git.sesse.net Git - remoteglot-book/blob - binloader.cpp

   1 //#define _GLIBCXX_PARALLEL
   2
   3 // Usage: ./binloader IN1 IN2 IN3 ... OUT NUM_BUCKETS
   4
   5 #include <stdio.h>
   6 #include <vector>
   7 #include <mtbl.h>
   8 #include <algorithm>
   9 #include <utility>
  10 #include <memory>
  11 #include <string>
  12 #include <unordered_set>
  13 #include <string.h>
  14 #include "count.pb.h"
  15 #include "hash.h"
  16
  17 #define DUMMY_TIMESTAMP 32503680000
  18
  19 using namespace std;
  20
  21 enum Result { WHITE = 0, DRAW, BLACK };
  22 struct Element {
  23         string bpfen;
  24         string move;
  25         Result result;
  26         int opening_num, white_elo, black_elo;
  27         time_t timestamp;
  28         long start_position;
  29
  30         bool operator< (const Element& other) const {
  31                 return bpfen < other.bpfen;
  32         }
  33 };
  34
  35 int main(int argc, char **argv)
  36 {
  37         int num_buckets = atoi(argv[argc - 1]);
  38
  39         vector<vector<Element>> elems;
  40         elems.resize(num_buckets);
  41
  42         size_t num_elems = 0;
  43         for (int i = 1; i < argc - 2; ++i) {
  44                 FILE *fp = fopen(argv[i], "rb");
  45                 if (fp == NULL) {
  46                         perror(argv[i]);
  47                         exit(1);
  48                 }
  49                 for ( ;; ) {
  50                         int l = getc(fp);
  51                         if (l == -1) {
  52                                 break;
  53                         }
  54
  55                         string bpfen;
  56                         bpfen.resize(l);
  57                         if (fread(&bpfen[0], l, 1, fp) != 1) {
  58                                 perror("fread()");
  59                 //              exit(1);
  60                                 break;
  61                         }
  62
  63                         int r = getc(fp);
  64                         if (r == -1) {
  65                                 perror("getc()");
  66                                 //exit(1);
  67                                 break;
  68                         }
  69
  70                         int opening_num, white_elo, black_elo;
  71                         time_t timestamp;
  72                         long start_position;
  73                         if (fread(&white_elo, sizeof(white_elo), 1, fp) != 1) {
  74                                 perror("fread()");
  75                                 //exit(1);
  76                                 break;
  77                         }
  78                         if (fread(&black_elo, sizeof(black_elo), 1, fp) != 1) {
  79                                 perror("fread()");
  80                                 //exit(1);
  81                                 break;
  82                         }
  83                         if (fread(&opening_num, sizeof(opening_num), 1, fp) != 1) {
  84                                 perror("fread()");
  85                                 //exit(1);
  86                                 break;
  87                         }
  88                         if (fread(&timestamp, sizeof(timestamp), 1, fp) != 1) {
  89                                 perror("fread()");
  90                                 //exit(1);
  91                                 break;
  92                         }
  93                         if (fread(&start_position, sizeof(start_position), 1, fp) != 1) {
  94                                 perror("fread()");
  95                                 //exit(1);
  96                                 break;
  97                         }
  98
  99
 100                         l = getc(fp);
 101                         if (l == -1) {
 102                                 break;
 103                         }
 104                         string move;
 105                         move.resize(l);
 106                         if (fread(&move[0], l, 1, fp) != 1) {
 107                                 perror("fread()");
 108                 //              exit(1);
 109                                 break;
 110                         }
 111
 112                         int bucket = hash_key_to_bucket(bpfen.data(), bpfen.size(), num_buckets);
 113                         elems[bucket].emplace_back(Element {std::move(bpfen), std::move(move), Result(r), opening_num, white_elo, black_elo, timestamp, start_position});
 114                         ++num_elems;
 115                 }
 116                 fclose(fp);
 117
 118                 printf("Read %ld elems\n", num_elems);
 119         }
 120
 121         printf("Sorting...\n");
 122         for (int i = 0; i < num_buckets; ++i) {
 123                 sort(elems[i].begin(), elems[i].end());
 124         }
 125
 126         printf("Writing SSTables...\n");
 127         string buf;  // Keep allocated.
 128         for (int i = 0; i < num_buckets; ++i) {
 129                 char filename[256];
 130                 snprintf(filename, sizeof(filename), "%s.part%04d", argv[argc - 2], i);
 131
 132                 mtbl_writer_options* wopt = mtbl_writer_options_init();
 133                 mtbl_writer_options_set_compression(wopt, MTBL_COMPRESSION_SNAPPY);
 134                 mtbl_writer* mtbl = mtbl_writer_init(filename, wopt);
 135                 Count c;
 136                 unordered_set<string> moves;
 137                 for (size_t j = 0; j < elems[i].size(); ++j) {
 138                         const Element &e = elems[i][j];
 139                         if (e.result == WHITE) {
 140                                 c.set_white(c.white() + 1);
 141                         } else if (e.result == DRAW) {
 142                                 c.set_draw(c.draw() + 1);
 143                         } else if (e.result == BLACK) {
 144                                 c.set_black(c.black() + 1);
 145                         }
 146                         if (e.white_elo >= 100 && e.black_elo >= 100) {
 147                                 c.set_sum_white_elo(c.sum_white_elo() + e.white_elo);
 148                                 c.set_sum_black_elo(c.sum_black_elo() + e.black_elo);
 149                                 c.set_num_elo(c.num_elo() + 1);
 150                         }
 151                         if (!c.has_first_timestamp() || e.timestamp < c.first_timestamp()) {
 152                                 if (e.timestamp != DUMMY_TIMESTAMP) {
 153                                         c.set_first_timestamp(e.timestamp);
 154                                 }
 155                                 c.set_opening_num(e.opening_num);
 156                                 c.set_pgn_start_position(e.start_position);
 157                         }
 158                         if (!moves.count(e.move)) {
 159                                 moves.insert(e.move);
 160                                 c.add_move(e.move);
 161                         }
 162                         if (j == elems[i].size() - 1 || e.bpfen != elems[i][j + 1].bpfen) {
 163                                 c.SerializeToString(&buf);
 164                                 mtbl_writer_add(mtbl,
 165                                         (const uint8_t *)e.bpfen.data(), e.bpfen.size(),
 166                                         (const uint8_t *)buf.data(), buf.size());
 167                                 c = Count();
 168                                 moves.clear();
 169                         }
 170                 }
 171                 mtbl_writer_destroy(&mtbl);
 172         }
 173 }