NetBSD/gnu/usr.bin/groff/indxbib/indxbib.cc
2000-01-24 02:03:37 +00:00

743 lines
17 KiB
C++

// -*- C++ -*-
/* Copyright (C) 1989, 1990, 1991, 1992 Free Software Foundation, Inc.
Written by James Clark (jjc@jclark.com)
This file is part of groff.
groff is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free
Software Foundation; either version 2, or (at your option) any later
version.
groff is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License along
with groff; see the file COPYING. If not, write to the Free Software
Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <assert.h>
#include <errno.h>
#include "posix.h"
#include "lib.h"
#include "errarg.h"
#include "error.h"
#include "stringclass.h"
#include "cset.h"
#include "cmap.h"
#include "defs.h"
#include "index.h"
extern "C" {
// Sun's stdlib.h fails to declare this.
char *mktemp(char *);
}
#define DEFAULT_HASH_TABLE_SIZE 997
#define TEMP_INDEX_TEMPLATE "indxbibXXXXXX"
// (2^n - MALLOC_OVERHEAD) should be a good argument for malloc().
#define MALLOC_OVERHEAD 16
#ifdef BLOCK_SIZE
#undef BLOCK_SIZE
#endif
const int BLOCK_SIZE = ((1024 - MALLOC_OVERHEAD - sizeof(struct block *)
- sizeof(int)) / sizeof(int));
struct block {
block *next;
int used;
int v[BLOCK_SIZE];
block(block *p = 0) : next(p), used(0) { }
};
struct block;
union table_entry {
block *ptr;
int count;
};
struct word_list {
word_list *next;
char *str;
int len;
word_list(const char *, int, word_list *);
};
table_entry *hash_table;
int hash_table_size = DEFAULT_HASH_TABLE_SIZE;
// We make this the same size as hash_table so we only have to do one
// mod per key.
static word_list **common_words_table = 0;
char *key_buffer;
FILE *indxfp;
int ntags = 0;
string filenames;
char *temp_index_file = 0;
const char *ignore_fields = "XYZ";
const char *common_words_file = COMMON_WORDS_FILE;
int n_ignore_words = 100;
int truncate_len = 6;
int shortest_len = 3;
int max_keys_per_item = 100;
static void usage();
static void write_hash_table();
static void init_hash_table();
static void read_common_words_file();
static int store_key(char *s, int len);
static void possibly_store_key(char *s, int len);
static int do_whole_file(const char *filename);
static int do_file(const char *filename);
static void store_reference(int filename_index, int pos, int len);
static void check_integer_arg(char opt, const char *arg, int min, int *res);
static void store_filename(const char *);
static void fwrite_or_die(const void *ptr, int size, int nitems, FILE *fp);
static char *get_cwd();
extern "C" {
void cleanup();
long dir_name_max(const char *);
void catch_fatal_signals();
void ignore_fatal_signals();
}
int main(int argc, char **argv)
{
program_name = argv[0];
static char stderr_buf[BUFSIZ];
setbuf(stderr, stderr_buf);
const char *basename = 0;
typedef int (*parser_t)(const char *);
parser_t parser = do_file;
const char *directory = 0;
const char *foption = 0;
int opt;
while ((opt = getopt(argc, argv, "c:o:h:i:k:l:t:n:c:d:f:vw")) != EOF)
switch (opt) {
case 'c':
common_words_file = optarg;
break;
case 'd':
directory = optarg;
break;
case 'f':
foption = optarg;
break;
case 'h':
check_integer_arg('h', optarg, 1, &hash_table_size);
if (!is_prime(hash_table_size)) {
while (!is_prime(++hash_table_size))
;
warning("%1 not prime: using %2 instead", optarg, hash_table_size);
}
break;
case 'i':
ignore_fields = optarg;
break;
case 'k':
check_integer_arg('k', optarg, 1, &max_keys_per_item);
break;
case 'l':
check_integer_arg('l', optarg, 0, &shortest_len);
break;
case 'n':
check_integer_arg('n', optarg, 0, &n_ignore_words);
break;
case 'o':
basename = optarg;
break;
case 't':
check_integer_arg('t', optarg, 1, &truncate_len);
break;
case 'w':
parser = do_whole_file;
break;
case 'v':
{
extern const char *version_string;
fprintf(stderr, "GNU indxbib version %s\n", version_string);
fflush(stderr);
break;
}
case '?':
usage();
break;
default:
assert(0);
break;
}
if (optind >= argc && foption == 0)
fatal("no files and no -f option");
if (!directory) {
char *path = get_cwd();
store_filename(path);
a_delete path;
}
else
store_filename(directory);
init_hash_table();
store_filename(common_words_file);
store_filename(ignore_fields);
key_buffer = new char[truncate_len];
read_common_words_file();
if (!basename)
basename = optind < argc ? argv[optind] : DEFAULT_INDEX_NAME;
const char *p = strrchr(basename, '/');
long name_max;
if (p) {
char *dir = strsave(basename);
dir[p - basename] = '\0';
name_max = dir_name_max(dir);
a_delete dir;
}
else
name_max = dir_name_max(".");
const char *filename = p ? p + 1 : basename;
if (name_max >= 0 && strlen(filename) + sizeof(INDEX_SUFFIX) - 1 > name_max)
fatal("`%1.%2' is too long for a filename", filename, INDEX_SUFFIX);
if (p) {
p++;
temp_index_file = new char[p - basename + sizeof(TEMP_INDEX_TEMPLATE)];
memcpy(temp_index_file, basename, p - basename);
strcpy(temp_index_file + (p - basename), TEMP_INDEX_TEMPLATE);
}
else {
temp_index_file = strsave(TEMP_INDEX_TEMPLATE);
}
catch_fatal_signals();
int fd = mkstemp(temp_index_file);
if (fd < 0)
fatal("can't create temporary index file: %1", strerror(errno));
indxfp = fdopen(fd, "w");
if (indxfp == 0)
fatal("fdopen failed");
if (fseek(indxfp, sizeof(index_header), 0) < 0)
fatal("can't seek past index header: %1", strerror(errno));
int failed = 0;
if (foption) {
FILE *fp = stdin;
if (strcmp(foption, "-") != 0) {
errno = 0;
fp = fopen(foption, "r");
if (!fp)
fatal("can't open `%1': %2", foption, strerror(errno));
}
string path;
int lineno = 1;
for (;;) {
int c;
for (c = getc(fp); c != '\n' && c != EOF; c = getc(fp)) {
if (c == '\0')
error_with_file_and_line(foption, lineno,
"nul character in pathname ignored");
else
path += c;
}
if (path.length() > 0) {
path += '\0';
if (!(*parser)(path.contents()))
failed = 1;
path.clear();
}
if (c == EOF)
break;
lineno++;
}
if (fp != stdin)
fclose(fp);
}
for (int i = optind; i < argc; i++)
if (!(*parser)(argv[i]))
failed = 1;
write_hash_table();
if (fclose(indxfp) < 0)
fatal("error closing temporary index file: %1", strerror(errno));
char *index_file = new char[strlen(basename) + sizeof(INDEX_SUFFIX)];
strcpy(index_file, basename);
strcat(index_file, INDEX_SUFFIX);
#ifdef HAVE_RENAME
if (rename(temp_index_file, index_file) < 0)
fatal("can't rename temporary index file: %1", strerror(errno));
#else /* not HAVE_RENAME */
ignore_fatal_signals();
if (unlink(index_file) < 0) {
if (errno != ENOENT)
fatal("can't unlink `%1': %2", index_file, strerror(errno));
}
if (link(temp_index_file, index_file) < 0)
fatal("can't link temporary index file: %1", strerror(errno));
if (unlink(temp_index_file) < 0)
fatal("can't unlink temporary index file: %1", strerror(errno));
#endif /* not HAVE_RENAME */
temp_index_file = 0;
return failed;
}
static void usage()
{
fprintf(stderr,
"usage: %s [-vw] [-c file] [-d dir] [-f file] [-h n] [-i XYZ] [-k n]\n"
" [-l n] [-n n] [-o base] [-t n] [files...]\n",
program_name);
exit(1);
}
static void check_integer_arg(char opt, const char *arg, int min, int *res)
{
char *ptr;
long n = strtol(arg, &ptr, 10);
if (n == 0 && ptr == arg)
error("argument to -%1 not an integer", opt);
else if (n < min)
error("argument to -%1 must not be less than %2", opt, min);
else {
if (n > INT_MAX)
error("argument to -%1 greater than maximum integer", opt);
else if (*ptr != '\0')
error("junk after integer argument to -%1", opt);
*res = int(n);
}
}
static char *get_cwd()
{
char *buf;
int size = 12;
for (;;) {
buf = new char[size];
if (getcwd(buf, size))
break;
if (errno != ERANGE)
fatal("cannot get current working directory: %1", strerror(errno));
a_delete buf;
if (size == INT_MAX)
fatal("current working directory longer than INT_MAX");
if (size > INT_MAX/2)
size = INT_MAX;
else
size *= 2;
}
return buf;
}
word_list::word_list(const char *s, int n, word_list *p)
: next(p), len(n)
{
str = new char[n];
memcpy(str, s, n);
}
static void read_common_words_file()
{
if (n_ignore_words <= 0)
return;
errno = 0;
FILE *fp = fopen(common_words_file, "r");
if (!fp)
fatal("can't open `%1': %2", common_words_file, strerror(errno));
common_words_table = new word_list * [hash_table_size];
for (int i = 0; i < hash_table_size; i++)
common_words_table[i] = 0;
int count = 0;
int key_len = 0;
for (;;) {
int c = getc(fp);
while (c != EOF && !csalnum(c))
c = getc(fp);
if (c == EOF)
break;
do {
if (key_len < truncate_len)
key_buffer[key_len++] = cmlower(c);
c = getc(fp);
} while (c != EOF && csalnum(c));
if (key_len >= shortest_len) {
int h = hash(key_buffer, key_len) % hash_table_size;
common_words_table[h] = new word_list(key_buffer, key_len,
common_words_table[h]);
}
if (++count >= n_ignore_words)
break;
key_len = 0;
if (c == EOF)
break;
}
n_ignore_words = count;
fclose(fp);
}
static int do_whole_file(const char *filename)
{
errno = 0;
FILE *fp = fopen(filename, "r");
if (!fp) {
error("can't open `%1': %2", filename, strerror(errno));
return 0;
}
int count = 0;
int key_len = 0;
int c;
while ((c = getc(fp)) != EOF) {
if (csalnum(c)) {
key_len = 1;
key_buffer[0] = c;
while ((c = getc(fp)) != EOF) {
if (!csalnum(c))
break;
if (key_len < truncate_len)
key_buffer[key_len++] = c;
}
if (store_key(key_buffer, key_len)) {
if (++count >= max_keys_per_item)
break;
}
if (c == EOF)
break;
}
}
store_reference(filenames.length(), 0, 0);
store_filename(filename);
fclose(fp);
return 1;
}
static int do_file(const char *filename)
{
errno = 0;
FILE *fp = fopen(filename, "r");
if (fp == 0) {
error("can't open `%1': %2", filename, strerror(errno));
return 0;
}
int filename_index = filenames.length();
store_filename(filename);
enum {
START, // at the start of the file; also in between references
BOL, // in the middle of a reference, at the beginning of the line
PERCENT, // seen a percent at the beginning of the line
IGNORE, // ignoring a field
IGNORE_BOL, // at the beginning of a line ignoring a field
KEY, // in the middle of a key
DISCARD, // after truncate_len bytes of a key
MIDDLE // in between keys
} state = START;
// In states START, BOL, IGNORE_BOL, space_count how many spaces at
// the beginning have been seen. In states PERCENT, IGNORE, KEY,
// MIDDLE space_count must be 0.
int space_count = 0;
int byte_count = 0; // bytes read
int key_len = 0;
int ref_start = -1; // position of start of current reference
for (;;) {
int c = getc(fp);
if (c == EOF)
break;
byte_count++;
switch (state) {
case START:
if (c == ' ' || c == '\t') {
space_count++;
break;
}
if (c == '\n') {
space_count = 0;
break;
}
ref_start = byte_count - space_count - 1;
space_count = 0;
if (c == '%')
state = PERCENT;
else if (csalnum(c)) {
state = KEY;
key_buffer[0] = c;
key_len = 1;
}
else
state = MIDDLE;
break;
case BOL:
switch (c) {
case '%':
if (space_count > 0) {
space_count = 0;
state = MIDDLE;
}
else
state = PERCENT;
break;
case ' ':
case '\t':
space_count++;
break;
case '\n':
store_reference(filename_index, ref_start,
byte_count - 1 - space_count - ref_start);
state = START;
space_count = 0;
break;
default:
space_count = 0;
if (csalnum(c)) {
state = KEY;
key_buffer[0] = c;
key_len = 1;
}
else
state = MIDDLE;
}
break;
case PERCENT:
if (strchr(ignore_fields, c) != 0)
state = IGNORE;
else if (c == '\n')
state = BOL;
else
state = MIDDLE;
break;
case IGNORE:
if (c == '\n')
state = IGNORE_BOL;
break;
case IGNORE_BOL:
switch (c) {
case '%':
if (space_count > 0) {
state = IGNORE;
space_count = 0;
}
else
state = PERCENT;
break;
case ' ':
case '\t':
space_count++;
break;
case '\n':
store_reference(filename_index, ref_start,
byte_count - 1 - space_count - ref_start);
state = START;
space_count = 0;
break;
default:
space_count = 0;
state = IGNORE;
}
break;
case KEY:
if (csalnum(c)) {
if (key_len < truncate_len)
key_buffer[key_len++] = c;
else
state = DISCARD;
}
else {
possibly_store_key(key_buffer, key_len);
key_len = 0;
if (c == '\n')
state = BOL;
else
state = MIDDLE;
}
break;
case DISCARD:
if (!csalnum(c)) {
possibly_store_key(key_buffer, key_len);
key_len = 0;
if (c == '\n')
state = BOL;
else
state = MIDDLE;
}
break;
case MIDDLE:
if (csalnum(c)) {
state = KEY;
key_buffer[0] = c;
key_len = 1;
}
else if (c == '\n')
state = BOL;
break;
default:
assert(0);
}
}
switch (state) {
case START:
break;
case DISCARD:
case KEY:
possibly_store_key(key_buffer, key_len);
// fall through
case BOL:
case PERCENT:
case IGNORE_BOL:
case IGNORE:
case MIDDLE:
store_reference(filename_index, ref_start,
byte_count - ref_start - space_count);
break;
default:
assert(0);
}
fclose(fp);
return 1;
}
static void store_reference(int filename_index, int pos, int len)
{
tag t;
t.filename_index = filename_index;
t.start = pos;
t.length = len;
fwrite_or_die(&t, sizeof(t), 1, indxfp);
ntags++;
}
static void store_filename(const char *fn)
{
filenames += fn;
filenames += '\0';
}
static void init_hash_table()
{
hash_table = new table_entry[hash_table_size];
for (int i = 0; i < hash_table_size; i++)
hash_table[i].ptr = 0;
}
static void possibly_store_key(char *s, int len)
{
static int last_tagno = -1;
static int key_count;
if (last_tagno != ntags) {
last_tagno = ntags;
key_count = 0;
}
if (key_count < max_keys_per_item) {
if (store_key(s, len))
key_count++;
}
}
static int store_key(char *s, int len)
{
if (len < shortest_len)
return 0;
int is_number = 1;
for (int i = 0; i < len; i++)
if (!csdigit(s[i])) {
is_number = 0;
s[i] = cmlower(s[i]);
}
if (is_number && !(len == 4 && s[0] == '1' && s[1] == '9'))
return 0;
int h = hash(s, len) % hash_table_size;
if (common_words_table) {
for (word_list *ptr = common_words_table[h]; ptr; ptr = ptr->next)
if (len == ptr->len && memcmp(s, ptr->str, len) == 0)
return 0;
}
table_entry *pp = hash_table + h;
if (!pp->ptr)
pp->ptr = new block;
else if (pp->ptr->v[pp->ptr->used - 1] == ntags)
return 1;
else if (pp->ptr->used >= BLOCK_SIZE)
pp->ptr = new block(pp->ptr);
pp->ptr->v[(pp->ptr->used)++] = ntags;
return 1;
}
static void write_hash_table()
{
const int minus_one = -1;
int li = 0;
for (int i = 0; i < hash_table_size; i++) {
block *ptr = hash_table[i].ptr;
if (!ptr)
hash_table[i].count = -1;
else {
hash_table[i].count = li;
block *rev = 0;
while (ptr) {
block *tem = ptr;
ptr = ptr->next;
tem->next = rev;
rev = tem;
}
while (rev) {
fwrite_or_die(rev->v, sizeof(int), rev->used, indxfp);
li += rev->used;
block *tem = rev;
rev = rev->next;
delete tem;
}
fwrite_or_die(&minus_one, sizeof(int), 1, indxfp);
li += 1;
}
}
if (sizeof(table_entry) == sizeof(int))
fwrite_or_die(hash_table, sizeof(int), hash_table_size, indxfp);
else {
// write it out word by word
for (int i = 0; i < hash_table_size; i++)
fwrite_or_die(&hash_table[i].count, sizeof(int), 1, indxfp);
}
fwrite_or_die(filenames.contents(), 1, filenames.length(), indxfp);
if (fseek(indxfp, 0, 0) < 0)
fatal("error seeking on index file: %1", strerror(errno));
index_header h;
h.magic = INDEX_MAGIC;
h.version = INDEX_VERSION;
h.tags_size = ntags;
h.lists_size = li;
h.table_size = hash_table_size;
h.strings_size = filenames.length();
h.truncate = truncate_len;
h.shortest = shortest_len;
h.common = n_ignore_words;
fwrite_or_die(&h, sizeof(h), 1, indxfp);
}
static void fwrite_or_die(const void *ptr, int size, int nitems, FILE *fp)
{
if (fwrite(ptr, size, nitems, fp) != nitems)
fatal("fwrite failed: %1", strerror(errno));
}
void fatal_error_exit()
{
cleanup();
exit(3);
}
extern "C" {
void cleanup()
{
if (temp_index_file)
unlink(temp_index_file);
}
}