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 static int memcmp_different_len(const void *s1, size_t n1, const void *s2, size_t n2)
  23 {
  24         int shared_len = min(n1, n2);
  25         int s = memcmp(s1, s2, shared_len);
  26         if (s != 0) {
  27                 return s;
  28         }
  29
  30         if (n1 < n2) {
  31                 return -1;
  32         } else if (n1 > n2) {
  33                 return 1;
  34         } else {
  35                 return 0;
  36         }
  37 }
  38
  39 enum Result { WHITE = 0, DRAW, BLACK };
  40 struct Element {
  41         char *bpfen;  // includes prev_board_hash
  42         int bpfen_len;
  43         char move[8];   // Na1xc3+
  44         Result result;
  45         int opening_num, white_elo, black_elo;
  46         int file_num;
  47         time_t timestamp;
  48         long start_position;
  49
  50         bool operator< (const Element& other) const {
  51                 int s = memcmp_different_len(bpfen, bpfen_len, other.bpfen, other.bpfen_len);
  52                 return s < 0;
  53         }
  54 };
  55
  56 struct ShardData {
  57         vector<Element> elems;
  58         unique_ptr<Arena> arena;  // Used to allocate bpfen.
  59         int num_written_subshards = 0;
  60 };
  61
  62 void write_subshard(const char *basename, ShardData* shard, int bucket)
  63 {
  64         string buf;  // Keep allocated.
  65         char filename[256];
  66         snprintf(filename, sizeof(filename), "%s.part%04d.subshard%04d",
  67                 basename, bucket, shard->num_written_subshards++);
  68         printf("Writing SSTable %s...\n", filename);
  69
  70         sort(shard->elems.begin(), shard->elems.end());
  71
  72         mtbl_writer_options* wopt = mtbl_writer_options_init();
  73         mtbl_writer_options_set_compression(wopt, MTBL_COMPRESSION_SNAPPY);
  74         mtbl_writer* mtbl = mtbl_writer_init(filename, wopt);
  75         Count c;
  76         unordered_set<string> moves;
  77         for (size_t i = 0; i < shard->elems.size(); ++i) {
  78                 const Element &e = shard->elems[i];
  79                 if (e.result == WHITE) {
  80                         c.set_white(c.white() + 1);
  81                 } else if (e.result == DRAW) {
  82                         c.set_draw(c.draw() + 1);
  83                 } else if (e.result == BLACK) {
  84                         c.set_black(c.black() + 1);
  85                 }
  86                 if (e.white_elo >= 100 && e.black_elo >= 100) {
  87                         c.set_sum_white_elo(c.sum_white_elo() + e.white_elo);
  88                         c.set_sum_black_elo(c.sum_black_elo() + e.black_elo);
  89                         c.set_num_elo(c.num_elo() + 1);
  90                 }
  91                 if (!c.has_first_timestamp() || e.timestamp < c.first_timestamp()) {
  92                         if (e.timestamp != DUMMY_TIMESTAMP) {
  93                                 c.set_first_timestamp(e.timestamp);
  94                         }
  95                         c.set_opening_num(e.opening_num);
  96                         c.set_pgn_file_num(e.file_num);
  97                         c.set_pgn_start_position(e.start_position);
  98                 }
  99                 if (!moves.count(e.move)) {
 100                         moves.insert(e.move);
 101                         c.add_move(e.move);
 102                 }
 103                 if (i == shard->elems.size() - 1 ||
 104                     e.bpfen_len != shard->elems[i + 1].bpfen_len ||
 105                     memcmp(e.bpfen, shard->elems[i + 1].bpfen, e.bpfen_len) != 0) {
 106                         c.SerializeToString(&buf);
 107                         mtbl_writer_add(mtbl,
 108                                 (const uint8_t *)e.bpfen, e.bpfen_len,
 109                                 (const uint8_t *)buf.data(), buf.size());
 110                         c = Count();
 111                         moves.clear();
 112                 }
 113         }
 114         mtbl_writer_destroy(&mtbl);
 115
 116         shard->elems.clear();
 117         shard->arena.reset(new Arena);
 118 }
 119
 120 int main(int argc, char **argv)
 121 {
 122         int num_buckets = atoi(argv[argc - 2]);
 123         size_t num_pos_per_subshard = atoi(argv[argc - 1]);  // 500000 is a reasonable value.
 124
 125         vector<ShardData> shards;
 126         shards.resize(num_buckets);
 127
 128         for (int i = 0; i < num_buckets; ++i) {
 129                 shards[i].elems.reserve(num_pos_per_subshard);
 130                 shards[i].arena.reset(new Arena);
 131         }
 132
 133         size_t num_elems = 0;
 134         for (int i = 1; i < argc - 3; ++i) {
 135                 FILE *fp;
 136                 if (strcmp(argv[i], "-") == 0) {
 137                         fp = stdin;
 138                 } else {
 139                         fp = fopen(argv[i], "rb");
 140                         if (fp == NULL) {
 141                                 perror(argv[i]);
 142                                 exit(1);
 143                         }
 144                 }
 145                 for ( ;; ) {
 146                         char bpfen[256];
 147
 148                         int bpfen_len = getc(fp);
 149                         if (bpfen_len == -1) {
 150                                 break;
 151                         }
 152                         if (bpfen_len >= int(sizeof(bpfen))) {
 153                                 fprintf(stderr, "Overlong BPFEN (%d bytes)\n", bpfen_len);
 154                 //              exit(1);
 155                                 break;
 156                         }
 157                         if (fread(bpfen, bpfen_len, 1, fp) != 1) {
 158                                 perror("fread()");
 159                 //              exit(1);
 160                                 break;
 161                         }
 162
 163                         int r = getc(fp);
 164                         if (r == -1) {
 165                                 perror("getc()");
 166                                 //exit(1);
 167                                 break;
 168                         }
 169
 170                         int opening_num, white_elo, black_elo, file_num;
 171                         time_t timestamp;
 172                         long start_position;
 173                         if (fread(&white_elo, sizeof(white_elo), 1, fp) != 1) {
 174                                 perror("fread()");
 175                                 //exit(1);
 176                                 break;
 177                         }
 178                         if (fread(&black_elo, sizeof(black_elo), 1, fp) != 1) {
 179                                 perror("fread()");
 180                                 //exit(1);
 181                                 break;
 182                         }
 183                         if (fread(&opening_num, sizeof(opening_num), 1, fp) != 1) {
 184                                 perror("fread()");
 185                                 //exit(1);
 186                                 break;
 187                         }
 188                         if (fread(&timestamp, sizeof(timestamp), 1, fp) != 1) {
 189                                 perror("fread()");
 190                                 //exit(1);
 191                                 break;
 192                         }
 193                         if (fread(&file_num, sizeof(file_num), 1, fp) != 1) {
 194                                 perror("fread()");
 195                                 //exit(1);
 196                                 break;
 197                         }
 198                         if (fread(&start_position, sizeof(start_position), 1, fp) != 1) {
 199                                 perror("fread()");
 200                                 //exit(1);
 201                                 break;
 202                         }
 203
 204
 205                         char move[8];
 206                         int l = getc(fp);
 207                         if (l == -1) {
 208                                 break;
 209                         }
 210                         if (l >= int(sizeof(move))) {
 211                                 fprintf(stderr, "Overlong move (%d bytes)\n", l);
 212                 //              exit(1);
 213                                 break;
 214                         }
 215                         if (fread(&move[0], l, 1, fp) != 1) {
 216                                 perror("fread()");
 217                 //              exit(1);
 218                                 break;
 219                         }
 220                         move[l] = 0;
 221
 222                         int bucket = hash_key_to_bucket(bpfen, bpfen_len, num_buckets);
 223                         Element e;
 224                         e.bpfen = shards[bucket].arena->alloc(bpfen_len);
 225                         memcpy(e.bpfen, bpfen, bpfen_len);
 226                         e.bpfen_len = bpfen_len;
 227                         strcpy(e.move, move);
 228                         e.result = Result(r);
 229                         e.opening_num = opening_num;
 230                         e.white_elo = white_elo;
 231                         e.black_elo = black_elo;
 232                         e.file_num = file_num;
 233                         e.timestamp = timestamp;
 234                         e.start_position = start_position;
 235                         shards[bucket].elems.push_back(e);
 236                         ++num_elems;
 237
 238                         if (shards[bucket].elems.size() >= num_pos_per_subshard) {
 239                                 write_subshard(argv[argc - 3], &shards[bucket], bucket);
 240                                 shards[bucket].elems.reserve(num_pos_per_subshard);
 241                         }
 242                 }
 243                 fclose(fp);
 244
 245                 printf("Read %ld elems\n", num_elems);
 246         }
 247
 248         for (int i = 0; i < num_buckets; ++i) {
 249                 write_subshard(argv[argc - 3], &shards[i], i);
 250         }
 251 }
 252