X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=plocate.cpp;h=a76c288a2bca86f0a11ebfec822d34721fa38335;hb=c427ecd63267946d66cf15808ed507d4f94c3566;hp=5b9db1807b6d1862b4520eddd961677103e782ff;hpb=69474d09da74fafcc90a5cd6919a1fd514458efb;p=plocate

diff --git a/plocate.cpp b/plocate.cpp
index 5b9db18..a76c288 100644
--- a/plocate.cpp
+++ b/plocate.cpp
@@ -1,4 +1,5 @@
 #include "db.h"
+#include "dprintf.h"
 #include "io_uring_engine.h"
 #include "parse_trigrams.h"
 #include "turbopfor.h"
@@ -11,6 +12,7 @@
 #include <fnmatch.h>
 #include <functional>
 #include <getopt.h>
+#include <inttypes.h>
 #include <iosfwd>
 #include <iterator>
 #include <limits>
@@ -33,13 +35,6 @@
 using namespace std;
 using namespace std::chrono;
 
-#define dprintf(...) \
-	do { \
-		if (use_debug) { \
-			fprintf(stderr, __VA_ARGS__); \
-		} \
-	} while (false)
-
 #define DEFAULT_DBPATH "/var/lib/mlocate/plocate.db"
 
 const char *dbpath = DEFAULT_DBPATH;
@@ -47,6 +42,8 @@ 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 {
@@ -96,7 +93,7 @@ void Serializer::release_current()
 
 struct Needle {
 	enum { STRSTR,
-	       REGEX,  // Not currently used.
+	       REGEX,
 	       GLOB } type;
 	string str;  // Filled in no matter what.
 	regex_t re;  // For REGEX.
@@ -400,11 +397,17 @@ void do_search_file(const vector<Needle> &needles, const char *filename)
 	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; },
@@ -430,7 +433,7 @@ void do_search_file(const vector<Needle> &needles, const char *filename)
 		// work for fairly unclear gain.)
 		uint64_t matched = scan_all_docids(needles, fd, corpus, &engine);
 		if (only_count) {
-			printf("%zu\n", matched);
+			printf("%" PRId64 "\n", matched);
 		}
 		return;
 	}
@@ -550,11 +553,11 @@ void do_search_file(const vector<Needle> &needles, const char *filename)
 	        1e3 * duration<float>(steady_clock::now() - start).count());
 
 	uint64_t matched = scan_docids(needles, cur_candidates, corpus, &engine);
-	dprintf("Done in %.1f ms, found %zu matches.\n",
+	dprintf("Done in %.1f ms, found %" PRId64 " matches.\n",
 	        1e3 * duration<float>(steady_clock::now() - start).count(), matched);
 
 	if (only_count) {
-		printf("%zu\n", matched);
+		printf("%" PRId64 "\n", matched);
 	}
 }
 
@@ -573,6 +576,26 @@ string unescape_glob_to_plain_string(const string &needle)
 	return unescaped;
 }
 
+regex_t compile_regex(const string &needle)
+{
+	regex_t re;
+	int flags = REG_NOSUB;
+	if (ignore_case) {
+		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 '%s': %s\n", needle.c_str(), errbuf);
+		exit(1);
+	}
+	return re;
+}
+
 void usage()
 {
 	printf(
@@ -584,6 +607,8 @@ void usage()
 		"  -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");
 }
@@ -600,6 +625,7 @@ void version()
 
 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' },
@@ -608,6 +634,8 @@ int main(int argc, char **argv)
 		{ "limit", required_argument, 0, 'l' },
 		{ "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 }
 	};
@@ -635,10 +663,21 @@ int main(int argc, char **argv)
 		case 'l':
 		case 'n':
 			limit_matches = atoll(optarg);
+			if (limit_matches <= 0) {
+				fprintf(stderr, "Error: limit must be a strictly positive number.\n");
+				exit(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;
@@ -675,7 +714,10 @@ int main(int argc, char **argv)
 			}
 		}
 
-		if (any_wildcard) {
+		if (patterns_are_regex) {
+			needle.type = Needle::REGEX;
+			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