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 "hash.h"
  16
  17 #define DUMMY_TIMESTAMP 32503680000
  18
  19 using namespace std;
  20
  21 static inline int memcmp_different_len(const void *s1, size_t n1, const void *s2, size_t n2)
  22 {
  23         size_t shared_len = min(n1, n2);
  24         if (shared_len >= 8) {
  25                 uint64_t a1 = *(const uint64_t *)s1;
  26                 uint64_t a2 = *(const uint64_t *)s2;
  27                 if (a1 != a2) {
  28                         a1 = __builtin_bswap64(a1);
  29                         a2 = __builtin_bswap64(a2);
  30                         return (a1 < a2) ? -1 : 1;
  31                 }
  32         }
  33
  34         int s = memcmp(s1, s2, shared_len);
  35         if (s != 0) {
  36                 return s;
  37         }
  38
  39         return n2 - n1;
  40 }
  41
  42 enum Result { WHITE = 0, DRAW, BLACK };
  43 struct Element {
  44         char bpfen[32];  // includes prev_board_hash
  45         int bpfen_len;
  46         char move[8];   // Na1xc3+
  47         Result result;
  48         int opening_num, white_elo, black_elo;
  49         int file_num;
  50         time_t timestamp;
  51         long start_position;
  52
  53         bool operator< (const Element& other) const {
  54                 int s = memcmp_different_len(bpfen, bpfen_len, other.bpfen, other.bpfen_len);
  55                 return s < 0;
  56         }
  57 };
  58
  59 struct ShardData {
  60         vector<Element> elems;
  61         int num_written_subshards = 0;
  62 };
  63
  64 void write_subshard(const char *basename, ShardData* shard, int bucket)
  65 {
  66         string buf;  // Keep allocated.
  67         char filename[256];
  68         snprintf(filename, sizeof(filename), "%s.part%04d.subshard%04d",
  69                 basename, bucket, shard->num_written_subshards++);
  70         printf("Writing SSTable %s...\n", filename);
  71
  72         sort(shard->elems.begin(), shard->elems.end());
  73
  74         mtbl_writer_options* wopt = mtbl_writer_options_init();
  75         mtbl_writer_options_set_compression(wopt, MTBL_COMPRESSION_SNAPPY);
  76         mtbl_writer* mtbl = mtbl_writer_init(filename, wopt);
  77         Count c;
  78         unordered_set<string> moves;
  79         for (size_t i = 0; i < shard->elems.size(); ++i) {
  80                 const Element &e = shard->elems[i];
  81                 if (e.result == WHITE) {
  82                         c.set_white(c.white() + 1);
  83                 } else if (e.result == DRAW) {
  84                         c.set_draw(c.draw() + 1);
  85                 } else if (e.result == BLACK) {
  86                         c.set_black(c.black() + 1);
  87                 }
  88                 if (e.white_elo >= 100 && e.black_elo >= 100) {
  89                         c.set_sum_white_elo(c.sum_white_elo() + e.white_elo);
  90                         c.set_sum_black_elo(c.sum_black_elo() + e.black_elo);
  91                         c.set_num_elo(c.num_elo() + 1);
  92                 }
  93                 if (!c.has_first_timestamp() || e.timestamp < c.first_timestamp()) {
  94                         if (e.timestamp != DUMMY_TIMESTAMP) {
  95                                 c.set_first_timestamp(e.timestamp);
  96                         }
  97                         c.set_opening_num(e.opening_num);
  98                         c.set_pgn_file_num(e.file_num);
  99                         c.set_pgn_start_position(e.start_position);
 100                 }
 101                 if (strlen(e.move) > 0 && !moves.count(e.move)) {
 102                         moves.insert(e.move);
 103                         c.add_move(e.move);
 104                 }
 105                 if (i == shard->elems.size() - 1 ||
 106                     e.bpfen_len != shard->elems[i + 1].bpfen_len ||
 107                     memcmp(e.bpfen, shard->elems[i + 1].bpfen, e.bpfen_len) != 0) {
 108                         c.SerializeToString(&buf);
 109                         mtbl_res res = mtbl_writer_add(mtbl,
 110                                 (const uint8_t *)e.bpfen, e.bpfen_len,
 111                                 (const uint8_t *)buf.data(), buf.size());
 112                         assert(res == mtbl_res_success);
 113                         c = Count();
 114                         moves.clear();
 115                 }
 116         }
 117         mtbl_writer_destroy(&mtbl);
 118
 119         shard->elems.clear();
 120 }
 121
 122 int main(int argc, char **argv)
 123 {
 124         int num_buckets = atoi(argv[argc - 2]);
 125         size_t num_pos_per_subshard = atoi(argv[argc - 1]);  // 500000 is a reasonable value.
 126
 127         vector<ShardData> shards;
 128         shards.resize(num_buckets);
 129
 130         for (int i = 0; i < num_buckets; ++i) {
 131                 shards[i].elems.reserve(num_pos_per_subshard);
 132         }
 133
 134         size_t num_elems = 0;
 135         for (int i = 1; i < argc - 3; ++i) {
 136                 FILE *fp;
 137                 if (strcmp(argv[i], "-") == 0) {
 138                         fp = stdin;
 139                 } else {
 140                         fp = fopen(argv[i], "rb");
 141                         if (fp == NULL) {
 142                                 perror(argv[i]);
 143                                 exit(1);
 144                         }
 145                 }
 146                 for ( ;; ) {
 147                         char bpfen[256];
 148
 149                         int bpfen_len = getc(fp);
 150                         if (bpfen_len == -1) {
 151                                 break;
 152                         }
 153                         if (bpfen_len <= 0) {
 154                                 fprintf(stderr, "Underlong BPFEN (%d bytes)\n", bpfen_len);
 155                                 exit(1);
 156                         }
 157                         if (bpfen_len >= 32) {
 158                                 fprintf(stderr, "Overlong BPFEN (%d bytes)\n", bpfen_len);
 159                                 exit(1);
 160                         }
 161                         if (fread(bpfen, bpfen_len, 1, fp) != 1) {
 162                                 perror("fread(bpfen)");
 163                                 exit(1);
 164                         }
 165
 166                         int r = getc(fp);
 167                         if (r == -1) {
 168                                 perror("getc()");
 169                                 exit(1);
 170                         }
 171
 172                         int opening_num, white_elo, black_elo, file_num;
 173                         time_t timestamp;
 174                         long start_position;
 175                         if (fread(&white_elo, sizeof(white_elo), 1, fp) != 1) {
 176                                 perror("fread(white_elo)");
 177                                 exit(1);
 178                         }
 179                         if (fread(&black_elo, sizeof(black_elo), 1, fp) != 1) {
 180                                 perror("fread(black_elo)");
 181                                 exit(1);
 182                         }
 183                         if (fread(&opening_num, sizeof(opening_num), 1, fp) != 1) {
 184                                 perror("fread(opening_num)");
 185                                 exit(1);
 186                         }
 187                         if (fread(&timestamp, sizeof(timestamp), 1, fp) != 1) {
 188                                 perror("fread(timestamp)");
 189                                 exit(1);
 190                         }
 191                         if (fread(&file_num, sizeof(file_num), 1, fp) != 1) {
 192                                 perror("fread(file_num)");
 193                                 exit(1);
 194                         }
 195                         if (fread(&start_position, sizeof(start_position), 1, fp) != 1) {
 196                                 perror("fread(start_position)");
 197                                 exit(1);
 198                         }
 199
 200
 201                         char move[8];
 202                         int l = getc(fp);
 203                         if (l == -1) {
 204                                 perror("getc(move_length)");
 205                                 break;
 206                         }
 207                         if (l >= int(sizeof(move))) {
 208                                 fprintf(stderr, "Overlong move (%d bytes)\n", l);
 209                                 exit(1);
 210                         }
 211                         if (l == 0) {
 212                                 move[0] = 0;
 213                         } else if (fread(&move[0], l, 1, fp) != 1) {
 214                                 perror("fread(move)");
 215                                 exit(1);
 216                         }
 217                         move[l] = 0;
 218
 219                         int bucket = hash_key_to_bucket(bpfen, bpfen_len, num_buckets);
 220                         Element e;
 221                         memcpy(e.bpfen, bpfen, bpfen_len);
 222                         e.bpfen_len = bpfen_len;
 223                         strcpy(e.move, move);
 224                         e.result = Result(r);
 225                         e.opening_num = opening_num;
 226                         e.white_elo = white_elo;
 227                         e.black_elo = black_elo;
 228                         e.file_num = file_num;
 229                         e.timestamp = timestamp;
 230                         e.start_position = start_position;
 231                         shards[bucket].elems.push_back(e);
 232                         ++num_elems;
 233
 234                         if (shards[bucket].elems.size() >= num_pos_per_subshard) {
 235                                 write_subshard(argv[argc - 3], &shards[bucket], bucket);
 236                                 shards[bucket].elems.reserve(num_pos_per_subshard);
 237                         }
 238                 }
 239                 fclose(fp);
 240
 241                 printf("Read %ld elems\n", num_elems);
 242         }
 243
 244         for (int i = 0; i < num_buckets; ++i) {
 245                 write_subshard(argv[argc - 3], &shards[i], i);
 246         }
 247 }
 248