git.sesse.net Git - plocate/blob - plocate-build.cpp

   1 #include <assert.h>
   2 #include <stdio.h>
   3 #include <string.h>
   4 #include <algorithm>
   5 #include <unordered_map>
   6 #include <string>
   7 #include <vector>
   8 #include <chrono>
   9 #include <unistd.h>
  10 #include <fcntl.h>
  11 #include <sys/mman.h>
  12 #include <arpa/inet.h>
  13 #include <endian.h>
  14 #include <sys/types.h>
  15 #include <sys/stat.h>
  16 #include <zstd.h>
  17
  18 #include "vp4.h"
  19
  20 #define P4NENC_BOUND(n) ((n+127)/128+(n+32)*sizeof(uint32_t))
  21 #define dprintf(...)
  22 //#define dprintf(...) fprintf(stderr, __VA_ARGS__);
  23
  24 using namespace std;
  25 using namespace std::chrono;
  26
  27 string zstd_compress(const string &src, string *tempbuf);
  28
  29 static inline uint32_t read_unigram(const string_view s, size_t idx)
  30 {
  31         if (idx < s.size()) {
  32                 return (unsigned char)s[idx];
  33         } else {
  34                 return 0;
  35         }
  36 }
  37
  38 static inline uint32_t read_trigram(const string_view s, size_t start)
  39 {
  40         return read_unigram(s, start) |
  41                 (read_unigram(s, start + 1) << 8) |
  42                 (read_unigram(s, start + 2) << 16);
  43 }
  44
  45 enum
  46 {
  47         DBE_NORMAL          = 0,    /* A non-directory file */
  48         DBE_DIRECTORY       = 1,    /* A directory */
  49         DBE_END             = 2   /* End of directory contents; contains no name */
  50 };
  51
  52 // From mlocate.
  53 struct db_header
  54 {
  55         uint8_t magic[8];
  56         uint32_t conf_size;
  57         uint8_t version;
  58         uint8_t check_visibility;
  59         uint8_t pad[2];
  60 };
  61
  62 // From mlocate.
  63 struct db_directory
  64 {
  65         uint64_t time_sec;
  66         uint32_t time_nsec;
  67         uint8_t pad[4];
  68 };
  69
  70 class PostingListBuilder {
  71 public:
  72         void add_docid(uint32_t docid);
  73         void finish();
  74
  75         string encoded;
  76         size_t num_docids = 0;
  77
  78 private:
  79         void write_header(uint32_t docid);
  80         void append_block();
  81
  82         vector<uint32_t> pending_docids;
  83
  84         uint32_t last_block_end;
  85 };
  86
  87 void PostingListBuilder::add_docid(uint32_t docid)
  88 {
  89         // Deduplicate against the last inserted value, if any.
  90         if (pending_docids.empty()) {
  91                 if (encoded.empty()) {
  92                         // Very first docid.
  93                         write_header(docid);
  94                         ++num_docids;
  95                         last_block_end = docid;
  96                         return;
  97                 } else if (docid == last_block_end) {
  98                         return;
  99                 }
 100         } else {
 101                 if (docid == pending_docids.back()) {
 102                         return;
 103                 }
 104         }
 105
 106         pending_docids.push_back(docid);
 107         if (pending_docids.size() == 128) {
 108                 append_block();
 109                 pending_docids.clear();
 110                 last_block_end = docid;
 111         }
 112         ++num_docids;
 113 }
 114
 115 void PostingListBuilder::finish()
 116 {
 117         if (pending_docids.empty()) {
 118                 return;
 119         }
 120
 121         assert(!encoded.empty());  // write_header() should already have run.
 122
 123         // No interleaving for partial blocks.
 124         unsigned char buf[P4NENC_BOUND(128)];
 125         unsigned char *end = p4d1enc32(pending_docids.data(), pending_docids.size(), buf, last_block_end);
 126         encoded.append(reinterpret_cast<char *>(buf), reinterpret_cast<char *>(end));
 127 }
 128
 129 void PostingListBuilder::append_block()
 130 {
 131         unsigned char buf[P4NENC_BOUND(128)];
 132         assert(pending_docids.size() == 128);
 133         unsigned char *end = p4d1enc128v32(pending_docids.data(), 128, buf, last_block_end);
 134         encoded.append(reinterpret_cast<char *>(buf), reinterpret_cast<char *>(end));
 135 }
 136
 137 void PostingListBuilder::write_header(uint32_t docid)
 138 {
 139         unsigned char buf[P4NENC_BOUND(1)];
 140         size_t bytes = p4nd1enc128v32(&docid, 1, buf);
 141         encoded.append(reinterpret_cast<char *>(buf), bytes);
 142 }
 143
 144 class Corpus {
 145 public:
 146         Corpus(size_t block_size) : block_size(block_size) {}
 147         void add_file(string filename);
 148         void flush_block();
 149
 150         vector<string> filename_blocks;
 151         unordered_map<uint32_t, PostingListBuilder> invindex;
 152         size_t num_files = 0, num_files_in_block = 0, num_blocks = 0;
 153
 154 private:
 155         string current_block;
 156         string tempbuf;
 157         const size_t block_size;
 158 };
 159
 160 void Corpus::add_file(string filename)
 161 {
 162         ++num_files;
 163         if (!current_block.empty()) {
 164                 current_block.push_back('\0');
 165         }
 166         current_block += filename;
 167         if (++num_files_in_block == block_size) {
 168                 flush_block();
 169         }
 170 }
 171
 172 void Corpus::flush_block()
 173 {
 174         if (current_block.empty()) {
 175                 return;
 176         }
 177
 178         uint32_t docid = num_blocks;
 179
 180         // Create trigrams.
 181         const char *ptr = current_block.c_str();
 182         while (ptr < current_block.c_str() + current_block.size()) {
 183                 string_view s(ptr);
 184                 if (s.size() >= 3) {
 185                         for (size_t j = 0; j < s.size() - 2; ++j) {
 186                                 uint32_t trgm = read_trigram(s, j);
 187                                 invindex[trgm].add_docid(docid);
 188                         }
 189                 }
 190                 ptr += s.size() + 1;
 191         }
 192
 193         // Compress and add the filename block.
 194         filename_blocks.push_back(zstd_compress(current_block, &tempbuf));
 195
 196         current_block.clear();
 197         num_files_in_block = 0;
 198         ++num_blocks;
 199 }
 200
 201 const char *handle_directory(const char *ptr, Corpus *corpus)
 202 {
 203         ptr += sizeof(db_directory);
 204
 205         string dir_path = ptr;
 206         ptr += dir_path.size() + 1;
 207         if (dir_path == "/") {
 208                 dir_path = "";
 209         }
 210
 211         for ( ;; ) {
 212                 uint8_t type = *ptr++;
 213                 if (type == DBE_NORMAL) {
 214                         string filename = ptr;
 215                         corpus->add_file(dir_path + "/" + filename);
 216                         ptr += filename.size() + 1;
 217                 } else if (type == DBE_DIRECTORY) {
 218                         string dirname = ptr;
 219                         corpus->add_file(dir_path + "/" + dirname);
 220                         ptr += dirname.size() + 1;
 221                 } else {
 222                         return ptr;
 223                 }
 224         }
 225 }
 226
 227 void read_mlocate(const char *filename, Corpus *corpus)
 228 {
 229         int fd = open(filename, O_RDONLY);
 230         if (fd == -1) {
 231                 perror(filename);
 232                 exit(1);
 233         }
 234         off_t len = lseek(fd, 0, SEEK_END);
 235         if (len == -1) {
 236                 perror("lseek");
 237                 exit(1);
 238         }
 239         const char *data = (char *)mmap(nullptr, len, PROT_READ, MAP_SHARED, fd, /*offset=*/0);
 240         if (data == MAP_FAILED) {
 241                 perror("mmap");
 242                 exit(1);
 243         }
 244
 245         const db_header *hdr = (const db_header *)data;
 246
 247         // TODO: Care about the base path.
 248         string path = data + sizeof(db_header);
 249         uint64_t offset = sizeof(db_header) + path.size() + 1 + ntohl(hdr->conf_size);
 250
 251         const char *ptr = data + offset;
 252         while (ptr < data + len) {
 253                 ptr = handle_directory(ptr, corpus);
 254         }
 255
 256         munmap((void *)data, len);
 257         close(fd);
 258 }
 259
 260 string zstd_compress(const string &src, string *tempbuf)
 261 {
 262         size_t max_size = ZSTD_compressBound(src.size());
 263         if (tempbuf->size() < max_size) {
 264                 tempbuf->resize(max_size);
 265         }
 266         size_t size = ZSTD_compress(&(*tempbuf)[0], max_size, src.data(), src.size(), /*level=*/6);
 267         return string(tempbuf->data(), size);
 268 }
 269
 270 void do_build(const char *infile, const char *outfile, int block_size)
 271 {
 272         //steady_clock::time_point start = steady_clock::now();
 273
 274         Corpus corpus(block_size);
 275
 276         read_mlocate(infile, &corpus);
 277         if (false) {  // To read a plain text file.
 278                 FILE *fp = fopen(infile, "r");
 279                 while (!feof(fp)) {
 280                         char buf[1024];
 281                         if (fgets(buf, 1024, fp) == nullptr || feof(fp)) {
 282                                 break;
 283                         }
 284                         string s(buf);
 285                         if (s.back() == '\n') s.pop_back();
 286                         corpus.add_file(move(s));
 287                 }
 288                 fclose(fp);
 289         }
 290         corpus.flush_block();
 291         dprintf("Read %zu files from %s\n", corpus.num_files, infile);
 292
 293         size_t trigrams = 0, longest_posting_list = 0;
 294         size_t bytes_used = 0;
 295         for (auto &[trigram, pl_builder] : corpus.invindex) {
 296                 pl_builder.finish();
 297                 longest_posting_list = max(longest_posting_list, pl_builder.num_docids);
 298                 trigrams += pl_builder.num_docids;
 299                 bytes_used += pl_builder.encoded.size();
 300         }
 301         dprintf("%zu files, %zu different trigrams, %zu entries, avg len %.2f, longest %zu\n",
 302                 corpus.num_files, corpus.invindex.size(), trigrams, double(trigrams) / corpus.invindex.size(), longest_posting_list);
 303         dprintf("%zu bytes used for posting lists (%.2f bits/entry)\n", bytes_used, 8 * bytes_used / double(trigrams));
 304
 305         //steady_clock::time_point end = steady_clock::now();
 306         dprintf("Building posting lists took %.1f ms.\n\n", 1e3 * duration<float>(end - start).count());
 307
 308         vector<uint32_t> all_trigrams;
 309         for (auto &[trigram, pl_builder] : corpus.invindex) {
 310                 all_trigrams.push_back(trigram);
 311         }
 312         sort(all_trigrams.begin(), all_trigrams.end());
 313
 314         umask(0027);
 315         FILE *outfp = fopen(outfile, "wb");
 316
 317         // Write the header.
 318         uint64_t num_trigrams = corpus.invindex.size();  // 64 to get alignment.
 319         fwrite(&num_trigrams, sizeof(num_trigrams), 1, outfp);
 320
 321         // Find out where the posting lists will end.
 322         uint64_t offset = sizeof(uint64_t) * 2 + 16 * all_trigrams.size();
 323         uint64_t filename_index_offset = offset;
 324         for (auto &[trigram, pl_builder] : corpus.invindex) {
 325                 filename_index_offset += pl_builder.encoded.size();
 326         }
 327         fwrite(&filename_index_offset, sizeof(filename_index_offset), 1, outfp);
 328
 329         size_t bytes_for_trigrams = 0, bytes_for_posting_lists = 0, bytes_for_filename_index = 0, bytes_for_filenames = 0;
 330         for (uint32_t trgm : all_trigrams) {
 331                 // 16 bytes: trgm (4), number of docids (4), byte offset (8)
 332                 fwrite(&trgm, sizeof(trgm), 1, outfp);  // One byte wasted, but OK.
 333                 const PostingListBuilder &pl_builder = corpus.invindex[trgm];
 334                 uint32_t num_docids = pl_builder.num_docids;
 335                 fwrite(&num_docids, sizeof(num_docids), 1, outfp);
 336                 fwrite(&offset, sizeof(offset), 1, outfp);
 337                 offset += pl_builder.encoded.size();
 338                 bytes_for_trigrams += 16;
 339         }
 340
 341         // Write the actual posting lists.
 342         for (uint32_t trgm : all_trigrams) {
 343                 const string &encoded = corpus.invindex[trgm].encoded;
 344                 fwrite(encoded.data(), encoded.size(), 1, outfp);
 345                 bytes_for_posting_lists += encoded.size();
 346         }
 347
 348         // Stick an empty block at the end as sentinel.
 349         corpus.filename_blocks.push_back("");
 350
 351         // Write the offsets to the filenames.
 352         offset = filename_index_offset + corpus.filename_blocks.size() * sizeof(offset);
 353         for (const string &filename : corpus.filename_blocks) {
 354                 fwrite(&offset, sizeof(offset), 1, outfp);
 355                 offset += filename.size();
 356                 bytes_for_filename_index += sizeof(offset);
 357                 bytes_for_filenames += filename.size();
 358         }
 359
 360         // Write the actual filenames.
 361         for (const string &filename : corpus.filename_blocks) {
 362                 fwrite(filename.data(), filename.size(), 1, outfp);
 363         }
 364
 365         fclose(outfp);
 366
 367         //size_t total_bytes = (bytes_for_trigrams + bytes_for_posting_lists + bytes_for_filename_index + bytes_for_filenames);
 368
 369         dprintf("Block size:     %7d files\n", block_size);
 370         dprintf("Trigrams:       %'7.1f MB\n", bytes_for_trigrams / 1048576.0);
 371         dprintf("Posting lists:  %'7.1f MB\n", bytes_for_posting_lists / 1048576.0);
 372         dprintf("Filename index: %'7.1f MB\n", bytes_for_filename_index / 1048576.0);
 373         dprintf("Filenames:      %'7.1f MB\n", bytes_for_filenames / 1048576.0);
 374         dprintf("Total:          %'7.1f MB\n", total_bytes / 1048576.0);
 375         dprintf("\n");
 376 }
 377
 378 int main(int argc, char **argv)
 379 {
 380         do_build(argv[1], argv[2], 32);
 381         exit(EXIT_SUCCESS);
 382 }