X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=database-builder.cpp;h=419e012ca94c2625471680de17b67f07716ecec8;hb=fd6198891d6fd9642effc0843fef6f23b991af3e;hp=75b05f6a2cf1a898c014fd1227b4c6471c4f0148;hpb=2ce489a666c51d516b1a1b40cd420397350c8f9f;p=plocate diff --git a/database-builder.cpp b/database-builder.cpp index 75b05f6..419e012 100644 --- a/database-builder.cpp +++ b/database-builder.cpp @@ -36,7 +36,8 @@ public: void finish(); vector encoded; - size_t get_num_docids() const { + size_t get_num_docids() const + { // Updated only when we flush, so check that we're finished. assert(pending_deltas.empty()); return num_docids; @@ -208,6 +209,7 @@ private: std::unique_ptr invindex; FILE *outfp; + off_t outfp_pos; // Cheaper than calling ftell(outfp) all the time. std::string current_block; std::string tempbuf; const size_t block_size; @@ -219,9 +221,8 @@ private: std::string dir_times_compressed; }; - EncodingCorpus::EncodingCorpus(FILE *outfp, size_t block_size, ZSTD_CDict *cdict, bool store_dir_times) - : invindex(new PostingListBuilder *[NUM_TRIGRAMS]), outfp(outfp), block_size(block_size), store_dir_times(store_dir_times), cdict(cdict) + : invindex(new PostingListBuilder *[NUM_TRIGRAMS]), outfp(outfp), outfp_pos(ftell(outfp)), block_size(block_size), store_dir_times(store_dir_times), cdict(cdict) { fill(invindex.get(), invindex.get() + NUM_TRIGRAMS, nullptr); if (store_dir_times) { @@ -318,7 +319,7 @@ void EncodingCorpus::flush_block() ptr += 3; continue; } - for ( ;; ) { + for (;;) { // NOTE: Will read one byte past the end of the trigram, but it's OK, // since we always call it from contexts where there's a terminating zero byte. uint32_t trgm; @@ -336,12 +337,13 @@ void EncodingCorpus::flush_block() } // Compress and add the filename block. - filename_blocks.push_back(ftell(outfp)); + filename_blocks.push_back(outfp_pos); string compressed = zstd_compress(current_block, cdict, &tempbuf); if (fwrite(compressed.data(), compressed.size(), 1, outfp) != 1) { perror("fwrite()"); exit(1); } + outfp_pos += compressed.size(); current_block.clear(); num_files_in_block = 0; @@ -488,7 +490,7 @@ DatabaseBuilder::DatabaseBuilder(const char *outfile, gid_t owner, int block_siz int fd = -1; #ifdef O_TMPFILE fd = open(path.c_str(), O_WRONLY | O_TMPFILE, 0640); - if (fd == -1 && errno != EOPNOTSUPP) { + if (fd == -1 && errno != EOPNOTSUPP && errno != EISDIR) { perror(path.c_str()); exit(1); }