=================================================================== RCS file: /cvs/mandoc/Attic/argv.c,v retrieving revision 1.23 retrieving revision 1.24 diff -u -p -r1.23 -r1.24 --- mandoc/Attic/argv.c 2009/01/20 22:55:46 1.23 +++ mandoc/Attic/argv.c 2009/01/21 17:56:32 1.24 @@ -1,4 +1,4 @@ -/* $Id: argv.c,v 1.23 2009/01/20 22:55:46 kristaps Exp $ */ +/* $Id: argv.c,v 1.24 2009/01/21 17:56:32 kristaps Exp $ */ /* * Copyright (c) 2008 Kristaps Dzonsons * @@ -32,7 +32,9 @@ * strange punctuation rules dictated by groff. */ -/* FIXME: .It called with -column and quoted arguments. */ +#define ARGS_QUOTED (1 << 0) +#define ARGS_DELIM (1 << 1) +#define ARGS_TABSEP (1 << 2) static int lookup(int, const char *); static int parse(struct mdoc *, int, @@ -43,10 +45,10 @@ static int parse_multi(struct mdoc *, int, struct mdoc_arg *, int *, char *); static int postparse(struct mdoc *, int, const struct mdoc_arg *, int); +static int pwarn(struct mdoc *, int, int, int); -#define ARGS_QUOTED (1 << 0) -#define ARGS_DELIM (1 << 1) -#define ARGS_TABSEP (1 << 2) +#define WQUOTPARM (0) +#define WARGVPARM (1) static int mdoc_argflags[MDOC_MAX] = { 0, /* \" */ @@ -158,43 +160,87 @@ static int mdoc_argflags[MDOC_MAX] = { }; +static int +pwarn(struct mdoc *mdoc, int line, int pos, int code) +{ + int c; + + switch (code) { + case (WQUOTPARM): + c = mdoc_pwarn(mdoc, line, pos, WARN_SYNTAX, + "unexpected quoted parameter"); + break; + case (WARGVPARM): + c = mdoc_pwarn(mdoc, line, pos, WARN_SYNTAX, + "argument-like parameter"); + break; + default: + abort(); + /* NOTREACHED */ + } + return(c); +} + + int mdoc_args(struct mdoc *mdoc, int line, int *pos, char *buf, int tok, char **v) { - int i, fl; + int i, c, fl; + char *p, *pp; struct mdoc_node *n; - fl = 0 == tok ? 0 : mdoc_argflags[tok]; - if (MDOC_It == tok) { - n = mdoc->last->parent; - /* FIXME: scan for ARGS_TABSEP. */ - } + assert(*pos > 0); if (0 == buf[*pos]) return(ARGS_EOLN); + fl = (0 == tok) ? 0 : mdoc_argflags[tok]; + if ('\"' == buf[*pos] && ! (fl & ARGS_QUOTED)) - if ( ! mdoc_pwarn(mdoc, line, *pos, WARN_SYNTAX, "unexpected quoted parameter")) + if ( ! pwarn(mdoc, line, *pos, WQUOTPARM)) return(ARGS_ERROR); if ('-' == buf[*pos]) - if ( ! mdoc_pwarn(mdoc, line, *pos, WARN_SYNTAX, "argument-like parameter")) + if ( ! pwarn(mdoc, line, *pos, WARGVPARM)) return(ARGS_ERROR); + /* + * First see if we should use TABSEP (Bl -column). This + * invalidates the use of ARGS_DELIM. + */ + + if (MDOC_It == tok) { + for (n = mdoc->last; n; n = n->parent) + if (MDOC_BLOCK == n->type) + if (MDOC_Bl == n->tok) + break; + assert(n); + c = (int)n->data.block.argc; + assert(c > 0); + for (i = 0; i < c; i++) { + if (MDOC_Column != n->data.block.argv[i].arg) + continue; + fl |= ARGS_TABSEP; + fl &= ~ARGS_DELIM; + } + } + + /* + * If the first character is a delimiter and we're to look for + * delimited strings, then pass down the buffer seeing if it + * follows the pattern of [[::delim::][ ]+]+. + */ + if ((fl & ARGS_DELIM) && mdoc_iscdelim(buf[*pos])) { - /* - * If ARGS_DELIM, return ARGS_PUNCT if only space-separated - * punctuation remains. - */ - for (i = *pos; buf[i]; ) { - if ( ! mdoc_iscdelim(buf[i])) + for (i = *pos; (c = buf[i]); ) { + if ( ! mdoc_iscdelim(c)) break; i++; - if (0 == buf[i] || ! isspace((int)buf[i])) + if (0 == buf[i] || ! isspace(c)) break; i++; - while (buf[i] && isspace((int)buf[i])) + while (buf[i] && isspace(c)) i++; } if (0 == buf[i]) { @@ -203,37 +249,105 @@ mdoc_args(struct mdoc *mdoc, int line, } } - /* Parse routine for non-quoted string. */ + /* First parse non-quoted strings. */ - assert(*pos > 0); if ('\"' != buf[*pos] || ! (ARGS_QUOTED & fl)) { *v = &buf[*pos]; - /* FIXME: UGLY tab-sep processing. */ + /* + * Thar be dragons here! If we're tab-separated, search + * ahead for either a tab or the `Ta' macro. If a tab + * is detected, it mustn't be escaped; if a `Ta' is + * detected, it must be space-buffered before and after. + * If either of these hold true, then prune out the + * extra spaces and call it an argument. + */ - if (ARGS_TABSEP & fl) - while (buf[*pos]) { - if ('\t' == buf[*pos]) + if (ARGS_TABSEP & fl) { + /* Scan ahead to unescaped tab. */ + + for (p = *v; ; p++) { + if (NULL == (p = strchr(p, '\t'))) break; - if ('T' == buf[*pos]) { - (*pos)++; - if (0 == buf[*pos]) - break; - if ('a' == buf[*pos]) { - buf[*pos - 1] = 0; - break; - } - } - (*pos)++; + if (p == *v) + break; + if ('\\' != *(p - 1)) + break; } - else { - while (buf[*pos]) { - if (isspace((int)buf[*pos])) + + /* Scan ahead to unescaped `Ta'. */ + + for (pp = *v; ; pp++) { + if (NULL == (pp = strstr(pp, "Ta"))) + break; + if (pp > *v && ' ' != *(pp - 1)) + continue; + if (' ' == *(pp + 2) || 0 == *(pp + 2)) + break; + } + + /* Choose delimiter tab/Ta. */ + + if (p && pp) + p = (p < pp ? p : pp); + else if ( ! p && pp) + p = pp; + + /* Strip delimiter's preceding whitespace. */ + + if (p && p > *v) { + pp = p - 1; + while (pp > *v && ' ' == *pp) + pp--; + if (pp == *v && ' ' == *pp) + *pp = 0; + else if (' ' == *pp) + *(pp + 1) = 0; + } + + /* ...in- and proceding whitespace. */ + + if (p && ('\t' != *p)) { + *p++ = 0; + *p++ = 0; + } else if (p) + *p++ = 0; + + if (p) { + while (' ' == *p) + p++; + if (0 != *p) + *(p - 1) = 0; + else if (0 == *p) + if ( ! mdoc_pwarn(mdoc, line, *pos, WARN_SYNTAX, "empty final token")) /* FIXME: verbiage */ + return(0); + *pos += p - *v; + } + + /* Configure the eoln case, too. */ + + if (NULL == p) { + p = strchr(*v, 0); + assert(p); + + /*if (p > *v && ' ' == *(p - 1)) + Warn about whitespace. */ + + *pos += p - *v; + } + + return(ARGS_WORD); + } + + /* Do non-tabsep look-ahead here. */ + + if ( ! (ARGS_TABSEP & fl)) + while ((c = buf[*pos])) { + if (isspace(c)) if ('\\' != buf[*pos - 1]) break; (*pos)++; } - } if (0 == buf[*pos]) return(ARGS_WORD);