1 /* $Id: mdoc.c,v 1.223 2014/08/06 15:09:05 schwarze Exp $ */
3 * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010, 2012, 2013, 2014 Ingo Schwarze <schwarze@openbsd.org>
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.
10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR 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.
22 #include <sys/types.h>
34 #include "mandoc_aux.h"
36 #include "libmandoc.h"
38 const char *const __mdoc_macronames[MDOC_MAX + 1] = {
39 "Ap", "Dd", "Dt", "Os",
40 "Sh", "Ss", "Pp", "D1",
41 "Dl", "Bd", "Ed", "Bl",
42 "El", "It", "Ad", "An",
43 "Ar", "Cd", "Cm", "Dv",
44 "Er", "Ev", "Ex", "Fa",
45 "Fd", "Fl", "Fn", "Ft",
46 "Ic", "In", "Li", "Nd",
47 "Nm", "Op", "Ot", "Pa",
48 "Rv", "St", "Va", "Vt",
49 "Xr", "%A", "%B", "%D",
50 "%I", "%J", "%N", "%O",
51 "%P", "%R", "%T", "%V",
52 "Ac", "Ao", "Aq", "At",
53 "Bc", "Bf", "Bo", "Bq",
54 "Bsx", "Bx", "Db", "Dc",
55 "Do", "Dq", "Ec", "Ef",
56 "Em", "Eo", "Fx", "Ms",
57 "No", "Ns", "Nx", "Ox",
58 "Pc", "Pf", "Po", "Pq",
59 "Qc", "Ql", "Qo", "Qq",
60 "Re", "Rs", "Sc", "So",
61 "Sq", "Sm", "Sx", "Sy",
62 "Tn", "Ux", "Xc", "Xo",
63 "Fo", "Fc", "Oo", "Oc",
64 "Bk", "Ek", "Bt", "Hf",
65 "Fr", "Ud", "Lb", "Lp",
66 "Lk", "Mt", "Brq", "Bro",
67 "Brc", "%C", "Es", "En",
68 "Dx", "%Q", "br", "sp",
69 "%U", "Ta", "ll", "text",
72 const char *const __mdoc_argnames[MDOC_ARG_MAX] = {
73 "split", "nosplit", "ragged",
74 "unfilled", "literal", "file",
75 "offset", "bullet", "dash",
76 "hyphen", "item", "enum",
77 "tag", "diag", "hang",
78 "ohang", "inset", "column",
79 "width", "compact", "std",
80 "filled", "words", "emphasis",
81 "symbolic", "nested", "centered"
84 const char * const *mdoc_macronames = __mdoc_macronames;
85 const char * const *mdoc_argnames = __mdoc_argnames;
87 static void mdoc_node_free(struct mdoc_node *);
88 static void mdoc_node_unlink(struct mdoc *,
90 static void mdoc_free1(struct mdoc *);
91 static void mdoc_alloc1(struct mdoc *);
92 static struct mdoc_node *node_alloc(struct mdoc *, int, int,
93 enum mdoct, enum mdoc_type);
94 static int node_append(struct mdoc *,
97 static int mdoc_preptext(struct mdoc *, int, char *, int);
99 static int mdoc_ptext(struct mdoc *, int, char *, int);
100 static int mdoc_pmacro(struct mdoc *, int, char *, int);
103 const struct mdoc_node *
104 mdoc_node(const struct mdoc *mdoc)
110 const struct mdoc_meta *
111 mdoc_meta(const struct mdoc *mdoc)
118 * Frees volatile resources (parse tree, meta-data, fields).
121 mdoc_free1(struct mdoc *mdoc)
125 mdoc_node_delete(mdoc, mdoc->first);
126 free(mdoc->meta.msec);
127 free(mdoc->meta.vol);
128 free(mdoc->meta.arch);
129 free(mdoc->meta.date);
130 free(mdoc->meta.title);
132 free(mdoc->meta.name);
136 * Allocate all volatile resources (parse tree, meta-data, fields).
139 mdoc_alloc1(struct mdoc *mdoc)
142 memset(&mdoc->meta, 0, sizeof(struct mdoc_meta));
144 mdoc->lastnamed = mdoc->lastsec = SEC_NONE;
145 mdoc->last = mandoc_calloc(1, sizeof(struct mdoc_node));
146 mdoc->first = mdoc->last;
147 mdoc->last->type = MDOC_ROOT;
148 mdoc->last->tok = MDOC_MAX;
149 mdoc->next = MDOC_NEXT_CHILD;
153 * Free up volatile resources (see mdoc_free1()) then re-initialises the
154 * data with mdoc_alloc1(). After invocation, parse data has been reset
155 * and the parser is ready for re-invocation on a new tree; however,
156 * cross-parse non-volatile data is kept intact.
159 mdoc_reset(struct mdoc *mdoc)
167 * Completely free up all volatile and non-volatile parse resources.
168 * After invocation, the pointer is no longer usable.
171 mdoc_free(struct mdoc *mdoc)
179 * Allocate volatile and non-volatile parse resources.
182 mdoc_alloc(struct roff *roff, struct mparse *parse,
183 const char *defos, int quick)
187 p = mandoc_calloc(1, sizeof(struct mdoc));
200 mdoc_endparse(struct mdoc *mdoc)
203 return(mdoc_macroend(mdoc));
207 mdoc_addeqn(struct mdoc *mdoc, const struct eqn *ep)
211 n = node_alloc(mdoc, ep->ln, ep->pos, MDOC_MAX, MDOC_EQN);
214 if ( ! node_append(mdoc, n))
217 mdoc->next = MDOC_NEXT_SIBLING;
222 mdoc_addspan(struct mdoc *mdoc, const struct tbl_span *sp)
226 n = node_alloc(mdoc, sp->line, 0, MDOC_MAX, MDOC_TBL);
229 if ( ! node_append(mdoc, n))
232 mdoc->next = MDOC_NEXT_SIBLING;
237 * Main parse routine. Parses a single line -- really just hands off to
238 * the macro (mdoc_pmacro()) or text parser (mdoc_ptext()).
241 mdoc_parseln(struct mdoc *mdoc, int ln, char *buf, int offs)
244 mdoc->flags |= MDOC_NEWLINE;
247 * Let the roff nS register switch SYNOPSIS mode early,
248 * such that the parser knows at all times
249 * whether this mode is on or off.
250 * Note that this mode is also switched by the Sh macro.
252 if (roff_getreg(mdoc->roff, "nS"))
253 mdoc->flags |= MDOC_SYNOPSIS;
255 mdoc->flags &= ~MDOC_SYNOPSIS;
257 return(roff_getcontrol(mdoc->roff, buf, &offs) ?
258 mdoc_pmacro(mdoc, ln, buf, offs) :
259 mdoc_ptext(mdoc, ln, buf, offs));
263 mdoc_macro(MACRO_PROT_ARGS)
265 assert(tok < MDOC_MAX);
267 if (mdoc->flags & MDOC_PBODY) {
268 if (tok == MDOC_Dt) {
269 mandoc_vmsg(MANDOCERR_DT_LATE,
270 mdoc->parse, line, ppos,
271 "Dt %s", buf + *pos);
274 } else if ( ! (mdoc_macros[tok].flags & MDOC_PROLOGUE)) {
275 if (mdoc->meta.title == NULL) {
276 mandoc_vmsg(MANDOCERR_DT_NOTITLE,
277 mdoc->parse, line, ppos, "%s %s",
278 mdoc_macronames[tok], buf + *pos);
279 mdoc->meta.title = mandoc_strdup("UNTITLED");
281 if (NULL == mdoc->meta.vol)
282 mdoc->meta.vol = mandoc_strdup("LOCAL");
283 mdoc->flags |= MDOC_PBODY;
286 return((*mdoc_macros[tok].fp)(mdoc, tok, line, ppos, pos, buf));
291 node_append(struct mdoc *mdoc, struct mdoc_node *p)
296 assert(MDOC_ROOT != p->type);
298 switch (mdoc->next) {
299 case MDOC_NEXT_SIBLING:
300 mdoc->last->next = p;
301 p->prev = mdoc->last;
302 p->parent = mdoc->last->parent;
304 case MDOC_NEXT_CHILD:
305 mdoc->last->child = p;
306 p->parent = mdoc->last;
316 * Copy over the normalised-data pointer of our parent. Not
317 * everybody has one, but copying a null pointer is fine.
322 if (ENDBODY_NOT != p->end)
328 p->norm = p->parent->norm;
334 if ( ! mdoc_valid_pre(mdoc, p))
339 assert(MDOC_BLOCK == p->parent->type);
343 assert(MDOC_BLOCK == p->parent->type);
349 assert(MDOC_BLOCK == p->parent->type);
362 if ( ! mdoc_valid_post(mdoc))
372 static struct mdoc_node *
373 node_alloc(struct mdoc *mdoc, int line, int pos,
374 enum mdoct tok, enum mdoc_type type)
378 p = mandoc_calloc(1, sizeof(struct mdoc_node));
379 p->sec = mdoc->lastsec;
388 if (MDOC_SYNOPSIS & mdoc->flags)
389 p->flags |= MDOC_SYNPRETTY;
391 p->flags &= ~MDOC_SYNPRETTY;
392 if (MDOC_NEWLINE & mdoc->flags)
393 p->flags |= MDOC_LINE;
394 mdoc->flags &= ~MDOC_NEWLINE;
400 mdoc_tail_alloc(struct mdoc *mdoc, int line, int pos, enum mdoct tok)
404 p = node_alloc(mdoc, line, pos, tok, MDOC_TAIL);
405 if ( ! node_append(mdoc, p))
407 mdoc->next = MDOC_NEXT_CHILD;
412 mdoc_head_alloc(struct mdoc *mdoc, int line, int pos, enum mdoct tok)
419 p = node_alloc(mdoc, line, pos, tok, MDOC_HEAD);
420 if ( ! node_append(mdoc, p))
422 mdoc->next = MDOC_NEXT_CHILD;
427 mdoc_body_alloc(struct mdoc *mdoc, int line, int pos, enum mdoct tok)
431 p = node_alloc(mdoc, line, pos, tok, MDOC_BODY);
432 if ( ! node_append(mdoc, p))
434 mdoc->next = MDOC_NEXT_CHILD;
439 mdoc_endbody_alloc(struct mdoc *mdoc, int line, int pos, enum mdoct tok,
440 struct mdoc_node *body, enum mdoc_endbody end)
444 p = node_alloc(mdoc, line, pos, tok, MDOC_BODY);
446 p->norm = body->norm;
448 if ( ! node_append(mdoc, p))
450 mdoc->next = MDOC_NEXT_SIBLING;
455 mdoc_block_alloc(struct mdoc *mdoc, int line, int pos,
456 enum mdoct tok, struct mdoc_arg *args)
460 p = node_alloc(mdoc, line, pos, tok, MDOC_BLOCK);
475 p->norm = mandoc_calloc(1, sizeof(union mdoc_data));
481 if ( ! node_append(mdoc, p))
483 mdoc->next = MDOC_NEXT_CHILD;
488 mdoc_elem_alloc(struct mdoc *mdoc, int line, int pos,
489 enum mdoct tok, struct mdoc_arg *args)
493 p = node_alloc(mdoc, line, pos, tok, MDOC_ELEM);
500 p->norm = mandoc_calloc(1, sizeof(union mdoc_data));
506 if ( ! node_append(mdoc, p))
508 mdoc->next = MDOC_NEXT_CHILD;
513 mdoc_word_alloc(struct mdoc *mdoc, int line, int pos, const char *p)
517 n = node_alloc(mdoc, line, pos, MDOC_MAX, MDOC_TEXT);
518 n->string = roff_strdup(mdoc->roff, p);
520 if ( ! node_append(mdoc, n))
523 mdoc->next = MDOC_NEXT_SIBLING;
528 mdoc_word_append(struct mdoc *mdoc, const char *p)
531 char *addstr, *newstr;
534 addstr = roff_strdup(mdoc->roff, p);
535 mandoc_asprintf(&newstr, "%s %s", n->string, addstr);
539 mdoc->next = MDOC_NEXT_SIBLING;
543 mdoc_node_free(struct mdoc_node *p)
546 if (MDOC_BLOCK == p->type || MDOC_ELEM == p->type)
551 mdoc_argv_free(p->args);
556 mdoc_node_unlink(struct mdoc *mdoc, struct mdoc_node *n)
559 /* Adjust siblings. */
562 n->prev->next = n->next;
564 n->next->prev = n->prev;
570 if (n->parent->child == n)
571 n->parent->child = n->prev ? n->prev : n->next;
572 if (n->parent->last == n)
573 n->parent->last = n->prev ? n->prev : NULL;
576 /* Adjust parse point, if applicable. */
578 if (mdoc && mdoc->last == n) {
580 mdoc->last = n->prev;
581 mdoc->next = MDOC_NEXT_SIBLING;
583 mdoc->last = n->parent;
584 mdoc->next = MDOC_NEXT_CHILD;
588 if (mdoc && mdoc->first == n)
593 mdoc_node_delete(struct mdoc *mdoc, struct mdoc_node *p)
598 mdoc_node_delete(mdoc, p->child);
600 assert(0 == p->nchild);
602 mdoc_node_unlink(mdoc, p);
607 mdoc_node_relink(struct mdoc *mdoc, struct mdoc_node *p)
610 mdoc_node_unlink(mdoc, p);
611 return(node_append(mdoc, p));
616 * Pre-treat a text line.
617 * Text lines can consist of equations, which must be handled apart from
619 * Thus, use this function to step through a line checking if it has any
620 * equations embedded in it.
621 * This must handle multiple equations AND equations that do not end at
622 * the end-of-line, i.e., will re-enter in the next roff parse.
625 mdoc_preptext(struct mdoc *mdoc, int line, char *buf, int offs)
630 while ('\0' != buf[offs]) {
631 /* Mark starting position if eqn is set. */
633 if ('\0' != (delim = roff_eqndelim(mdoc->roff)))
634 if (NULL != (start = strchr(buf + offs, delim)))
637 /* Parse text as normal. */
638 if ( ! mdoc_ptext(mdoc, line, buf, offs))
641 /* Continue only if an equation exists. */
645 /* Read past the end of the equation. */
646 offs += start - (buf + offs);
647 assert(start == &buf[offs]);
648 if (NULL != (end = strchr(buf + offs, delim))) {
654 /* Parse the equation itself. */
655 roff_openeqn(mdoc->roff, NULL, line, offs, buf);
657 /* Process a finished equation? */
658 if (roff_closeeqn(mdoc->roff))
659 if ( ! mdoc_addeqn(mdoc, roff_eqn(mdoc->roff)))
661 offs += (end - (buf + offs));
669 * Parse free-form text, that is, a line that does not begin with the
673 mdoc_ptext(struct mdoc *mdoc, int line, char *buf, int offs)
682 * Divert directly to list processing if we're encountering a
683 * columnar MDOC_BLOCK with or without a prior MDOC_BLOCK entry
684 * (a MDOC_BODY means it's already open, in which case we should
685 * process within its context in the normal way).
688 if (MDOC_Bl == n->tok && MDOC_BODY == n->type &&
689 LIST_column == n->norm->Bl.type) {
690 /* `Bl' is open without any children. */
691 mdoc->flags |= MDOC_FREECOL;
692 return(mdoc_macro(mdoc, MDOC_It, line, offs, &offs, buf));
695 if (MDOC_It == n->tok && MDOC_BLOCK == n->type &&
697 MDOC_Bl == n->parent->tok &&
698 LIST_column == n->parent->norm->Bl.type) {
699 /* `Bl' has block-level `It' children. */
700 mdoc->flags |= MDOC_FREECOL;
701 return(mdoc_macro(mdoc, MDOC_It, line, offs, &offs, buf));
705 * Search for the beginning of unescaped trailing whitespace (ws)
706 * and for the first character not to be output (end).
709 /* FIXME: replace with strcspn(). */
711 for (c = end = buf + offs; *c; c++) {
719 * Always warn about trailing tabs,
720 * even outside literal context,
721 * where they should be put on the next line.
726 * Strip trailing tabs in literal context only;
727 * outside, they affect the next line.
729 if (MDOC_LITERAL & mdoc->flags)
733 /* Skip the escaped character, too, if any. */
746 mandoc_msg(MANDOCERR_SPACE_EOL, mdoc->parse,
747 line, (int)(ws-buf), NULL);
749 if ('\0' == buf[offs] && ! (MDOC_LITERAL & mdoc->flags)) {
750 mandoc_msg(MANDOCERR_FI_BLANK, mdoc->parse,
751 line, (int)(c - buf), NULL);
754 * Insert a `sp' in the case of a blank line. Technically,
755 * blank lines aren't allowed, but enough manuals assume this
756 * behaviour that we want to work around it.
758 if ( ! mdoc_elem_alloc(mdoc, line, offs, MDOC_sp, NULL))
761 mdoc->next = MDOC_NEXT_SIBLING;
763 return(mdoc_valid_post(mdoc));
766 if ( ! mdoc_word_alloc(mdoc, line, offs, buf+offs))
769 if (MDOC_LITERAL & mdoc->flags)
773 * End-of-sentence check. If the last character is an unescaped
774 * EOS character, then flag the node as being the end of a
775 * sentence. The front-end will know how to interpret this.
780 if (mandoc_eos(buf+offs, (size_t)(end-buf-offs)))
781 mdoc->last->flags |= MDOC_EOS;
787 * Parse a macro line, that is, a line beginning with the control
791 mdoc_pmacro(struct mdoc *mdoc, int ln, char *buf, int offs)
798 /* Empty post-control lines are ignored. */
800 if ('"' == buf[offs]) {
801 mandoc_msg(MANDOCERR_COMMENT_BAD, mdoc->parse,
804 } else if ('\0' == buf[offs])
810 * Copy the first word into a nil-terminated buffer.
811 * Stop copying when a tab, space, or eoln is encountered.
815 while (i < 4 && '\0' != buf[offs] && ' ' != buf[offs] &&
817 mac[i++] = buf[offs++];
821 tok = (i > 1 && i < 4) ? mdoc_hash_find(mac) : MDOC_MAX;
823 if (MDOC_MAX == tok) {
824 mandoc_msg(MANDOCERR_MACRO, mdoc->parse,
825 ln, sv, buf + sv - 1);
829 /* Disregard the first trailing tab, if applicable. */
831 if ('\t' == buf[offs])
834 /* Jump to the next non-whitespace word. */
836 while (buf[offs] && ' ' == buf[offs])
840 * Trailing whitespace. Note that tabs are allowed to be passed
841 * into the parser as "text", so we only warn about spaces here.
844 if ('\0' == buf[offs] && ' ' == buf[offs - 1])
845 mandoc_msg(MANDOCERR_SPACE_EOL, mdoc->parse,
849 * If an initial macro or a list invocation, divert directly
850 * into macro processing.
853 if (NULL == mdoc->last || MDOC_It == tok || MDOC_El == tok)
854 return(mdoc_macro(mdoc, tok, ln, sv, &offs, buf));
860 * If the first macro of a `Bl -column', open an `It' block
861 * context around the parsed macro.
864 if (MDOC_Bl == n->tok && MDOC_BODY == n->type &&
865 LIST_column == n->norm->Bl.type) {
866 mdoc->flags |= MDOC_FREECOL;
867 return(mdoc_macro(mdoc, MDOC_It, ln, sv, &sv, buf));
871 * If we're following a block-level `It' within a `Bl -column'
872 * context (perhaps opened in the above block or in ptext()),
873 * then open an `It' block context around the parsed macro.
876 if (MDOC_It == n->tok && MDOC_BLOCK == n->type &&
878 MDOC_Bl == n->parent->tok &&
879 LIST_column == n->parent->norm->Bl.type) {
880 mdoc->flags |= MDOC_FREECOL;
881 return(mdoc_macro(mdoc, MDOC_It, ln, sv, &sv, buf));
884 /* Normal processing of a macro. */
886 if ( ! mdoc_macro(mdoc, tok, ln, sv, &offs, buf))
889 /* In quick mode (for mandocdb), abort after the NAME section. */
891 if (mdoc->quick && MDOC_Sh == tok &&
892 SEC_NAME != mdoc->last->sec)
899 mdoc_isdelim(const char *p)
912 return(DELIM_MIDDLE);
936 if (0 == strcmp(p + 1, "."))
938 if (0 == strcmp(p + 1, "fR|\\fP"))
939 return(DELIM_MIDDLE);
945 mdoc_deroff(char **dest, const struct mdoc_node *n)
950 if (MDOC_TEXT != n->type) {
951 for (n = n->child; n; n = n->next)
952 mdoc_deroff(dest, n);
956 /* Skip leading whitespace. */
958 for (cp = n->string; '\0' != *cp; cp++)
959 if (0 == isspace((unsigned char)*cp))
962 /* Skip trailing whitespace. */
964 for (sz = strlen(cp); sz; sz--)
965 if (0 == isspace((unsigned char)cp[sz-1]))
968 /* Skip empty strings. */
974 *dest = mandoc_strndup(cp, sz);
978 mandoc_asprintf(&cp, "%s %*s", *dest, (int)sz, cp);