=================================================================== RCS file: /cvs/mandoc/Attic/macro.c,v retrieving revision 1.32 retrieving revision 1.48 diff -u -p -r1.32 -r1.48 --- mandoc/Attic/macro.c 2009/01/12 12:52:21 1.32 +++ mandoc/Attic/macro.c 2009/01/20 20:56:21 1.48 @@ -1,4 +1,4 @@ -/* $Id: macro.c,v 1.32 2009/01/12 12:52:21 kristaps Exp $ */ +/* $Id: macro.c,v 1.48 2009/01/20 20:56:21 kristaps Exp $ */ /* * Copyright (c) 2008 Kristaps Dzonsons * @@ -25,18 +25,72 @@ #include #endif +/* + * This has scanning/parsing routines, each of which extract a macro and + * its arguments and parameters, then know how to progress to the next + * macro. Macros are parsed according as follows: + * + * ELEMENT: TEXT | epsilon + * BLOCK: HEAD PUNCT BODY PUNCT BLOCK_TAIL PUNCT + * BLOCK_TAIL: TAIL | epsilon + * HEAD: ELEMENT | TEXT | BLOCK | epsilon + * BODY: ELEMENT | TEXT | BLOCK | epsilon + * TAIL: TEXT | epsilon + * PUNCT: TEXT (delimiters) | epsilon + * + * These are arranged into a parse tree, an example of which follows: + * + * ROOT + * BLOCK (.Sh) + * HEAD + * TEXT (`NAME') + * BODY + * ELEMENT (.Nm) + * TEXT (`mdocml') + * ELEMENT (.Nd) + * TEXT (`mdoc macro compiler') + * BLOCK (.Op) + * HEAD + * ELEMENT (.Fl) + * TEXT (`v') + * BLOCK (.Op) + * HEAD + * ELEMENT (.Fl) + * TEXT (`v') + * ELEMENT (.Fl) + * TEXT (`W') + * ELEMENT (.Ns) + * ELEMENT (.Ar) + * TEXT (`err...') + * + * These types are always per-line except for block bodies, which may + * span multiple lines. Macros are assigned a parsing routine, which + * corresponds to the type, in the mdoc_macros table. + * + * Note that types are general: there can be several parsing routines + * corresponding to a single type. The macro_text function, for + * example, parses an ELEMENT type (see the function definition for + * details) that may be interrupted by further macros; the + * macro_constant function, on the other hand, parses an ELEMENT type + * spanning a single line. + */ + #include "private.h" -/* FIXME: maxlineargs should be per LINE, no per TOKEN. */ - +#define REWIND_REWIND (1 << 0) +#define REWIND_NOHALT (1 << 1) +#define REWIND_HALT (1 << 2) +static int rewind_dohalt(int, enum mdoc_type, + const struct mdoc_node *); +static int rewind_alt(int); +static int rewind_dobreak(int, const struct mdoc_node *); static int rewind_elem(struct mdoc *, int); -static int rewind_impblock(struct mdoc *, int); -static int rewind_expblock(struct mdoc *, int); -static int rewind_head(struct mdoc *, int); -static int rewind_body(struct mdoc *, int); +static int rewind_impblock(struct mdoc *, int, int, int); +static int rewind_expblock(struct mdoc *, int, int, int); +static int rewind_subblock(enum mdoc_type, + struct mdoc *, int, int, int); static int rewind_last(struct mdoc *, struct mdoc_node *); -static int append_delims(struct mdoc *, - int, int, int *, char *); +static int append_delims(struct mdoc *, int, int *, char *); static int lookup(struct mdoc *, int, int, int, const char *); @@ -50,8 +104,7 @@ lookup(struct mdoc *mdoc, int line, int pos, int from, return(res); if (MDOC_MAX == res) return(res); - - if ( ! mdoc_pwarn(mdoc, line, pos, WARN_SYNTAX_MACLIKE)) + if ( ! mdoc_pwarn(mdoc, line, pos, WARN_SYNTAX, "macro-like parameter")) return(-1); return(MDOC_MAX); } @@ -63,24 +116,234 @@ rewind_last(struct mdoc *mdoc, struct mdoc_node *to) assert(to); mdoc->next = MDOC_NEXT_SIBLING; - if (mdoc->last == to) { + + while (mdoc->last != to) { if ( ! mdoc_valid_post(mdoc)) return(0); if ( ! mdoc_action_post(mdoc)) return(0); + mdoc->last = mdoc->last->parent; + assert(mdoc->last); + } + + if ( ! mdoc_valid_post(mdoc)) + return(0); + return(mdoc_action_post(mdoc)); +} + + +static int +rewind_alt(int tok) +{ + switch (tok) { + case (MDOC_Ac): + return(MDOC_Ao); + case (MDOC_Bc): + return(MDOC_Bo); + case (MDOC_Dc): + return(MDOC_Do); + case (MDOC_Ec): + return(MDOC_Eo); + case (MDOC_Ed): + return(MDOC_Bd); + case (MDOC_Ef): + return(MDOC_Bf); + case (MDOC_Ek): + return(MDOC_Bk); + case (MDOC_El): + return(MDOC_Bl); + case (MDOC_Fc): + return(MDOC_Fo); + case (MDOC_Oc): + return(MDOC_Oo); + case (MDOC_Pc): + return(MDOC_Po); + case (MDOC_Qc): + return(MDOC_Qo); + case (MDOC_Re): + return(MDOC_Rs); + case (MDOC_Sc): + return(MDOC_So); + case (MDOC_Xc): + return(MDOC_Xo); + default: + break; + } + abort(); + /* NOTREACHED */ +} + + +static int +rewind_dohalt(int tok, enum mdoc_type type, const struct mdoc_node *p) +{ + + if (MDOC_ROOT == p->type) + return(REWIND_HALT); + if (MDOC_VALID & p->flags) + return(REWIND_NOHALT); + + switch (tok) { + /* One-liner implicit-scope. */ + case (MDOC_Aq): + /* FALLTHROUGH */ + case (MDOC_Bq): + /* FALLTHROUGH */ + case (MDOC_D1): + /* FALLTHROUGH */ + case (MDOC_Dl): + /* FALLTHROUGH */ + case (MDOC_Dq): + /* FALLTHROUGH */ + case (MDOC_Op): + /* FALLTHROUGH */ + case (MDOC_Pq): + /* FALLTHROUGH */ + case (MDOC_Ql): + /* FALLTHROUGH */ + case (MDOC_Qq): + /* FALLTHROUGH */ + case (MDOC_Sq): + assert(MDOC_HEAD != type); + assert(MDOC_TAIL != type); + if (type == p->type && tok == p->tok) + return(REWIND_REWIND); + break; + + /* Multi-line implicit-scope. */ + case (MDOC_It): + assert(MDOC_TAIL != type); + if (type == p->type && tok == p->tok) + return(REWIND_REWIND); + if (MDOC_BODY == p->type && MDOC_Bl == p->tok) + return(REWIND_HALT); + break; + case (MDOC_Sh): + if (type == p->type && tok == p->tok) + return(REWIND_REWIND); + break; + case (MDOC_Ss): + assert(MDOC_TAIL != type); + if (type == p->type && tok == p->tok) + return(REWIND_REWIND); + if (MDOC_BODY == p->type && MDOC_Sh == p->tok) + return(REWIND_HALT); + break; + + /* Multi-line explicit scope start. */ + case (MDOC_Ao): + /* FALLTHROUGH */ + case (MDOC_Bd): + /* FALLTHROUGH */ + case (MDOC_Bf): + /* FALLTHROUGH */ + case (MDOC_Bk): + /* FALLTHROUGH */ + case (MDOC_Bl): + /* FALLTHROUGH */ + case (MDOC_Bo): + /* FALLTHROUGH */ + case (MDOC_Do): + /* FALLTHROUGH */ + case (MDOC_Eo): + /* FALLTHROUGH */ + case (MDOC_Fo): + /* FALLTHROUGH */ + case (MDOC_Oo): + /* FALLTHROUGH */ + case (MDOC_Po): + /* FALLTHROUGH */ + case (MDOC_Qo): + /* FALLTHROUGH */ + case (MDOC_Rs): + /* FALLTHROUGH */ + case (MDOC_So): + /* FALLTHROUGH */ + case (MDOC_Xo): + if (type == p->type && tok == p->tok) + return(REWIND_REWIND); + break; + + /* Multi-line explicit scope close. */ + case (MDOC_Ac): + /* FALLTHROUGH */ + case (MDOC_Bc): + /* FALLTHROUGH */ + case (MDOC_Dc): + /* FALLTHROUGH */ + case (MDOC_Ec): + /* FALLTHROUGH */ + case (MDOC_Ed): + /* FALLTHROUGH */ + case (MDOC_Ek): + /* FALLTHROUGH */ + case (MDOC_El): + /* FALLTHROUGH */ + case (MDOC_Fc): + /* FALLTHROUGH */ + case (MDOC_Ef): + /* FALLTHROUGH */ + case (MDOC_Oc): + /* FALLTHROUGH */ + case (MDOC_Pc): + /* FALLTHROUGH */ + case (MDOC_Qc): + /* FALLTHROUGH */ + case (MDOC_Re): + /* FALLTHROUGH */ + case (MDOC_Sc): + /* FALLTHROUGH */ + case (MDOC_Xc): + if (type == p->type && rewind_alt(tok) == p->tok) + return(REWIND_REWIND); + break; + default: + abort(); + /* NOTREACHED */ + } + + return(REWIND_NOHALT); +} + + +static int +rewind_dobreak(int tok, const struct mdoc_node *p) +{ + + assert(MDOC_ROOT != p->type); + if (MDOC_ELEM == p->type) return(1); + if (MDOC_TEXT == p->type) + return(1); + if (MDOC_VALID & p->flags) + return(1); + + switch (tok) { + /* Implicit rules. */ + case (MDOC_It): + return(MDOC_It == p->tok); + case (MDOC_Ss): + return(MDOC_Ss == p->tok); + case (MDOC_Sh): + if (MDOC_Ss == p->tok) + return(1); + return(MDOC_Sh == p->tok); + + /* Extra scope rules. */ + case (MDOC_El): + if (MDOC_It == p->tok) + return(1); + break; + default: + break; } - do { - mdoc->last = mdoc->last->parent; - assert(mdoc->last); - if ( ! mdoc_valid_post(mdoc)) - return(0); - if ( ! mdoc_action_post(mdoc)) - return(0); - } while (mdoc->last != to); + if (MDOC_EXPLICIT & mdoc_macros[tok].flags) + return(p->tok == rewind_alt(tok)); + else if (MDOC_BLOCK == p->type) + return(1); - return(1); + return(tok == p->tok); } @@ -100,22 +363,22 @@ rewind_elem(struct mdoc *mdoc, int tok) static int -rewind_body(struct mdoc *mdoc, int tok) +rewind_subblock(enum mdoc_type type, struct mdoc *mdoc, + int tok, int line, int ppos) { struct mdoc_node *n; - int t; + int c; - assert(mdoc->last); - /* LINTED */ for (n = mdoc->last; n; n = n->parent) { - if (MDOC_BODY != n->type) - continue; - if (tok == (t = n->tok)) + c = rewind_dohalt(tok, type, n); + if (REWIND_HALT == c) + return(1); + if (REWIND_REWIND == c) break; - if ( ! (MDOC_EXPLICIT & mdoc_macros[t].flags)) + else if (rewind_dobreak(tok, n)) continue; - return(mdoc_verr(mdoc, n, ERR_SCOPE_BREAK)); + return(mdoc_perr(mdoc, line, ppos, "scope breaks prior %s", mdoc_node2a(n))); } assert(n); @@ -124,22 +387,21 @@ rewind_body(struct mdoc *mdoc, int tok) static int -rewind_head(struct mdoc *mdoc, int tok) +rewind_expblock(struct mdoc *mdoc, int tok, int line, int ppos) { struct mdoc_node *n; - int t; + int c; - assert(mdoc->last); - /* LINTED */ for (n = mdoc->last; n; n = n->parent) { - if (MDOC_HEAD != n->type) - continue; - if (tok == (t = n->tok)) + c = rewind_dohalt(tok, MDOC_BLOCK, n); + if (REWIND_HALT == c) + return(mdoc_perr(mdoc, line, ppos, "closing macro has no context")); + if (REWIND_REWIND == c) break; - if ( ! (MDOC_EXPLICIT & mdoc_macros[t].flags)) + else if (rewind_dobreak(tok, n)) continue; - return(mdoc_verr(mdoc, n, ERR_SCOPE_BREAK)); + return(mdoc_perr(mdoc, line, ppos, "scope breaks prior %s", mdoc_node2a(n))); } assert(n); @@ -148,22 +410,21 @@ rewind_head(struct mdoc *mdoc, int tok) static int -rewind_expblock(struct mdoc *mdoc, int tok) +rewind_impblock(struct mdoc *mdoc, int tok, int line, int ppos) { struct mdoc_node *n; - int t; + int c; - n = mdoc->last ? mdoc->last->parent : NULL; - /* LINTED */ - for ( ; n; n = n->parent) { - if (MDOC_BLOCK != n->type) - continue; - if (tok == (t = n->tok)) + for (n = mdoc->last; n; n = n->parent) { + c = rewind_dohalt(tok, MDOC_BLOCK, n); + if (REWIND_HALT == c) + return(1); + else if (REWIND_REWIND == c) break; - if (MDOC_NESTED & mdoc_macros[t].flags) + else if (rewind_dobreak(tok, n)) continue; - return(mdoc_verr(mdoc, n, ERR_SCOPE_BREAK)); + return(mdoc_perr(mdoc, line, ppos, "scope breaks prior %s", mdoc_node2a(n))); } assert(n); @@ -172,36 +433,8 @@ rewind_expblock(struct mdoc *mdoc, int tok) static int -rewind_impblock(struct mdoc *mdoc, int tok) +append_delims(struct mdoc *mdoc, int line, int *pos, char *buf) { - int t; - struct mdoc_node *n; - - n = mdoc->last ? mdoc->last->parent : NULL; - - /* LINTED */ - for ( ; n; n = n->parent) { - if (MDOC_BLOCK != n->type) - continue; - if (tok == (t = n->tok)) - break; - if ( ! (MDOC_EXPLICIT & mdoc_macros[t].flags)) - continue; - if (MDOC_NESTED & mdoc_macros[tok].flags) - return(1); - return(mdoc_verr(mdoc, n, ERR_SCOPE_BREAK)); - } - - if (NULL == n) - return(1); - return(rewind_last(mdoc, n)); -} - - -static int -append_delims(struct mdoc *mdoc, int tok, - int line, int *pos, char *buf) -{ int c, lastarg; char *p; @@ -225,66 +458,18 @@ append_delims(struct mdoc *mdoc, int tok, } -/* ARGSUSED */ +/* + * Close out an explicit scope. This optionally parses a TAIL type with + * a set number of TEXT children. + */ int -macro_close_explicit(MACRO_PROT_ARGS) +macro_scoped_close(MACRO_PROT_ARGS) { int tt, j, c, lastarg, maxargs, flushed; char *p; switch (tok) { - case (MDOC_Ac): - tt = MDOC_Ao; - break; - case (MDOC_Bc): - tt = MDOC_Bo; - break; - case (MDOC_Dc): - tt = MDOC_Do; - break; case (MDOC_Ec): - tt = MDOC_Eo; - break; - case (MDOC_Ed): - tt = MDOC_Bd; - break; - case (MDOC_Ef): - tt = MDOC_Bf; - break; - case (MDOC_Ek): - tt = MDOC_Bk; - break; - case (MDOC_El): - tt = MDOC_Bl; - break; - case (MDOC_Fc): - tt = MDOC_Fo; - break; - case (MDOC_Oc): - tt = MDOC_Oo; - break; - case (MDOC_Pc): - tt = MDOC_Po; - break; - case (MDOC_Qc): - tt = MDOC_Qo; - break; - case (MDOC_Re): - tt = MDOC_Rs; - break; - case (MDOC_Sc): - tt = MDOC_So; - break; - case (MDOC_Xc): - tt = MDOC_Xo; - break; - default: - abort(); - /* NOTREACHED */ - } - - switch (tok) { - case (MDOC_Ec): maxargs = 1; break; default: @@ -292,13 +477,21 @@ macro_close_explicit(MACRO_PROT_ARGS) break; } + tt = rewind_alt(tok); + + mdoc_msg(mdoc, "parse: %s closing %s", + mdoc_macronames[tok], mdoc_macronames[tt]); + if ( ! (MDOC_CALLABLE & mdoc_macros[tok].flags)) { - if (0 == buf[*pos]) - return(rewind_expblock(mdoc, tt)); - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_EQ0)); + if (0 == buf[*pos]) { + if ( ! rewind_subblock(MDOC_BODY, mdoc, tok, line, ppos)) + return(0); + return(rewind_expblock(mdoc, tok, line, ppos)); + } + return(mdoc_perr(mdoc, line, ppos, "macro expects no parameters")); } - if ( ! rewind_body(mdoc, tt)) + if ( ! rewind_subblock(MDOC_BODY, mdoc, tok, line, ppos)) return(0); lastarg = ppos; @@ -310,16 +503,16 @@ macro_close_explicit(MACRO_PROT_ARGS) mdoc->next = MDOC_NEXT_CHILD; } - for (j = 0; j < MDOC_LINEARG_MAX; j++) { + for (j = 0; /* No sentinel. */; j++) { lastarg = *pos; if (j == maxargs && ! flushed) { - if ( ! rewind_expblock(mdoc, tt)) + if ( ! rewind_expblock(mdoc, tok, line, ppos)) return(0); flushed = 1; } - c = mdoc_args(mdoc, line, pos, buf, ARGS_DELIM, &p); + c = mdoc_args(mdoc, line, pos, buf, tok, &p); if (ARGS_ERROR == c) return(0); if (ARGS_PUNCT == c) @@ -331,7 +524,7 @@ macro_close_explicit(MACRO_PROT_ARGS) return(0); else if (MDOC_MAX != c) { if ( ! flushed) { - if ( ! rewind_expblock(mdoc, tt)) + if ( ! rewind_expblock(mdoc, tok, line, ppos)) return(0); flushed = 1; } @@ -345,31 +538,43 @@ macro_close_explicit(MACRO_PROT_ARGS) mdoc->next = MDOC_NEXT_SIBLING; } - if (MDOC_LINEARG_MAX == j) - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_MANY)); - - if ( ! flushed && ! rewind_expblock(mdoc, tt)) + if ( ! flushed && ! rewind_expblock(mdoc, tok, line, ppos)) return(0); if (ppos > 1) return(1); - return(append_delims(mdoc, tok, line, pos, buf)); + return(append_delims(mdoc, line, pos, buf)); } /* - * A general text domain macro. When invoked, this opens a scope that - * accepts words until either end-of-line, only-punctuation, or a - * callable macro. If the word is punctuation (not only-punctuation), - * then the scope is closed out, the punctuation appended, then the - * scope opened again. If any terminating conditions are met, the scope - * is closed out. If this is the first macro in the line and - * only-punctuation remains, this punctuation is flushed. + * A general text macro. This is a complex case because of punctuation. + * If a text macro is followed by words, then punctuation, the macro is + * "stopped" and "reopened" following the punctuation. Thus, the + * following arises: + * + * .Fl a ; b + * + * ELEMENT (.Fl) + * TEXT (`a') + * TEXT (`;') + * ELEMENT (.Fl) + * TEXT (`b') + * + * This must handle the following situations: + * + * .Fl Ar b ; ; + * + * ELEMENT (.Fl) + * ELEMENT (.Ar) + * TEXT (`b') + * TEXT (`;') + * TEXT (`;') */ int macro_text(MACRO_PROT_ARGS) { - int la, lastpunct, c, sz, fl, argc; + int la, lastpunct, c, w, argc; struct mdoc_arg argv[MDOC_LINEARG_MAX]; char *p; @@ -378,22 +583,25 @@ macro_text(MACRO_PROT_ARGS) for (argc = 0; argc < MDOC_LINEARG_MAX; argc++) { la = *pos; - c = mdoc_argv(mdoc, line, tok, &argv[argc], pos, buf); - if (ARGV_EOLN == c || ARGV_WORD == c) + if (ARGV_EOLN == c) break; - else if (ARGV_ARG == c) + if (ARGV_WORD == c) { + *pos = la; + break; + } else if (ARGV_ARG == c) continue; + mdoc_argv_free(argc, argv); return(0); } if (MDOC_LINEARG_MAX == argc) { - mdoc_argv_free(argc, argv); - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_MANY)); + mdoc_argv_free(argc - 1, argv); + return(mdoc_perr(mdoc, line, ppos, "parameter hard-limit exceeded")); } - c = mdoc_elem_alloc(mdoc, line, la, tok, argc, argv); + c = mdoc_elem_alloc(mdoc, line, ppos, tok, argc, argv); if (0 == c) { mdoc_argv_free(argc, argv); @@ -402,59 +610,57 @@ macro_text(MACRO_PROT_ARGS) mdoc->next = MDOC_NEXT_CHILD; - fl = ARGS_DELIM; - if (MDOC_QUOTABLE & mdoc_macros[tok].flags) - fl |= ARGS_QUOTED; - - for (lastpunct = sz = 0; sz + argc < MDOC_LINEARG_MAX; sz++) { + lastpunct = 0; + for (;;) { la = *pos; - - if (lastpunct) { - c = mdoc_elem_alloc(mdoc, line, - la, tok, argc, argv); - if (0 == c) { - mdoc_argv_free(argc, argv); - return(0); - } - mdoc->next = MDOC_NEXT_CHILD; - lastpunct = 0; - } - - c = mdoc_args(mdoc, line, pos, buf, fl, &p); - if (ARGS_ERROR == c) { + w = mdoc_args(mdoc, line, pos, buf, tok, &p); + if (ARGS_ERROR == w) { mdoc_argv_free(argc, argv); return(0); } - if (ARGS_EOLN == c) + if (ARGS_EOLN == w) break; - if (ARGS_PUNCT == c) + if (ARGS_PUNCT == w) break; - if (-1 == (c = lookup(mdoc, line, la, tok, p))) - return(0); - else if (MDOC_MAX != c) { - if ( ! rewind_elem(mdoc, tok)) { + c = ARGS_QWORD == w ? MDOC_MAX : + lookup(mdoc, line, la, tok, p); + + if (MDOC_MAX != c && -1 != c) { + if (0 == lastpunct && ! rewind_elem(mdoc, tok)) { mdoc_argv_free(argc, argv); return(0); } mdoc_argv_free(argc, argv); - c = mdoc_macro(mdoc, c, line, la, pos, buf); if (0 == c) return(0); if (ppos > 1) return(1); - return(append_delims(mdoc, tok, line, pos, buf)); + return(append_delims(mdoc, line, pos, buf)); + } else if (-1 == c) { + mdoc_argv_free(argc, argv); + return(0); } - if (mdoc_isdelim(p)) { - if ( ! rewind_elem(mdoc, tok)) { + if (ARGS_QWORD != w && mdoc_isdelim(p)) { + if (0 == lastpunct && ! rewind_elem(mdoc, tok)) { mdoc_argv_free(argc, argv); return(0); } lastpunct = 1; + } else if (lastpunct) { + c = mdoc_elem_alloc(mdoc, line, + ppos, tok, argc, argv); + if (0 == c) { + mdoc_argv_free(argc, argv); + return(0); + } + mdoc->next = MDOC_NEXT_CHILD; + lastpunct = 0; } + if ( ! mdoc_word_alloc(mdoc, line, la, p)) return(0); mdoc->next = MDOC_NEXT_SIBLING; @@ -462,47 +668,75 @@ macro_text(MACRO_PROT_ARGS) mdoc_argv_free(argc, argv); - if (sz == MDOC_LINEARG_MAX) - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_MANY)); - - if ( ! rewind_elem(mdoc, tok)) + if (0 == lastpunct && ! rewind_elem(mdoc, tok)) return(0); if (ppos > 1) return(1); - return(append_delims(mdoc, tok, line, pos, buf)); + return(append_delims(mdoc, line, pos, buf)); } /* - * Implicit- or explicit-end multi-line scoped macro. + * Handle explicit-scope (having a different closure token) and implicit + * scope (closing out prior scopes when re-invoked) macros. These + * constitute the BLOCK type and usually span multiple lines. These + * always have HEAD and sometimes have BODY types. In the multi-line + * case: + * + * .Bd -ragged + * Text. + * .Fl macro + * Another. + * .Ed + * + * BLOCK (.Bd) + * HEAD + * BODY + * TEXT (`Text.') + * ELEMENT (.Fl) + * TEXT (`macro') + * TEXT (`Another.') + * + * Note that the `.It' macro, possibly the most difficult (as it has + * embedded scope, etc.) is handled by this routine. */ int macro_scoped(MACRO_PROT_ARGS) { - int c, lastarg, argc, j; + int c, lastarg, argc; struct mdoc_arg argv[MDOC_LINEARG_MAX]; char *p; assert ( ! (MDOC_CALLABLE & mdoc_macros[tok].flags)); - if ( ! (MDOC_EXPLICIT & mdoc_macros[tok].flags)) - if ( ! rewind_impblock(mdoc, tok)) + /* First rewind extant implicit scope. */ + + if ( ! (MDOC_EXPLICIT & mdoc_macros[tok].flags)) { + if ( ! rewind_subblock(MDOC_BODY, mdoc, tok, line, ppos)) return(0); + if ( ! rewind_impblock(mdoc, tok, line, ppos)) + return(0); + } + /* Parse arguments. */ + for (argc = 0; argc < MDOC_LINEARG_MAX; argc++) { lastarg = *pos; c = mdoc_argv(mdoc, line, tok, &argv[argc], pos, buf); - if (ARGV_EOLN == c || ARGV_WORD == c) + if (ARGV_EOLN == c) break; - else if (ARGV_ARG == c) + if (ARGV_WORD == c) { + *pos = lastarg; + break; + } else if (ARGV_ARG == c) continue; mdoc_argv_free(argc, argv); return(0); } if (MDOC_LINEARG_MAX == argc) { - mdoc_argv_free(argc, argv); - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_MANY)); + mdoc_argv_free(argc - 1, argv); + return(mdoc_perr(mdoc, line, ppos, "parameter hard-limit exceeded")); } c = mdoc_block_alloc(mdoc, line, ppos, @@ -517,7 +751,7 @@ macro_scoped(MACRO_PROT_ARGS) if (0 == buf[*pos]) { if ( ! mdoc_head_alloc(mdoc, line, ppos, tok)) return(0); - if ( ! rewind_head(mdoc, tok)) + if ( ! rewind_subblock(MDOC_HEAD, mdoc, tok, line, ppos)) return(0); if ( ! mdoc_body_alloc(mdoc, line, ppos, tok)) return(0); @@ -529,9 +763,9 @@ macro_scoped(MACRO_PROT_ARGS) return(0); mdoc->next = MDOC_NEXT_CHILD; - for (j = 0; j < MDOC_LINEARG_MAX; j++) { + for (;;) { lastarg = *pos; - c = mdoc_args(mdoc, line, pos, buf, ARGS_DELIM, &p); + c = mdoc_args(mdoc, line, pos, buf, tok, &p); if (ARGS_ERROR == c) return(0); @@ -539,10 +773,14 @@ macro_scoped(MACRO_PROT_ARGS) break; if (ARGS_EOLN == c) break; + + /* FIXME: if .It -column, the lookup must be for a + * sub-line component. BLAH. */ if (-1 == (c = lookup(mdoc, line, lastarg, tok, p))) return(0); - else if (MDOC_MAX == c) { + + if (MDOC_MAX == c) { if ( ! mdoc_word_alloc(mdoc, line, lastarg, p)) return(0); mdoc->next = MDOC_NEXT_SIBLING; @@ -553,13 +791,10 @@ macro_scoped(MACRO_PROT_ARGS) return(0); break; } - - if (j == MDOC_LINEARG_MAX) - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_MANY)); - - if ( ! rewind_head(mdoc, tok)) + + if ( ! rewind_subblock(MDOC_HEAD, mdoc, tok, line, ppos)) return(0); - if (1 == ppos && ! append_delims(mdoc, tok, line, pos, buf)) + if (1 == ppos && ! append_delims(mdoc, line, pos, buf)) return(0); if ( ! mdoc_body_alloc(mdoc, line, ppos, tok)) @@ -571,14 +806,28 @@ macro_scoped(MACRO_PROT_ARGS) /* - * When scoped to a line, a macro encompasses all of the contents. This - * differs from constants or text macros, where a new macro will - * terminate the existing context. + * This handles a case of implicitly-scoped macro (BLOCK) limited to a + * single line. Instead of being closed out by a subsequent call to + * another macro, the scope is closed at the end of line. These don't + * have BODY or TAIL types. Notice that the punctuation falls outside + * of the HEAD type. + * + * .Qq a Fl b Ar d ; ; + * + * BLOCK (Qq) + * HEAD + * TEXT (`a') + * ELEMENT (.Fl) + * TEXT (`b') + * ELEMENT (.Ar) + * TEXT (`d') + * TEXT (`;') + * TEXT (`;') */ int macro_scoped_line(MACRO_PROT_ARGS) { - int lastarg, c, j; + int lastarg, c; char *p; if ( ! mdoc_block_alloc(mdoc, line, ppos, tok, 0, NULL)) @@ -587,13 +836,17 @@ macro_scoped_line(MACRO_PROT_ARGS) if ( ! mdoc_head_alloc(mdoc, line, ppos, tok)) return(0); + mdoc->next = MDOC_NEXT_SIBLING; + if ( ! mdoc_body_alloc(mdoc, line, ppos, tok)) + return(0); mdoc->next = MDOC_NEXT_CHILD; /* XXX - no known argument macros. */ - for (lastarg = ppos, j = 0; j < MDOC_LINEARG_MAX; j++) { + lastarg = ppos; + for (;;) { lastarg = *pos; - c = mdoc_args(mdoc, line, pos, buf, ARGS_DELIM, &p); + c = mdoc_args(mdoc, line, pos, buf, tok, &p); if (ARGS_ERROR == c) return(0); @@ -616,22 +869,31 @@ macro_scoped_line(MACRO_PROT_ARGS) break; } - if (j == MDOC_LINEARG_MAX) - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_MANY)); - if (1 == ppos) { - if ( ! rewind_head(mdoc, tok)) + if ( ! rewind_subblock(MDOC_BODY, mdoc, tok, line, ppos)) return(0); - if ( ! append_delims(mdoc, tok, line, pos, buf)) + if ( ! append_delims(mdoc, line, pos, buf)) return(0); - } - return(rewind_impblock(mdoc, tok)); + } else if ( ! rewind_subblock(MDOC_BODY, mdoc, tok, line, ppos)) + return(0); + return(rewind_impblock(mdoc, tok, line, ppos)); } /* - * Constant-scope macros accept a fixed number of arguments and behave - * like constant macros except that they're scoped across lines. + * A constant-scoped macro is like a simple-scoped macro (mdoc_scoped) + * except that it doesn't handle implicit scopes and explicit ones have + * a fixed number of TEXT children to the BODY. + * + * .Fl a So b Sc ; + * + * ELEMENT (.Fl) + * TEXT (`a') + * BLOCK (.So) + * HEAD + * BODY + * TEXT (`b') + * TEXT (';') */ int macro_constant_scoped(MACRO_PROT_ARGS) @@ -658,7 +920,7 @@ macro_constant_scoped(MACRO_PROT_ARGS) if (0 == maxargs) { if ( ! mdoc_head_alloc(mdoc, line, ppos, tok)) return(0); - if ( ! rewind_head(mdoc, tok)) + if ( ! rewind_subblock(MDOC_HEAD, mdoc, tok, line, ppos)) return(0); if ( ! mdoc_body_alloc(mdoc, line, ppos, tok)) return(0); @@ -668,11 +930,11 @@ macro_constant_scoped(MACRO_PROT_ARGS) mdoc->next = MDOC_NEXT_CHILD; - for (j = 0; j < MDOC_LINEARG_MAX; j++) { + for (j = 0; /* No sentinel. */; j++) { lastarg = *pos; if (j == maxargs && ! flushed) { - if ( ! rewind_head(mdoc, tok)) + if ( ! rewind_subblock(MDOC_HEAD, mdoc, tok, line, ppos)) return(0); flushed = 1; if ( ! mdoc_body_alloc(mdoc, line, ppos, tok)) @@ -680,7 +942,7 @@ macro_constant_scoped(MACRO_PROT_ARGS) mdoc->next = MDOC_NEXT_CHILD; } - c = mdoc_args(mdoc, line, pos, buf, ARGS_DELIM, &p); + c = mdoc_args(mdoc, line, pos, buf, tok, &p); if (ARGS_ERROR == c) return(0); if (ARGS_PUNCT == c) @@ -692,7 +954,7 @@ macro_constant_scoped(MACRO_PROT_ARGS) return(0); else if (MDOC_MAX != c) { if ( ! flushed) { - if ( ! rewind_head(mdoc, tok)) + if ( ! rewind_subblock(MDOC_HEAD, mdoc, tok, line, ppos)) return(0); flushed = 1; if ( ! mdoc_body_alloc(mdoc, line, ppos, tok)) @@ -705,7 +967,7 @@ macro_constant_scoped(MACRO_PROT_ARGS) } if ( ! flushed && mdoc_isdelim(p)) { - if ( ! rewind_head(mdoc, tok)) + if ( ! rewind_subblock(MDOC_HEAD, mdoc, tok, line, ppos)) return(0); flushed = 1; if ( ! mdoc_body_alloc(mdoc, line, ppos, tok)) @@ -718,11 +980,8 @@ macro_constant_scoped(MACRO_PROT_ARGS) mdoc->next = MDOC_NEXT_SIBLING; } - if (MDOC_LINEARG_MAX == j) - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_MANY)); - if ( ! flushed) { - if ( ! rewind_head(mdoc, tok)) + if ( ! rewind_subblock(MDOC_HEAD, mdoc, tok, line, ppos)) return(0); if ( ! mdoc_body_alloc(mdoc, line, ppos, tok)) return(0); @@ -731,15 +990,22 @@ macro_constant_scoped(MACRO_PROT_ARGS) if (ppos > 1) return(1); - return(append_delims(mdoc, tok, line, pos, buf)); + return(append_delims(mdoc, line, pos, buf)); } /* - * Delimited macros are like text macros except that, should punctuation - * be encountered, the macro isn't re-started with remaining tokens - * (it's only emitted once). Delimited macros can have a maximum number - * of arguments. + * A delimited constant is very similar to the macros parsed by + * macro_text except that, in the event of punctuation, the macro isn't + * "re-opened" as it is in macro_text. Also, these macros have a fixed + * number of parameters. + * + * .Fl a No b + * + * ELEMENT (.Fl) + * TEXT (`a') + * ELEMENT (.No) + * TEXT (`b') */ int macro_constant_delimited(MACRO_PROT_ARGS) @@ -756,6 +1022,8 @@ macro_constant_delimited(MACRO_PROT_ARGS) /* FALLTHROUGH */ case (MDOC_Ns): /* FALLTHROUGH */ + case (MDOC_Pf): + /* FALLTHROUGH */ case (MDOC_Ux): /* FALLTHROUGH */ case (MDOC_St): @@ -769,15 +1037,23 @@ macro_constant_delimited(MACRO_PROT_ARGS) for (argc = 0; argc < MDOC_LINEARG_MAX; argc++) { lastarg = *pos; c = mdoc_argv(mdoc, line, tok, &argv[argc], pos, buf); - if (ARGV_EOLN == c || ARGV_WORD == c) + if (ARGV_EOLN == c) break; - else if (ARGV_ARG == c) + if (ARGV_WORD == c) { + *pos = lastarg; + break; + } else if (ARGV_ARG == c) continue; mdoc_argv_free(argc, argv); return(0); } - c = mdoc_elem_alloc(mdoc, line, lastarg, tok, argc, argv); + if (MDOC_LINEARG_MAX == argc) { + mdoc_argv_free(argc - 1, argv); + return(mdoc_perr(mdoc, line, ppos, "parameter hard-limit exceeded")); + } + + c = mdoc_elem_alloc(mdoc, line, ppos, tok, argc, argv); mdoc_argv_free(argc, argv); if (0 == c) @@ -785,7 +1061,7 @@ macro_constant_delimited(MACRO_PROT_ARGS) mdoc->next = MDOC_NEXT_CHILD; - for (j = 0; j < MDOC_LINEARG_MAX; j++) { + for (j = 0; /* No sentinel. */; j++) { lastarg = *pos; if (j == maxargs && ! flushed) { @@ -794,7 +1070,7 @@ macro_constant_delimited(MACRO_PROT_ARGS) flushed = 1; } - c = mdoc_args(mdoc, line, pos, buf, ARGS_DELIM, &p); + c = mdoc_args(mdoc, line, pos, buf, tok, &p); if (ARGS_ERROR == c) return(0); if (ARGS_PUNCT == c) @@ -824,49 +1100,48 @@ macro_constant_delimited(MACRO_PROT_ARGS) mdoc->next = MDOC_NEXT_SIBLING; } - if (MDOC_LINEARG_MAX == j) - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_MANY)); - - if ( ! flushed && rewind_elem(mdoc, tok)) + if ( ! flushed && ! rewind_elem(mdoc, tok)) return(0); if (ppos > 1) return(1); - return(append_delims(mdoc, tok, line, pos, buf)); + return(append_delims(mdoc, line, pos, buf)); } /* - * Constant macros span an entire line: they constitute a macro and all - * of its arguments and child data. + * A constant macro is the simplest classification. It spans an entire + * line. */ int macro_constant(MACRO_PROT_ARGS) { - int c, lastarg, argc, sz, fl; - struct mdoc_arg argv[MDOC_LINEARG_MAX]; - char *p; + int c, w, la, argc; + struct mdoc_arg argv[MDOC_LINEARG_MAX]; + char *p; - /* FIXME: parsing macros! */ + assert( ! (MDOC_CALLABLE & mdoc_macros[tok].flags)); - fl = 0; - if (MDOC_QUOTABLE & mdoc_macros[tok].flags) - fl = ARGS_QUOTED; - for (argc = 0; argc < MDOC_LINEARG_MAX; argc++) { - lastarg = *pos; + la = *pos; c = mdoc_argv(mdoc, line, tok, &argv[argc], pos, buf); if (ARGV_EOLN == c) break; - else if (ARGV_ARG == c) - continue; - else if (ARGV_WORD == c) + if (ARGV_WORD == c) { + *pos = la; break; + } else if (ARGV_ARG == c) + continue; mdoc_argv_free(argc, argv); return(0); } + if (MDOC_LINEARG_MAX == argc) { + mdoc_argv_free(argc - 1, argv); + return(mdoc_perr(mdoc, line, ppos, "parameter hard-limit exceeded")); + } + c = mdoc_elem_alloc(mdoc, line, ppos, tok, argc, argv); mdoc_argv_free(argc, argv); @@ -875,25 +1150,29 @@ macro_constant(MACRO_PROT_ARGS) mdoc->next = MDOC_NEXT_CHILD; - if (MDOC_LINEARG_MAX == argc) - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_MANY)); - - for (sz = 0; sz + argc < MDOC_LINEARG_MAX; sz++) { - lastarg = *pos; - c = mdoc_args(mdoc, line, pos, buf, fl, &p); - if (ARGS_ERROR == c) + for (;;) { + la = *pos; + w = mdoc_args(mdoc, line, pos, buf, tok, &p); + if (ARGS_ERROR == w) return(0); - if (ARGS_EOLN == c) + if (ARGS_EOLN == w) break; - if ( ! mdoc_word_alloc(mdoc, line, lastarg, p)) + c = ARGS_QWORD == w ? MDOC_MAX : + lookup(mdoc, line, la, tok, p); + + if (MDOC_MAX != c && -1 != c) { + if ( ! rewind_elem(mdoc, tok)) + return(0); + return(mdoc_macro(mdoc, c, line, la, pos, buf)); + } else if (-1 == c) return(0); + + if ( ! mdoc_word_alloc(mdoc, line, la, p)) + return(0); mdoc->next = MDOC_NEXT_SIBLING; } - if (MDOC_LINEARG_MAX == sz + argc) - return(mdoc_perr(mdoc, line, ppos, ERR_ARGS_MANY)); - return(rewind_elem(mdoc, tok)); } @@ -903,15 +1182,35 @@ int macro_obsolete(MACRO_PROT_ARGS) { - return(mdoc_pwarn(mdoc, line, ppos, WARN_IGN_OBSOLETE)); + return(mdoc_pwarn(mdoc, line, ppos, WARN_SYNTAX, "macro is obsolete")); } +/* + * This is called at the end of parsing. It must traverse up the tree, + * closing out open [implicit] scopes. Obviously, open explicit scopes + * are errors. + */ int macro_end(struct mdoc *mdoc) { + struct mdoc_node *n; assert(mdoc->first); assert(mdoc->last); + + /* Scan for open explicit scopes. */ + + n = MDOC_VALID & mdoc->last->flags ? + mdoc->last->parent : mdoc->last; + + for ( ; n; n = n->parent) { + if (MDOC_BLOCK != n->type) + continue; + if ( ! (MDOC_EXPLICIT & mdoc_macros[n->tok].flags)) + continue; + return(mdoc_nerr(mdoc, n, "macro scope still open on exit")); + } + return(rewind_last(mdoc, mdoc->first)); }