1 /* $Id: man.c,v 1.96 2011/01/03 11:31:26 kristaps Exp $ */
3 * Copyright (c) 2008, 2009, 2010 Kristaps Dzonsons <kristaps@bsd.lv>
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.
9 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
10 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
11 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
12 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
13 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
14 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
15 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
21 #include <sys/types.h>
31 #include "libmandoc.h"
33 const char *const __man_macronames[MAN_MAX] = {
34 "br", "TH", "SH", "SS",
35 "TP", "LP", "PP", "P",
36 "IP", "HP", "SM", "SB",
37 "BI", "IB", "BR", "RB",
39 "RI", "na", "sp", "nf",
40 "fi", "RE", "RS", "DT",
41 "UC", "PD", "AT", "in",
45 const char * const *man_macronames = __man_macronames;
47 static struct man_node *man_node_alloc(int, int,
48 enum man_type, enum mant);
49 static int man_node_append(struct man *,
51 static int man_span_alloc(struct man *,
52 const struct tbl_span *);
53 static void man_node_free(struct man_node *);
54 static void man_node_unlink(struct man *,
56 static int man_ptext(struct man *, int, char *, int);
57 static int man_pmacro(struct man *, int, char *, int);
58 static void man_free1(struct man *);
59 static void man_alloc1(struct man *);
60 static int man_descope(struct man *, int, int);
63 const struct man_node *
64 man_node(const struct man *m)
67 assert( ! (MAN_HALT & m->flags));
72 const struct man_meta *
73 man_meta(const struct man *m)
76 assert( ! (MAN_HALT & m->flags));
82 man_reset(struct man *man)
91 man_free(struct man *man)
100 man_alloc(struct regset *regs, void *data, mandocmsg msg)
104 p = mandoc_calloc(1, sizeof(struct man));
117 man_endparse(struct man *m)
120 assert( ! (MAN_HALT & m->flags));
123 m->flags |= MAN_HALT;
129 man_parseln(struct man *m, int ln, char *buf, int offs)
132 assert( ! (MAN_HALT & m->flags));
133 return(('.' == buf[offs] || '\'' == buf[offs]) ?
134 man_pmacro(m, ln, buf, offs) :
135 man_ptext(m, ln, buf, offs));
140 man_free1(struct man *man)
144 man_node_delete(man, man->first);
146 free(man->meta.title);
147 if (man->meta.source)
148 free(man->meta.source);
149 if (man->meta.rawdate)
150 free(man->meta.rawdate);
154 free(man->meta.msec);
159 man_alloc1(struct man *m)
162 memset(&m->meta, 0, sizeof(struct man_meta));
164 m->last = mandoc_calloc(1, sizeof(struct man_node));
166 m->last->type = MAN_ROOT;
167 m->last->tok = MAN_MAX;
168 m->next = MAN_NEXT_CHILD;
173 man_node_append(struct man *man, struct man_node *p)
178 assert(MAN_ROOT != p->type);
181 case (MAN_NEXT_SIBLING):
184 p->parent = man->last->parent;
186 case (MAN_NEXT_CHILD):
187 man->last->child = p;
188 p->parent = man->last;
198 if ( ! man_valid_pre(man, p))
203 assert(MAN_BLOCK == p->parent->type);
207 assert(MAN_BLOCK == p->parent->type);
220 if ( ! man_valid_post(man))
231 static struct man_node *
232 man_node_alloc(int line, int pos, enum man_type type, enum mant tok)
236 p = mandoc_calloc(1, sizeof(struct man_node));
246 man_elem_alloc(struct man *m, int line, int pos, enum mant tok)
250 p = man_node_alloc(line, pos, MAN_ELEM, tok);
251 if ( ! man_node_append(m, p))
253 m->next = MAN_NEXT_CHILD;
259 man_head_alloc(struct man *m, int line, int pos, enum mant tok)
263 p = man_node_alloc(line, pos, MAN_HEAD, tok);
264 if ( ! man_node_append(m, p))
266 m->next = MAN_NEXT_CHILD;
272 man_body_alloc(struct man *m, int line, int pos, enum mant tok)
276 p = man_node_alloc(line, pos, MAN_BODY, tok);
277 if ( ! man_node_append(m, p))
279 m->next = MAN_NEXT_CHILD;
285 man_block_alloc(struct man *m, int line, int pos, enum mant tok)
289 p = man_node_alloc(line, pos, MAN_BLOCK, tok);
290 if ( ! man_node_append(m, p))
292 m->next = MAN_NEXT_CHILD;
297 man_span_alloc(struct man *m, const struct tbl_span *span)
301 /* FIXME: grab from span */
302 n = man_node_alloc(0, 0, MAN_TBL, MAN_MAX);
305 if ( ! man_node_append(m, n))
308 m->next = MAN_NEXT_SIBLING;
313 man_word_alloc(struct man *m, int line, int pos, const char *word)
320 n = man_node_alloc(line, pos, MAN_TEXT, MAN_MAX);
321 n->string = mandoc_malloc(len + 1);
322 sv = strlcpy(n->string, word, len + 1);
324 /* Prohibit truncation. */
325 assert(sv < len + 1);
327 if ( ! man_node_append(m, n))
330 m->next = MAN_NEXT_SIBLING;
336 * Free all of the resources held by a node. This does NOT unlink a
337 * node from its context; for that, see man_node_unlink().
340 man_node_free(struct man_node *p)
350 man_node_delete(struct man *m, struct man_node *p)
354 man_node_delete(m, p->child);
356 man_node_unlink(m, p);
362 man_addspan(struct man *m, const struct tbl_span *sp)
365 assert( ! (MAN_HALT & m->flags));
366 if ( ! man_span_alloc(m, sp))
368 return(man_descope(m, 0, 0));
372 man_descope(struct man *m, int line, int offs)
375 * Co-ordinate what happens with having a next-line scope open:
376 * first close out the element scope (if applicable), then close
377 * out the block scope (also if applicable).
380 if (MAN_ELINE & m->flags) {
381 m->flags &= ~MAN_ELINE;
382 if ( ! man_unscope(m, m->last->parent, MANDOCERR_MAX))
386 if ( ! (MAN_BLINE & m->flags))
388 m->flags &= ~MAN_BLINE;
390 if ( ! man_unscope(m, m->last->parent, MANDOCERR_MAX))
392 return(man_body_alloc(m, line, offs, m->last->tok));
397 man_ptext(struct man *m, int line, char *buf, int offs)
401 /* Ignore bogus comments. */
403 if ('\\' == buf[offs] &&
404 '.' == buf[offs + 1] &&
405 '"' == buf[offs + 2]) {
406 man_pmsg(m, line, offs, MANDOCERR_BADCOMMENT);
410 /* Literal free-form text whitespace is preserved. */
412 if (MAN_LITERAL & m->flags) {
413 if ( ! man_word_alloc(m, line, offs, buf + offs))
415 return(man_descope(m, line, offs));
418 /* Pump blank lines directly into the backend. */
420 for (i = offs; ' ' == buf[i]; i++)
421 /* Skip leading whitespace. */ ;
423 if ('\0' == buf[i]) {
424 /* Allocate a blank entry. */
425 if ( ! man_word_alloc(m, line, offs, ""))
427 return(man_descope(m, line, offs));
431 * Warn if the last un-escaped character is whitespace. Then
432 * strip away the remaining spaces (tabs stay!).
435 i = (int)strlen(buf);
438 if (' ' == buf[i - 1] || '\t' == buf[i - 1]) {
439 if (i > 1 && '\\' != buf[i - 2])
440 man_pmsg(m, line, i - 1, MANDOCERR_EOLNSPACE);
442 for (--i; i && ' ' == buf[i]; i--)
443 /* Spin back to non-space. */ ;
445 /* Jump ahead of escaped whitespace. */
446 i += '\\' == buf[i] ? 2 : 1;
451 if ( ! man_word_alloc(m, line, offs, buf + offs))
455 * End-of-sentence check. If the last character is an unescaped
456 * EOS character, then flag the node as being the end of a
457 * sentence. The front-end will know how to interpret this.
461 if (mandoc_eos(buf, (size_t)i, 0))
462 m->last->flags |= MAN_EOS;
464 return(man_descope(m, line, offs));
469 man_pmacro(struct man *m, int ln, char *buf, int offs)
476 /* Comments and empties are quickly ignored. */
480 if ('\0' == buf[offs])
486 * Skip whitespace between the control character and initial
487 * text. "Whitespace" is both spaces and tabs.
490 if (' ' == buf[i] || '\t' == buf[i]) {
492 while (buf[i] && (' ' == buf[i] || '\t' == buf[i]))
501 * Copy the first word into a nil-terminated buffer.
502 * Stop copying when a tab, space, or eoln is encountered.
506 while (j < 4 && '\0' != buf[i] && ' ' != buf[i] && '\t' != buf[i])
510 tok = (j > 0 && j < 4) ? man_hash_find(mac) : MAN_MAX;
511 if (MAN_MAX == tok) {
512 man_vmsg(m, MANDOCERR_MACRO, ln, ppos, "%s", buf + ppos - 1);
516 /* The macro is sane. Jump to the next word. */
518 while (buf[i] && ' ' == buf[i])
522 * Trailing whitespace. Note that tabs are allowed to be passed
523 * into the parser as "text", so we only warn about spaces here.
526 if ('\0' == buf[i] && ' ' == buf[i - 1])
527 man_pmsg(m, ln, i - 1, MANDOCERR_EOLNSPACE);
530 * Remove prior ELINE macro, as it's being clobbered by a new
531 * macro. Note that NSCOPED macros do not close out ELINE
532 * macros---they don't print text---so we let those slip by.
535 if ( ! (MAN_NSCOPED & man_macros[tok].flags) &&
536 m->flags & MAN_ELINE) {
538 assert(MAN_TEXT != n->type);
540 /* Remove repeated NSCOPED macros causing ELINE. */
542 if (MAN_NSCOPED & man_macros[n->tok].flags)
545 man_vmsg(m, MANDOCERR_LINESCOPE, n->line, n->pos,
546 "%s", man_macronames[n->tok]);
548 man_node_delete(m, n);
549 m->flags &= ~MAN_ELINE;
553 * Save the fact that we're in the next-line for a block. In
554 * this way, embedded roff instructions can "remember" state
558 if (MAN_BLINE & m->flags)
559 m->flags |= MAN_BPLINE;
561 /* Call to handler... */
563 assert(man_macros[tok].fp);
564 if ( ! (*man_macros[tok].fp)(m, tok, ln, ppos, &i, buf))
569 * We weren't in a block-line scope when entering the
570 * above-parsed macro, so return.
573 if ( ! (MAN_BPLINE & m->flags)) {
574 m->flags &= ~MAN_ILINE;
577 m->flags &= ~MAN_BPLINE;
580 * If we're in a block scope, then allow this macro to slip by
581 * without closing scope around it.
584 if (MAN_ILINE & m->flags) {
585 m->flags &= ~MAN_ILINE;
590 * If we've opened a new next-line element scope, then return
591 * now, as the next line will close out the block scope.
594 if (MAN_ELINE & m->flags)
597 /* Close out the block scope opened in the prior line. */
599 assert(MAN_BLINE & m->flags);
600 m->flags &= ~MAN_BLINE;
602 if ( ! man_unscope(m, m->last->parent, MANDOCERR_MAX))
604 return(man_body_alloc(m, ln, offs, m->last->tok));
606 err: /* Error out. */
608 m->flags |= MAN_HALT;
614 man_vmsg(struct man *man, enum mandocerr t,
615 int ln, int pos, const char *fmt, ...)
621 vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
623 return((*man->msg)(t, man->data, ln, pos, buf));
628 * Unlink a node from its context. If "m" is provided, the last parse
629 * point will also be adjusted accordingly.
632 man_node_unlink(struct man *m, struct man_node *n)
635 /* Adjust siblings. */
638 n->prev->next = n->next;
640 n->next->prev = n->prev;
646 if (n->parent->child == n)
647 n->parent->child = n->prev ? n->prev : n->next;
650 /* Adjust parse point, if applicable. */
652 if (m && m->last == n) {
653 /*XXX: this can occur when bailing from validation. */
654 /*assert(NULL == n->next);*/
657 m->next = MAN_NEXT_SIBLING;
660 m->next = MAN_NEXT_CHILD;
664 if (m && m->first == n)