#include "db.h"
#include "io_uring_engine.h"
#include "parse_trigrams.h"
+#include "turbopfor.h"
#include "unique_sort.h"
#include <algorithm>
#include <assert.h>
#include <chrono>
#include <fcntl.h>
+#include <fnmatch.h>
#include <functional>
#include <getopt.h>
#include <iosfwd>
using namespace std;
using namespace std::chrono;
-#define dprintf(...)
-//#define dprintf(...) fprintf(stderr, __VA_ARGS__);
+#define dprintf(...) \
+ do { \
+ if (use_debug) { \
+ fprintf(stderr, __VA_ARGS__); \
+ } \
+ } while (false)
-#include "turbopfor.h"
+#define DEFAULT_DBPATH "/var/lib/mlocate/plocate.db"
-const char *dbpath = "/var/lib/mlocate/plocate.db";
+const char *dbpath = DEFAULT_DBPATH;
bool ignore_case = false;
bool only_count = false;
bool print_nul = false;
+bool use_debug = false;
+bool patterns_are_regex = false;
+bool use_extended_regex = false;
int64_t limit_matches = numeric_limits<int64_t>::max();
class Serializer {
struct Needle {
enum { STRSTR,
- REGEX } type;
+ REGEX,
+ GLOB } type;
string str; // Filled in no matter what.
regex_t re; // For REGEX.
};
{
if (needle.type == Needle::STRSTR) {
return strstr(haystack, needle.str.c_str()) != nullptr;
+ } else if (needle.type == Needle::GLOB) {
+ int flags = ignore_case ? FNM_CASEFOLD : 0;
+ return fnmatch(needle.str.c_str(), haystack, flags) == 0;
} else {
assert(needle.type == Needle::REGEX);
return regexec(&needle.re, haystack, /*nmatch=*/0, /*pmatch=*/nullptr, /*flags=*/0) == 0;
// Need to wait for more.
if (ignore_case) {
dprintf(" ... %u reads left in OR group %u (%zu docids in list)\n",
- td->remaining_trigrams_to_read, td->index, td->docids.size());
+ td->remaining_trigrams_to_read, td->index, td->docids.size());
}
return false;
}
if (cur_candidates->empty()) {
if (ignore_case) {
dprintf(" ... all reads done for OR group %u (%zu docids)\n",
- td->index, td->docids.size());
+ td->index, td->docids.size());
}
*cur_candidates = move(td->docids);
} else {
if (ignore_case) {
if (cur_candidates->empty()) {
dprintf(" ... all reads done for OR group %u (%zu docids), intersected (none left, search is done)\n",
- td->index, td->docids.size());
+ td->index, td->docids.size());
return true;
} else {
dprintf(" ... all reads done for OR group %u (%zu docids), intersected (%zu left)\n",
- td->index, td->docids.size(), cur_candidates->size());
+ td->index, td->docids.size(), cur_candidates->size());
}
}
}
dprintf("Corpus init done after %.1f ms.\n", 1e3 * duration<float>(steady_clock::now() - start).count());
vector<TrigramDisjunction> trigram_groups;
- for (const Needle &needle : needles) {
- if (needle.str.size() < 3)
- continue;
- parse_trigrams(needle.str, ignore_case, &trigram_groups);
+ if (patterns_are_regex) {
+ // We could parse the regex to find trigrams that have to be there
+ // (there are actually known algorithms to deal with disjunctions
+ // and such, too), but for now, we just go brute force.
+ // Using locate with regexes is pretty niche.
+ } else {
+ for (const Needle &needle : needles) {
+ parse_trigrams(needle.str, ignore_case, &trigram_groups);
+ }
}
+
unique_sort(
&trigram_groups,
[](const TrigramDisjunction &a, const TrigramDisjunction &b) { return a.trigram_alternatives < b.trigram_alternatives; },
}
}
-regex_t needle_to_regex(const string &needle)
+string unescape_glob_to_plain_string(const string &needle)
{
- string escaped_needle;
- for (char ch : needle) {
- switch (ch) {
- // Directly from what regex(7) considers an “atom”.
- case '^':
- case '.':
- case '[':
- case '$':
- case '(':
- case ')':
- case '|':
- case '*':
- case '+':
- case '?':
- case '{':
- case '\\':
- escaped_needle.push_back('\\');
- // Fall through.
- default:
- escaped_needle.push_back(ch);
+ string unescaped;
+ for (size_t i = 0; i < needle.size(); i += read_unigram(needle, i).second) {
+ uint32_t ch = read_unigram(needle, i).first;
+ assert(ch != WILDCARD_UNIGRAM);
+ if (ch == PREMATURE_END_UNIGRAM) {
+ fprintf(stderr, "Pattern '%s' ended prematurely\n", needle.c_str());
+ exit(1);
}
+ unescaped.push_back(ch);
}
+ return unescaped;
+}
+
+regex_t compile_regex(const string &needle)
+{
regex_t re;
- int err;
+ int flags = REG_NOSUB;
if (ignore_case) {
- err = regcomp(&re, escaped_needle.c_str(), REG_NOSUB | REG_ICASE);
- } else {
- err = regcomp(&re, escaped_needle.c_str(), REG_NOSUB);
+ flags |= REG_ICASE;
}
+ if (use_extended_regex) {
+ flags |= REG_EXTENDED;
+ }
+ int err = regcomp(&re, needle.c_str(), flags);
if (err != 0) {
char errbuf[256];
regerror(err, &re, errbuf, sizeof(errbuf));
- fprintf(stderr, "Error when compiling regex for '%s': %s\n", needle.c_str(), errbuf);
+ fprintf(stderr, "Error when compiling regex '%s': %s\n", needle.c_str(), errbuf);
exit(1);
}
return re;
void usage()
{
- // The help text comes from mlocate.
- printf("Usage: plocate [OPTION]... PATTERN...\n");
- printf("\n");
- printf(" -c, --count only print number of found entries\n");
- printf(" -d, --database DBPATH use DBPATH instead of default database (which is\n");
- printf(" %s)\n", dbpath);
- printf(" -h, --help print this help\n");
- printf(" -i, --ignore-case ignore case distinctions when matching patterns\n");
- printf(" -l, --limit, -n LIMIT limit output (or counting) to LIMIT entries\n");
- printf(" -0, --null separate entries with NUL on output\n");
+ printf(
+ "Usage: plocate [OPTION]... PATTERN...\n"
+ "\n"
+ " -c, --count print number of matches instead of the matches\n"
+ " -d, --database DBPATH search for files in DBPATH\n"
+ " (default is " DEFAULT_DBPATH ")\n"
+ " -i, --ignore-case search case-insensitively\n"
+ " -l, --limit LIMIT stop after LIMIT matches\n"
+ " -0, --null delimit matches by NUL instead of newline\n"
+ " -r, --regexp interpret patterns as basic regexps (slow)\n"
+ " --regex interpret patterns as extended regexps (slow)\n"
+ " --help print this help\n"
+ " --version print version information\n");
+}
+
+void version()
+{
+ printf("plocate %s\n", PLOCATE_VERSION);
+ printf("Copyright 2020 Steinar H. Gunderson\n");
+ printf("License GPLv2+: GNU GPL version 2 or later <https://gnu.org/licenses/gpl.html>.\n");
+ printf("This is free software: you are free to change and redistribute it.\n");
+ printf("There is NO WARRANTY, to the extent permitted by law.\n");
+ exit(0);
}
int main(int argc, char **argv)
{
+ constexpr int EXTENDED_REGEX = 1000;
static const struct option long_options[] = {
{ "help", no_argument, 0, 'h' },
{ "count", no_argument, 0, 'c' },
{ "database", required_argument, 0, 'd' },
{ "ignore-case", no_argument, 0, 'i' },
{ "limit", required_argument, 0, 'l' },
- { nullptr, required_argument, 0, 'n' },
{ "null", no_argument, 0, '0' },
+ { "version", no_argument, 0, 'V' },
+ { "regexp", no_argument, 0, 'r' },
+ { "regex", no_argument, 0, EXTENDED_REGEX },
+ { "debug", no_argument, 0, 'D' }, // Not documented.
{ 0, 0, 0, 0 }
};
setlocale(LC_ALL, "");
for (;;) {
int option_index = 0;
- int c = getopt_long(argc, argv, "cd:hil:n:0", long_options, &option_index);
+ int c = getopt_long(argc, argv, "cd:hil:n:0VD", long_options, &option_index);
if (c == -1) {
break;
}
case '0':
print_nul = true;
break;
+ case 'r':
+ patterns_are_regex = true;
+ break;
+ case EXTENDED_REGEX:
+ patterns_are_regex = true;
+ use_extended_regex = true;
+ break;
+ case 'D':
+ use_debug = true;
+ break;
+ case 'V':
+ version();
+ break;
default:
exit(1);
}
}
+ if (use_debug) {
+ // Debug information would leak information about which files exist,
+ // so drop setgid before we open the file; one would either need to run
+ // as root, or use a locally-built file.
+ if (setgid(getgid()) != 0) {
+ perror("setgid");
+ exit(EXIT_FAILURE);
+ }
+ }
+
vector<Needle> needles;
for (int i = optind; i < argc; ++i) {
Needle needle;
needle.str = argv[i];
- if (ignore_case) {
- // strcasestr() doesn't handle locales correctly (even though LSB
- // claims it should), but somehow, the glibc regex engine does.
- // It's much slower than strstr() for non-case-sensitive searches, though
- // (even though it really ought to be faster, since it can precompile),
- // so only use it for that.
+
+
+ // See if there are any wildcard characters, which indicates we should treat it
+ // as an (anchored) glob.
+ bool any_wildcard = false;
+ for (size_t i = 0; i < needle.str.size(); i += read_unigram(needle.str, i).second) {
+ if (read_unigram(needle.str, i).first == WILDCARD_UNIGRAM) {
+ any_wildcard = true;
+ break;
+ }
+ }
+
+ if (patterns_are_regex) {
needle.type = Needle::REGEX;
- needle.re = needle_to_regex(argv[i]);
+ needle.re = compile_regex(needle.str);
+ } else if (any_wildcard) {
+ needle.type = Needle::GLOB;
+ } else if (ignore_case) {
+ // strcasestr() doesn't handle locales correctly (even though LSB
+ // claims it should), but somehow, fnmatch() does, and it's about
+ // the same speed as using a regex.
+ needle.type = Needle::GLOB;
+ needle.str = "*" + needle.str + "*";
} else {
needle.type = Needle::STRSTR;
+ needle.str = unescape_glob_to_plain_string(needle.str);
}
needles.push_back(move(needle));
}