53 schwarze 1273:
1274: if ( ! indisplay)
1275: puts(".Bd -literal");
1276: else if (wantsp)
1277: putchar('\n');
1278: indisplay = 1;
1279: wantsp = 0;
1280:
1281: for (last = '\n'; start < end; start++) {
1.8 kristaps 1282: /*
1283: * Handle accidental macros (newline starting with
1284: * control character) and escapes.
1285: */
1.53 schwarze 1286: if ('\n' == last) {
1287: if ('\n' == buf[start])
1288: goto again;
1.7 kristaps 1289: if ('.' == buf[start] || '\'' == buf[start])
1290: printf("\\&");
1.53 schwarze 1291: }
1.8 kristaps 1292: putchar(last = buf[start]);
1293: if ('\\' == buf[start])
1294: printf("e");
1.7 kristaps 1295: }
1.53 schwarze 1296: if ('\n' != last)
1297: putchar('\n');
1298: if (indisplay)
1299: puts(".Ed");
1.1 schwarze 1300: }
1301:
1302: /*
1.13 kristaps 1303: * See dosynopsisop().
1304: */
1305: static int
1306: hasmatch(const char *buf, size_t start, size_t end)
1307: {
1308: size_t stack;
1309:
1310: for (stack = 0; start < end; start++)
1311: if (buf[start] == '[')
1312: stack++;
1313: else if (buf[start] == ']' && 0 == stack)
1314: return(1);
1315: else if (buf[start] == ']')
1316: stack--;
1317: return(0);
1318: }
1319:
1320: /*
1321: * If we're in the SYNOPSIS section and we've encounter braces in an
1322: * ordinary paragraph, then try to see whether we're an [-option].
1323: * Do this, if we're an opening bracket, by first seeing if we have a
1324: * matching end via hasmatch().
1325: * If we're an ending bracket, see if we have a stack already.
1326: */
1327: static int
1.32 schwarze 1328: dosynopsisop(struct state *st, const char *buf,
1329: size_t *start, size_t end, size_t *opstack)
1.13 kristaps 1330: {
1331:
1332: assert('[' == buf[*start] || ']' == buf[*start]);
1333:
1334: if ('[' == buf[*start] && hasmatch(buf, *start + 1, end)) {
1.32 schwarze 1335: mdoc_newln(st);
1.13 kristaps 1336: puts(".Oo");
1337: (*opstack)++;
1338: } else if ('[' == buf[*start])
1339: return(0);
1340:
1341: if (']' == buf[*start] && *opstack > 0) {
1.32 schwarze 1342: mdoc_newln(st);
1.13 kristaps 1343: puts(".Oc");
1344: (*opstack)--;
1345: } else if (']' == buf[*start])
1346: return(0);
1347:
1348: (*start)++;
1.31 schwarze 1349: last = '\n';
1.13 kristaps 1350: while (' ' == buf[*start])
1351: (*start)++;
1352: return(1);
1353: }
1354:
1355: /*
1.17 kristaps 1356: * Format multiple "Nm" manpage names in the NAME section.
1.32 schwarze 1357: * From the perspective of external callers,
1358: * always stays in OUST_NL/wantws mode,
1359: * but its children do use OUST_MAC.
1.17 kristaps 1360: */
1361: static void
1362: donamenm(struct state *st, const char *buf, size_t *start, size_t end)
1363: {
1364: size_t word;
1365:
1.32 schwarze 1366: assert(OUST_NL == st->oust);
1367: assert(st->wantws);
1368:
1.47 schwarze 1369: while (*start < end && isspace((unsigned char)buf[*start]))
1.17 kristaps 1370: (*start)++;
1371:
1372: if (end == *start) {
1373: puts(".Nm unknown");
1374: return;
1375: }
1376:
1377: while (*start < end) {
1378: for (word = *start; word < end; word++)
1379: if (',' == buf[word])
1380: break;
1.32 schwarze 1381: formatcodeln(st, "Nm", buf, start, word, 1);
1.17 kristaps 1382: if (*start == end) {
1.32 schwarze 1383: mdoc_newln(st);
1384: break;
1.17 kristaps 1385: }
1386: assert(',' == buf[*start]);
1.32 schwarze 1387: printf(" ,");
1388: mdoc_newln(st);
1.17 kristaps 1389: (*start)++;
1.47 schwarze 1390: while (*start < end && isspace((unsigned char)buf[*start]))
1.17 kristaps 1391: (*start)++;
1392: }
1393: }
1394:
1395: /*
1.1 schwarze 1396: * Ordinary paragraph.
1397: * Well, this is really the hardest--POD seems to assume that, for
1398: * example, a leading space implies a newline, and so on.
1399: * Lots of other snakes in the grass: escaping a newline followed by a
1400: * period (accidental mdoc(7) control), double-newlines after macro
1401: * passages, etc.
1.32 schwarze 1402: *
1403: * Uses formatcode() to go to OUST_MAC mode
1404: * and outbuf_flush() to go to OUST_TXT mode.
1.40 schwarze 1405: * In text mode, wantws requests white space before the text
1406: * currently contained in the outbuf, not before upcoming text.
1.32 schwarze 1407: * Must make sure to go back to OUST_NL/wantws mode before returning.
1.1 schwarze 1408: */
1409: static void
1410: ordinary(struct state *st, const char *buf, size_t start, size_t end)
1411: {
1.44 schwarze 1412: size_t i, j, opstack, wend;
1.43 schwarze 1413: enum mdoc_type mtype;
1.44 schwarze 1414: int eos, noeos, seq;
1.49 schwarze 1415: char savechar;
1.1 schwarze 1416:
1417: if ( ! st->parsing || st->paused)
1418: return;
1419:
1420: /*
1421: * Special-case: the NAME section.
1422: * If we find a "-" when searching from the end, assume that
1423: * we're in "name - description" format.
1424: * To wit, print out a "Nm" and "Nd" in that format.
1425: */
1.11 kristaps 1426: if (SECT_NAME == st->sect) {
1.15 kristaps 1427: for (i = end - 2; i > start; i--)
1.47 schwarze 1428: if ('-' == buf[i] &&
1429: isspace((unsigned char)buf[i + 1]))
1.1 schwarze 1430: break;
1431: if ('-' == buf[i]) {
1432: j = i;
1433: /* Roll over multiple "-". */
1434: for ( ; i > start; i--)
1435: if ('-' != buf[i])
1436: break;
1.17 kristaps 1437: donamenm(st, buf, &start, i + 1);
1.5 kristaps 1438: start = j + 1;
1.47 schwarze 1439: while (start < end &&
1440: isspace((unsigned char)buf[start]))
1.17 kristaps 1441: start++;
1.32 schwarze 1442: formatcodeln(st, "Nd", buf, &start, end, 1);
1443: mdoc_newln(st);
1.1 schwarze 1444: return;
1445: }
1446: }
1447:
1448: if ( ! st->haspar)
1449: puts(".Pp");
1450:
1451: st->haspar = 0;
1452: last = '\n';
1.13 kristaps 1453: opstack = 0;
1.1 schwarze 1454:
1.15 kristaps 1455: for (seq = 0; start < end; seq++) {
1.1 schwarze 1456: /*
1457: * Loop til we get either to a newline or escape.
1458: * Escape initial control characters.
1459: */
1460: while (start < end) {
1.34 schwarze 1461: if (start < end - 1 && '<' == buf[start + 1] &&
1462: 'A' <= buf[start] && 'Z' >= buf[start])
1.1 schwarze 1463: break;
1464: else if ('\n' == buf[start])
1465: break;
1466: else if ('\n' == last && '.' == buf[start])
1.31 schwarze 1467: outbuf_addstr(st, "\\&");
1.1 schwarze 1468: else if ('\n' == last && '\'' == buf[start])
1.31 schwarze 1469: outbuf_addstr(st, "\\&");
1.12 kristaps 1470: /*
1471: * If we're in the SYNOPSIS, have square
1472: * brackets indicate that we're opening and
1473: * closing an optional context.
1474: */
1.32 schwarze 1475:
1.13 kristaps 1476: if (SECT_SYNOPSIS == st->sect &&
1477: ('[' == buf[start] ||
1478: ']' == buf[start]) &&
1.32 schwarze 1479: dosynopsisop(st, buf,
1480: &start, end, &opstack))
1.13 kristaps 1481: continue;
1.32 schwarze 1482:
1.42 schwarze 1483: /* Merely buffer non-whitespace. */
1.32 schwarze 1484:
1.31 schwarze 1485: last = buf[start++];
1.44 schwarze 1486: if ( ! isspace(last))
1.37 schwarze 1487: outbuf_addchar(st);
1.44 schwarze 1488: if (start < end &&
1.52 schwarze 1489: ! isspace((unsigned char)buf[start - 1]) &&
1.44 schwarze 1490: ! isspace((unsigned char)buf[start]))
1.37 schwarze 1491: continue;
1492:
1.44 schwarze 1493: /*
1494: * Found the end of a word.
1495: * Rewind trailing delimiters.
1496: */
1497:
1498: eos = noeos = 0;
1499: for (wend = st->outbuflen; wend; wend--)
1500: if ('.' == st->outbuf[wend - 1] ||
1501: '!' == st->outbuf[wend - 1] ||
1502: '?' == st->outbuf[wend - 1])
1503: eos = 1;
1504: else if ('|' == st->outbuf[wend - 1] ||
1505: ',' == st->outbuf[wend - 1] ||
1506: ';' == st->outbuf[wend - 1] ||
1507: ':' == st->outbuf[wend - 1])
1508: noeos = 1;
1509: else if ('\'' != st->outbuf[wend - 1] &&
1510: '"' != st->outbuf[wend - 1] &&
1511: ')' != st->outbuf[wend - 1] &&
1512: ']' != st->outbuf[wend - 1])
1513: break;
1514: eos &= ! noeos;
1515:
1516: /*
1517: * Detect function names.
1518: */
1.42 schwarze 1519:
1.43 schwarze 1520: mtype = MDOC_Fa;
1.49 schwarze 1521: savechar = '\0';
1.44 schwarze 1522: if (wend && ')' == st->outbuf[wend] &&
1523: '(' == st->outbuf[wend - 1]) {
1524: mtype = dict_get(st->outbuf, --wend);
1.49 schwarze 1525: if (MDOC_Dv == mtype)
1526: mtype = MDOC_Fo;
1.43 schwarze 1527: if (MDOC_Fo == mtype || MDOC_MAX == mtype) {
1.44 schwarze 1528: st->outbuflen = wend;
1529: st->outbuf[wend] = '\0';
1.43 schwarze 1530: mdoc_newln(st);
1531: if (MDOC_Fo == mtype)
1532: fputs(".Fn ", stdout);
1533: else
1534: fputs(".Xr ", stdout);
1535: st->oust = OUST_MAC;
1536: }
1.49 schwarze 1537: } else {
1538: mtype = dict_get(st->outbuf, wend);
1539: if (MDOC_Dv == mtype) {
1540: savechar = st->outbuf[wend];
1541: st->outbuf[wend] = '\0';
1542: mdoc_newln(st);
1543: fputs(".Dv ", stdout);
1544: st->oust = OUST_MAC;
1545: } else
1546: mtype = MDOC_Fa;
1.37 schwarze 1547: }
1548:
1.42 schwarze 1549: /*
1550: * On whitespace, flush the output buffer
1551: * and allow breaking to a macro line.
1552: */
1553:
1.37 schwarze 1554: outbuf_flush(st);
1.42 schwarze 1555:
1556: /*
1557: * End macro lines, and
1558: * end text lines at the end of sentences.
1559: */
1560:
1.44 schwarze 1561: if (OUST_MAC == st->oust || (eos && wend > 1 &&
1562: islower((unsigned char)st->outbuf[wend - 1]))) {
1.43 schwarze 1563: if (MDOC_MAX == mtype)
1564: fputs(" 3", stdout);
1.49 schwarze 1565: if (MDOC_Fa != mtype) {
1566: if (MDOC_Dv == mtype)
1567: st->outbuf[wend] = savechar;
1568: else
1569: wend += 2;
1570: while ('\0' != st->outbuf[wend])
1.44 schwarze 1571: printf(" %c",
1.49 schwarze 1572: st->outbuf[wend++]);
1573: }
1.40 schwarze 1574: mdoc_newln(st);
1.43 schwarze 1575: }
1.42 schwarze 1576:
1577: /* Advance to the next word. */
1578:
1.44 schwarze 1579: while ('\n' != buf[start] &&
1580: isspace((unsigned char)buf[start]))
1.42 schwarze 1581: start++;
1582: st->wantws = 1;
1.1 schwarze 1583: }
1584:
1.34 schwarze 1585: if (start < end - 1 && '<' == buf[start + 1] &&
1586: 'A' <= buf[start] && 'Z' >= buf[start]) {
1.32 schwarze 1587: formatcode(st, buf, &start, end, 0, seq);
1588: if (OUST_MAC == st->oust) {
1.30 schwarze 1589: /*
1590: * Let mdoc(7) handle trailing punctuation.
1591: * XXX Some punctuation characters
1592: * are not handled yet.
1593: */
1.51 schwarze 1594: if ((start == end - 1 ||
1595: (start < end - 1 &&
1596: (' ' == buf[start + 1] ||
1597: '\n' == buf[start + 1]))) &&
1598: NULL != strchr("|.,;:?!)]", buf[start])) {
1.16 kristaps 1599: putchar(' ');
1600: putchar(buf[start++]);
1601: }
1.32 schwarze 1602:
1603: if (st->wantws ||
1604: ' ' == buf[start] ||
1605: '\n' == buf[start])
1606: mdoc_newln(st);
1607:
1.30 schwarze 1608: /*
1609: * Consume all whitespace
1610: * so we don't accidentally start
1611: * an implicit literal line.
1612: */
1.32 schwarze 1613:
1.6 kristaps 1614: while (start < end && ' ' == buf[start])
1615: start++;
1.32 schwarze 1616:
1617: /*
1618: * Some text is following.
1619: * Implement requested spacing.
1620: */
1621:
1622: if ( ! st->wantws && start < end &&
1.34 schwarze 1623: ('<' != buf[start + 1] ||
1624: 'A' > buf[start] ||
1625: 'Z' < buf[start])) {
1.32 schwarze 1626: printf(" Ns ");
1627: st->wantws = 1;
1628: }
1.6 kristaps 1629: }
1.1 schwarze 1630: } else if (start < end && '\n' == buf[start]) {
1.32 schwarze 1631: outbuf_flush(st);
1632: mdoc_newln(st);
1.1 schwarze 1633: if (++start >= end)
1634: continue;
1635: /*
1636: * If we have whitespace next, eat it to prevent
1637: * mdoc(7) from thinking that it's meant for
1638: * verbatim text.
1639: * It is--but if we start with that, we can't
1640: * have a macro subsequent it, which may be
1641: * possible if we have an escape next.
1642: */
1.31 schwarze 1643: if (' ' == buf[start] || '\t' == buf[start])
1.1 schwarze 1644: puts(".br");
1645: for ( ; start < end; start++)
1646: if (' ' != buf[start] && '\t' != buf[start])
1647: break;
1.12 kristaps 1648: }
1.1 schwarze 1649: }
1.32 schwarze 1650: outbuf_flush(st);
1651: mdoc_newln(st);
1.1 schwarze 1652: }
1653:
1654: /*
1655: * There are three kinds of paragraphs: verbatim (starts with whitespace
1656: * of some sort), ordinary (starts without "=" marker), or a command
1657: * (default: starts with "=").
1658: */
1659: static void
1.35 schwarze 1660: dopar(struct state *st, char *buf, size_t start, size_t end)
1.1 schwarze 1661: {
1662:
1.32 schwarze 1663: assert(OUST_NL == st->oust);
1664: assert(st->wantws);
1665:
1.1 schwarze 1666: if (end == start)
1667: return;
1668: if (' ' == buf[start] || '\t' == buf[start])
1669: verbatim(st, buf, start, end);
1670: else if ('=' != buf[start])
1671: ordinary(st, buf, start, end);
1672: else
1673: command(st, buf, start, end);
1674: }
1675:
1676: /*
1677: * Loop around paragraphs within a document, processing each one in the
1678: * POD way.
1679: */
1680: static void
1681: dofile(const struct args *args, const char *fname,
1.35 schwarze 1682: const struct tm *tm, char *buf, size_t sz)
1.1 schwarze 1683: {
1.29 schwarze 1684: char datebuf[64];
1.1 schwarze 1685: struct state st;
1.46 schwarze 1686: const char *fbase, *fext, *section, *date, *format;
1.1 schwarze 1687: char *title, *cp;
1.53 schwarze 1688: size_t cur, end;
1689: int verb;
1.1 schwarze 1690:
1691: if (0 == sz)
1692: return;
1693:
1.29 schwarze 1694: /*
1695: * Parsing the filename is almost always required,
1696: * except when both the title and the section
1697: * are provided on the command line.
1698: */
1699:
1700: if (NULL == args->title || NULL == args->section) {
1701: fbase = strrchr(fname, '/');
1702: if (NULL == fbase)
1703: fbase = fname;
1704: else
1705: fbase++;
1706: fext = strrchr(fbase, '.');
1707: } else
1708: fext = NULL;
1709:
1710: /*
1711: * The title will be converted to uppercase,
1712: * so it needs to be copied.
1713: */
1714:
1715: title = (NULL != args->title) ? strdup(args->title) :
1716: (NULL != fext) ? strndup(fbase, fext - fbase) :
1717: strdup(fbase);
1.1 schwarze 1718:
1719: if (NULL == title) {
1720: perror(NULL);
1721: exit(EXIT_FAILURE);
1722: }
1723:
1724: /* Section is 1 unless suffix is "pm". */
1725:
1.29 schwarze 1726: section = (NULL != args->section) ? args->section :
1727: (NULL == fext || strcmp(fext + 1, "pm")) ? "1" :
1728: PERL_SECTION;
1.1 schwarze 1729:
1730: /* Date. Or the given "tm" if not supplied. */
1731:
1.46 schwarze 1732: date = args->date;
1733: format = (NULL == date) ? "%B %d, %Y" :
1.48 schwarze 1734: strcmp(date, "Mdocdate") ? NULL : "$" "Mdocdate: %B %d %Y $";
1.46 schwarze 1735:
1736: if (NULL != format) {
1737: strftime(datebuf, sizeof(datebuf), format, tm);
1.1 schwarze 1738: date = datebuf;
1739: }
1740:
1741: for (cp = title; '\0' != *cp; cp++)
1742: *cp = toupper((int)*cp);
1743:
1744: /* The usual mdoc(7) preamble. */
1745:
1746: printf(".Dd %s\n", date);
1747: printf(".Dt %s %s\n", title, section);
1748: puts(".Os");
1749:
1750: free(title);
1751:
1.37 schwarze 1752: dict_init();
1.1 schwarze 1753: memset(&st, 0, sizeof(struct state));
1.32 schwarze 1754: st.oust = OUST_NL;
1755: st.wantws = 1;
1756:
1.1 schwarze 1757: assert(sz > 0);
1758:
1759: /* Main loop over file contents. */
1760:
1.53 schwarze 1761: cur = 0;
1762: for (;;) {
1763: while (cur < sz && '\n' == buf[cur])
1764: cur++;
1765: if (cur >= sz)
1766: break;
1767:
1768: verb = isspace((unsigned char)buf[cur]);
1769:
1.1 schwarze 1770: /* Read until next paragraph. */
1.53 schwarze 1771:
1772: for (end = cur + 1; end + 1 < sz; end++)
1773: if ('\n' == buf[end] && '\n' == buf[end + 1] &&
1774: !(verb && end + 2 < sz &&
1775: isspace((unsigned char)buf[end + 2])))
1.1 schwarze 1776: break;
1777:
1778: /* Adjust end marker for EOF. */
1.53 schwarze 1779:
1780: if (end < sz && '\n' != buf[end])
1781: end++;
1.1 schwarze 1782:
1783: /* Process paragraph and adjust start. */
1.53 schwarze 1784:
1.1 schwarze 1785: dopar(&st, buf, cur, end);
1.53 schwarze 1786: cur = end + 2;
1.1 schwarze 1787: }
1.37 schwarze 1788: dict_destroy();
1.1 schwarze 1789: }
1790:
1791: /*
1792: * Read a single file fully into memory.
1793: * If the file is "-", do it from stdin.
1794: * If successfully read, send the input buffer to dofile() for further
1795: * processing.
1796: */
1797: static int
1798: readfile(const struct args *args, const char *fname)
1799: {
1800: int fd;
1801: char *buf;
1802: size_t bufsz, cur;
1803: ssize_t ssz;
1804: struct tm *tm;
1805: time_t ttm;
1806: struct stat st;
1807:
1808: fd = 0 != strcmp("-", fname) ?
1809: open(fname, O_RDONLY, 0) : STDIN_FILENO;
1810:
1811: if (-1 == fd) {
1812: perror(fname);
1813: return(0);
1814: }
1815:
1816: if (STDIN_FILENO == fd || -1 == fstat(fd, &st)) {
1817: ttm = time(NULL);
1818: tm = localtime(&ttm);
1819: } else
1820: tm = localtime(&st.st_mtime);
1821:
1822: /*
1823: * Arbitrarily-sized initial buffer.
1824: * Should be big enough for most files...
1825: */
1826: cur = 0;
1827: bufsz = 1 << 14;
1828: if (NULL == (buf = malloc(bufsz))) {
1829: perror(NULL);
1830: exit(EXIT_FAILURE);
1831: }
1832:
1833: while ((ssz = read(fd, buf + cur, bufsz - cur)) > 0) {
1834: /* Double buffer size on fill. */
1835: if ((size_t)ssz == bufsz - cur) {
1836: bufsz *= 2;
1837: if (NULL == (buf = realloc(buf, bufsz))) {
1838: perror(NULL);
1839: exit(EXIT_FAILURE);
1840: }
1841: }
1842: cur += (size_t)ssz;
1843: }
1844: if (ssz < 0) {
1845: perror(fname);
1846: free(buf);
1847: return(0);
1848: }
1849:
1850: dofile(args, STDIN_FILENO == fd ?
1851: "STDIN" : fname, tm, buf, cur);
1852: free(buf);
1853: if (STDIN_FILENO != fd)
1854: close(fd);
1855: return(1);
1856: }
1857:
1858: int
1859: main(int argc, char *argv[])
1860: {
1861: const char *fname, *name;
1862: struct args args;
1863: int c;
1864:
1865: name = strrchr(argv[0], '/');
1866: if (name == NULL)
1867: name = argv[0];
1868: else
1869: ++name;
1870:
1871: memset(&args, 0, sizeof(struct args));
1872: fname = "-";
1873:
1874: /* Accept no arguments for now. */
1875:
1876: while (-1 != (c = getopt(argc, argv, "c:d:hln:oq:rs:uv")))
1877: switch (c) {
1878: case ('h'):
1879: /* FALLTHROUGH */
1880: case ('l'):
1881: /* FALLTHROUGH */
1882: case ('c'):
1883: /* FALLTHROUGH */
1884: case ('o'):
1885: /* FALLTHROUGH */
1886: case ('q'):
1887: /* FALLTHROUGH */
1888: case ('r'):
1889: /* FALLTHROUGH */
1890: case ('u'):
1891: /* FALLTHROUGH */
1892: case ('v'):
1893: /* Ignore these. */
1894: break;
1895: case ('d'):
1896: args.date = optarg;
1897: break;
1898: case ('n'):
1899: args.title = optarg;
1900: break;
1901: case ('s'):
1902: args.section = optarg;
1903: break;
1904: default:
1905: goto usage;
1906: }
1907:
1908: argc -= optind;
1909: argv += optind;
1910:
1911: /* Accept only a single input file. */
1912:
1.25 schwarze 1913: if (argc > 1)
1914: goto usage;
1.1 schwarze 1915: else if (1 == argc)
1916: fname = *argv;
1917:
1918: return(readfile(&args, fname) ?
1919: EXIT_SUCCESS : EXIT_FAILURE);
1920:
1921: usage:
1922: fprintf(stderr, "usage: %s [-d date] "
1.25 schwarze 1923: "[-n title] [-s section] [file]\n", name);
1.1 schwarze 1924:
1925: return(EXIT_FAILURE);
1926: }
CVSweb