1 /* $Id: mdoc_macro.c,v 1.109 2011/04/30 10:18:24 kristaps Exp $ */
3 * Copyright (c) 2008, 2009, 2010 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.
32 #include "libmandoc.h"
34 enum rew { /* see rew_dohalt() */
43 static int blk_full(MACRO_PROT_ARGS);
44 static int blk_exp_close(MACRO_PROT_ARGS);
45 static int blk_part_exp(MACRO_PROT_ARGS);
46 static int blk_part_imp(MACRO_PROT_ARGS);
47 static int ctx_synopsis(MACRO_PROT_ARGS);
48 static int in_line_eoln(MACRO_PROT_ARGS);
49 static int in_line_argn(MACRO_PROT_ARGS);
50 static int in_line(MACRO_PROT_ARGS);
51 static int obsolete(MACRO_PROT_ARGS);
52 static int phrase_ta(MACRO_PROT_ARGS);
54 static int dword(struct mdoc *, int, int,
55 const char *, enum mdelim);
56 static int append_delims(struct mdoc *,
58 static enum mdoct lookup(enum mdoct, const char *);
59 static enum mdoct lookup_raw(const char *);
60 static int make_pending(struct mdoc_node *, enum mdoct,
61 struct mdoc *, int, int);
62 static int phrase(struct mdoc *, int, int, char *);
63 static enum mdoct rew_alt(enum mdoct);
64 static enum rew rew_dohalt(enum mdoct, enum mdoc_type,
65 const struct mdoc_node *);
66 static int rew_elem(struct mdoc *, enum mdoct);
67 static int rew_last(struct mdoc *,
68 const struct mdoc_node *);
69 static int rew_sub(enum mdoc_type, struct mdoc *,
70 enum mdoct, int, int);
72 const struct mdoc_macro __mdoc_macros[MDOC_MAX] = {
73 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ap */
74 { in_line_eoln, MDOC_PROLOGUE }, /* Dd */
75 { in_line_eoln, MDOC_PROLOGUE }, /* Dt */
76 { in_line_eoln, MDOC_PROLOGUE }, /* Os */
77 { blk_full, 0 }, /* Sh */
78 { blk_full, 0 }, /* Ss */
79 { in_line_eoln, 0 }, /* Pp */
80 { blk_part_imp, MDOC_PARSED }, /* D1 */
81 { blk_part_imp, MDOC_PARSED }, /* Dl */
82 { blk_full, MDOC_EXPLICIT }, /* Bd */
83 { blk_exp_close, MDOC_EXPLICIT }, /* Ed */
84 { blk_full, MDOC_EXPLICIT }, /* Bl */
85 { blk_exp_close, MDOC_EXPLICIT }, /* El */
86 { blk_full, MDOC_PARSED }, /* It */
87 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ad */
88 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* An */
89 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ar */
90 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Cd */
91 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Cm */
92 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Dv */
93 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Er */
94 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ev */
95 { in_line_eoln, 0 }, /* Ex */
96 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fa */
97 { in_line_eoln, 0 }, /* Fd */
98 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fl */
99 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fn */
100 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ft */
101 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ic */
102 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* In */
103 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Li */
104 { blk_full, 0 }, /* Nd */
105 { ctx_synopsis, MDOC_CALLABLE | MDOC_PARSED }, /* Nm */
106 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Op */
107 { obsolete, 0 }, /* Ot */
108 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Pa */
109 { in_line_eoln, 0 }, /* Rv */
110 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* St */
111 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Va */
112 { ctx_synopsis, MDOC_CALLABLE | MDOC_PARSED }, /* Vt */
113 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Xr */
114 { in_line_eoln, 0 }, /* %A */
115 { in_line_eoln, 0 }, /* %B */
116 { in_line_eoln, 0 }, /* %D */
117 { in_line_eoln, 0 }, /* %I */
118 { in_line_eoln, 0 }, /* %J */
119 { in_line_eoln, 0 }, /* %N */
120 { in_line_eoln, 0 }, /* %O */
121 { in_line_eoln, 0 }, /* %P */
122 { in_line_eoln, 0 }, /* %R */
123 { in_line_eoln, 0 }, /* %T */
124 { in_line_eoln, 0 }, /* %V */
125 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Ac */
126 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Ao */
127 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Aq */
128 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* At */
129 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Bc */
130 { blk_full, MDOC_EXPLICIT }, /* Bf */
131 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Bo */
132 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Bq */
133 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Bsx */
134 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Bx */
135 { in_line_eoln, 0 }, /* Db */
136 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Dc */
137 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Do */
138 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Dq */
139 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Ec */
140 { blk_exp_close, MDOC_EXPLICIT }, /* Ef */
141 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Em */
142 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Eo */
143 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Fx */
144 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ms */
145 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* No */
146 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* Ns */
147 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Nx */
148 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ox */
149 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Pc */
150 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* Pf */
151 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Po */
152 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Pq */
153 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Qc */
154 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Ql */
155 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Qo */
156 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Qq */
157 { blk_exp_close, MDOC_EXPLICIT }, /* Re */
158 { blk_full, MDOC_EXPLICIT }, /* Rs */
159 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Sc */
160 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* So */
161 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Sq */
162 { in_line_eoln, 0 }, /* Sm */
163 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Sx */
164 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Sy */
165 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Tn */
166 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ux */
167 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Xc */
168 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Xo */
169 { blk_full, MDOC_EXPLICIT | MDOC_CALLABLE }, /* Fo */
170 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Fc */
171 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Oo */
172 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Oc */
173 { blk_full, MDOC_EXPLICIT }, /* Bk */
174 { blk_exp_close, MDOC_EXPLICIT }, /* Ek */
175 { in_line_eoln, 0 }, /* Bt */
176 { in_line_eoln, 0 }, /* Hf */
177 { obsolete, 0 }, /* Fr */
178 { in_line_eoln, 0 }, /* Ud */
179 { in_line, 0 }, /* Lb */
180 { in_line_eoln, 0 }, /* Lp */
181 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Lk */
182 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Mt */
183 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Brq */
184 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Bro */
185 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Brc */
186 { in_line_eoln, 0 }, /* %C */
187 { obsolete, 0 }, /* Es */
188 { obsolete, 0 }, /* En */
189 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Dx */
190 { in_line_eoln, 0 }, /* %Q */
191 { in_line_eoln, 0 }, /* br */
192 { in_line_eoln, 0 }, /* sp */
193 { in_line_eoln, 0 }, /* %U */
194 { phrase_ta, MDOC_CALLABLE | MDOC_PARSED }, /* Ta */
197 const struct mdoc_macro * const mdoc_macros = __mdoc_macros;
201 * This is called at the end of parsing. It must traverse up the tree,
202 * closing out open [implicit] scopes. Obviously, open explicit scopes
206 mdoc_macroend(struct mdoc *m)
210 /* Scan for open explicit scopes. */
212 n = MDOC_VALID & m->last->flags ? m->last->parent : m->last;
214 for ( ; n; n = n->parent)
215 if (MDOC_BLOCK == n->type &&
216 MDOC_EXPLICIT & mdoc_macros[n->tok].flags)
217 mdoc_nmsg(m, n, MANDOCERR_SCOPEEXIT);
219 /* Rewind to the first. */
221 return(rew_last(m, m->first));
226 * Look up a macro from within a subsequent context.
229 lookup(enum mdoct from, const char *p)
231 /* FIXME: make -diag lists be un-PARSED. */
233 if ( ! (MDOC_PARSED & mdoc_macros[from].flags))
235 return(lookup_raw(p));
240 * Lookup a macro following the initial line macro.
243 lookup_raw(const char *p)
247 if (MDOC_MAX == (res = mdoc_hash_find(p)))
249 if (MDOC_CALLABLE & mdoc_macros[res].flags)
256 rew_last(struct mdoc *mdoc, const struct mdoc_node *to)
258 struct mdoc_node *n, *np;
261 mdoc->next = MDOC_NEXT_SIBLING;
264 while (mdoc->last != to) {
266 * Save the parent here, because we may delete the
267 * m->last node in the post-validation phase and reset
268 * it to m->last->parent, causing a step in the closing
271 np = mdoc->last->parent;
272 if ( ! mdoc_valid_post(mdoc))
277 mdoc->last->last = n;
280 return(mdoc_valid_post(mdoc));
285 * For a block closing macro, return the corresponding opening one.
286 * Otherwise, return the macro itself.
289 rew_alt(enum mdoct tok)
332 * Rewinding to tok, how do we have to handle *p?
333 * REWIND_NONE: *p would delimit tok, but no tok scope is open
334 * inside *p, so there is no need to rewind anything at all.
335 * REWIND_THIS: *p matches tok, so rewind *p and nothing else.
336 * REWIND_MORE: *p is implicit, rewind it and keep searching for tok.
337 * REWIND_FORCE: *p is explicit, but tok is full, force rewinding *p.
338 * REWIND_LATER: *p is explicit and still open, postpone rewinding.
339 * REWIND_ERROR: No tok block is open at all.
342 rew_dohalt(enum mdoct tok, enum mdoc_type type,
343 const struct mdoc_node *p)
347 * No matching token, no delimiting block, no broken block.
348 * This can happen when full implicit macros are called for
349 * the first time but try to rewind their previous
352 if (MDOC_ROOT == p->type)
353 return(MDOC_BLOCK == type &&
354 MDOC_EXPLICIT & mdoc_macros[tok].flags ?
355 REWIND_ERROR : REWIND_NONE);
358 * When starting to rewind, skip plain text
359 * and nodes that have already been rewound.
361 if (MDOC_TEXT == p->type || MDOC_VALID & p->flags)
365 * The easiest case: Found a matching token.
366 * This applies to both blocks and elements.
370 return(p->end ? REWIND_NONE :
371 type == p->type ? REWIND_THIS : REWIND_MORE);
374 * While elements do require rewinding for themselves,
375 * they never affect rewinding of other nodes.
377 if (MDOC_ELEM == p->type)
381 * Blocks delimited by our target token get REWIND_MORE.
382 * Blocks delimiting our target token get REWIND_NONE.
386 if (MDOC_It == p->tok)
390 if (MDOC_BODY == p->type && MDOC_Bl == p->tok)
394 * XXX Badly nested block handling still fails badly
395 * when one block is breaking two blocks of the same type.
396 * This is an incomplete and extremely ugly workaround,
397 * required to let the OpenBSD tree build.
400 if (MDOC_Op == p->tok)
408 if (MDOC_BODY == p->type && MDOC_Sh == p->tok)
412 if (MDOC_Nd == p->tok || MDOC_Ss == p->tok ||
421 * Default block rewinding rules.
422 * In particular, always skip block end markers,
423 * and let all blocks rewind Nm children.
425 if (ENDBODY_NOT != p->end || MDOC_Nm == p->tok ||
426 (MDOC_BLOCK == p->type &&
427 ! (MDOC_EXPLICIT & mdoc_macros[tok].flags)))
431 * By default, closing out full blocks
432 * forces closing of broken explicit blocks,
433 * while closing out partial blocks
434 * allows delayed rewinding by default.
436 return (&blk_full == mdoc_macros[tok].fp ?
437 REWIND_FORCE : REWIND_LATER);
442 rew_elem(struct mdoc *mdoc, enum mdoct tok)
447 if (MDOC_ELEM != n->type)
449 assert(MDOC_ELEM == n->type);
450 assert(tok == n->tok);
452 return(rew_last(mdoc, n));
457 * We are trying to close a block identified by tok,
458 * but the child block *broken is still open.
459 * Thus, postpone closing the tok block
460 * until the rew_sub call closing *broken.
463 make_pending(struct mdoc_node *broken, enum mdoct tok,
464 struct mdoc *m, int line, int ppos)
466 struct mdoc_node *breaker;
469 * Iterate backwards, searching for the block matching tok,
470 * that is, the block breaking the *broken block.
472 for (breaker = broken->parent; breaker; breaker = breaker->parent) {
475 * If the *broken block had already been broken before
476 * and we encounter its breaker, make the tok block
477 * pending on the inner breaker.
478 * Graphically, "[A breaker=[B broken=[C->B B] tok=A] C]"
479 * becomes "[A broken=[B [C->B B] tok=A] C]"
480 * and finally "[A [B->A [C->B B] A] C]".
482 if (breaker == broken->pending) {
487 if (REWIND_THIS != rew_dohalt(tok, MDOC_BLOCK, breaker))
489 if (MDOC_BODY == broken->type)
490 broken = broken->parent;
494 * If another, outer breaker is already pending on
495 * the *broken block, we must not clobber the link
496 * to the outer breaker, but make it pending on the
497 * new, now inner breaker.
498 * Graphically, "[A breaker=[B broken=[C->A A] tok=B] C]"
499 * becomes "[A breaker=[B->A broken=[C A] tok=B] C]"
500 * and finally "[A [B->A [C->B A] B] C]".
502 if (broken->pending) {
503 struct mdoc_node *taker;
506 * If the breaker had also been broken before,
507 * it cannot take on the outer breaker itself,
508 * but must hand it on to its own breakers.
509 * Graphically, this is the following situation:
510 * "[A [B breaker=[C->B B] broken=[D->A A] tok=C] D]"
511 * "[A taker=[B->A breaker=[C->B B] [D->C A] C] D]"
514 while (taker->pending)
515 taker = taker->pending;
516 taker->pending = broken->pending;
518 broken->pending = breaker;
519 mandoc_vmsg(MANDOCERR_SCOPENEST, m->parse, line, ppos,
520 "%s breaks %s", mdoc_macronames[tok],
521 mdoc_macronames[broken->tok]);
526 * Found no matching block for tok.
527 * Are you trying to close a block that is not open?
534 rew_sub(enum mdoc_type t, struct mdoc *m,
535 enum mdoct tok, int line, int ppos)
541 switch (rew_dohalt(tok, t, n)) {
547 mandoc_vmsg(MANDOCERR_SCOPEBROKEN, m->parse,
548 line, ppos, "%s breaks %s",
549 mdoc_macronames[tok],
550 mdoc_macronames[n->tok]);
556 if (make_pending(n, tok, m, line, ppos) ||
561 mdoc_pmsg(m, line, ppos, MANDOCERR_NOSCOPE);
568 if ( ! rew_last(m, n))
572 * The current block extends an enclosing block.
573 * Now that the current block ends, close the enclosing block, too.
575 while (NULL != (n = n->pending)) {
576 if ( ! rew_last(m, n))
578 if (MDOC_HEAD == n->type &&
579 ! mdoc_body_alloc(m, n->line, n->pos, n->tok))
587 * Allocate a word and check whether it's punctuation or not.
588 * Punctuation consists of those tokens found in mdoc_isdelim().
591 dword(struct mdoc *m, int line,
592 int col, const char *p, enum mdelim d)
598 if ( ! mdoc_word_alloc(m, line, col, p))
602 m->last->flags |= MDOC_DELIMO;
605 * Closing delimiters only suppress the preceding space
606 * when they follow something, not when they start a new
607 * block or element, and not when they follow `No'.
609 * XXX Explicitly special-casing MDOC_No here feels
610 * like a layering violation. Find a better way
611 * and solve this in the code related to `No'!
614 else if (DELIM_CLOSE == d && m->last->prev &&
615 m->last->prev->tok != MDOC_No)
616 m->last->flags |= MDOC_DELIMC;
622 append_delims(struct mdoc *m, int line, int *pos, char *buf)
628 if ('\0' == buf[*pos])
633 ac = mdoc_zargs(m, line, pos, buf, &p);
635 if (ARGS_ERROR == ac)
637 else if (ARGS_EOLN == ac)
640 dword(m, line, la, p, DELIM_MAX);
643 * If we encounter end-of-sentence symbols, then trigger
646 * XXX: it's easy to allow this to propagate outward to
647 * the last symbol, such that `. )' will cause the
648 * correct double-spacing. However, (1) groff isn't
649 * smart enough to do this and (2) it would require
650 * knowing which symbols break this behaviour, for
651 * example, `. ;' shouldn't propagate the double-space.
653 if (mandoc_eos(p, strlen(p), 0))
654 m->last->flags |= MDOC_EOS;
662 * Close out block partial/full explicit.
665 blk_exp_close(MACRO_PROT_ARGS)
667 struct mdoc_node *body; /* Our own body. */
668 struct mdoc_node *later; /* A sub-block starting later. */
669 struct mdoc_node *n; /* For searching backwards. */
671 int j, lastarg, maxargs, flushed, nl;
673 enum mdoct atok, ntok;
676 nl = MDOC_NEWLINE & m->flags;
688 * Search backwards for beginnings of blocks,
689 * both of our own and of pending sub-blocks.
693 for (n = m->last; n; n = n->parent) {
694 if (MDOC_VALID & n->flags)
697 /* Remember the start of our own body. */
698 if (MDOC_BODY == n->type && atok == n->tok) {
699 if (ENDBODY_NOT == n->end)
704 if (MDOC_BLOCK != n->type || MDOC_Nm == n->tok)
706 if (atok == n->tok) {
710 * Found the start of our own block.
711 * When there is no pending sub block,
712 * just proceed to closing out.
718 * When there is a pending sub block,
719 * postpone closing out the current block
720 * until the rew_sub() closing out the sub-block.
722 make_pending(later, tok, m, line, ppos);
725 * Mark the place where the formatting - but not
726 * the scope - of the current block ends.
728 if ( ! mdoc_endbody_alloc(m, line, ppos,
729 atok, body, ENDBODY_SPACE))
735 * When finding an open sub block, remember the last
736 * open explicit block, or, in case there are only
737 * implicit ones, the first open implicit block.
740 MDOC_EXPLICIT & mdoc_macros[later->tok].flags)
742 if (MDOC_CALLABLE & mdoc_macros[n->tok].flags)
746 if ( ! (MDOC_CALLABLE & mdoc_macros[tok].flags)) {
747 /* FIXME: do this in validate */
749 mdoc_pmsg(m, line, ppos, MANDOCERR_ARGSLOST);
751 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos))
753 return(rew_sub(MDOC_BLOCK, m, tok, line, ppos));
756 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos))
759 if (NULL == later && maxargs > 0)
760 if ( ! mdoc_tail_alloc(m, line, ppos, rew_alt(tok)))
763 for (flushed = j = 0; ; j++) {
766 if (j == maxargs && ! flushed) {
767 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
772 ac = mdoc_args(m, line, pos, buf, tok, &p);
774 if (ARGS_ERROR == ac)
776 if (ARGS_PUNCT == ac)
781 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
783 if (MDOC_MAX == ntok) {
784 if ( ! dword(m, line, lastarg, p, DELIM_MAX))
790 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
794 if ( ! mdoc_macro(m, ntok, line, lastarg, pos, buf))
799 if ( ! flushed && ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
804 return(append_delims(m, line, pos, buf));
809 in_line(MACRO_PROT_ARGS)
811 int la, scope, cnt, nc, nl;
816 struct mdoc_arg *arg;
819 nl = MDOC_NEWLINE & m->flags;
822 * Whether we allow ignored elements (those without content,
823 * usually because of reserved words) to squeak by.
845 for (arg = NULL;; ) {
847 av = mdoc_argv(m, line, tok, &arg, pos, buf);
849 if (ARGV_WORD == av) {
862 for (cnt = scope = 0;; ) {
864 ac = mdoc_args(m, line, pos, buf, tok, &p);
866 if (ARGS_ERROR == ac)
870 if (ARGS_PUNCT == ac)
873 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
876 * In this case, we've located a submacro and must
877 * execute it. Close out scope, if open. If no
878 * elements have been generated, either create one (nc)
879 * or raise a warning.
882 if (MDOC_MAX != ntok) {
883 if (scope && ! rew_elem(m, tok))
885 if (nc && 0 == cnt) {
886 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg))
888 if ( ! rew_last(m, m->last))
890 } else if ( ! nc && 0 == cnt) {
892 mdoc_pmsg(m, line, ppos, MANDOCERR_MACROEMPTY);
895 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
899 return(append_delims(m, line, pos, buf));
903 * Non-quote-enclosed punctuation. Set up our scope, if
904 * a word; rewind the scope, if a delimiter; then append
908 d = ARGS_QWORD == ac ? DELIM_NONE : mdoc_isdelim(p);
910 if (DELIM_NONE != d) {
912 * If we encounter closing punctuation, no word
913 * has been omitted, no scope is open, and we're
914 * allowed to have an empty element, then start
915 * a new scope. `Ar', `Fl', and `Li', only do
916 * this once per invocation. There may be more
917 * of these (all of them?).
919 if (0 == cnt && (nc || MDOC_Li == tok) &&
920 DELIM_CLOSE == d && ! scope) {
921 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg))
923 if (MDOC_Ar == tok || MDOC_Li == tok ||
929 * Close out our scope, if one is open, before
932 if (scope && ! rew_elem(m, tok))
935 } else if ( ! scope) {
936 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg))
944 if ( ! dword(m, line, la, p, d))
948 * `Fl' macros have their scope re-opened with each new
949 * word so that the `-' can be added to each one without
950 * having to parse out spaces.
952 if (scope && MDOC_Fl == tok) {
953 if ( ! rew_elem(m, tok))
959 if (scope && ! rew_elem(m, tok))
963 * If no elements have been collected and we're allowed to have
964 * empties (nc), open a scope and close it out. Otherwise,
968 if (nc && 0 == cnt) {
969 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg))
971 if ( ! rew_last(m, m->last))
973 } else if ( ! nc && 0 == cnt) {
975 mdoc_pmsg(m, line, ppos, MANDOCERR_MACROEMPTY);
980 return(append_delims(m, line, pos, buf));
985 blk_full(MACRO_PROT_ARGS)
988 struct mdoc_arg *arg;
989 struct mdoc_node *head; /* save of head macro */
990 struct mdoc_node *body; /* save of body macro */
994 enum margserr ac, lac;
998 nl = MDOC_NEWLINE & m->flags;
1000 /* Close out prior implicit scope. */
1002 if ( ! (MDOC_EXPLICIT & mdoc_macros[tok].flags)) {
1003 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos))
1005 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
1010 * This routine accommodates implicitly- and explicitly-scoped
1011 * macro openings. Implicit ones first close out prior scope
1012 * (seen above). Delay opening the head until necessary to
1013 * allow leading punctuation to print. Special consideration
1014 * for `It -column', which has phrase-part syntax instead of
1015 * regular child nodes.
1018 for (arg = NULL;; ) {
1020 av = mdoc_argv(m, line, tok, &arg, pos, buf);
1022 if (ARGV_WORD == av) {
1027 if (ARGV_EOLN == av)
1032 mdoc_argv_free(arg);
1036 if ( ! mdoc_block_alloc(m, line, ppos, tok, arg))
1042 * The `Nd' macro has all arguments in its body: it's a hybrid
1043 * of block partial-explicit and full-implicit. Stupid.
1046 if (MDOC_Nd == tok) {
1047 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1050 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos))
1052 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1061 /* Initialise last-phrase-type with ARGS_PEND. */
1062 lac = ARGS_ERROR == ac ? ARGS_PEND : ac;
1063 ac = mdoc_args(m, line, pos, buf, tok, &p);
1065 if (ARGS_PUNCT == ac)
1068 if (ARGS_ERROR == ac)
1071 if (ARGS_EOLN == ac) {
1072 if (ARGS_PPHRASE != lac && ARGS_PHRASE != lac)
1075 * This is necessary: if the last token on a
1076 * line is a `Ta' or tab, then we'll get
1077 * ARGS_EOLN, so we must be smart enough to
1078 * reopen our scope if the last parse was a
1079 * phrase or partial phrase.
1081 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos))
1083 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1090 * Emit leading punctuation (i.e., punctuation before
1091 * the MDOC_HEAD) for non-phrase types.
1096 ARGS_PHRASE != ac &&
1097 ARGS_PPHRASE != ac &&
1099 DELIM_OPEN == mdoc_isdelim(p)) {
1100 if ( ! dword(m, line, la, p, DELIM_OPEN))
1105 /* Open a head if one hasn't been opened. */
1108 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1113 if (ARGS_PHRASE == ac ||
1115 ARGS_PPHRASE == ac) {
1117 * If we haven't opened a body yet, rewind the
1118 * head; if we have, rewind that instead.
1121 mtt = body ? MDOC_BODY : MDOC_HEAD;
1122 if ( ! rew_sub(mtt, m, tok, line, ppos))
1125 /* Then allocate our body context. */
1127 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1132 * Process phrases: set whether we're in a
1133 * partial-phrase (this effects line handling)
1134 * then call down into the phrase parser.
1137 if (ARGS_PPHRASE == ac)
1138 m->flags |= MDOC_PPHRASE;
1139 if (ARGS_PEND == ac && ARGS_PPHRASE == lac)
1140 m->flags |= MDOC_PPHRASE;
1142 if ( ! phrase(m, line, la, buf))
1145 m->flags &= ~MDOC_PPHRASE;
1149 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1151 if (MDOC_MAX == ntok) {
1152 if ( ! dword(m, line, la, p, DELIM_MAX))
1157 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
1163 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1168 if (nl && ! append_delims(m, line, pos, buf))
1171 /* If we've already opened our body, exit now. */
1177 * If there is an open (i.e., unvalidated) sub-block requiring
1178 * explicit close-out, postpone switching the current block from
1179 * head to body until the rew_sub() call closing out that
1182 for (n = m->last; n && n != head; n = n->parent) {
1183 if (MDOC_BLOCK == n->type &&
1184 MDOC_EXPLICIT & mdoc_macros[n->tok].flags &&
1185 ! (MDOC_VALID & n->flags)) {
1191 /* Close out scopes to remain in a consistent state. */
1193 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos))
1195 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1199 if ( ! (MDOC_FREECOL & m->flags))
1202 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos))
1204 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
1207 m->flags &= ~MDOC_FREECOL;
1213 blk_part_imp(MACRO_PROT_ARGS)
1219 struct mdoc_node *blk; /* saved block context */
1220 struct mdoc_node *body; /* saved body context */
1221 struct mdoc_node *n;
1223 nl = MDOC_NEWLINE & m->flags;
1226 * A macro that spans to the end of the line. This is generally
1227 * (but not necessarily) called as the first macro. The block
1228 * has a head as the immediate child, which is always empty,
1229 * followed by zero or more opening punctuation nodes, then the
1230 * body (which may be empty, depending on the macro), then zero
1231 * or more closing punctuation nodes.
1234 if ( ! mdoc_block_alloc(m, line, ppos, tok, NULL))
1239 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1241 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos))
1245 * Open the body scope "on-demand", that is, after we've
1246 * processed all our the leading delimiters (open parenthesis,
1250 for (body = NULL; ; ) {
1252 ac = mdoc_args(m, line, pos, buf, tok, &p);
1254 if (ARGS_ERROR == ac)
1256 if (ARGS_EOLN == ac)
1258 if (ARGS_PUNCT == ac)
1261 if (NULL == body && ARGS_QWORD != ac &&
1262 DELIM_OPEN == mdoc_isdelim(p)) {
1263 if ( ! dword(m, line, la, p, DELIM_OPEN))
1269 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1274 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1276 if (MDOC_MAX == ntok) {
1277 if ( ! dword(m, line, la, p, DELIM_MAX))
1282 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
1287 /* Clean-ups to leave in a consistent state. */
1290 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1295 for (n = body->child; n && n->next; n = n->next)
1299 * End of sentence spacing: if the last node is a text node and
1300 * has a trailing period, then mark it as being end-of-sentence.
1303 if (n && MDOC_TEXT == n->type && n->string)
1304 if (mandoc_eos(n->string, strlen(n->string), 1))
1305 n->flags |= MDOC_EOS;
1307 /* Up-propagate the end-of-space flag. */
1309 if (n && (MDOC_EOS & n->flags)) {
1310 body->flags |= MDOC_EOS;
1311 body->parent->flags |= MDOC_EOS;
1315 * If there is an open sub-block requiring explicit close-out,
1316 * postpone closing out the current block
1317 * until the rew_sub() call closing out the sub-block.
1319 for (n = m->last; n && n != body && n != blk->parent; n = n->parent) {
1320 if (MDOC_BLOCK == n->type &&
1321 MDOC_EXPLICIT & mdoc_macros[n->tok].flags &&
1322 ! (MDOC_VALID & n->flags)) {
1323 make_pending(n, tok, m, line, ppos);
1324 if ( ! mdoc_endbody_alloc(m, line, ppos,
1325 tok, body, ENDBODY_NOSPACE))
1332 * If we can't rewind to our body, then our scope has already
1333 * been closed by another macro (like `Oc' closing `Op'). This
1334 * is ugly behaviour nodding its head to OpenBSD's overwhelming
1335 * crufty use of `Op' breakage.
1338 mandoc_vmsg(MANDOCERR_SCOPENEST, m->parse, line, ppos,
1339 "%s broken", mdoc_macronames[tok]);
1341 if (n && ! rew_sub(MDOC_BODY, m, tok, line, ppos))
1344 /* Standard appending of delimiters. */
1346 if (nl && ! append_delims(m, line, pos, buf))
1349 /* Rewind scope, if applicable. */
1351 if (n && ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
1359 blk_part_exp(MACRO_PROT_ARGS)
1363 struct mdoc_node *head; /* keep track of head */
1364 struct mdoc_node *body; /* keep track of body */
1368 nl = MDOC_NEWLINE & m->flags;
1371 * The opening of an explicit macro having zero or more leading
1372 * punctuation nodes; a head with optional single element (the
1373 * case of `Eo'); and a body that may be empty.
1376 if ( ! mdoc_block_alloc(m, line, ppos, tok, NULL))
1379 for (head = body = NULL; ; ) {
1381 ac = mdoc_args(m, line, pos, buf, tok, &p);
1383 if (ARGS_ERROR == ac)
1385 if (ARGS_PUNCT == ac)
1387 if (ARGS_EOLN == ac)
1390 /* Flush out leading punctuation. */
1392 if (NULL == head && ARGS_QWORD != ac &&
1393 DELIM_OPEN == mdoc_isdelim(p)) {
1394 assert(NULL == body);
1395 if ( ! dword(m, line, la, p, DELIM_OPEN))
1401 assert(NULL == body);
1402 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1408 * `Eo' gobbles any data into the head, but most other
1409 * macros just immediately close out and begin the body.
1414 /* No check whether it's a macro! */
1416 if ( ! dword(m, line, la, p, DELIM_MAX))
1419 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos))
1421 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1429 assert(NULL != head && NULL != body);
1431 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1433 if (MDOC_MAX == ntok) {
1434 if ( ! dword(m, line, la, p, DELIM_MAX))
1439 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
1444 /* Clean-up to leave in a consistent state. */
1447 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1453 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos))
1455 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1460 /* Standard appending of delimiters. */
1464 return(append_delims(m, line, pos, buf));
1470 in_line_argn(MACRO_PROT_ARGS)
1472 int la, flushed, j, maxargs, nl;
1475 struct mdoc_arg *arg;
1479 nl = MDOC_NEWLINE & m->flags;
1482 * A line macro that has a fixed number of arguments (maxargs).
1483 * Only open the scope once the first non-leading-punctuation is
1484 * found (unless MDOC_IGNDELIM is noted, like in `Pf'), then
1485 * keep it open until the maximum number of arguments are
1509 for (arg = NULL; ; ) {
1511 av = mdoc_argv(m, line, tok, &arg, pos, buf);
1513 if (ARGV_WORD == av) {
1518 if (ARGV_EOLN == av)
1523 mdoc_argv_free(arg);
1527 for (flushed = j = 0; ; ) {
1529 ac = mdoc_args(m, line, pos, buf, tok, &p);
1531 if (ARGS_ERROR == ac)
1533 if (ARGS_PUNCT == ac)
1535 if (ARGS_EOLN == ac)
1538 if ( ! (MDOC_IGNDELIM & mdoc_macros[tok].flags) &&
1539 ARGS_QWORD != ac && 0 == j &&
1540 DELIM_OPEN == mdoc_isdelim(p)) {
1541 if ( ! dword(m, line, la, p, DELIM_OPEN))
1545 if ( ! mdoc_elem_alloc(m, line, la, tok, arg))
1548 if (j == maxargs && ! flushed) {
1549 if ( ! rew_elem(m, tok))
1554 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1556 if (MDOC_MAX != ntok) {
1557 if ( ! flushed && ! rew_elem(m, tok))
1560 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
1566 if ( ! (MDOC_IGNDELIM & mdoc_macros[tok].flags) &&
1569 DELIM_NONE != mdoc_isdelim(p)) {
1570 if ( ! rew_elem(m, tok))
1576 * XXX: this is a hack to work around groff's ugliness
1577 * as regards `Xr' and extraneous arguments. It should
1578 * ideally be deprecated behaviour, but because this is
1579 * code is no here, it's unlikely to be removed.
1583 if (MDOC_Xr == tok && j == maxargs) {
1584 if ( ! mdoc_elem_alloc(m, line, la, MDOC_Ns, NULL))
1586 if ( ! rew_elem(m, MDOC_Ns))
1591 if ( ! dword(m, line, la, p, DELIM_MAX))
1596 if (0 == j && ! mdoc_elem_alloc(m, line, la, tok, arg))
1599 /* Close out in a consistent state. */
1601 if ( ! flushed && ! rew_elem(m, tok))
1605 return(append_delims(m, line, pos, buf));
1610 in_line_eoln(MACRO_PROT_ARGS)
1615 struct mdoc_arg *arg;
1619 assert( ! (MDOC_PARSED & mdoc_macros[tok].flags));
1622 rew_sub(MDOC_BLOCK, m, MDOC_Nm, line, ppos);
1624 /* Parse macro arguments. */
1626 for (arg = NULL; ; ) {
1628 av = mdoc_argv(m, line, tok, &arg, pos, buf);
1630 if (ARGV_WORD == av) {
1634 if (ARGV_EOLN == av)
1639 mdoc_argv_free(arg);
1643 /* Open element scope. */
1645 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg))
1648 /* Parse argument terms. */
1652 ac = mdoc_args(m, line, pos, buf, tok, &p);
1654 if (ARGS_ERROR == ac)
1656 if (ARGS_EOLN == ac)
1659 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1661 if (MDOC_MAX == ntok) {
1662 if ( ! dword(m, line, la, p, DELIM_MAX))
1667 if ( ! rew_elem(m, tok))
1669 return(mdoc_macro(m, ntok, line, la, pos, buf));
1672 /* Close out (no delimiters). */
1674 return(rew_elem(m, tok));
1680 ctx_synopsis(MACRO_PROT_ARGS)
1684 nl = MDOC_NEWLINE & m->flags;
1686 /* If we're not in the SYNOPSIS, go straight to in-line. */
1687 if ( ! (MDOC_SYNOPSIS & m->flags))
1688 return(in_line(m, tok, line, ppos, pos, buf));
1690 /* If we're a nested call, same place. */
1692 return(in_line(m, tok, line, ppos, pos, buf));
1695 * XXX: this will open a block scope; however, if later we end
1696 * up formatting the block scope, then child nodes will inherit
1697 * the formatting. Be careful.
1700 return(blk_full(m, tok, line, ppos, pos, buf));
1701 assert(MDOC_Vt == tok);
1702 return(blk_part_imp(m, tok, line, ppos, pos, buf));
1708 obsolete(MACRO_PROT_ARGS)
1711 mdoc_pmsg(m, line, ppos, MANDOCERR_MACROOBS);
1717 * Phrases occur within `Bl -column' entries, separated by `Ta' or tabs.
1718 * They're unusual because they're basically free-form text until a
1719 * macro is encountered.
1722 phrase(struct mdoc *m, int line, int ppos, char *buf)
1729 for (pos = ppos; ; ) {
1732 ac = mdoc_zargs(m, line, &pos, buf, &p);
1734 if (ARGS_ERROR == ac)
1736 if (ARGS_EOLN == ac)
1739 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup_raw(p);
1741 if (MDOC_MAX == ntok) {
1742 if ( ! dword(m, line, la, p, DELIM_MAX))
1747 if ( ! mdoc_macro(m, ntok, line, la, &pos, buf))
1749 return(append_delims(m, line, &pos, buf));
1758 phrase_ta(MACRO_PROT_ARGS)
1766 * FIXME: this is overly restrictive: if the `Ta' is unexpected,
1767 * it should simply error out with ARGSLOST.
1770 if ( ! rew_sub(MDOC_BODY, m, MDOC_It, line, ppos))
1772 if ( ! mdoc_body_alloc(m, line, ppos, MDOC_It))
1777 ac = mdoc_zargs(m, line, pos, buf, &p);
1779 if (ARGS_ERROR == ac)
1781 if (ARGS_EOLN == ac)
1784 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup_raw(p);
1786 if (MDOC_MAX == ntok) {
1787 if ( ! dword(m, line, la, p, DELIM_MAX))
1792 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
1794 return(append_delims(m, line, pos, buf));