4 #include <unordered_map>
11 #include <arpa/inet.h>
17 #define P4NENC_BOUND(n) ((n+127)/128+(n+32)*sizeof(uint32_t))
20 using namespace std::chrono;
23 //#define dprintf(...) fprintf(stderr, __VA_ARGS__);
25 static inline uint32_t read_unigram(const string &s, size_t idx)
28 return (unsigned char)s[idx];
34 static inline uint32_t read_trigram(const string &s, size_t start)
36 return read_unigram(s, start) |
37 (read_unigram(s, start + 1) << 8) |
38 (read_unigram(s, start + 2) << 16);
41 bool has_access(const char *filename, unordered_map<string, bool> *access_rx_cache)
43 const char *end = strchr(filename + 1, '/');
44 while (end != nullptr) {
45 string parent_path(filename, end);
46 auto it = access_rx_cache->find(parent_path);
48 if (it == access_rx_cache->end()) {
49 ok = access(parent_path.c_str(), R_OK | X_OK) == 0;
50 access_rx_cache->emplace(move(parent_path), ok);
57 end = strchr(end + 1, '/');
73 const Trigram *find_trigram(uint32_t trgm) const;
74 const unsigned char *get_compressed_posting_list(const Trigram *trigram) const;
75 string_view get_compressed_filename_block(uint32_t docid) const;
81 const uint64_t *filename_offsets;
82 const Trigram *trgm_begin, *trgm_end;
85 Corpus::Corpus(int fd)
88 len = lseek(fd, 0, SEEK_END);
93 data = (char *)mmap(nullptr, len, PROT_READ, MAP_SHARED, fd, /*offset=*/0);
94 if (data == MAP_FAILED) {
99 uint64_t num_trigrams = *(const uint64_t *)data;
100 uint64_t filename_index_offset = *(const uint64_t *)(data + sizeof(uint64_t));
101 filename_offsets = (const uint64_t *)(data + filename_index_offset);
103 trgm_begin = (Trigram *)(data + sizeof(uint64_t) * 2);
104 trgm_end = trgm_begin + num_trigrams;
109 munmap((void *)data, len);
113 const Trigram *Corpus::find_trigram(uint32_t trgm) const
115 const Trigram *trgmptr = lower_bound(trgm_begin, trgm_end, trgm, [](const Trigram &trgm, uint32_t t) {
116 return trgm.trgm < t;
118 if (trgmptr == trgm_end || trgmptr->trgm != trgm) {
124 const unsigned char *Corpus::get_compressed_posting_list(const Trigram *trgmptr) const
126 return reinterpret_cast<const unsigned char *>(data + trgmptr->offset);
129 string_view Corpus::get_compressed_filename_block(uint32_t docid) const
131 const char *compressed = (const char *)(data + filename_offsets[docid]);
132 size_t compressed_size = filename_offsets[docid + 1] - filename_offsets[docid]; // Allowed we have a sentinel block at the end.
133 return {compressed, compressed_size};
136 size_t scan_docid(const string &needle, uint32_t docid, const Corpus &corpus, unordered_map<string, bool> *access_rx_cache)
138 string_view compressed = corpus.get_compressed_filename_block(docid);
142 block.resize(ZSTD_getFrameContentSize(compressed.data(), compressed.size()) + 1);
144 ZSTD_decompress(&block[0], block.size(), compressed.data(), compressed.size());
145 block[block.size() - 1] = '\0';
147 for (const char *filename = block.data();
148 filename != block.data() + block.size();
149 filename += strlen(filename) + 1) {
150 if (strstr(filename, needle.c_str()) == nullptr) {
153 if (has_access(filename, access_rx_cache)) {
155 printf("%s\n", filename);
161 void do_search_file(const string &needle, const char *filename)
163 int fd = open(filename, O_RDONLY);
170 if (setgid(getgid()) != 0) {
175 //steady_clock::time_point start = steady_clock::now();
176 if (access("/", R_OK | X_OK)) {
177 // We can't find anything, no need to bother...
183 vector<const Trigram *> trigrams;
184 for (size_t i = 0; i < needle.size() - 2; ++i) {
185 uint32_t trgm = read_trigram(needle, i);
186 const Trigram *trgmptr = corpus.find_trigram(trgm);
187 if (trgmptr == nullptr) {
188 dprintf("trigram %06x isn't found, we abort the search\n", trgm);
191 trigrams.push_back(trgmptr);
193 sort(trigrams.begin(), trigrams.end());
195 auto last = unique(trigrams.begin(), trigrams.end());
196 trigrams.erase(last, trigrams.end());
198 sort(trigrams.begin(), trigrams.end(), [&](const Trigram *a, const Trigram *b) {
199 return a->num_docids < b->num_docids;
202 vector<uint32_t> in1, in2, out;
203 for (const Trigram *trgmptr : trigrams) {
204 //uint32_t trgm = trgmptr->trgm;
205 size_t num = trgmptr->num_docids;
206 const unsigned char *pldata = corpus.get_compressed_posting_list(trgmptr);
208 in1.resize(num + 128);
209 p4nd1dec128v32(const_cast<unsigned char *>(pldata), num, &in1[0]);
211 dprintf("trigram '%c%c%c' decoded to %zu entries\n", trgm & 0xff, (trgm >> 8) & 0xff, (trgm >> 16) & 0xff, num);
213 if (num > in1.size() * 100) {
214 dprintf("trigram '%c%c%c' has %zu entries, ignoring the rest (will weed out false positives later)\n",
215 trgm & 0xff, (trgm >> 8) & 0xff, (trgm >> 16) & 0xff, num);
219 if (in2.size() < num + 128) {
220 in2.resize(num + 128);
222 p4nd1dec128v32(const_cast<unsigned char *>(pldata), num, &in2[0]);
225 set_intersection(in1.begin(), in1.end(), in2.begin(), in2.begin() + num, back_inserter(out));
227 dprintf("trigram '%c%c%c' decoded to %zu entries, %zu left\n", trgm & 0xff, (trgm >> 8) & 0xff, (trgm >> 16) & 0xff, num, in1.size());
229 dprintf("no matches (intersection list is empty)\n");
234 steady_clock::time_point end = steady_clock::now();
236 dprintf("Intersection took %.1f ms. Doing final verification and printing:\n",
237 1e3 * duration<float>(end - start).count());
239 unordered_map<string, bool> access_rx_cache;
242 for (uint32_t docid : in1) {
243 matched += scan_docid(needle, docid, corpus, &access_rx_cache);
245 end = steady_clock::now();
246 dprintf("Done in %.1f ms, found %d matches.\n",
247 1e3 * duration<float>(end - start).count(), matched);
250 int main(int argc, char **argv)
252 do_search_file(argv[1], "/var/lib/mlocate/plocate.db");