From 179cbff3ea0ec44e732bf8f748222d79c649cdac Mon Sep 17 00:00:00 2001 From: Kristaps Dzonsons Date: Sun, 20 Mar 2011 11:41:24 +0000 Subject: Split the document parsing sequence out of main.c and into read.c, putting the interface into mandoc.h. This effectively makes the function of main.c be command-line handling, invoking the parser, and sending its output to the output handler. The sequence of parsing (pfile(), pdesc(), etc.) has changed very little but for clean-up of some state variables (curp->fd, etc.). --- read.c | 611 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 611 insertions(+) create mode 100644 read.c (limited to 'read.c') diff --git a/read.c b/read.c new file mode 100644 index 00000000..d6e102d9 --- /dev/null +++ b/read.c @@ -0,0 +1,611 @@ +/* $Id: read.c,v 1.1 2011/03/20 11:41:24 kristaps Exp $ */ +/* + * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons + * Copyright (c) 2010, 2011 Ingo Schwarze + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ +#include +#include + +#include +#include +#include +#include +#include +#include +#include + +#include "mandoc.h" +#include "mdoc.h" +#include "man.h" +#include "roff.h" + +#ifndef MAP_FILE +#define MAP_FILE 0 +#endif + +#define REPARSE_LIMIT 1000 + +struct buf { + char *buf; /* binary input buffer */ + size_t sz; /* size of binary buffer */ +}; + +struct mparse { + enum mandoclevel file_status; /* status of current parse */ + int line; /* line number in the file */ + enum mparset inttype; /* which parser to use */ + struct man *pman; /* persistent man parser */ + struct mdoc *pmdoc; /* persistent mdoc parser */ + struct man *man; /* man parser */ + struct mdoc *mdoc; /* mdoc parser */ + struct roff *roff; /* roff parser (!NULL) */ + struct regset regs; /* roff registers */ + int reparse_count; /* finite interp. stack */ + mandocmsg mmsg; /* warning/error message handler */ + void *arg; /* argument to mmsg */ + mevt_open evt_open; /* file-open event */ + mevt_close evt_close; /* file-close event */ + const char *svfile; +}; + +static void resize_buf(struct buf *, size_t); +static void mparse_buf_r(struct mparse *, struct buf, int); +static void mparse_readfd_r(struct mparse *, int, const char *, int); +static void pset(const char *, int, struct mparse *); +static void pdesc(struct mparse *, const char *, int); +static int read_whole_file(const char *, int, struct buf *, int *); +static void mparse_end(struct mparse *); + +static void +resize_buf(struct buf *buf, size_t initial) +{ + + buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial; + buf->buf = mandoc_realloc(buf->buf, buf->sz); +} + +static void +pset(const char *buf, int pos, struct mparse *curp) +{ + int i; + + /* + * Try to intuit which kind of manual parser should be used. If + * passed in by command-line (-man, -mdoc), then use that + * explicitly. If passed as -mandoc, then try to guess from the + * line: either skip dot-lines, use -mdoc when finding `.Dt', or + * default to -man, which is more lenient. + * + * Separate out pmdoc/pman from mdoc/man: the first persists + * through all parsers, while the latter is used per-parse. + */ + + if ('.' == buf[0] || '\'' == buf[0]) { + for (i = 1; buf[i]; i++) + if (' ' != buf[i] && '\t' != buf[i]) + break; + if ('\0' == buf[i]) + return; + } + + switch (curp->inttype) { + case (MPARSE_MDOC): + if (NULL == curp->pmdoc) + curp->pmdoc = mdoc_alloc + (&curp->regs, curp->arg, curp->mmsg); + assert(curp->pmdoc); + curp->mdoc = curp->pmdoc; + return; + case (MPARSE_MAN): + if (NULL == curp->pman) + curp->pman = man_alloc + (&curp->regs, curp->arg, curp->mmsg); + assert(curp->pman); + curp->man = curp->pman; + return; + default: + break; + } + + if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) { + if (NULL == curp->pmdoc) + curp->pmdoc = mdoc_alloc + (&curp->regs, curp->arg, curp->mmsg); + assert(curp->pmdoc); + curp->mdoc = curp->pmdoc; + return; + } + + if (NULL == curp->pman) + curp->pman = man_alloc + (&curp->regs, curp->arg, curp->mmsg); + assert(curp->pman); + curp->man = curp->pman; +} + +/* + * Main parse routine for an opened file. This is called for each + * opened file and simply loops around the full input file, possibly + * nesting (i.e., with `so'). + */ +static void +mparse_buf_r(struct mparse *curp, struct buf blk, int start) +{ + const struct tbl_span *span; + struct buf ln; + enum rofferr rr; + int i, of, rc; + int pos; /* byte number in the ln buffer */ + int lnn; /* line number in the real file */ + unsigned char c; + + memset(&ln, 0, sizeof(struct buf)); + + lnn = curp->line; + pos = 0; + + for (i = 0; i < (int)blk.sz; ) { + if (0 == pos && '\0' == blk.buf[i]) + break; + + if (start) { + curp->line = lnn; + curp->reparse_count = 0; + } + + while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) { + + /* + * When finding an unescaped newline character, + * leave the character loop to process the line. + * Skip a preceding carriage return, if any. + */ + + if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz && + '\n' == blk.buf[i + 1]) + ++i; + if ('\n' == blk.buf[i]) { + ++i; + ++lnn; + break; + } + + /* + * Warn about bogus characters. If you're using + * non-ASCII encoding, you're screwing your + * readers. Since I'd rather this not happen, + * I'll be helpful and drop these characters so + * we don't display gibberish. Note to manual + * writers: use special characters. + */ + + c = (unsigned char) blk.buf[i]; + + if ( ! (isascii(c) && + (isgraph(c) || isblank(c)))) { + curp->mmsg(MANDOCERR_BADCHAR, curp->arg, + curp->line, pos, "ignoring byte"); + i++; + continue; + } + + /* Trailing backslash = a plain char. */ + + if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) { + if (pos >= (int)ln.sz) + resize_buf(&ln, 256); + ln.buf[pos++] = blk.buf[i++]; + continue; + } + + /* + * Found escape and at least one other character. + * When it's a newline character, skip it. + * When there is a carriage return in between, + * skip that one as well. + */ + + if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz && + '\n' == blk.buf[i + 2]) + ++i; + if ('\n' == blk.buf[i + 1]) { + i += 2; + ++lnn; + continue; + } + + if ('"' == blk.buf[i + 1]) { + i += 2; + /* Comment, skip to end of line */ + for (; i < (int)blk.sz; ++i) { + if ('\n' == blk.buf[i]) { + ++i; + ++lnn; + break; + } + } + + /* Backout trailing whitespaces */ + for (; pos > 0; --pos) { + if (ln.buf[pos - 1] != ' ') + break; + if (pos > 2 && ln.buf[pos - 2] == '\\') + break; + } + break; + } + + /* Some other escape sequence, copy & cont. */ + + if (pos + 1 >= (int)ln.sz) + resize_buf(&ln, 256); + + ln.buf[pos++] = blk.buf[i++]; + ln.buf[pos++] = blk.buf[i++]; + } + + if (pos >= (int)ln.sz) + resize_buf(&ln, 256); + + ln.buf[pos] = '\0'; + + /* + * A significant amount of complexity is contained by + * the roff preprocessor. It's line-oriented but can be + * expressed on one line, so we need at times to + * readjust our starting point and re-run it. The roff + * preprocessor can also readjust the buffers with new + * data, so we pass them in wholesale. + */ + + of = 0; + +rerun: + rr = roff_parseln + (curp->roff, curp->line, + &ln.buf, &ln.sz, of, &of); + + switch (rr) { + case (ROFF_REPARSE): + if (REPARSE_LIMIT >= ++curp->reparse_count) + mparse_buf_r(curp, ln, 0); + else + curp->mmsg(MANDOCERR_ROFFLOOP, curp->arg, + curp->line, pos, NULL); + pos = 0; + continue; + case (ROFF_APPEND): + pos = (int)strlen(ln.buf); + continue; + case (ROFF_RERUN): + goto rerun; + case (ROFF_IGN): + pos = 0; + continue; + case (ROFF_ERR): + assert(MANDOCLEVEL_FATAL <= curp->file_status); + break; + case (ROFF_SO): + mparse_readfd_r(curp, -1, ln.buf + of, 1); + if (MANDOCLEVEL_FATAL <= curp->file_status) + break; + pos = 0; + continue; + default: + break; + } + + /* + * If we encounter errors in the recursive parse, make + * sure we don't continue parsing. + */ + + if (MANDOCLEVEL_FATAL <= curp->file_status) + break; + + /* + * If input parsers have not been allocated, do so now. + * We keep these instanced betwen parsers, but set them + * locally per parse routine since we can use different + * parsers with each one. + */ + + if ( ! (curp->man || curp->mdoc)) + pset(ln.buf + of, pos - of, curp); + + /* + * Lastly, push down into the parsers themselves. One + * of these will have already been set in the pset() + * routine. + * If libroff returns ROFF_TBL, then add it to the + * currently open parse. Since we only get here if + * there does exist data (see tbl_data.c), we're + * guaranteed that something's been allocated. + * Do the same for ROFF_EQN. + */ + + rc = -1; + + if (ROFF_TBL == rr) + while (NULL != (span = roff_span(curp->roff))) { + rc = curp->man ? + man_addspan(curp->man, span) : + mdoc_addspan(curp->mdoc, span); + if (0 == rc) + break; + } + else if (ROFF_EQN == rr) + rc = curp->mdoc ? + mdoc_addeqn(curp->mdoc, + roff_eqn(curp->roff)) : + man_addeqn(curp->man, + roff_eqn(curp->roff)); + else if (curp->man || curp->mdoc) + rc = curp->man ? + man_parseln(curp->man, + curp->line, ln.buf, of) : + mdoc_parseln(curp->mdoc, + curp->line, ln.buf, of); + + if (0 == rc) { + assert(MANDOCLEVEL_FATAL <= curp->file_status); + break; + } + + /* Temporary buffers typically are not full. */ + + if (0 == start && '\0' == blk.buf[i]) + break; + + /* Start the next input line. */ + + pos = 0; + } + + free(ln.buf); +} + +static void +pdesc(struct mparse *curp, const char *file, int fd) +{ + struct buf blk; + int with_mmap; + + /* + * Run for each opened file; may be called more than once for + * each full parse sequence if the opened file is nested (i.e., + * from `so'). Simply sucks in the whole file and moves into + * the parse phase for the file. + */ + + if ( ! read_whole_file(file, fd, &blk, &with_mmap)) { + curp->file_status = MANDOCLEVEL_SYSERR; + return; + } + + /* Line number is per-file. */ + + curp->line = 1; + + mparse_buf_r(curp, blk, 1); + + if (with_mmap) + munmap(blk.buf, blk.sz); + else + free(blk.buf); +} + +static int +read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap) +{ + struct stat st; + size_t off; + ssize_t ssz; + + if (-1 == fstat(fd, &st)) { + perror(file); + return(0); + } + + /* + * If we're a regular file, try just reading in the whole entry + * via mmap(). This is faster than reading it into blocks, and + * since each file is only a few bytes to begin with, I'm not + * concerned that this is going to tank any machines. + */ + + if (S_ISREG(st.st_mode)) { + if (st.st_size >= (1U << 31)) { + fprintf(stderr, "%s: input too large\n", file); + return(0); + } + *with_mmap = 1; + fb->sz = (size_t)st.st_size; + fb->buf = mmap(NULL, fb->sz, PROT_READ, + MAP_FILE|MAP_SHARED, fd, 0); + if (fb->buf != MAP_FAILED) + return(1); + } + + /* + * If this isn't a regular file (like, say, stdin), then we must + * go the old way and just read things in bit by bit. + */ + + *with_mmap = 0; + off = 0; + fb->sz = 0; + fb->buf = NULL; + for (;;) { + if (off == fb->sz) { + if (fb->sz == (1U << 31)) { + fprintf(stderr, "%s: input too large\n", file); + break; + } + resize_buf(fb, 65536); + } + ssz = read(fd, fb->buf + (int)off, fb->sz - off); + if (ssz == 0) { + fb->sz = off; + return(1); + } + if (ssz == -1) { + perror(file); + break; + } + off += (size_t)ssz; + } + + free(fb->buf); + fb->buf = NULL; + return(0); +} + +static void +mparse_end(struct mparse *curp) +{ + + if (MANDOCLEVEL_FATAL <= curp->file_status) + return; + + if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) { + assert(MANDOCLEVEL_FATAL <= curp->file_status); + return; + } + + if (curp->man && ! man_endparse(curp->man)) { + assert(MANDOCLEVEL_FATAL <= curp->file_status); + return; + } + +#if 0 + /* NOTE a parser may not have been assigned, yet. */ + + if ( ! (curp->man || curp->mdoc)) { + /* FIXME: make into an mandoc.h error. */ + fprintf(stderr, "%s: Not a manual\n", curp->file); + curp->file_status = MANDOCLEVEL_FATAL; + goto cleanup; + } +#endif + + roff_endparse(curp->roff); +} + +static void +mparse_readfd_r(struct mparse *curp, int fd, const char *file, int re) +{ + const char *svfile; + + if ( ! (*curp->evt_open)(curp->arg, file)) { + curp->file_status = MANDOCLEVEL_SYSERR; + return; + } + + if (-1 == fd) + if (-1 == (fd = open(file, O_RDONLY, 0))) { + perror(file); + curp->file_status = MANDOCLEVEL_SYSERR; + return; + } + + svfile = curp->svfile; + curp->svfile = file; + + pdesc(curp, file, fd); + + if (0 == re && MANDOCLEVEL_FATAL > curp->file_status) + mparse_end(curp); + + if (STDIN_FILENO != fd && -1 == close(fd)) + perror(file); + + (*curp->evt_close)(curp->arg, svfile); + curp->svfile = svfile; +} + +enum mandoclevel +mparse_readfd(struct mparse *curp, int fd, const char *file) +{ + + mparse_readfd_r(curp, fd, file, 0); + return(curp->file_status); +} + +void +mparse_setstatus(struct mparse *curp, enum mandoclevel lvl) +{ + + if (curp->file_status < lvl) + curp->file_status = lvl; +} + +struct mparse * +mparse_alloc(enum mparset inttype, mevt_open eopen, + mevt_close eclose, mandocmsg mmsg, void *arg) +{ + struct mparse *curp; + + curp = mandoc_calloc(1, sizeof(struct mparse)); + + curp->mmsg = mmsg; + curp->arg = arg; + curp->inttype = inttype; + curp->evt_open = eopen; + curp->evt_close = eclose; + + curp->roff = roff_alloc(&curp->regs, arg, mmsg); + return(curp); +} + +void +mparse_reset(struct mparse *curp) +{ + + memset(&curp->regs, 0, sizeof(struct regset)); + + roff_reset(curp->roff); + + if (curp->mdoc) + mdoc_reset(curp->mdoc); + if (curp->man) + man_reset(curp->man); + + curp->file_status = MANDOCLEVEL_OK; + curp->mdoc = NULL; + curp->man = NULL; +} + +void +mparse_free(struct mparse *curp) +{ + + if (curp->pmdoc) + mdoc_free(curp->pmdoc); + if (curp->pman) + man_free(curp->pman); + if (curp->roff) + roff_free(curp->roff); + + free(curp); +} + +void +mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man) +{ + + *mdoc = curp->mdoc; + *man = curp->man; +} -- cgit v1.2.3-56-ge451