=================================================================== RCS file: /cvs/mandoc/roff.c,v retrieving revision 1.79 retrieving revision 1.85 diff -u -p -r1.79 -r1.85 --- mandoc/roff.c 2010/05/17 00:06:36 1.79 +++ mandoc/roff.c 2010/05/29 19:41:47 1.85 @@ -1,4 +1,4 @@ -/* $Id: roff.c,v 1.79 2010/05/17 00:06:36 kristaps Exp $ */ +/* $Id: roff.c,v 1.85 2010/05/29 19:41:47 kristaps Exp $ */ /* * Copyright (c) 2010 Kristaps Dzonsons * @@ -19,6 +19,7 @@ #endif #include +#include #include #include #include @@ -26,36 +27,43 @@ #include "mandoc.h" #include "roff.h" +#define RSTACK_MAX 128 + #define ROFF_CTL(c) \ ('.' == (c) || '\'' == (c)) enum rofft { - ROFF_if, - ROFF_ig, - ROFF_cblock, - ROFF_ccond, -#if 0 ROFF_am, ROFF_ami, + ROFF_am1, ROFF_de, ROFF_dei, - ROFF_ie, + ROFF_de1, + ROFF_ds, ROFF_el, -#endif + ROFF_ie, + ROFF_if, + ROFF_ig, + ROFF_rm, + ROFF_tr, + ROFF_cblock, + ROFF_ccond, ROFF_MAX }; +enum roffrule { + ROFFRULE_ALLOW, + ROFFRULE_DENY +}; + struct roff { struct roffnode *last; /* leaf of stack */ mandocmsg msg; /* err/warn/fatal messages */ void *data; /* privdata for messages */ + enum roffrule rstack[RSTACK_MAX]; /* stack of !`ie' rules */ + int rstackpos; /* position in rstack */ }; -enum roffrule { - ROFFRULE_ALLOW, - ROFFRULE_DENY -}; - struct roffnode { enum rofft tok; /* type of node */ struct roffnode *parent; /* up one in stack */ @@ -63,7 +71,7 @@ struct roffnode { int col; /* parse col */ char *end; /* end-rules: custom token */ int endspan; /* end-rules: next-line or infty */ - enum roffrule rule; + enum roffrule rule; /* current evaluation rule */ }; #define ROFF_ARGS struct roff *r, /* parse ctx */ \ @@ -84,33 +92,79 @@ struct roffmac { roffproc sub; /* process as child of macro */ int flags; #define ROFFMAC_STRUCT (1 << 0) /* always interpret */ + struct roffmac *next; }; -static enum rofferr roff_if(ROFF_ARGS); -static enum rofferr roff_if_text(ROFF_ARGS); -static enum rofferr roff_if_sub(ROFF_ARGS); -static enum rofferr roff_ig(ROFF_ARGS); -static enum rofferr roff_ig_text(ROFF_ARGS); -static enum rofferr roff_ig_sub(ROFF_ARGS); +static enum rofferr roff_block(ROFF_ARGS); +static enum rofferr roff_block_text(ROFF_ARGS); +static enum rofferr roff_block_sub(ROFF_ARGS); static enum rofferr roff_cblock(ROFF_ARGS); static enum rofferr roff_ccond(ROFF_ARGS); +static enum rofferr roff_cond(ROFF_ARGS); +static enum rofferr roff_cond_text(ROFF_ARGS); +static enum rofferr roff_cond_sub(ROFF_ARGS); +static enum rofferr roff_line(ROFF_ARGS); -const struct roffmac roffs[ROFF_MAX] = { - { "if", roff_if, roff_if_text, roff_if_sub, ROFFMAC_STRUCT }, - { "ig", roff_ig, roff_ig_text, roff_ig_sub, 0 }, - { ".", roff_cblock, NULL, NULL, 0 }, - { "\\}", roff_ccond, NULL, NULL, 0 }, +/* See roff_hash_find() */ + +#define ASCII_HI 126 +#define ASCII_LO 33 +#define HASHWIDTH (ASCII_HI - ASCII_LO + 1) + +static struct roffmac *hash[HASHWIDTH]; + +static struct roffmac roffs[ROFF_MAX] = { + { "am", roff_block, roff_block_text, roff_block_sub, 0, NULL }, + { "ami", roff_block, roff_block_text, roff_block_sub, 0, NULL }, + { "am1", roff_block, roff_block_text, roff_block_sub, 0, NULL }, + { "de", roff_block, roff_block_text, roff_block_sub, 0, NULL }, + { "dei", roff_block, roff_block_text, roff_block_sub, 0, NULL }, + { "de1", roff_block, roff_block_text, roff_block_sub, 0, NULL }, + { "ds", roff_line, NULL, NULL, 0, NULL }, + { "el", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL }, + { "ie", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL }, + { "if", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL }, + { "ig", roff_block, roff_block_text, roff_block_sub, 0, NULL }, + { "rm", roff_line, NULL, NULL, 0, NULL }, + { "tr", roff_line, NULL, NULL, 0, NULL }, + { ".", roff_cblock, NULL, NULL, 0, NULL }, + { "\\}", roff_ccond, NULL, NULL, 0, NULL }, }; static void roff_free1(struct roff *); static enum rofft roff_hash_find(const char *); +static void roff_hash_init(void); static void roffnode_cleanscope(struct roff *); static int roffnode_push(struct roff *, enum rofft, int, int); static void roffnode_pop(struct roff *); static enum rofft roff_parse(const char *, int *); +/* See roff_hash_find() */ +#define ROFF_HASH(p) (p[0] - ASCII_LO) +static void +roff_hash_init(void) +{ + struct roffmac *n; + int buc, i; + + for (i = 0; i < (int)ROFF_MAX; i++) { + assert(roffs[i].name[0] >= ASCII_LO); + assert(roffs[i].name[0] <= ASCII_HI); + + buc = ROFF_HASH(roffs[i].name); + + if (NULL != (n = hash[buc])) { + for ( ; n->next; n = n->next) + /* Do nothing. */ ; + n->next = &roffs[i]; + } else + hash[buc] = &roffs[i]; + } +} + + /* * Look up a roff token by its name. Returns ROFF_MAX if no macro by * the nil-terminated string name could be found. @@ -118,14 +172,27 @@ static enum rofft roff_parse(const char *, int *); static enum rofft roff_hash_find(const char *p) { - int i; + int buc; + struct roffmac *n; - /* FIXME: make this be fast and efficient. */ + /* + * libroff has an extremely simple hashtable, for the time + * being, which simply keys on the first character, which must + * be printable, then walks a chain. It works well enough until + * optimised. + */ - for (i = 0; i < (int)ROFF_MAX; i++) - if (0 == strcmp(roffs[i].name, p)) - return((enum rofft)i); + if (p[0] < ASCII_LO || p[0] > ASCII_HI) + return(ROFF_MAX); + buc = ROFF_HASH(p); + + if (NULL == (n = hash[buc])) + return(ROFF_MAX); + for ( ; n; n = n->next) + if (0 == strcmp(n->name, p)) + return((enum rofft)(n - roffs)); + return(ROFF_MAX); } @@ -141,6 +208,11 @@ roffnode_pop(struct roff *r) assert(r->last); p = r->last; + + if (ROFF_el == p->tok) + if (r->rstackpos > -1) + r->rstackpos--; + r->last = r->last->parent; if (p->end) free(p->end); @@ -211,6 +283,9 @@ roff_alloc(const mandocmsg msg, void *data) r->msg = msg; r->data = data; + r->rstackpos = -1; + + roff_hash_init(); return(r); } @@ -298,7 +373,7 @@ roff_parse(const char *buf, int *pos) for (j = 0; j < 4; j++, (*pos)++) if ('\0' == (mac[j] = buf[*pos])) break; - else if (' ' == buf[*pos]) + else if (' ' == buf[*pos] || (j && '\\' == buf[*pos])) break; if (j == 4 || j < 1) @@ -332,7 +407,22 @@ roff_cblock(ROFF_ARGS) return(ROFF_IGN); } - if (ROFF_ig != r->last->tok) { + switch (r->last->tok) { + case (ROFF_am): + /* FALLTHROUGH */ + case (ROFF_ami): + /* FALLTHROUGH */ + case (ROFF_am1): + /* FALLTHROUGH */ + case (ROFF_de): + /* FALLTHROUGH */ + case (ROFF_dei): + /* FALLTHROUGH */ + case (ROFF_de1): + /* FALLTHROUGH */ + case (ROFF_ig): + break; + default: if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) return(ROFF_ERR); return(ROFF_IGN); @@ -372,7 +462,14 @@ roff_ccond(ROFF_ARGS) return(ROFF_IGN); } - if (ROFF_if != r->last->tok) { + switch (r->last->tok) { + case (ROFF_el): + /* FALLTHROUGH */ + case (ROFF_ie): + /* FALLTHROUGH */ + case (ROFF_if): + break; + default: if ( ! (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL)) return(ROFF_ERR); return(ROFF_IGN); @@ -396,11 +493,22 @@ roff_ccond(ROFF_ARGS) /* ARGSUSED */ static enum rofferr -roff_ig(ROFF_ARGS) +roff_block(ROFF_ARGS) { int sv; size_t sz; + if (ROFF_ig != tok && '\0' == (*bufp)[pos]) { + if ( ! (*r->msg)(MANDOCERR_NOARGS, r->data, ln, ppos, NULL)) + return(ROFF_ERR); + return(ROFF_IGN); + } else if (ROFF_ig != tok) { + while ((*bufp)[pos] && ' ' != (*bufp)[pos]) + pos++; + while (' ' == (*bufp)[pos]) + pos++; + } + if ( ! roffnode_push(r, tok, ln, ppos)) return(ROFF_ERR); @@ -444,38 +552,8 @@ roff_ig(ROFF_ARGS) /* ARGSUSED */ static enum rofferr -roff_if_sub(ROFF_ARGS) +roff_block_sub(ROFF_ARGS) { - enum rofft t; - enum roffrule rr; - - ppos = pos; - rr = r->last->rule; - roffnode_cleanscope(r); - - if (ROFF_MAX == (t = roff_parse(*bufp, &pos))) - return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT); - - /* - * A denied conditional must evaluate its children if and only - * if they're either structurally required (such as loops and - * conditionals) or a closing macro. - */ - if (ROFFRULE_DENY == rr) - if ( ! (ROFFMAC_STRUCT & roffs[t].flags)) - if (ROFF_ccond != t) - return(ROFF_IGN); - - assert(roffs[t].proc); - return((*roffs[t].proc) - (r, t, bufp, szp, ln, ppos, pos, offs)); -} - - -/* ARGSUSED */ -static enum rofferr -roff_ig_sub(ROFF_ARGS) -{ enum rofft t; int i, j; @@ -530,7 +608,7 @@ roff_ig_sub(ROFF_ARGS) /* ARGSUSED */ static enum rofferr -roff_ig_text(ROFF_ARGS) +roff_block_text(ROFF_ARGS) { return(ROFF_IGN); @@ -539,40 +617,92 @@ roff_ig_text(ROFF_ARGS) /* ARGSUSED */ static enum rofferr -roff_if_text(ROFF_ARGS) +roff_cond_sub(ROFF_ARGS) { + enum rofft t; + enum roffrule rr; + + ppos = pos; + rr = r->last->rule; + + roff_cond_text(r, tok, bufp, szp, ln, ppos, pos, offs); + + if (ROFF_MAX == (t = roff_parse(*bufp, &pos))) + return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT); + + /* + * A denied conditional must evaluate its children if and only + * if they're either structurally required (such as loops and + * conditionals) or a closing macro. + */ + if (ROFFRULE_DENY == rr) + if ( ! (ROFFMAC_STRUCT & roffs[t].flags)) + if (ROFF_ccond != t) + return(ROFF_IGN); + + assert(roffs[t].proc); + return((*roffs[t].proc) + (r, t, bufp, szp, ln, ppos, pos, offs)); +} + + +/* ARGSUSED */ +static enum rofferr +roff_cond_text(ROFF_ARGS) +{ char *ep, *st; + enum roffrule rr; + rr = r->last->rule; + + /* + * We display the value of the text if out current evaluation + * scope permits us to do so. + */ + st = &(*bufp)[pos]; if (NULL == (ep = strstr(st, "\\}"))) { roffnode_cleanscope(r); - return(ROFF_IGN); + return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT); } - if (ep > st && '\\' != *(ep - 1)) + if (ep > st && '\\' != *(ep - 1)) { + ep = '\0'; roffnode_pop(r); + } roffnode_cleanscope(r); - return(ROFF_IGN); + return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT); } /* ARGSUSED */ static enum rofferr -roff_if(ROFF_ARGS) +roff_cond(ROFF_ARGS) { + int cpos; /* position of the condition */ int sv; - /* - * Read ahead past the conditional. - * FIXME: this does not work, as conditionals don't end on - * whitespace, but are parsed according to a formal grammar. - * It's good enough for now, however. - */ + /* Stack overflow! */ - while ((*bufp)[pos] && ' ' != (*bufp)[pos]) - pos++; + if (ROFF_ie == tok && r->rstackpos == RSTACK_MAX - 1) { + (*r->msg)(MANDOCERR_MEM, r->data, ln, ppos, NULL); + return(ROFF_ERR); + } + cpos = pos; + + if (ROFF_if == tok || ROFF_ie == tok) { + /* + * Read ahead past the conditional. FIXME: this does + * not work, as conditionals don't end on whitespace, + * but are parsed according to a formal grammar. It's + * good enough for now, however. + */ + while ((*bufp)[pos] && ' ' != (*bufp)[pos]) + pos++; + } + sv = pos; while (' ' == (*bufp)[pos]) pos++; @@ -583,7 +713,6 @@ roff_if(ROFF_ARGS) * really doing anything. Warn about this. It's probably * wrong. */ - if ('\0' == (*bufp)[pos] && sv != pos) { if ( ! (*r->msg)(MANDOCERR_NOARGS, r->data, ln, ppos, NULL)) return(ROFF_ERR); @@ -593,8 +722,35 @@ roff_if(ROFF_ARGS) if ( ! roffnode_push(r, tok, ln, ppos)) return(ROFF_ERR); - /* Don't evaluate: just assume NO. */ + /* XXX: Implement more conditionals. */ + if (ROFF_if == tok || ROFF_ie == tok) + r->last->rule = 'n' == (*bufp)[cpos] ? + ROFFRULE_ALLOW : ROFFRULE_DENY; + else if (ROFF_el == tok) { + /* + * An `.el' will get the value of the current rstack + * entry set in prior `ie' calls or defaults to DENY. + */ + if (r->rstackpos < 0) + r->last->rule = ROFFRULE_DENY; + else + r->last->rule = r->rstack[r->rstackpos]; + } + if (ROFF_ie == tok) { + /* + * An if-else will put the NEGATION of the current + * evaluated conditional into the stack. + */ + r->rstackpos++; + if (ROFFRULE_DENY == r->last->rule) + r->rstack[r->rstackpos] = ROFFRULE_ALLOW; + else + r->rstack[r->rstackpos] = ROFFRULE_DENY; + } + if (r->last->parent && ROFFRULE_DENY == r->last->parent->rule) + r->last->rule = ROFFRULE_DENY; + r->last->endspan = 1; if ('\\' == (*bufp)[pos] && '{' == (*bufp)[pos + 1]) { @@ -614,4 +770,13 @@ roff_if(ROFF_ARGS) *offs = pos; return(ROFF_RERUN); +} + + +/* ARGSUSED */ +static enum rofferr +roff_line(ROFF_ARGS) +{ + + return(ROFF_IGN); }