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