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 #define VINT_IN
  19 #include "vint.h"
  20 #include "vp4.h"
  21
  22 #define P4NENC_BOUND(n) ((n+127)/128+(n+32)*sizeof(uint32_t))
  23 #define dprintf(...)
  24 //#define dprintf(...) fprintf(stderr, __VA_ARGS__);
  25
  26 using namespace std;
  27 using namespace std::chrono;
  28
  29 string zstd_compress(const string &src, string *tempbuf);
  30
  31 static inline uint32_t read_unigram(const string_view s, size_t idx)
  32 {
  33         if (idx < s.size()) {
  34                 return (unsigned char)s[idx];
  35         } else {
  36                 return 0;
  37         }
  38 }
  39
  40 static inline uint32_t read_trigram(const string_view s, size_t start)
  41 {
  42         return read_unigram(s, start) |
  43                 (read_unigram(s, start + 1) << 8) |
  44                 (read_unigram(s, start + 2) << 16);
  45 }
  46
  47 enum
  48 {
  49         DBE_NORMAL          = 0,    /* A non-directory file */
  50         DBE_DIRECTORY       = 1,    /* A directory */
  51         DBE_END             = 2   /* End of directory contents; contains no name */
  52 };
  53
  54 // From mlocate.
  55 struct db_header
  56 {
  57         uint8_t magic[8];
  58         uint32_t conf_size;
  59         uint8_t version;
  60         uint8_t check_visibility;
  61         uint8_t pad[2];
  62 };
  63
  64 // From mlocate.
  65 struct db_directory
  66 {
  67         uint64_t time_sec;
  68         uint32_t time_nsec;
  69         uint8_t pad[4];
  70 };
  71
  72 class PostingListBuilder {
  73 public:
  74         void add_docid(uint32_t docid);
  75         void finish();
  76
  77         string encoded;
  78         size_t num_docids = 0;
  79
  80 private:
  81         void write_header(uint32_t docid);
  82         void append_block();
  83
  84         uint32_t pending_docids[128];
  85         unsigned num_pending_docids = 0;
  86
  87         uint32_t last_block_end;
  88 };
  89
  90 void PostingListBuilder::add_docid(uint32_t docid)
  91 {
  92         // Deduplicate against the last inserted value, if any.
  93         if (num_pending_docids == 0) {
  94                 if (encoded.empty()) {
  95                         // Very first docid.
  96                         write_header(docid);
  97                         ++num_docids;
  98                         last_block_end = docid;
  99                         return;
 100                 } else if (docid == last_block_end) {
 101                         return;
 102                 }
 103         } else {
 104                 if (docid == pending_docids[num_pending_docids - 1]) {
 105                         return;
 106                 }
 107         }
 108
 109         pending_docids[num_pending_docids++] = docid;
 110         if (num_pending_docids == 128) {
 111                 append_block();
 112                 num_pending_docids = 0;
 113                 last_block_end = docid;
 114         }
 115         ++num_docids;
 116 }
 117
 118 void PostingListBuilder::finish()
 119 {
 120         if (num_pending_docids == 0) {
 121                 return;
 122         }
 123
 124         assert(!encoded.empty());  // write_header() should already have run.
 125
 126         // No interleaving for partial blocks.
 127         unsigned char buf[P4NENC_BOUND(128)];
 128         unsigned char *end = p4d1enc32(pending_docids, num_pending_docids, buf, last_block_end);
 129         encoded.append(reinterpret_cast<char *>(buf), reinterpret_cast<char *>(end));
 130 }
 131
 132 void PostingListBuilder::append_block()
 133 {
 134         unsigned char buf[P4NENC_BOUND(128)];
 135         assert(num_pending_docids == 128);
 136         unsigned char *end = p4d1enc128v32(pending_docids, 128, buf, last_block_end);
 137         encoded.append(reinterpret_cast<char *>(buf), reinterpret_cast<char *>(end));
 138 }
 139
 140 void PostingListBuilder::write_header(uint32_t docid)
 141 {
 142         unsigned char buf[P4NENC_BOUND(1)];
 143         size_t bytes = p4nd1enc128v32(&docid, 1, buf);
 144         encoded.append(reinterpret_cast<char *>(buf), bytes);
 145 }
 146
 147 class Corpus {
 148 public:
 149         Corpus(size_t block_size) : block_size(block_size) {}
 150         void add_file(string filename);
 151         void flush_block();
 152
 153         vector<string> filename_blocks;
 154         unordered_map<uint32_t, PostingListBuilder> invindex;
 155         size_t num_files = 0, num_files_in_block = 0, num_blocks = 0;
 156
 157 private:
 158         string current_block;
 159         string tempbuf;
 160         const size_t block_size;
 161 };
 162
 163 void Corpus::add_file(string filename)
 164 {
 165         ++num_files;
 166         if (!current_block.empty()) {
 167                 current_block.push_back('\0');
 168         }
 169         current_block += filename;
 170         if (++num_files_in_block == block_size) {
 171                 flush_block();
 172         }
 173 }
 174
 175 void Corpus::flush_block()
 176 {
 177         if (current_block.empty()) {
 178                 return;
 179         }
 180
 181         uint32_t docid = num_blocks;
 182
 183         // Create trigrams.
 184         const char *ptr = current_block.c_str();
 185         while (ptr < current_block.c_str() + current_block.size()) {
 186                 string_view s(ptr);
 187                 if (s.size() >= 3) {
 188                         for (size_t j = 0; j < s.size() - 2; ++j) {
 189                                 uint32_t trgm = read_trigram(s, j);
 190                                 invindex[trgm].add_docid(docid);
 191                         }
 192                 }
 193                 ptr += s.size() + 1;
 194         }
 195
 196         // Compress and add the filename block.
 197         filename_blocks.push_back(zstd_compress(current_block, &tempbuf));
 198
 199         current_block.clear();
 200         num_files_in_block = 0;
 201         ++num_blocks;
 202 }
 203
 204 const char *handle_directory(const char *ptr, Corpus *corpus)
 205 {
 206         ptr += sizeof(db_directory);
 207
 208         string dir_path = ptr;
 209         ptr += dir_path.size() + 1;
 210         if (dir_path == "/") {
 211                 dir_path = "";
 212         }
 213
 214         for ( ;; ) {
 215                 uint8_t type = *ptr++;
 216                 if (type == DBE_NORMAL) {
 217                         string filename = ptr;
 218                         corpus->add_file(dir_path + "/" + filename);
 219                         ptr += filename.size() + 1;
 220                 } else if (type == DBE_DIRECTORY) {
 221                         string dirname = ptr;
 222                         corpus->add_file(dir_path + "/" + dirname);
 223                         ptr += dirname.size() + 1;
 224                 } else {
 225                         return ptr;
 226                 }
 227         }
 228 }
 229
 230 void read_mlocate(const char *filename, Corpus *corpus)
 231 {
 232         int fd = open(filename, O_RDONLY);
 233         if (fd == -1) {
 234                 perror(filename);
 235                 exit(1);
 236         }
 237         off_t len = lseek(fd, 0, SEEK_END);
 238         if (len == -1) {
 239                 perror("lseek");
 240                 exit(1);
 241         }
 242         const char *data = (char *)mmap(nullptr, len, PROT_READ, MAP_SHARED, fd, /*offset=*/0);
 243         if (data == MAP_FAILED) {
 244                 perror("mmap");
 245                 exit(1);
 246         }
 247
 248         const db_header *hdr = (const db_header *)data;
 249
 250         // TODO: Care about the base path.
 251         string path = data + sizeof(db_header);
 252         uint64_t offset = sizeof(db_header) + path.size() + 1 + ntohl(hdr->conf_size);
 253
 254         const char *ptr = data + offset;
 255         while (ptr < data + len) {
 256                 ptr = handle_directory(ptr, corpus);
 257         }
 258
 259         munmap((void *)data, len);
 260         close(fd);
 261 }
 262
 263 string zstd_compress(const string &src, string *tempbuf)
 264 {
 265         size_t max_size = ZSTD_compressBound(src.size());
 266         if (tempbuf->size() < max_size) {
 267                 tempbuf->resize(max_size);
 268         }
 269         size_t size = ZSTD_compress(&(*tempbuf)[0], max_size, src.data(), src.size(), /*level=*/6);
 270         return string(tempbuf->data(), size);
 271 }
 272
 273 void do_build(const char *infile, const char *outfile, int block_size)
 274 {
 275         //steady_clock::time_point start = steady_clock::now();
 276
 277         Corpus corpus(block_size);
 278
 279         read_mlocate(infile, &corpus);
 280         if (false) {  // To read a plain text file.
 281                 FILE *fp = fopen(infile, "r");
 282                 while (!feof(fp)) {
 283                         char buf[1024];
 284                         if (fgets(buf, 1024, fp) == nullptr || feof(fp)) {
 285                                 break;
 286                         }
 287                         string s(buf);
 288                         if (s.back() == '\n') s.pop_back();
 289                         corpus.add_file(move(s));
 290                 }
 291                 fclose(fp);
 292         }
 293         corpus.flush_block();
 294         dprintf("Read %zu files from %s\n", corpus.num_files, infile);
 295
 296         size_t trigrams = 0, longest_posting_list = 0;
 297         size_t bytes_used = 0;
 298         for (auto &[trigram, pl_builder] : corpus.invindex) {
 299                 pl_builder.finish();
 300                 longest_posting_list = max(longest_posting_list, pl_builder.num_docids);
 301                 trigrams += pl_builder.num_docids;
 302                 bytes_used += pl_builder.encoded.size();
 303         }
 304         dprintf("%zu files, %zu different trigrams, %zu entries, avg len %.2f, longest %zu\n",
 305                 corpus.num_files, corpus.invindex.size(), trigrams, double(trigrams) / corpus.invindex.size(), longest_posting_list);
 306         dprintf("%zu bytes used for posting lists (%.2f bits/entry)\n", bytes_used, 8 * bytes_used / double(trigrams));
 307
 308         //steady_clock::time_point end = steady_clock::now();
 309         dprintf("Building posting lists took %.1f ms.\n\n", 1e3 * duration<float>(end - start).count());
 310
 311         vector<uint32_t> all_trigrams;
 312         for (auto &[trigram, pl_builder] : corpus.invindex) {
 313                 all_trigrams.push_back(trigram);
 314         }
 315         sort(all_trigrams.begin(), all_trigrams.end());
 316
 317         umask(0027);
 318         FILE *outfp = fopen(outfile, "wb");
 319
 320         // Write the header.
 321         uint64_t num_trigrams = corpus.invindex.size();  // 64 to get alignment.
 322         fwrite(&num_trigrams, sizeof(num_trigrams), 1, outfp);
 323
 324         // Find out where the posting lists will end.
 325         uint64_t offset = sizeof(uint64_t) * 2 + 16 * all_trigrams.size();
 326         uint64_t filename_index_offset = offset;
 327         for (auto &[trigram, pl_builder] : corpus.invindex) {
 328                 filename_index_offset += pl_builder.encoded.size();
 329         }
 330         fwrite(&filename_index_offset, sizeof(filename_index_offset), 1, outfp);
 331
 332         size_t bytes_for_trigrams = 0, bytes_for_posting_lists = 0, bytes_for_filename_index = 0, bytes_for_filenames = 0;
 333         for (uint32_t trgm : all_trigrams) {
 334                 // 16 bytes: trgm (4), number of docids (4), byte offset (8)
 335                 fwrite(&trgm, sizeof(trgm), 1, outfp);  // One byte wasted, but OK.
 336                 const PostingListBuilder &pl_builder = corpus.invindex[trgm];
 337                 uint32_t num_docids = pl_builder.num_docids;
 338                 fwrite(&num_docids, sizeof(num_docids), 1, outfp);
 339                 fwrite(&offset, sizeof(offset), 1, outfp);
 340                 offset += pl_builder.encoded.size();
 341                 bytes_for_trigrams += 16;
 342         }
 343
 344         // Write the actual posting lists.
 345         for (uint32_t trgm : all_trigrams) {
 346                 const string &encoded = corpus.invindex[trgm].encoded;
 347                 fwrite(encoded.data(), encoded.size(), 1, outfp);
 348                 bytes_for_posting_lists += encoded.size();
 349         }
 350
 351         // Stick an empty block at the end as sentinel.
 352         corpus.filename_blocks.push_back("");
 353
 354         // Write the offsets to the filenames.
 355         offset = filename_index_offset + corpus.filename_blocks.size() * sizeof(offset);
 356         for (const string &filename : corpus.filename_blocks) {
 357                 fwrite(&offset, sizeof(offset), 1, outfp);
 358                 offset += filename.size();
 359                 bytes_for_filename_index += sizeof(offset);
 360                 bytes_for_filenames += filename.size();
 361         }
 362
 363         // Write the actual filenames.
 364         for (const string &filename : corpus.filename_blocks) {
 365                 fwrite(filename.data(), filename.size(), 1, outfp);
 366         }
 367
 368         fclose(outfp);
 369
 370         //size_t total_bytes = (bytes_for_trigrams + bytes_for_posting_lists + bytes_for_filename_index + bytes_for_filenames);
 371
 372         dprintf("Block size:     %7d files\n", block_size);
 373         dprintf("Trigrams:       %'7.1f MB\n", bytes_for_trigrams / 1048576.0);
 374         dprintf("Posting lists:  %'7.1f MB\n", bytes_for_posting_lists / 1048576.0);
 375         dprintf("Filename index: %'7.1f MB\n", bytes_for_filename_index / 1048576.0);
 376         dprintf("Filenames:      %'7.1f MB\n", bytes_for_filenames / 1048576.0);
 377         dprintf("Total:          %'7.1f MB\n", total_bytes / 1048576.0);
 378         dprintf("\n");
 379 }
 380
 381 int main(int argc, char **argv)
 382 {
 383         do_build(argv[1], argv[2], 32);
 384         exit(EXIT_SUCCESS);
 385 }