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

   1 //#define _GLIBCXX_PARALLEL
   2
   3 // Usage: ./binloader IN1 IN2 IN3 ... OUT NUM_BUCKETS NUM_POS_PER_SUBSHARD
   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 "arena.h"
  16 #include "hash.h"
  17
  18 #define DUMMY_TIMESTAMP 32503680000
  19
  20 using namespace std;
  21
  22 Arena arena;
  23
  24 enum Result { WHITE = 0, DRAW, BLACK };
  25 struct Element {
  26         char *bpfen;
  27         int bpfen_len;
  28         char move[8];   // Na1xc3+
  29         Result result;
  30         int opening_num, white_elo, black_elo;
  31         int file_num;
  32         time_t timestamp;
  33         long start_position;
  34
  35         bool operator< (const Element& other) const {
  36                 int shared_len = min(bpfen_len, other.bpfen_len);
  37                 int s = memcmp(bpfen, other.bpfen, shared_len);
  38                 if (s < 0) {
  39                         return true;
  40                 } else if (s > 0) {
  41                         return false;
  42                 } else {
  43                         return bpfen_len < other.bpfen_len;
  44                 }
  45         }
  46 };
  47
  48 struct ShardData {
  49         vector<Element> elems;
  50         unique_ptr<Arena> arena;  // Used to allocate bpfen.
  51         int num_written_subshards = 0;
  52 };
  53
  54 void write_subshard(const char *basename, ShardData* shard, int bucket)
  55 {
  56         string buf;  // Keep allocated.
  57         char filename[256];
  58         snprintf(filename, sizeof(filename), "%s.part%04d.subshard%04d",
  59                 basename, bucket, shard->num_written_subshards++);
  60         printf("Writing SSTable %s...\n", filename);
  61
  62         sort(shard->elems.begin(), shard->elems.end());
  63
  64         mtbl_writer_options* wopt = mtbl_writer_options_init();
  65         mtbl_writer_options_set_compression(wopt, MTBL_COMPRESSION_SNAPPY);
  66         mtbl_writer* mtbl = mtbl_writer_init(filename, wopt);
  67         Count c;
  68         unordered_set<string> moves;
  69         for (size_t i = 0; i < shard->elems.size(); ++i) {
  70                 const Element &e = shard->elems[i];
  71                 if (e.result == WHITE) {
  72                         c.set_white(c.white() + 1);
  73                 } else if (e.result == DRAW) {
  74                         c.set_draw(c.draw() + 1);
  75                 } else if (e.result == BLACK) {
  76                         c.set_black(c.black() + 1);
  77                 }
  78                 if (e.white_elo >= 100 && e.black_elo >= 100) {
  79                         c.set_sum_white_elo(c.sum_white_elo() + e.white_elo);
  80                         c.set_sum_black_elo(c.sum_black_elo() + e.black_elo);
  81                         c.set_num_elo(c.num_elo() + 1);
  82                 }
  83                 if (!c.has_first_timestamp() || e.timestamp < c.first_timestamp()) {
  84                         if (e.timestamp != DUMMY_TIMESTAMP) {
  85                                 c.set_first_timestamp(e.timestamp);
  86                         }
  87                         c.set_opening_num(e.opening_num);
  88                         c.set_pgn_file_num(e.file_num);
  89                         c.set_pgn_start_position(e.start_position);
  90                 }
  91                 if (!moves.count(e.move)) {
  92                         moves.insert(e.move);
  93                         c.add_move(e.move);
  94                 }
  95                 if (i == shard->elems.size() - 1 ||
  96                     e.bpfen_len != shard->elems[i + 1].bpfen_len ||
  97                     memcmp(e.bpfen, shard->elems[i + 1].bpfen, e.bpfen_len) != 0) {
  98                         c.SerializeToString(&buf);
  99                         mtbl_writer_add(mtbl,
 100                                 (const uint8_t *)e.bpfen, e.bpfen_len,
 101                                 (const uint8_t *)buf.data(), buf.size());
 102                         c = Count();
 103                         moves.clear();
 104                 }
 105         }
 106         mtbl_writer_destroy(&mtbl);
 107
 108         shard->elems.clear();
 109         shard->arena.reset(new Arena);
 110 }
 111
 112 int main(int argc, char **argv)
 113 {
 114         int num_buckets = atoi(argv[argc - 2]);
 115         size_t num_pos_per_subshard = atoi(argv[argc - 1]);  // 500000 is a reasonable value.
 116
 117         vector<ShardData> shards;
 118         shards.resize(num_buckets);
 119
 120         for (int i = 0; i < num_buckets; ++i) {
 121                 shards[i].elems.reserve(num_pos_per_subshard);
 122                 shards[i].arena.reset(new Arena);
 123         }
 124
 125         size_t num_elems = 0;
 126         for (int i = 1; i < argc - 3; ++i) {
 127                 FILE *fp;
 128                 if (strcmp(argv[i], "-") == 0) {
 129                         fp = stdin;
 130                 } else {
 131                         fp = fopen(argv[i], "rb");
 132                         if (fp == NULL) {
 133                                 perror(argv[i]);
 134                                 exit(1);
 135                         }
 136                 }
 137                 for ( ;; ) {
 138                         char bpfen[256];
 139
 140                         int bpfen_len = getc(fp);
 141                         if (bpfen_len == -1) {
 142                                 break;
 143                         }
 144                         if (bpfen_len >= int(sizeof(bpfen))) {
 145                                 fprintf(stderr, "Overlong BPFEN (%d bytes)\n", bpfen_len);
 146                 //              exit(1);
 147                                 break;
 148                         }
 149                         if (fread(bpfen, bpfen_len, 1, fp) != 1) {
 150                                 perror("fread()");
 151                 //              exit(1);
 152                                 break;
 153                         }
 154
 155                         int r = getc(fp);
 156                         if (r == -1) {
 157                                 perror("getc()");
 158                                 //exit(1);
 159                                 break;
 160                         }
 161
 162                         int opening_num, white_elo, black_elo, file_num;
 163                         time_t timestamp;
 164                         long start_position;
 165                         if (fread(&white_elo, sizeof(white_elo), 1, fp) != 1) {
 166                                 perror("fread()");
 167                                 //exit(1);
 168                                 break;
 169                         }
 170                         if (fread(&black_elo, sizeof(black_elo), 1, fp) != 1) {
 171                                 perror("fread()");
 172                                 //exit(1);
 173                                 break;
 174                         }
 175                         if (fread(&opening_num, sizeof(opening_num), 1, fp) != 1) {
 176                                 perror("fread()");
 177                                 //exit(1);
 178                                 break;
 179                         }
 180                         if (fread(&timestamp, sizeof(timestamp), 1, fp) != 1) {
 181                                 perror("fread()");
 182                                 //exit(1);
 183                                 break;
 184                         }
 185                         if (fread(&file_num, sizeof(file_num), 1, fp) != 1) {
 186                                 perror("fread()");
 187                                 //exit(1);
 188                                 break;
 189                         }
 190                         if (fread(&start_position, sizeof(start_position), 1, fp) != 1) {
 191                                 perror("fread()");
 192                                 //exit(1);
 193                                 break;
 194                         }
 195
 196
 197                         char move[8];
 198                         int l = getc(fp);
 199                         if (l == -1) {
 200                                 break;
 201                         }
 202                         if (l >= int(sizeof(move))) {
 203                                 fprintf(stderr, "Overlong move (%d bytes)\n", l);
 204                 //              exit(1);
 205                                 break;
 206                         }
 207                         if (fread(&move[0], l, 1, fp) != 1) {
 208                                 perror("fread()");
 209                 //              exit(1);
 210                                 break;
 211                         }
 212                         move[l] = 0;
 213
 214                         int bucket = hash_key_to_bucket(bpfen, bpfen_len, num_buckets);
 215                         Element e;
 216                         e.bpfen = shards[bucket].arena->alloc(bpfen_len);
 217                         memcpy(e.bpfen, bpfen, bpfen_len);
 218                         e.bpfen_len = bpfen_len;
 219                         strcpy(e.move, move);
 220                         e.result = Result(r);
 221                         e.opening_num = opening_num;
 222                         e.white_elo = white_elo;
 223                         e.black_elo = black_elo;
 224                         e.file_num = file_num;
 225                         e.timestamp = timestamp;
 226                         e.start_position = start_position;
 227                         shards[bucket].elems.push_back(e);
 228                         ++num_elems;
 229
 230                         if (shards[bucket].elems.size() >= num_pos_per_subshard) {
 231                                 write_subshard(argv[argc - 3], &shards[bucket], bucket);
 232                                 shards[bucket].elems.reserve(num_pos_per_subshard);
 233                         }
 234                 }
 235                 fclose(fp);
 236
 237                 printf("Read %ld elems\n", num_elems);
 238         }
 239
 240         for (int i = 0; i < num_buckets; ++i) {
 241                 write_subshard(argv[argc - 3], &shards[i], i);
 242         }
 243 }
 244