//#define _GLIBCXX_PARALLEL
-// Usage: ./binloader IN1 IN2 IN3 ... OUT NUM_BUCKETS
+// Usage: ./binloader IN1 IN2 IN3 ... OUT NUM_BUCKETS NUM_POS_PER_SUBSHARD
#include <stdio.h>
#include <vector>
#include <unordered_set>
#include <string.h>
#include "count.pb.h"
+#include "arena.h"
#include "hash.h"
#define DUMMY_TIMESTAMP 32503680000
using namespace std;
+Arena arena;
+
enum Result { WHITE = 0, DRAW, BLACK };
struct Element {
- string bpfen;
- string move;
+ char *bpfen;
+ int bpfen_len;
+ char move[8]; // Na1xc3+
Result result;
int opening_num, white_elo, black_elo;
+ int file_num;
time_t timestamp;
long start_position;
bool operator< (const Element& other) const {
- return bpfen < other.bpfen;
+ int shared_len = min(bpfen_len, other.bpfen_len);
+ int s = memcmp(bpfen, other.bpfen, shared_len);
+ if (s < 0) {
+ return true;
+ } else if (s > 0) {
+ return false;
+ } else {
+ return bpfen_len < other.bpfen_len;
+ }
}
};
+struct ShardData {
+ vector<Element> elems;
+ unique_ptr<Arena> arena; // Used to allocate bpfen.
+ int num_written_subshards = 0;
+};
+
+void write_subshard(const char *basename, ShardData* shard, int bucket)
+{
+ string buf; // Keep allocated.
+ char filename[256];
+ snprintf(filename, sizeof(filename), "%s.part%04d.subshard%04d",
+ basename, bucket, shard->num_written_subshards++);
+ printf("Writing SSTable %s...\n", filename);
+
+ sort(shard->elems.begin(), shard->elems.end());
+
+ mtbl_writer_options* wopt = mtbl_writer_options_init();
+ mtbl_writer_options_set_compression(wopt, MTBL_COMPRESSION_SNAPPY);
+ mtbl_writer* mtbl = mtbl_writer_init(filename, wopt);
+ Count c;
+ unordered_set<string> moves;
+ for (size_t i = 0; i < shard->elems.size(); ++i) {
+ const Element &e = shard->elems[i];
+ if (e.result == WHITE) {
+ c.set_white(c.white() + 1);
+ } else if (e.result == DRAW) {
+ c.set_draw(c.draw() + 1);
+ } else if (e.result == BLACK) {
+ c.set_black(c.black() + 1);
+ }
+ if (e.white_elo >= 100 && e.black_elo >= 100) {
+ c.set_sum_white_elo(c.sum_white_elo() + e.white_elo);
+ c.set_sum_black_elo(c.sum_black_elo() + e.black_elo);
+ c.set_num_elo(c.num_elo() + 1);
+ }
+ if (!c.has_first_timestamp() || e.timestamp < c.first_timestamp()) {
+ if (e.timestamp != DUMMY_TIMESTAMP) {
+ c.set_first_timestamp(e.timestamp);
+ }
+ c.set_opening_num(e.opening_num);
+ c.set_pgn_file_num(e.file_num);
+ c.set_pgn_start_position(e.start_position);
+ }
+ if (!moves.count(e.move)) {
+ moves.insert(e.move);
+ c.add_move(e.move);
+ }
+ if (i == shard->elems.size() - 1 ||
+ e.bpfen_len != shard->elems[i + 1].bpfen_len ||
+ memcmp(e.bpfen, shard->elems[i + 1].bpfen, e.bpfen_len) != 0) {
+ c.SerializeToString(&buf);
+ mtbl_writer_add(mtbl,
+ (const uint8_t *)e.bpfen, e.bpfen_len,
+ (const uint8_t *)buf.data(), buf.size());
+ c = Count();
+ moves.clear();
+ }
+ }
+ mtbl_writer_destroy(&mtbl);
+
+ shard->elems.clear();
+ shard->arena.reset(new Arena);
+}
+
int main(int argc, char **argv)
{
- int num_buckets = atoi(argv[argc - 1]);
+ int num_buckets = atoi(argv[argc - 2]);
+ size_t num_pos_per_subshard = atoi(argv[argc - 1]); // 500000 is a reasonable value.
- vector<vector<Element>> elems;
- elems.resize(num_buckets);
+ vector<ShardData> shards;
+ shards.resize(num_buckets);
+
+ for (int i = 0; i < num_buckets; ++i) {
+ shards[i].elems.reserve(num_pos_per_subshard);
+ shards[i].arena.reset(new Arena);
+ }
size_t num_elems = 0;
- for (int i = 1; i < argc - 2; ++i) {
- FILE *fp = fopen(argv[i], "rb");
- if (fp == NULL) {
- perror(argv[i]);
- exit(1);
+ for (int i = 1; i < argc - 3; ++i) {
+ FILE *fp;
+ if (strcmp(argv[i], "-") == 0) {
+ fp = stdin;
+ } else {
+ fp = fopen(argv[i], "rb");
+ if (fp == NULL) {
+ perror(argv[i]);
+ exit(1);
+ }
}
for ( ;; ) {
- int l = getc(fp);
- if (l == -1) {
+ char bpfen[256];
+
+ int bpfen_len = getc(fp);
+ if (bpfen_len == -1) {
+ break;
+ }
+ if (bpfen_len >= int(sizeof(bpfen))) {
+ fprintf(stderr, "Overlong BPFEN (%d bytes)\n", bpfen_len);
+ // exit(1);
break;
}
-
- string bpfen;
- bpfen.resize(l);
- if (fread(&bpfen[0], l, 1, fp) != 1) {
+ if (fread(bpfen, bpfen_len, 1, fp) != 1) {
perror("fread()");
// exit(1);
break;
break;
}
- int opening_num, white_elo, black_elo;
+ int opening_num, white_elo, black_elo, file_num;
time_t timestamp;
long start_position;
if (fread(&white_elo, sizeof(white_elo), 1, fp) != 1) {
//exit(1);
break;
}
+ if (fread(&file_num, sizeof(file_num), 1, fp) != 1) {
+ perror("fread()");
+ //exit(1);
+ break;
+ }
if (fread(&start_position, sizeof(start_position), 1, fp) != 1) {
perror("fread()");
//exit(1);
}
- l = getc(fp);
+ char move[8];
+ int l = getc(fp);
if (l == -1) {
break;
}
- string move;
- move.resize(l);
+ if (l >= int(sizeof(move))) {
+ fprintf(stderr, "Overlong move (%d bytes)\n", l);
+ // exit(1);
+ break;
+ }
if (fread(&move[0], l, 1, fp) != 1) {
perror("fread()");
// exit(1);
break;
}
-
- int bucket = hash_key_to_bucket(bpfen.data(), bpfen.size(), num_buckets);
- elems[bucket].emplace_back(Element {std::move(bpfen), std::move(move), Result(r), opening_num, white_elo, black_elo, timestamp, start_position});
+ move[l] = 0;
+
+ int bucket = hash_key_to_bucket(bpfen, bpfen_len, num_buckets);
+ Element e;
+ e.bpfen = shards[bucket].arena->alloc(bpfen_len);
+ memcpy(e.bpfen, bpfen, bpfen_len);
+ e.bpfen_len = bpfen_len;
+ strcpy(e.move, move);
+ e.result = Result(r);
+ e.opening_num = opening_num;
+ e.white_elo = white_elo;
+ e.black_elo = black_elo;
+ e.file_num = file_num;
+ e.timestamp = timestamp;
+ e.start_position = start_position;
+ shards[bucket].elems.push_back(e);
++num_elems;
+
+ if (shards[bucket].elems.size() >= num_pos_per_subshard) {
+ write_subshard(argv[argc - 3], &shards[bucket], bucket);
+ shards[bucket].elems.reserve(num_pos_per_subshard);
+ }
}
fclose(fp);
printf("Read %ld elems\n", num_elems);
}
- printf("Sorting...\n");
for (int i = 0; i < num_buckets; ++i) {
- sort(elems[i].begin(), elems[i].end());
- }
-
- printf("Writing SSTables...\n");
- string buf; // Keep allocated.
- for (int i = 0; i < num_buckets; ++i) {
- char filename[256];
- snprintf(filename, sizeof(filename), "%s.part%04d", argv[argc - 2], i);
-
- mtbl_writer_options* wopt = mtbl_writer_options_init();
- mtbl_writer_options_set_compression(wopt, MTBL_COMPRESSION_SNAPPY);
- mtbl_writer* mtbl = mtbl_writer_init(filename, wopt);
- Count c;
- unordered_set<string> moves;
- for (size_t j = 0; j < elems[i].size(); ++j) {
- const Element &e = elems[i][j];
- if (e.result == WHITE) {
- c.set_white(c.white() + 1);
- } else if (e.result == DRAW) {
- c.set_draw(c.draw() + 1);
- } else if (e.result == BLACK) {
- c.set_black(c.black() + 1);
- }
- if (e.white_elo >= 100 && e.black_elo >= 100) {
- c.set_sum_white_elo(c.sum_white_elo() + e.white_elo);
- c.set_sum_black_elo(c.sum_black_elo() + e.black_elo);
- c.set_num_elo(c.num_elo() + 1);
- }
- if (!c.has_first_timestamp() || e.timestamp < c.first_timestamp()) {
- if (e.timestamp != DUMMY_TIMESTAMP) {
- c.set_first_timestamp(e.timestamp);
- }
- c.set_opening_num(e.opening_num);
- c.set_pgn_start_position(e.start_position);
- }
- if (!moves.count(e.move)) {
- moves.insert(e.move);
- c.add_move(e.move);
- }
- if (j == elems[i].size() - 1 || e.bpfen != elems[i][j + 1].bpfen) {
- c.SerializeToString(&buf);
- mtbl_writer_add(mtbl,
- (const uint8_t *)e.bpfen.data(), e.bpfen.size(),
- (const uint8_t *)buf.data(), buf.size());
- c = Count();
- moves.clear();
- }
- }
- mtbl_writer_destroy(&mtbl);
+ write_subshard(argv[argc - 3], &shards[i], i);
}
}
+
projects / remoteglot-book / blobdiff