1 /* $Id: mdoc.c,v 1.196 2011/09/30 00:13:28 schwarze Exp $ */
3 * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010 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 "libmandoc.h"
36 const char *const __mdoc_macronames[MDOC_MAX] = {
37 "Ap", "Dd", "Dt", "Os",
38 "Sh", "Ss", "Pp", "D1",
39 "Dl", "Bd", "Ed", "Bl",
40 "El", "It", "Ad", "An",
41 "Ar", "Cd", "Cm", "Dv",
42 "Er", "Ev", "Ex", "Fa",
43 "Fd", "Fl", "Fn", "Ft",
44 "Ic", "In", "Li", "Nd",
45 "Nm", "Op", "Ot", "Pa",
46 "Rv", "St", "Va", "Vt",
48 "Xr", "%A", "%B", "%D",
50 "%I", "%J", "%N", "%O",
52 "%P", "%R", "%T", "%V",
53 "Ac", "Ao", "Aq", "At",
54 "Bc", "Bf", "Bo", "Bq",
55 "Bsx", "Bx", "Db", "Dc",
56 "Do", "Dq", "Ec", "Ef",
57 "Em", "Eo", "Fx", "Ms",
58 "No", "Ns", "Nx", "Ox",
59 "Pc", "Pf", "Po", "Pq",
60 "Qc", "Ql", "Qo", "Qq",
61 "Re", "Rs", "Sc", "So",
62 "Sq", "Sm", "Sx", "Sy",
63 "Tn", "Ux", "Xc", "Xo",
64 "Fo", "Fc", "Oo", "Oc",
65 "Bk", "Ek", "Bt", "Hf",
66 "Fr", "Ud", "Lb", "Lp",
67 "Lk", "Mt", "Brq", "Bro",
69 "Brc", "%C", "Es", "En",
71 "Dx", "%Q", "br", "sp",
76 const char *const __mdoc_argnames[MDOC_ARG_MAX] = {
77 "split", "nosplit", "ragged",
78 "unfilled", "literal", "file",
79 "offset", "bullet", "dash",
80 "hyphen", "item", "enum",
81 "tag", "diag", "hang",
82 "ohang", "inset", "column",
83 "width", "compact", "std",
84 "filled", "words", "emphasis",
85 "symbolic", "nested", "centered"
88 const char * const *mdoc_macronames = __mdoc_macronames;
89 const char * const *mdoc_argnames = __mdoc_argnames;
91 static void mdoc_node_free(struct mdoc_node *);
92 static void mdoc_node_unlink(struct mdoc *,
94 static void mdoc_free1(struct mdoc *);
95 static void mdoc_alloc1(struct mdoc *);
96 static struct mdoc_node *node_alloc(struct mdoc *, int, int,
97 enum mdoct, enum mdoc_type);
98 static int node_append(struct mdoc *,
101 static int mdoc_preptext(struct mdoc *, int, char *, int);
103 static int mdoc_ptext(struct mdoc *, int, char *, int);
104 static int mdoc_pmacro(struct mdoc *, int, char *, int);
106 const struct mdoc_node *
107 mdoc_node(const struct mdoc *m)
110 assert( ! (MDOC_HALT & m->flags));
115 const struct mdoc_meta *
116 mdoc_meta(const struct mdoc *m)
119 assert( ! (MDOC_HALT & m->flags));
125 * Frees volatile resources (parse tree, meta-data, fields).
128 mdoc_free1(struct mdoc *mdoc)
132 mdoc_node_delete(mdoc, mdoc->first);
133 if (mdoc->meta.title)
134 free(mdoc->meta.title);
138 free(mdoc->meta.name);
140 free(mdoc->meta.arch);
142 free(mdoc->meta.vol);
144 free(mdoc->meta.msec);
146 free(mdoc->meta.date);
151 * Allocate all volatile resources (parse tree, meta-data, fields).
154 mdoc_alloc1(struct mdoc *mdoc)
157 memset(&mdoc->meta, 0, sizeof(struct mdoc_meta));
159 mdoc->lastnamed = mdoc->lastsec = SEC_NONE;
160 mdoc->last = mandoc_calloc(1, sizeof(struct mdoc_node));
161 mdoc->first = mdoc->last;
162 mdoc->last->type = MDOC_ROOT;
163 mdoc->last->tok = MDOC_MAX;
164 mdoc->next = MDOC_NEXT_CHILD;
169 * Free up volatile resources (see mdoc_free1()) then re-initialises the
170 * data with mdoc_alloc1(). After invocation, parse data has been reset
171 * and the parser is ready for re-invocation on a new tree; however,
172 * cross-parse non-volatile data is kept intact.
175 mdoc_reset(struct mdoc *mdoc)
184 * Completely free up all volatile and non-volatile parse resources.
185 * After invocation, the pointer is no longer usable.
188 mdoc_free(struct mdoc *mdoc)
197 * Allocate volatile and non-volatile parse resources.
200 mdoc_alloc(struct roff *roff, struct mparse *parse)
204 p = mandoc_calloc(1, sizeof(struct mdoc));
216 * Climb back up the parse tree, validating open scopes. Mostly calls
217 * through to macro_end() in macro.c.
220 mdoc_endparse(struct mdoc *m)
223 assert( ! (MDOC_HALT & m->flags));
224 if (mdoc_macroend(m))
226 m->flags |= MDOC_HALT;
231 mdoc_addeqn(struct mdoc *m, const struct eqn *ep)
235 assert( ! (MDOC_HALT & m->flags));
237 /* No text before an initial macro. */
239 if (SEC_NONE == m->lastnamed) {
240 mdoc_pmsg(m, ep->ln, ep->pos, MANDOCERR_NOTEXT);
244 n = node_alloc(m, ep->ln, ep->pos, MDOC_MAX, MDOC_EQN);
247 if ( ! node_append(m, n))
250 m->next = MDOC_NEXT_SIBLING;
255 mdoc_addspan(struct mdoc *m, const struct tbl_span *sp)
259 assert( ! (MDOC_HALT & m->flags));
261 /* No text before an initial macro. */
263 if (SEC_NONE == m->lastnamed) {
264 mdoc_pmsg(m, sp->line, 0, MANDOCERR_NOTEXT);
268 n = node_alloc(m, sp->line, 0, MDOC_MAX, MDOC_TBL);
271 if ( ! node_append(m, n))
274 m->next = MDOC_NEXT_SIBLING;
280 * Main parse routine. Parses a single line -- really just hands off to
281 * the macro (mdoc_pmacro()) or text parser (mdoc_ptext()).
284 mdoc_parseln(struct mdoc *m, int ln, char *buf, int offs)
287 assert( ! (MDOC_HALT & m->flags));
289 m->flags |= MDOC_NEWLINE;
292 * Let the roff nS register switch SYNOPSIS mode early,
293 * such that the parser knows at all times
294 * whether this mode is on or off.
295 * Note that this mode is also switched by the Sh macro.
297 if (roff_regisset(m->roff, REG_nS)) {
298 if (roff_regget(m->roff, REG_nS))
299 m->flags |= MDOC_SYNOPSIS;
301 m->flags &= ~MDOC_SYNOPSIS;
304 return(mandoc_getcontrol(buf, &offs) ?
305 mdoc_pmacro(m, ln, buf, offs) :
306 mdoc_ptext(m, ln, buf, offs));
310 mdoc_macro(MACRO_PROT_ARGS)
312 assert(tok < MDOC_MAX);
314 /* If we're in the body, deny prologue calls. */
316 if (MDOC_PROLOGUE & mdoc_macros[tok].flags &&
317 MDOC_PBODY & m->flags) {
318 mdoc_pmsg(m, line, ppos, MANDOCERR_BADBODY);
322 /* If we're in the prologue, deny "body" macros. */
324 if ( ! (MDOC_PROLOGUE & mdoc_macros[tok].flags) &&
325 ! (MDOC_PBODY & m->flags)) {
326 mdoc_pmsg(m, line, ppos, MANDOCERR_BADPROLOG);
327 if (NULL == m->meta.msec)
328 m->meta.msec = mandoc_strdup("1");
329 if (NULL == m->meta.title)
330 m->meta.title = mandoc_strdup("UNKNOWN");
331 if (NULL == m->meta.vol)
332 m->meta.vol = mandoc_strdup("LOCAL");
333 if (NULL == m->meta.os)
334 m->meta.os = mandoc_strdup("LOCAL");
335 if (NULL == m->meta.date)
336 m->meta.date = mandoc_normdate
337 (m->parse, NULL, line, ppos);
338 m->flags |= MDOC_PBODY;
341 return((*mdoc_macros[tok].fp)(m, tok, line, ppos, pos, buf));
346 node_append(struct mdoc *mdoc, struct mdoc_node *p)
351 assert(MDOC_ROOT != p->type);
353 switch (mdoc->next) {
354 case (MDOC_NEXT_SIBLING):
355 mdoc->last->next = p;
356 p->prev = mdoc->last;
357 p->parent = mdoc->last->parent;
359 case (MDOC_NEXT_CHILD):
360 mdoc->last->child = p;
361 p->parent = mdoc->last;
371 * Copy over the normalised-data pointer of our parent. Not
372 * everybody has one, but copying a null pointer is fine.
381 p->norm = p->parent->norm;
387 if ( ! mdoc_valid_pre(mdoc, p))
392 assert(MDOC_BLOCK == p->parent->type);
396 assert(MDOC_BLOCK == p->parent->type);
402 assert(MDOC_BLOCK == p->parent->type);
415 if ( ! mdoc_valid_post(mdoc))
426 static struct mdoc_node *
427 node_alloc(struct mdoc *m, int line, int pos,
428 enum mdoct tok, enum mdoc_type type)
432 p = mandoc_calloc(1, sizeof(struct mdoc_node));
441 if (MDOC_SYNOPSIS & m->flags)
442 p->flags |= MDOC_SYNPRETTY;
444 p->flags &= ~MDOC_SYNPRETTY;
445 if (MDOC_NEWLINE & m->flags)
446 p->flags |= MDOC_LINE;
447 m->flags &= ~MDOC_NEWLINE;
454 mdoc_tail_alloc(struct mdoc *m, int line, int pos, enum mdoct tok)
458 p = node_alloc(m, line, pos, tok, MDOC_TAIL);
459 if ( ! node_append(m, p))
461 m->next = MDOC_NEXT_CHILD;
467 mdoc_head_alloc(struct mdoc *m, int line, int pos, enum mdoct tok)
474 p = node_alloc(m, line, pos, tok, MDOC_HEAD);
475 if ( ! node_append(m, p))
477 m->next = MDOC_NEXT_CHILD;
483 mdoc_body_alloc(struct mdoc *m, int line, int pos, enum mdoct tok)
487 p = node_alloc(m, line, pos, tok, MDOC_BODY);
488 if ( ! node_append(m, p))
490 m->next = MDOC_NEXT_CHILD;
496 mdoc_endbody_alloc(struct mdoc *m, int line, int pos, enum mdoct tok,
497 struct mdoc_node *body, enum mdoc_endbody end)
501 p = node_alloc(m, line, pos, tok, MDOC_BODY);
504 if ( ! node_append(m, p))
506 m->next = MDOC_NEXT_SIBLING;
512 mdoc_block_alloc(struct mdoc *m, int line, int pos,
513 enum mdoct tok, struct mdoc_arg *args)
517 p = node_alloc(m, line, pos, tok, MDOC_BLOCK);
530 p->norm = mandoc_calloc(1, sizeof(union mdoc_data));
536 if ( ! node_append(m, p))
538 m->next = MDOC_NEXT_CHILD;
544 mdoc_elem_alloc(struct mdoc *m, int line, int pos,
545 enum mdoct tok, struct mdoc_arg *args)
549 p = node_alloc(m, line, pos, tok, MDOC_ELEM);
556 p->norm = mandoc_calloc(1, sizeof(union mdoc_data));
562 if ( ! node_append(m, p))
564 m->next = MDOC_NEXT_CHILD;
569 mdoc_word_alloc(struct mdoc *m, int line, int pos, const char *p)
573 n = node_alloc(m, line, pos, MDOC_MAX, MDOC_TEXT);
574 n->string = roff_strdup(m->roff, p);
576 if ( ! node_append(m, n))
579 m->next = MDOC_NEXT_SIBLING;
585 mdoc_node_free(struct mdoc_node *p)
588 if (MDOC_BLOCK == p->type || MDOC_ELEM == p->type)
593 mdoc_argv_free(p->args);
599 mdoc_node_unlink(struct mdoc *m, struct mdoc_node *n)
602 /* Adjust siblings. */
605 n->prev->next = n->next;
607 n->next->prev = n->prev;
613 if (n->parent->child == n)
614 n->parent->child = n->prev ? n->prev : n->next;
615 if (n->parent->last == n)
616 n->parent->last = n->prev ? n->prev : NULL;
619 /* Adjust parse point, if applicable. */
621 if (m && m->last == n) {
624 m->next = MDOC_NEXT_SIBLING;
627 m->next = MDOC_NEXT_CHILD;
631 if (m && m->first == n)
637 mdoc_node_delete(struct mdoc *m, struct mdoc_node *p)
642 mdoc_node_delete(m, p->child);
644 assert(0 == p->nchild);
646 mdoc_node_unlink(m, p);
652 * Pre-treat a text line.
653 * Text lines can consist of equations, which must be handled apart from
655 * Thus, use this function to step through a line checking if it has any
656 * equations embedded in it.
657 * This must handle multiple equations AND equations that do not end at
658 * the end-of-line, i.e., will re-enter in the next roff parse.
661 mdoc_preptext(struct mdoc *m, int line, char *buf, int offs)
666 while ('\0' != buf[offs]) {
667 /* Mark starting position if eqn is set. */
669 if ('\0' != (delim = roff_eqndelim(m->roff)))
670 if (NULL != (start = strchr(buf + offs, delim)))
673 /* Parse text as normal. */
674 if ( ! mdoc_ptext(m, line, buf, offs))
677 /* Continue only if an equation exists. */
681 /* Read past the end of the equation. */
682 offs += start - (buf + offs);
683 assert(start == &buf[offs]);
684 if (NULL != (end = strchr(buf + offs, delim))) {
690 /* Parse the equation itself. */
691 roff_openeqn(m->roff, NULL, line, offs, buf);
693 /* Process a finished equation? */
694 if (roff_closeeqn(m->roff))
695 if ( ! mdoc_addeqn(m, roff_eqn(m->roff)))
697 offs += (end - (buf + offs));
705 * Parse free-form text, that is, a line that does not begin with the
709 mdoc_ptext(struct mdoc *m, int line, char *buf, int offs)
714 /* No text before an initial macro. */
716 if (SEC_NONE == m->lastnamed) {
717 mdoc_pmsg(m, line, offs, MANDOCERR_NOTEXT);
725 * Divert directly to list processing if we're encountering a
726 * columnar MDOC_BLOCK with or without a prior MDOC_BLOCK entry
727 * (a MDOC_BODY means it's already open, in which case we should
728 * process within its context in the normal way).
731 if (MDOC_Bl == n->tok && MDOC_BODY == n->type &&
732 LIST_column == n->norm->Bl.type) {
733 /* `Bl' is open without any children. */
734 m->flags |= MDOC_FREECOL;
735 return(mdoc_macro(m, MDOC_It, line, offs, &offs, buf));
738 if (MDOC_It == n->tok && MDOC_BLOCK == n->type &&
740 MDOC_Bl == n->parent->tok &&
741 LIST_column == n->parent->norm->Bl.type) {
742 /* `Bl' has block-level `It' children. */
743 m->flags |= MDOC_FREECOL;
744 return(mdoc_macro(m, MDOC_It, line, offs, &offs, buf));
748 * Search for the beginning of unescaped trailing whitespace (ws)
749 * and for the first character not to be output (end).
752 /* FIXME: replace with strcspn(). */
754 for (c = end = buf + offs; *c; c++) {
762 * Always warn about trailing tabs,
763 * even outside literal context,
764 * where they should be put on the next line.
769 * Strip trailing tabs in literal context only;
770 * outside, they affect the next line.
772 if (MDOC_LITERAL & m->flags)
776 /* Skip the escaped character, too, if any. */
789 mdoc_pmsg(m, line, (int)(ws-buf), MANDOCERR_EOLNSPACE);
791 if ('\0' == buf[offs] && ! (MDOC_LITERAL & m->flags)) {
792 mdoc_pmsg(m, line, (int)(c-buf), MANDOCERR_NOBLANKLN);
795 * Insert a `sp' in the case of a blank line. Technically,
796 * blank lines aren't allowed, but enough manuals assume this
797 * behaviour that we want to work around it.
799 if ( ! mdoc_elem_alloc(m, line, offs, MDOC_sp, NULL))
802 m->next = MDOC_NEXT_SIBLING;
806 if ( ! mdoc_word_alloc(m, line, offs, buf+offs))
809 if (MDOC_LITERAL & m->flags)
813 * End-of-sentence check. If the last character is an unescaped
814 * EOS character, then flag the node as being the end of a
815 * sentence. The front-end will know how to interpret this.
820 if (mandoc_eos(buf+offs, (size_t)(end-buf-offs), 0))
821 m->last->flags |= MDOC_EOS;
828 * Parse a macro line, that is, a line beginning with the control
832 mdoc_pmacro(struct mdoc *m, int ln, char *buf, int offs)
839 /* Empty post-control lines are ignored. */
841 if ('"' == buf[offs]) {
842 mdoc_pmsg(m, ln, offs, MANDOCERR_BADCOMMENT);
844 } else if ('\0' == buf[offs])
850 * Copy the first word into a nil-terminated buffer.
851 * Stop copying when a tab, space, or eoln is encountered.
855 while (i < 4 && '\0' != buf[offs] &&
856 ' ' != buf[offs] && '\t' != buf[offs])
857 mac[i++] = buf[offs++];
861 tok = (i > 1 || i < 4) ? mdoc_hash_find(mac) : MDOC_MAX;
863 if (MDOC_MAX == tok) {
864 mandoc_vmsg(MANDOCERR_MACRO, m->parse,
865 ln, sv, "%s", buf + sv - 1);
869 /* Disregard the first trailing tab, if applicable. */
871 if ('\t' == buf[offs])
874 /* Jump to the next non-whitespace word. */
876 while (buf[offs] && ' ' == buf[offs])
880 * Trailing whitespace. Note that tabs are allowed to be passed
881 * into the parser as "text", so we only warn about spaces here.
884 if ('\0' == buf[offs] && ' ' == buf[offs - 1])
885 mdoc_pmsg(m, ln, offs - 1, MANDOCERR_EOLNSPACE);
888 * If an initial macro or a list invocation, divert directly
889 * into macro processing.
892 if (NULL == m->last || MDOC_It == tok || MDOC_El == tok) {
893 if ( ! mdoc_macro(m, tok, ln, sv, &offs, buf))
902 * If the first macro of a `Bl -column', open an `It' block
903 * context around the parsed macro.
906 if (MDOC_Bl == n->tok && MDOC_BODY == n->type &&
907 LIST_column == n->norm->Bl.type) {
908 m->flags |= MDOC_FREECOL;
909 if ( ! mdoc_macro(m, MDOC_It, ln, sv, &sv, buf))
915 * If we're following a block-level `It' within a `Bl -column'
916 * context (perhaps opened in the above block or in ptext()),
917 * then open an `It' block context around the parsed macro.
920 if (MDOC_It == n->tok && MDOC_BLOCK == n->type &&
922 MDOC_Bl == n->parent->tok &&
923 LIST_column == n->parent->norm->Bl.type) {
924 m->flags |= MDOC_FREECOL;
925 if ( ! mdoc_macro(m, MDOC_It, ln, sv, &sv, buf))
930 /* Normal processing of a macro. */
932 if ( ! mdoc_macro(m, tok, ln, sv, &offs, buf))
937 err: /* Error out. */
939 m->flags |= MDOC_HALT;
944 mdoc_isdelim(const char *p)
957 return(DELIM_MIDDLE);
981 if (0 == strcmp(p + 1, "."))
983 if (0 == strcmp(p + 1, "*(Ba"))
984 return(DELIM_MIDDLE);