=================================================================== RCS file: /cvs/texi2mdoc/util.c,v retrieving revision 1.3 retrieving revision 1.23 diff -u -p -r1.3 -r1.23 --- texi2mdoc/util.c 2015/02/21 12:44:45 1.3 +++ texi2mdoc/util.c 2015/03/02 18:12:53 1.23 @@ -1,4 +1,4 @@ -/* $Id: util.c,v 1.3 2015/02/21 12:44:45 kristaps Exp $ */ +/* $Id: util.c,v 1.23 2015/03/02 18:12:53 kristaps Exp $ */ /* * Copyright (c) 2015 Kristaps Dzonsons * @@ -20,8 +20,6 @@ #include #include #include -#include -#include #include #include #include @@ -43,9 +41,30 @@ texifilepop(struct texi *p) assert(p->filepos > 0); f = &p->files[--p->filepos]; - munmap(f->map, f->mapsz); + free(f->map); } +static void +teximacrofree(struct teximacro *p) +{ + size_t i; + + for (i = 0; i < p->argsz; i++) + free(p->args[i]); + + free(p->args); + free(p->key); + free(p->value); +} + +static void +texivaluefree(struct texivalue *p) +{ + + free(p->key); + free(p->value); +} + /* * Unmap all files that we're currently using and free all resources * that we've allocated during the parse. @@ -58,21 +77,26 @@ texiexit(struct texi *p) /* Make sure we're newline-terminated. */ if (p->outcol) - putchar('\n'); + fputc('\n', p->outfile); + if (NULL != p->chapters) + teximdocclose(p, 1); /* Unmap all files. */ while (p->filepos > 0) texifilepop(p); + for (i = 0; i < p->macrosz; i++) + teximacrofree(&p->macros[i]); for (i = 0; i < p->dirsz; i++) free(p->dirs[i]); + for (i = 0; i < p->indexsz; i++) + free(p->indexs[i]); + for (i = 0; i < p->valsz; i++) + texivaluefree(&p->vals[i]); - for (i = 0; i < p->valsz; i++) { - free(p->vals[i].value); - free(p->vals[i].key); - } - + free(p->macros); free(p->vals); + free(p->indexs); free(p->dirs); free(p->subtitle); free(p->title); @@ -98,12 +122,19 @@ texiabort(struct texi *p, const char *errstring) void texiwarn(const struct texi *p, const char *fmt, ...) { - va_list ap; + va_list ap; + const struct texifile *f; - fprintf(stderr, "%s:%zu:%zu: warning: ", - p->files[p->filepos - 1].name, - p->files[p->filepos - 1].line + 1, - p->files[p->filepos - 1].col + 1); + f = &p->files[p->filepos - 1]; + + if (f->insplice) + fprintf(stderr, "%s:%zu:%zu (%zuB left in splice): " + "warning: ", f->name, f->line + 1, + f->col + 1, f->insplice); + else + fprintf(stderr, "%s:%zu:%zu: warning: ", + f->name, f->line + 1, f->col + 1); + va_start(ap, fmt); vfprintf(stderr, fmt, ap); va_end(ap); @@ -117,12 +148,19 @@ texiwarn(const struct texi *p, const char *fmt, ...) void texierr(struct texi *p, const char *fmt, ...) { - va_list ap; + va_list ap; + struct texifile *f; - fprintf(stderr, "%s:%zu:%zu: error: ", - p->files[p->filepos - 1].name, - p->files[p->filepos - 1].line + 1, - p->files[p->filepos - 1].col + 1); + f = &p->files[p->filepos - 1]; + + if (f->insplice) + fprintf(stderr, "%s:%zu:%zu: (%zuB left in splice): " + "error: ", f->name, f->line + 1, + f->col + 1, f->insplice); + else + fprintf(stderr, "%s:%zu:%zu: error: ", + f->name, f->line + 1, f->col + 1); + va_start(ap, fmt); vfprintf(stderr, fmt, ap); va_end(ap); @@ -133,9 +171,7 @@ texierr(struct texi *p, const char *fmt, ...) /* * Put a single data character to the output if we're not ignoring. - * Makes sure we don't spurriously start a macro. - * Adjusts our output status. - * This shouldn't be called for macros: just for ordinary text. + * Escape starting a line with a control character and slashes. */ void texiputchar(struct texi *p, char c) @@ -143,11 +179,17 @@ texiputchar(struct texi *p, char c) if (p->ign) return; - if ('.' == c && 0 == p->outcol) - fputs("\\&", stdout); + fputs("\\&", p->outfile); + if ('\'' == c && 0 == p->outcol) + fputs("\\&", p->outfile); - putchar(c); + if (p->uppercase) + fputc(toupper((unsigned int)c), p->outfile); + else + fputc(c, p->outfile); + if ('\\' == c) + fputc('e', p->outfile); p->seenvs = 0; if ('\n' == c) { p->outcol = 0; @@ -157,18 +199,43 @@ texiputchar(struct texi *p, char c) } /* - * Put multiple characters (see texiputchar()). - * This shouldn't be called for macros: just for ordinary text. + * Put an opaque series of characters. + * Characters starting a line with a control character are escaped, but + * that's it, so don't use this for non-controlled sequences of text. */ void texiputchars(struct texi *p, const char *s) { - while ('\0' != *s) - texiputchar(p, *s++); + if (p->ign) + return; + if ('.' == *s && 0 == p->outcol) + fputs("\\&", p->outfile); + if ('\'' == *s && 0 == p->outcol) + fputs("\\&", p->outfile); + if (p->uppercase) + for ( ; '\0' != *s; s++) + p->outcol += fputc(toupper + ((unsigned int)*s), p->outfile); + else + p->outcol += fputs(s, p->outfile); + p->seenvs = 0; } /* + * This puts all characters onto the output stream but makes sure to + * escape mdoc(7) slashes. + * FIXME: useless. + */ +void +texiputbuf(struct texi *p, size_t start, size_t end) +{ + + for ( ; start < end; start++) + texiputchar(p, BUF(p)[start]); +} + +/* * Close an mdoc(7) macro opened with teximacroopen(). * If there are no more macros on the line, prints a newline. */ @@ -180,7 +247,7 @@ teximacroclose(struct texi *p) return; if (0 == --p->outmacro) { - putchar('\n'); + fputc('\n', p->outfile); p->outcol = p->seenws = 0; } } @@ -200,19 +267,19 @@ teximacroopen(struct texi *p, const char *s) return; if (p->outcol && 0 == p->outmacro) { - putchar('\n'); + fputc('\n', p->outfile); p->outcol = 0; } if (0 == p->outmacro) - putchar('.'); + fputc('.', p->outfile); else - putchar(' '); + fputc(' ', p->outfile); - if (EOF != (rc = fputs(s, stdout))) + if (EOF != (rc = fputs(s, p->outfile))) p->outcol += rc; - putchar(' '); + fputc(' ', p->outfile); p->outcol++; p->outmacro++; p->seenws = 0; @@ -234,10 +301,11 @@ teximacro(struct texi *p, const char *s) texierr(p, "\"%s\" in a literal scope!?", s); if (p->outcol) - putchar('\n'); + fputc('\n', p->outfile); - putchar('.'); - puts(s); + fputc('.', p->outfile); + fputs(s, p->outfile); + fputc('\n', p->outfile); p->outcol = p->seenws = 0; } @@ -248,7 +316,7 @@ void texivspace(struct texi *p) { - if (p->seenvs) + if (p->seenvs || TEXILIST_TABLE == p->list) return; teximacro(p, "Pp"); p->seenvs = 1; @@ -259,15 +327,24 @@ texivspace(struct texi *p) * in the current input file. */ void -advance(struct texi *p, const char *buf, size_t *pos) +advance(struct texi *p, size_t *pos) { + struct texifile *f; - if ('\n' == buf[*pos]) { - p->files[p->filepos - 1].line++; - p->files[p->filepos - 1].col = 0; - } else - p->files[p->filepos - 1].col++; + f = &p->files[p->filepos - 1]; + if (0 == f->insplice) { + if ('\n' == BUF(p)[*pos]) { + f->line++; + f->col = 0; + } else + f->col++; + } else { + --f->insplice; + if (0 == f->insplice) + f->depth = 0; + } + (*pos)++; } @@ -279,20 +356,21 @@ advance(struct texi *p, const char *buf, size_t *pos) * appropriately flush-left punctuation alongside the macro. */ void -texipunctuate(struct texi *p, const char *buf, size_t sz, size_t *pos) +texipunctuate(struct texi *p, size_t *pos) { size_t start, end; if (1 != p->outmacro) return; - for (start = end = *pos; end < sz; end++) { - switch (buf[end]) { + for (start = end = *pos; end < BUFSZ(p); end++) { + switch (BUF(p)[end]) { case (','): case (')'): case ('.'): case ('"'): case (':'): + case (';'): case ('!'): case ('?'): continue; @@ -303,11 +381,12 @@ texipunctuate(struct texi *p, const char *buf, size_t } if (end == *pos) return; - if (end + 1 == sz || ' ' == buf[end] || '\n' == buf[end]) { + if (end + 1 == BUFSZ(p) || ' ' == BUF(p)[end] || + '\n' == BUF(p)[end]) { for ( ; start < end; start++) { texiputchar(p, ' '); - texiputchar(p, buf[start]); - advance(p, buf, pos); + texiputchar(p, BUF(p)[start]); + advance(p, pos); } } } @@ -318,48 +397,47 @@ texipunctuate(struct texi *p, const char *buf, size_t * doing so. */ static size_t -advancenext(struct texi *p, const char *buf, size_t sz, size_t *pos) +advancenext(struct texi *p, size_t *pos) { if (p->literal) { - while (*pos < sz && ismspace(buf[*pos])) { - if (*pos && '\n' == buf[*pos] && - '\\' == buf[*pos - 1]) - texiputchar(p, 'e'); - texiputchar(p, buf[*pos]); - advance(p, buf, pos); + while (*pos < BUFSZ(p) && ismspace(BUF(p)[*pos])) { + texiputchar(p, BUF(p)[*pos]); + advance(p, pos); } return(*pos); } - while (*pos < sz && ismspace(buf[*pos])) { + while (*pos < BUFSZ(p) && ismspace(BUF(p)[*pos])) { p->seenws = 1; /* * If it looks like we've printed a double-line, then * output a paragraph. * FIXME: this is stupid. */ - if (*pos && '\n' == buf[*pos] && '\n' == buf[*pos - 1]) + if (*pos && '\n' == BUF(p)[*pos] && '\n' == BUF(p)[*pos - 1]) texivspace(p); - advance(p, buf, pos); + advance(p, pos); } return(*pos); } /* * Advance to the EOLN in the input stream. - * NOTE: THIS SHOULD NOT BE CALLED ON BLANK TEXT, as it will read up to - * the @\n. + * This will skip over '@' markers in an effort to ignore escaped + * newlines. */ size_t -advanceeoln(struct texi *p, const char *buf, - size_t sz, size_t *pos, int consumenl) +advanceeoln(struct texi *p, size_t *pos, int consumenl) { - while (*pos < sz && '\n' != buf[*pos]) - advance(p, buf, pos); - if (*pos < sz && consumenl) - advance(p, buf, pos); + while (*pos < BUFSZ(p) && '\n' != BUF(p)[*pos]) { + if ('@' == BUF(p)[*pos]) + advance(p, pos); + advance(p, pos); + } + if (*pos < BUFSZ(p) && consumenl) + advance(p, pos); return(*pos); } @@ -368,22 +446,127 @@ advanceeoln(struct texi *p, const char *buf, * current buffer greater than or equal to the current position. */ void -advanceto(struct texi *p, const char *buf, size_t *pos, size_t end) +advanceto(struct texi *p, size_t *pos, size_t end) { assert(*pos <= end); while (*pos < end) - advance(p, buf, pos); + advance(p, pos); } +static void +texiexecmacro(struct texi *p, struct teximacro *m, size_t sv, size_t *pos) +{ + size_t valsz, realsz, aasz, asz, + ssz, i, j, k, start, end; + char *val; + char **args; + const char *cp; + + /* Disregard empty macros. */ + if (0 == (valsz = realsz = strlen(m->value))) { + args = argparse(p, pos, &asz, m->argsz); + for (i = 0; i < asz; i++) + free(args[i]); + free(args); + return; + } + + /* + * This is important: it protect us from macros that invoke more + * macros, possibly going on infinitely. + * We use "sv" instead of the current position because we might + * be invoked at the end of the macro (i.e., insplice == 0). + * The "sv" value was initialised at the start of the macro. + */ + if (sv > 0) + if (++p->files[p->filepos].depth > 64) + texierr(p, "maximium recursive depth"); + + args = argparse(p, pos, &asz, m->argsz); + if (asz != m->argsz) + texiwarn(p, "invalid macro argument length"); + aasz = asz < m->argsz ? asz : m->argsz; + + if (0 == aasz) { + texisplice(p, m->value, valsz, *pos); + return; + } + + val = strdup(m->value); + + for (i = j = 0; i < realsz; i++) { + /* Parse blindly til the backslash delimiter. */ + if ('\\' != m->value[i]) { + val[j++] = m->value[i]; + val[j] = '\0'; + continue; + } else if (i == realsz - 1) + texierr(p, "trailing argument name delimiter"); + + /* Double-backslash is escaped. */ + if ('\\' == m->value[i + 1]) { + val[j++] = m->value[i++]; + val[j] = '\0'; + continue; + } + + assert('\\' == m->value[i] && i < realsz - 1); + + /* Parse to terminating delimiter. */ + /* FIXME: embedded, escaped delimiters? */ + for (start = end = i + 1; end < realsz; end++) + if ('\\' == m->value[end]) + break; + if (end == realsz) + texierr(p, "unterminated argument name"); + + for (k = 0; k < aasz; k++) { + if ((ssz = strlen(m->args[k])) != (end - start)) + continue; + if (strncmp(&m->value[start], m->args[k], ssz)) + continue; + break; + } + + /* + * Argument didn't exist in argument table. + * Just ignore it. + */ + if (k == aasz) { + i = end; + continue; + } + + if (strlen(args[k]) > ssz) { + valsz += strlen(args[k]); + val = realloc(val, valsz + 1); + if (NULL == val) + texiabort(p, NULL); + } + + for (cp = args[k]; '\0' != *cp; cp++) + val[j++] = *cp; + + val[j] = '\0'; + i = end; + } + + texisplice(p, val, strlen(val), *pos); + + for (i = 0; i < asz; i++) + free(args[i]); + free(args); + free(val); +} + /* * Output a free-form word in the input stream, progressing to the next * command or white-space. * This also will advance the input stream. */ static void -texiword(struct texi *p, const char *buf, - size_t sz, size_t *pos, char extra) +parseword(struct texi *p, size_t *pos, char extra) { if (p->seenws && 0 == p->outmacro && @@ -395,28 +578,28 @@ texiword(struct texi *p, const char *buf, p->seenws = 0; - while (*pos < sz && ! ismspace(buf[*pos])) { - switch (buf[*pos]) { + while (*pos < BUFSZ(p) && ! ismspace(BUF(p)[*pos])) { + switch (BUF(p)[*pos]) { case ('@'): case ('}'): case ('{'): return; } - if ('\0' != extra && buf[*pos] == extra) + if ('\0' != extra && BUF(p)[*pos] == extra) return; - if (*pos < sz - 1 && - '`' == buf[*pos] && - '`' == buf[*pos + 1]) { + if (*pos < BUFSZ(p) - 1 && + '`' == BUF(p)[*pos] && + '`' == BUF(p)[*pos + 1]) { texiputchars(p, "\\(lq"); - advance(p, buf, pos); - } else if (*pos < sz - 1 && - '\'' == buf[*pos] && - '\'' == buf[*pos + 1]) { + advance(p, pos); + } else if (*pos < BUFSZ(p) - 1 && + '\'' == BUF(p)[*pos] && + '\'' == BUF(p)[*pos + 1]) { texiputchars(p, "\\(rq"); - advance(p, buf, pos); + advance(p, pos); } else - texiputchar(p, buf[*pos]); - advance(p, buf, pos); + texiputchar(p, BUF(p)[*pos]); + advance(p, pos); } } @@ -426,46 +609,71 @@ texiword(struct texi *p, const char *buf, * index after the command name. */ enum texicmd -texicmd(struct texi *p, const char *buf, - size_t pos, size_t sz, size_t *end) +texicmd(const struct texi *p, size_t pos, size_t *end, struct teximacro **macro) { - size_t i, len; + size_t i, len, toksz; - assert('@' == buf[pos]); + assert('@' == BUF(p)[pos]); - if ((*end = pos) == sz) + if (NULL != macro) + *macro = NULL; + + if ((*end = pos) == BUFSZ(p)) return(TEXICMD__MAX); - else if ((*end = ++pos) == sz) + else if ((*end = ++pos) == BUFSZ(p)) return(TEXICMD__MAX); /* Alphabetic commands are special. */ - if ( ! isalpha(buf[pos])) { - if ((*end = pos + 1) == sz) + if ( ! isalpha((unsigned int)BUF(p)[pos])) { + if ((*end = pos + 1) == BUFSZ(p)) return(TEXICMD__MAX); for (i = 0; i < TEXICMD__MAX; i++) { if (1 != texitoks[i].len) continue; - if (0 == strncmp(texitoks[i].tok, &buf[pos], 1)) + if (0 == strncmp(texitoks[i].tok, &BUF(p)[pos], 1)) return(i); } - texiwarn(p, "bad command: @%c", buf[pos]); + texiwarn(p, "bad command: @%c", BUF(p)[pos]); return(TEXICMD__MAX); } - for (*end = pos; *end < sz && ! ismspace(buf[*end]); (*end)++) - if ((*end > pos && ('@' == buf[*end] || - '{' == buf[*end] || '}' == buf[*end]))) + /* Scan to the end of the possible command name. */ + for (*end = pos; *end < BUFSZ(p) && ! ismspace(BUF(p)[*end]); (*end)++) + if ((*end > pos && ('@' == BUF(p)[*end] || + '{' == BUF(p)[*end] || '}' == BUF(p)[*end]))) break; + /* Look for the command. */ len = *end - pos; for (i = 0; i < TEXICMD__MAX; i++) { if (len != texitoks[i].len) continue; - if (0 == strncmp(texitoks[i].tok, &buf[pos], len)) + if (0 == strncmp(texitoks[i].tok, &BUF(p)[pos], len)) return(i); } - texiwarn(p, "bad command: @%.*s", (int)len, &buf[pos]); + /* Look for it in our indices. */ + for (i = 0; i < p->indexsz; i++) { + toksz = strlen(p->indexs[i]); + if (len != 5 + toksz) + continue; + if (strncmp(&BUF(p)[pos], p->indexs[i], toksz)) + continue; + if (0 == strncmp(&BUF(p)[pos + toksz], "index", 5)) + return(TEXICMD_USER_INDEX); + } + + for (i = 0; i < p->macrosz; i++) { + if (len != strlen(p->macros[i].key)) + continue; + if (strncmp(&BUF(p)[pos], p->macros[i].key, len)) + continue; + if (NULL != macro) + *macro = &p->macros[i]; + return(TEXICMD__MAX); + } + + texiwarn(p, "bad command: @%.*s", (int)len, &BUF(p)[pos]); return(TEXICMD__MAX); } @@ -478,45 +686,48 @@ texicmd(struct texi *p, const char *buf, * bracket for the zeroth parse. */ int -parsearg(struct texi *p, const char *buf, - size_t sz, size_t *pos, size_t num) +parsearg(struct texi *p, size_t *pos, size_t num) { - size_t end; - enum texicmd cmd; + size_t end, sv; + enum texicmd cmd; + struct teximacro *macro; - while (*pos < sz && ismspace(buf[*pos])) - advance(p, buf, pos); - if (*pos == sz || (0 == num && '{' != buf[*pos])) + while (*pos < BUFSZ(p) && ismspace(BUF(p)[*pos])) + advance(p, pos); + if (*pos == BUFSZ(p) || (0 == num && '{' != BUF(p)[*pos])) return(0); if (0 == num) - advance(p, buf, pos); + advance(p, pos); - while ((*pos = advancenext(p, buf, sz, pos)) < sz) { - switch (buf[*pos]) { + while ((*pos = advancenext(p, pos)) < BUFSZ(p)) { + switch (BUF(p)[*pos]) { case (','): - advance(p, buf, pos); + advance(p, pos); return(1); case ('}'): - advance(p, buf, pos); + advance(p, pos); return(0); case ('{'): if (0 == p->ign) texiwarn(p, "unexpected \"{\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('@'): break; default: - texiword(p, buf, sz, pos, ','); + parseword(p, pos, ','); continue; } - cmd = texicmd(p, buf, *pos, sz, &end); - advanceto(p, buf, pos, end); + sv = p->files[p->filepos - 1].insplice; + cmd = texicmd(p, *pos, &end, ¯o); + advanceto(p, pos, end); + if (NULL != macro) + texiexecmacro(p, macro, sv, pos); if (TEXICMD__MAX == cmd) continue; if (NULL != texitoks[cmd].fp) - (*texitoks[cmd].fp)(p, cmd, buf, sz, pos); + (*texitoks[cmd].fp)(p, cmd, pos); } return(0); } @@ -526,41 +737,58 @@ parsearg(struct texi *p, const char *buf, * This will stop in the event of EOF or if we're not at a bracket. */ void -parsebracket(struct texi *p, const char *buf, size_t sz, size_t *pos) +parsebracket(struct texi *p, size_t *pos, int dostack) { - size_t end; - enum texicmd cmd; + size_t end, sv, stack; + enum texicmd cmd; + struct teximacro *macro; - while (*pos < sz && ismspace(buf[*pos])) - advance(p, buf, pos); + while (*pos < BUFSZ(p) && ismspace(BUF(p)[*pos])) + advance(p, pos); - if (*pos == sz || '{' != buf[*pos]) + if (*pos == BUFSZ(p) || '{' != BUF(p)[*pos]) return; - advance(p, buf, pos); + advance(p, pos); - while ((*pos = advancenext(p, buf, sz, pos)) < sz) { - switch (buf[*pos]) { + stack = 0; + while ((*pos = advancenext(p, pos)) < BUFSZ(p)) { + switch (BUF(p)[*pos]) { case ('}'): - advance(p, buf, pos); + if (stack > 0) { + stack--; + advance(p, pos); + texiputchar(p, '}'); + continue; + } + advance(p, pos); return; case ('{'): + if (dostack) { + stack++; + advance(p, pos); + texiputchar(p, '{'); + continue; + } if (0 == p->ign) texiwarn(p, "unexpected \"{\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('@'): break; default: - texiword(p, buf, sz, pos, '\0'); + parseword(p, pos, '\0'); continue; } - cmd = texicmd(p, buf, *pos, sz, &end); - advanceto(p, buf, pos, end); + sv = p->files[p->filepos - 1].insplice; + cmd = texicmd(p, *pos, &end, ¯o); + advanceto(p, pos, end); + if (NULL != macro) + texiexecmacro(p, macro, sv, pos); if (TEXICMD__MAX == cmd) continue; if (NULL != texitoks[cmd].fp) - (*texitoks[cmd].fp)(p, cmd, buf, sz, pos); + (*texitoks[cmd].fp)(p, cmd, pos); } } @@ -570,82 +798,110 @@ parsebracket(struct texi *p, const char *buf, size_t s * the way. */ void -parseeoln(struct texi *p, const char *buf, size_t sz, size_t *pos) +parseeoln(struct texi *p, size_t *pos) { - size_t end; - enum texicmd cmd; + size_t end, sv; + enum texicmd cmd; + struct teximacro *macro; - while (*pos < sz && '\n' != buf[*pos]) { - while (*pos < sz && isws(buf[*pos])) { + while (*pos < BUFSZ(p) && '\n' != BUF(p)[*pos]) { + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) { p->seenws = 1; if (p->literal) - texiputchar(p, buf[*pos]); - advance(p, buf, pos); + texiputchar(p, BUF(p)[*pos]); + advance(p, pos); } - switch (buf[*pos]) { + switch (BUF(p)[*pos]) { case ('}'): if (0 == p->ign) texiwarn(p, "unexpected \"}\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('{'): if (0 == p->ign) texiwarn(p, "unexpected \"{\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('@'): break; default: - texiword(p, buf, sz, pos, '\0'); + parseword(p, pos, '\0'); continue; } - cmd = texicmd(p, buf, *pos, sz, &end); - advanceto(p, buf, pos, end); + sv = p->files[p->filepos - 1].insplice; + cmd = texicmd(p, *pos, &end, ¯o); + advanceto(p, pos, end); + if (NULL != macro) + texiexecmacro(p, macro, sv, pos); if (TEXICMD__MAX == cmd) continue; if (NULL != texitoks[cmd].fp) - (*texitoks[cmd].fp)(p, cmd, buf, sz, pos); + (*texitoks[cmd].fp)(p, cmd, pos); } + + if (*pos < BUFSZ(p) && '\n' == BUF(p)[*pos]) + advance(p, pos); } /* + * Peek to see if there's a command after subsequent whitespace. + * If so, return the macro identifier. + * This DOES NOT work with user-defined macros. + */ +enum texicmd +peekcmd(const struct texi *p, size_t pos) +{ + size_t end; + + while (pos < BUFSZ(p) && ismspace(BUF(p)[pos])) + pos++; + if (pos == BUFSZ(p) || '@' != BUF(p)[pos]) + return(TEXICMD__MAX); + return(texicmd(p, pos, &end, NULL)); +} + +/* * Parse a single word or command. * This will return immediately at the EOF. */ -void -parsesingle(struct texi *p, const char *buf, size_t sz, size_t *pos) +static void +parsesingle(struct texi *p, size_t *pos) { - size_t end; - enum texicmd cmd; + size_t end, sv; + enum texicmd cmd; + struct teximacro *macro; - if ((*pos = advancenext(p, buf, sz, pos)) >= sz) + if ((*pos = advancenext(p, pos)) >= BUFSZ(p)) return; - switch (buf[*pos]) { + switch (BUF(p)[*pos]) { case ('}'): if (0 == p->ign) texiwarn(p, "unexpected \"}\""); - advance(p, buf, pos); + advance(p, pos); return; case ('{'): if (0 == p->ign) texiwarn(p, "unexpected \"{\""); - advance(p, buf, pos); + advance(p, pos); return; case ('@'): break; default: - texiword(p, buf, sz, pos, '\0'); + parseword(p, pos, '\0'); return; } - cmd = texicmd(p, buf, *pos, sz, &end); - advanceto(p, buf, pos, end); + sv = p->files[p->filepos - 1].insplice; + cmd = texicmd(p, *pos, &end, ¯o); + advanceto(p, pos, end); + if (NULL != macro) + texiexecmacro(p, macro, sv, pos); if (TEXICMD__MAX == cmd) return; if (NULL != texitoks[cmd].fp) - (*texitoks[cmd].fp)(p, cmd, buf, sz, pos); + (*texitoks[cmd].fp)(p, cmd, pos); } /* @@ -657,18 +913,18 @@ parsesingle(struct texi *p, const char *buf, size_t sz * line or 1 otherwise. */ int -parselinearg(struct texi *p, const char *buf, size_t sz, size_t *pos) +parselinearg(struct texi *p, size_t *pos) { - while (*pos < sz && isws(buf[*pos])) { + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) { p->seenws = 1; - advance(p, buf, pos); + advance(p, pos); } - if (*pos < sz && '{' == buf[*pos]) - parsebracket(p, buf, sz, pos); - else if (*pos < sz && '\n' != buf[*pos]) - parsesingle(p, buf, sz, pos); + if (*pos < BUFSZ(p) && '{' == BUF(p)[*pos]) + parsebracket(p, pos, 0); + else if (*pos < BUFSZ(p) && '\n' != BUF(p)[*pos]) + parsesingle(p, pos); else return(0); @@ -678,75 +934,143 @@ parselinearg(struct texi *p, const char *buf, size_t s /* * Parse til the end of the buffer. */ -void -parseeof(struct texi *p, const char *buf, size_t sz) +static void +parseeof(struct texi *p) { size_t pos; - for (pos = 0; pos < sz; ) - parsesingle(p, buf, sz, &pos); + for (pos = 0; pos < BUFSZ(p); ) + parsesingle(p, &pos); } +void +texisplice(struct texi *p, const char *buf, size_t sz, size_t pos) +{ + char *cp; + struct texifile *f; + + assert(p->filepos > 0); + f = &p->files[p->filepos - 1]; + + if (f->mapsz + sz > f->mapmaxsz) { + f->mapmaxsz = f->mapsz + sz + 1024; + cp = realloc(f->map, f->mapmaxsz); + if (NULL == cp) + texiabort(p, NULL); + f->map = cp; + } + + f->insplice += sz; + memmove(f->map + pos + sz, f->map + pos, f->mapsz - pos); + memcpy(f->map + pos, buf, sz); + f->mapsz += sz; +} + /* * Parse a block sequence until we have the "@end endtoken" command * invocation. * This will return immediately at EOF. */ void -parseto(struct texi *p, const char *buf, - size_t sz, size_t *pos, const char *endtoken) +parseto(struct texi *p, size_t *pos, const char *endtoken) { - size_t end; - enum texicmd cmd; - size_t endtoksz; + size_t end, sv; + enum texicmd cmd; + size_t endtoksz; + struct teximacro *macro; endtoksz = strlen(endtoken); assert(endtoksz > 0); - while ((*pos = advancenext(p, buf, sz, pos)) < sz) { - switch (buf[*pos]) { + while ((*pos = advancenext(p, pos)) < BUFSZ(p)) { + switch (BUF(p)[*pos]) { case ('}'): if (0 == p->ign) texiwarn(p, "unexpected \"}\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('{'): if (0 == p->ign) texiwarn(p, "unexpected \"{\""); - advance(p, buf, pos); + advance(p, pos); continue; case ('@'): break; default: - texiword(p, buf, sz, pos, '\0'); + parseword(p, pos, '\0'); continue; } - cmd = texicmd(p, buf, *pos, sz, &end); - advanceto(p, buf, pos, end); + sv = p->files[p->filepos - 1].insplice; + cmd = texicmd(p, *pos, &end, ¯o); + advanceto(p, pos, end); if (TEXICMD_END == cmd) { - while (*pos < sz && isws(buf[*pos])) - advance(p, buf, pos); + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) + advance(p, pos); /* * FIXME: check the full word, not just its * initial substring! */ - if (sz - *pos >= endtoksz && 0 == strncmp - (&buf[*pos], endtoken, endtoksz)) { - advanceeoln(p, buf, sz, pos, 0); + if (BUFSZ(p) - *pos >= endtoksz && 0 == strncmp + (&BUF(p)[*pos], endtoken, endtoksz)) { + advanceeoln(p, pos, 0); break; } if (0 == p->ign) texiwarn(p, "unexpected \"end\""); - advanceeoln(p, buf, sz, pos, 0); + advanceeoln(p, pos, 0); continue; - } else if (TEXICMD__MAX != cmd) - if (NULL != texitoks[cmd].fp) - (*texitoks[cmd].fp)(p, cmd, buf, sz, pos); + } + if (NULL != macro) + texiexecmacro(p, macro, sv, pos); + if (TEXICMD__MAX == cmd) + continue; + if (NULL != texitoks[cmd].fp) + (*texitoks[cmd].fp)(p, cmd, pos); } } /* + * Like parsefile() but used for reading from stdandard input. + * This can only be called for the first file! + */ +void +parsestdin(struct texi *p) +{ + struct texifile *f; + ssize_t ssz; + + assert(0 == p->filepos); + f = &p->files[p->filepos]; + memset(f, 0, sizeof(struct texifile)); + + f->type = TEXISRC_STDIN; + f->name = ""; + + for (f->mapsz = 0; ; f->mapsz += (size_t)ssz) { + if (f->mapsz == f->mapmaxsz) { + if (f->mapmaxsz == (1U << 31)) + texierr(p, "stdin buffer too long"); + f->mapmaxsz = f->mapmaxsz > 65536 / 2 ? + 2 * f->mapmaxsz : 65536; + f->map = realloc(f->map, f->mapmaxsz); + if (NULL == f->map) + texiabort(p, NULL); + } + ssz = read(STDIN_FILENO, f->map + + (int)f->mapsz, f->mapmaxsz - f->mapsz); + if (0 == ssz) + break; + else if (-1 == ssz) + texiabort(p, NULL); + } + + p->filepos++; + parseeof(p); + texifilepop(p); +} + +/* * Memory-map the file "fname" and begin parsing it unless "parse" is * zero, in which case we just dump the file to stdout (making sure it * doesn't trip up mdoc(7) along the way). @@ -759,11 +1083,14 @@ parsefile(struct texi *p, const char *fname, int parse int fd; struct stat st; size_t i; + char *map; - assert(p->filepos < 64); + if (64 == p->filepos) + texierr(p, "too many open files"); f = &p->files[p->filepos]; memset(f, 0, sizeof(struct texifile)); + f->type = TEXISRC_FILE; f->name = fname; if (-1 == (fd = open(fname, O_RDONLY, 0))) { texiabort(p, fname); @@ -772,32 +1099,28 @@ parsefile(struct texi *p, const char *fname, int parse texiabort(p, fname); } - f->mapsz = st.st_size; - f->map = mmap(NULL, f->mapsz, + f->mapsz = f->mapmaxsz = st.st_size; + map = mmap(NULL, f->mapsz, PROT_READ, MAP_SHARED, fd, 0); close(fd); - if (MAP_FAILED == f->map) + if (MAP_FAILED == map) texiabort(p, fname); - p->filepos++; if ( ! parse) { - /* - * We're printing verbatim output. - * Make sure it doesn't get interpreted as mdoc by - * escaping escapes and making sure leading dots don't - * trigger mdoc(7) expansion. - */ - for (i = 0; i < f->mapsz; i++) { - if (i > 0 && '.' == f->map[i]) - if ('\n' == f->map[i - 1]) - fputs("\\&", stdout); - putchar(f->map[i]); - if ('\\' == f->map[i]) - putchar('e'); - } - } else - parseeof(p, f->map, f->mapsz); + for (i = 0; i < f->mapsz; i++) + texiputchar(p, map[i]); + if (p->outcol) + texiputchar(p, '\n'); + munmap(map, f->mapsz); + return; + } + + p->filepos++; + f->map = malloc(f->mapsz); + memcpy(f->map, map, f->mapsz); + munmap(map, f->mapsz); + parseeof(p); texifilepop(p); } @@ -809,8 +1132,7 @@ parsefile(struct texi *p, const char *fname, int parse * The pointer can point to NULL if the value has been unset. */ static char ** -valuequery(const struct texi *p, - const char *buf, size_t start, size_t end) +valuequery(const struct texi *p, size_t start, size_t end) { size_t i, sz, len; @@ -822,7 +1144,7 @@ valuequery(const struct texi *p, sz = strlen(p->vals[i].key); if (sz != len) continue; - if (0 == strncmp(p->vals[i].key, &buf[start], len)) + if (0 == strncmp(p->vals[i].key, &BUF(p)[start], len)) return(&p->vals[i].value); } return(NULL); @@ -833,45 +1155,45 @@ valuequery(const struct texi *p, * pointer to its value via valuequery(). */ static char ** -valuelquery(struct texi *p, const char *buf, size_t sz, size_t *pos) +valuelquery(struct texi *p, size_t *pos) { size_t start, end; char **ret; - while (*pos < sz && isws(buf[*pos])) - advance(p, buf, pos); - if (*pos == sz) + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) + advance(p, pos); + if (*pos == BUFSZ(p)) return(NULL); - for (start = end = *pos; end < sz; end++) - if ('\n' == buf[end]) + for (start = end = *pos; end < BUFSZ(p); end++) + if ('\n' == BUF(p)[end]) break; - advanceto(p, buf, pos, end); - if (*pos < sz) { - assert('\n' == buf[*pos]); - advance(p, buf, pos); + advanceto(p, pos, end); + if (*pos < BUFSZ(p)) { + assert('\n' == BUF(p)[*pos]); + advance(p, pos); } - if (NULL == (ret = valuequery(p, buf, start, end))) + if (NULL == (ret = valuequery(p, start, end))) return(NULL); return(ret); } void -valuelclear(struct texi *p, const char *buf, size_t sz, size_t *pos) +valuelclear(struct texi *p, size_t *pos) { char **ret; - if (NULL == (ret = valuelquery(p, buf, sz, pos))) + if (NULL == (ret = valuelquery(p, pos))) return; free(*ret); *ret = NULL; } const char * -valuellookup(struct texi *p, const char *buf, size_t sz, size_t *pos) +valuellookup(struct texi *p, size_t *pos) { char **ret; - if (NULL == (ret = valuelquery(p, buf, sz, pos))) + if (NULL == (ret = valuelquery(p, pos))) return(NULL); return(*ret); } @@ -884,25 +1206,25 @@ valuellookup(struct texi *p, const char *buf, size_t s * value had previously been unset. */ const char * -valueblookup(struct texi *p, const char *buf, size_t sz, size_t *pos) +valueblookup(struct texi *p, size_t *pos) { size_t start, end; char **ret; - while (*pos < sz && isws(buf[*pos])) - advance(p, buf, pos); - if (*pos == sz || '{' != buf[*pos]) + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) + advance(p, pos); + if (*pos == BUFSZ(p) || '{' != BUF(p)[*pos]) return(NULL); - advance(p, buf, pos); - for (start = end = *pos; end < sz; end++) - if ('}' == buf[end]) + advance(p, pos); + for (start = end = *pos; end < BUFSZ(p); end++) + if ('}' == BUF(p)[end]) break; - advanceto(p, buf, pos, end); - if (*pos < sz) { - assert('}' == buf[*pos]); - advance(p, buf, pos); + advanceto(p, pos, end); + if (*pos < BUFSZ(p)) { + assert('}' == BUF(p)[*pos]); + advance(p, pos); } - if (NULL == (ret = valuequery(p, buf, start, end))) + if (NULL == (ret = valuequery(p, start, end))) return(NULL); return(*ret); } @@ -924,15 +1246,238 @@ valueadd(struct texi *p, char *key, char *val) free(p->vals[i].value); p->vals[i].value = val; } else { + /* FIXME: reallocarray() */ p->vals = realloc(p->vals, (p->valsz + 1) * sizeof(struct texivalue)); - if (NULL == p->vals) { - perror(NULL); - exit(EXIT_FAILURE); - } + if (NULL == p->vals) + texiabort(p, NULL); p->vals[p->valsz].key = key; p->vals[p->valsz].value = val; p->valsz++; } } + +/* + * Take the arguments to a macro, e.g., @foo{bar, baz, xyzzy} (or the + * declaration form, @macro foo {arg1, ...}) and textually convert it to + * an array of arguments of size "argsz". + * These need to be freed individually and as a whole. + * NOTE: this will puke on @, or @} macros, which can trick it into + * stopping argument parsing earlier. + * Ergo, textual: this doesn't interpret the arguments in any way. + */ +char ** +argparse(struct texi *p, size_t *pos, size_t *argsz, size_t hint) +{ + char **args; + size_t start, end, stack; + + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) + advance(p, pos); + + args = NULL; + *argsz = 0; + + if (*pos == BUFSZ(p)) + return(args); + + if ('{' != BUF(p)[*pos] && hint) { + /* + * Special case: if we encounter an unbracketed argument + * and we're being invoked with non-zero arguments + * (versus being set, i.e., hint>0), then parse until + * the end of line. + */ + *argsz = 1; + args = calloc(1, sizeof(char *)); + if (NULL == args) + texiabort(p, NULL); + start = *pos; + while (*pos < BUFSZ(p)) { + if ('\n' == BUF(p)[*pos]) + break; + advance(p, pos); + } + args[0] = malloc(*pos - start + 1); + memcpy(args[0], &BUF(p)[start], *pos - start); + args[0][*pos - start] = '\0'; + if (*pos < BUFSZ(p) && '\n' == BUF(p)[*pos]) + advance(p, pos); + return(args); + } else if ('{' != BUF(p)[*pos]) + return(args); + + assert('{' == BUF(p)[*pos]); + + /* Parse til the closing '}', putting into the array. */ + advance(p, pos); + while (*pos < BUFSZ(p)) { + while (*pos < BUFSZ(p) && isws(BUF(p)[*pos])) + advance(p, pos); + start = *pos; + stack = 0; + while (*pos < BUFSZ(p)) { + /* + * According to the manual, commas within + * embedded commands are escaped. + * We keep track of embedded-ness in the "stack" + * state anyway, so this is free. + */ + if (',' == BUF(p)[*pos] && 0 == stack && 1 != hint) + break; + else if (0 == stack && '}' == BUF(p)[*pos]) + break; + else if (0 != stack && '}' == BUF(p)[*pos]) + stack--; + else if ('{' == BUF(p)[*pos]) + stack++; + advance(p, pos); + } + if (stack) + texiwarn(p, "unterminated macro " + "in macro arguments"); + if ((end = *pos) == BUFSZ(p)) + break; + /* Test for zero-length '{ }'. */ + if (start == end && '}' == BUF(p)[*pos] && 0 == *argsz) + break; + /* FIXME: use reallocarray. */ + args = realloc + (args, sizeof(char *) * + (*argsz + 1)); + if (NULL == args) + texiabort(p, NULL); + args[*argsz] = malloc(end - start + 1); + if (NULL == args[*argsz]) + texiabort(p, NULL); + memcpy(args[*argsz], + &BUF(p)[start], end - start); + args[*argsz][end - start] = '\0'; + (*argsz)++; + if ('}' == BUF(p)[*pos]) + break; + advance(p, pos); + } + + if (*pos == BUFSZ(p)) + texierr(p, "unterminated arguments"); + assert('}' == BUF(p)[*pos]); + advance(p, pos); + return(args); +} + +/* + * If we're printing chapters, then do some naviation here and then + * close our outfile. + * I want to call this the SEE ALSO section, but that's not really what + * it is: we'll refer to the "initial" (top) node and the next and + * previous chapters. + */ +void +teximdocclose(struct texi *p, int last) +{ + char buf[PATH_MAX]; + + if (NULL == p->chapters || 0 == p->chapnum) + return; + + teximacro(p, "Sh INFO NAVIGATION"); + + /* Print a reference to the "top" node. */ + if (p->chapnum > 1) { + texiputchars(p, "Top node,"); + snprintf(buf, sizeof(buf), "node1 7"); + teximacroopen(p, "Xr "); + texiputchars(p, buf); + texiputchars(p, " ;"); + teximacroclose(p); + } + + /* Print a reference to the previous node. */ + if (p->chapnum > 2) { + texiputchars(p, "previous node,"); + snprintf(buf, sizeof(buf), + "node%zu 7", p->chapnum - 1); + teximacroopen(p, "Xr "); + texiputchars(p, buf); + if ( ! last) + texiputchars(p, " ;"); + teximacroclose(p); + } + + /* Print a reference to the next node. */ + if ( ! last) { + if (1 == p->chapnum) + texiputchars(p, "Next node,"); + else + texiputchars(p, "next node,"); + snprintf(buf, sizeof(buf), + "node%zu 7", p->chapnum + 1); + teximacroopen(p, "Xr "); + texiputchars(p, buf); + teximacroclose(p); + } + + fclose(p->outfile); +} + +/* + * Open a mdoc(7) context. + * If we're printing chapters, then open the outfile here, too. + * Otherwise just print the mdoc(7) prologue. + */ +void +teximdocopen(struct texi *p, size_t *pos) +{ + const char *cp; + time_t t; + char date[32]; + char fname[PATH_MAX]; + + if (NULL != p->chapters) { + snprintf(fname, sizeof(fname), "%s/node%zu.7", + p->chapters, ++p->chapnum); + p->outfile = fopen(fname, "w"); + if (NULL == p->outfile) + texiabort(p, fname); + } + + /* + * Here we print our standard mdoc(7) prologue. + * We use the title set with @settitle for the `Nd' description + * and the source document filename (the first one as invoked on + * the command line) for the title. + * The date is set to the current date. + */ + t = time(NULL); + strftime(date, sizeof(date), "%F", localtime(&t)); + + teximacroopen(p, "Dd"); + texiputchars(p, date); + teximacroclose(p); + teximacroopen(p, "Dt"); + for (cp = p->title; '\0' != *cp; cp++) + texiputchar(p, toupper((unsigned int)*cp)); + texiputchars(p, " 7"); + teximacroclose(p); + teximacro(p, "Os"); + teximacro(p, "Sh NAME"); + teximacroopen(p, "Nm"); + for (cp = p->title; '\0' != *cp; cp++) + texiputchar(p, *cp); + teximacroclose(p); + teximacroopen(p, "Nd"); + /* + * The subtitle `Nd' can consist of arbitrary macros, so paste + * it and parse to the end of the line. + */ + if (NULL != p->subtitle) { + texisplice(p, p->subtitle, strlen(p->subtitle), *pos); + parseeoln(p, pos); + } else + texiputchars(p, "Unknown description"); + teximacroclose(p); + p->seenvs = 1; +} +