X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=plocate.cpp;h=4f88b4bd74c2ffe230052997939a1cbe30a5ffb0;hb=ddc5817ee81be45ce5eab251e5773284150da416;hp=0d59076e8fc99efb4829384da1894e99c4399867;hpb=d5ba26d705460a7e37213eeb4954b2efed8bebf0;p=plocate diff --git a/plocate.cpp b/plocate.cpp index 0d59076..4f88b4b 100644 --- a/plocate.cpp +++ b/plocate.cpp @@ -1,7 +1,9 @@ #include "access_rx_cache.h" +#include "complete_pread.h" #include "db.h" #include "dprintf.h" #include "io_uring_engine.h" +#include "needle.h" #include "parse_trigrams.h" #include "serializer.h" #include "turbopfor.h" @@ -14,7 +16,6 @@ #include #include #include -#include #include #include #include @@ -42,44 +43,22 @@ using namespace std; using namespace std::chrono; -#define DEFAULT_DBPATH "/var/lib/mlocate/plocate.db" - -const char *dbpath = DEFAULT_DBPATH; +const char *dbpath = DBFILE; bool ignore_case = false; bool only_count = false; bool print_nul = false; bool use_debug = false; +bool flush_cache = false; bool patterns_are_regex = false; bool use_extended_regex = false; +bool match_basename = false; int64_t limit_matches = numeric_limits::max(); int64_t limit_left = numeric_limits::max(); +bool stdout_is_tty = false; steady_clock::time_point start; ZSTD_DDict *ddict = nullptr; -regex_t compile_regex(const string &needle); - -struct Needle { - enum { STRSTR, - REGEX, - GLOB } type; - string str; // Filled in no matter what. - regex_t re; // For REGEX. -}; - -bool matches(const Needle &needle, const char *haystack) -{ - 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; - } -} - class Corpus { public: Corpus(int fd, IOUringEngine *engine); @@ -103,8 +82,7 @@ public: Corpus::Corpus(int fd, IOUringEngine *engine) : fd(fd), engine(engine) { - // Enable to test cold-cache behavior (except for access()). - if (false) { + if (flush_cache) { off_t len = lseek(fd, 0, SEEK_END); if (len == -1) { perror("lseek"); @@ -127,6 +105,10 @@ Corpus::Corpus(int fd, IOUringEngine *engine) hdr.zstd_dictionary_offset_bytes = 0; hdr.zstd_dictionary_length_bytes = 0; } + if (hdr.max_version < 2) { + // This too. (We ignore the other max_version 2 fields.) + hdr.require_visibility = true; + } } Corpus::~Corpus() @@ -217,9 +199,19 @@ void scan_file_block(const vector &needles, string_view compressed, for (const char *filename = block.data(); filename != block.data() + block.size(); filename += strlen(filename) + 1) { + const char *haystack = filename; + if (match_basename) { + haystack = strrchr(filename, '/'); + if (haystack == nullptr) { + haystack = filename; + } else { + ++haystack; + } + } + bool found = true; for (const Needle &needle : needles) { - if (!matches(needle, filename)) { + if (!matches(needle, haystack)) { found = false; break; } @@ -242,7 +234,7 @@ void scan_file_block(const vector &needles, string_view compressed, size_t scan_docids(const vector &needles, const vector &docids, const Corpus &corpus, IOUringEngine *engine) { Serializer docids_in_order; - AccessRXCache access_rx_cache(engine); + AccessRXCache access_rx_cache(engine, corpus.get_hdr().require_visibility); atomic matched{ 0 }; for (size_t i = 0; i < docids.size(); ++i) { uint32_t docid = docids[i]; @@ -318,7 +310,7 @@ uint64_t scan_all_docids(const vector &needles, int fd, const Corpus &co } } - AccessRXCache access_rx_cache(nullptr); + AccessRXCache access_rx_cache(nullptr, corpus.get_hdr().require_visibility); Serializer serializer; uint32_t num_blocks = corpus.get_num_filename_blocks(); unique_ptr offsets(new uint64_t[num_blocks + 1]); @@ -512,6 +504,8 @@ void do_search_file(const vector &needles, const char *filename) // the pattern and done a union of them, but that's a lot of // work for fairly unclear gain.) uint64_t matched = scan_all_docids(needles, fd, corpus); + dprintf("Done in %.1f ms, found %" PRId64 " matches.\n", + 1e3 * duration(steady_clock::now() - start).count(), matched); if (only_count) { printf("%" PRId64 "\n", matched); } @@ -611,7 +605,7 @@ void do_search_file(const vector &needles, const char *filename) if (done) return; - uint32_t trgm __attribute__((unused)) = trgmptr.trgm; + uint32_t trgm = trgmptr.trgm; const unsigned char *pldata = reinterpret_cast(s.data()); size_t num = trgmptr.num_docids; decoded.resize(num); @@ -656,61 +650,28 @@ void do_search_file(const vector &needles, const char *filename) } } -string unescape_glob_to_plain_string(const string &needle) -{ - 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 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( "Usage: plocate [OPTION]... PATTERN...\n" "\n" + " -b, --basename search only the file name portion of path names\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" + " (default is " DBFILE ")\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" + " -w, --wholename search the entire path name (default; see -b)\n" " --help print this help\n" " --version print version information\n"); } void version() { - printf("plocate %s\n", PLOCATE_VERSION); + printf("%s %s\n", PACKAGE_NAME, PACKAGE_VERSION); printf("Copyright 2020 Steinar H. Gunderson\n"); printf("License GPLv2+: GNU GPL version 2 or later .\n"); printf("This is free software: you are free to change and redistribute it.\n"); @@ -721,9 +682,11 @@ void version() int main(int argc, char **argv) { constexpr int EXTENDED_REGEX = 1000; + constexpr int FLUSH_CACHE = 1001; static const struct option long_options[] = { { "help", no_argument, 0, 'h' }, { "count", no_argument, 0, 'c' }, + { "basename", no_argument, 0, 'b' }, { "database", required_argument, 0, 'd' }, { "ignore-case", no_argument, 0, 'i' }, { "limit", required_argument, 0, 'l' }, @@ -731,18 +694,24 @@ int main(int argc, char **argv) { "version", no_argument, 0, 'V' }, { "regexp", no_argument, 0, 'r' }, { "regex", no_argument, 0, EXTENDED_REGEX }, + { "wholename", no_argument, 0, 'w' }, { "debug", no_argument, 0, 'D' }, // Not documented. + // Enable to test cold-cache behavior (except for access()). Not documented. + { "flush-cache", no_argument, 0, FLUSH_CACHE }, { 0, 0, 0, 0 } }; setlocale(LC_ALL, ""); for (;;) { int option_index = 0; - int c = getopt_long(argc, argv, "cd:hil:n:0VD", long_options, &option_index); + int c = getopt_long(argc, argv, "bcd:hil:n:0rwVD", long_options, &option_index); if (c == -1) { break; } switch (c) { + case 'b': + match_basename = true; + break; case 'c': only_count = true; break; @@ -773,9 +742,15 @@ int main(int argc, char **argv) patterns_are_regex = true; use_extended_regex = true; break; + case 'w': + match_basename = false; // No-op unless -b is given first. + break; case 'D': use_debug = true; break; + case FLUSH_CACHE: + flush_cache = true; + break; case 'V': version(); break; @@ -784,16 +759,22 @@ int main(int argc, char **argv) } } - if (use_debug) { + if (use_debug || flush_cache) { // 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. + // as root, or use a locally-built file. Doing the same thing for + // flush_cache is mostly paranoia, in an attempt to prevent random users + // from making plocate slow for everyone else. if (setgid(getgid()) != 0) { perror("setgid"); exit(EXIT_FAILURE); } } + if (!print_nul) { + stdout_is_tty = isatty(1); + } + vector needles; for (int i = optind; i < argc; ++i) { Needle needle;