rework dataset function and class structure

This commit is contained in:
2026-04-05 02:21:15 +02:00
parent 066a93f755
commit 332e86e3cc
9 changed files with 318 additions and 248 deletions
+45 -84
View File
@@ -1,50 +1,16 @@
from collections import Counter, defaultdict
import argparse, glob, json, re
import argparse, json, re
from datetime import datetime
from pathlib import Path
from server.dataset.DatasetIO import DatasetIO
class DatasetStats:
DAY_PATTERN = re.compile(r"(\d{4}-\d{2}-\d{2})")
def __init__(self, input_files:list[str]):
self.input_files = input_files
def _resolve_input_files(self):
resolved = []
seen = set()
for item in self.input_files:
path = Path(item)
if path.is_dir():
for file_path in sorted(path.rglob("*.jsonl")):
key = str(file_path.resolve())
if key in seen:
continue
seen.add(key)
resolved.append(file_path)
continue
if any(ch in item for ch in "*?[]"):
for match in sorted(glob.glob(item)):
file_path = Path(match)
if not file_path.is_file():
continue
key = str(file_path.resolve())
if key in seen:
continue
seen.add(key)
resolved.append(file_path)
continue
if path.is_file():
key = str(path.resolve())
if key in seen:
continue
seen.add(key)
resolved.append(path)
return resolved
def _infer_day(self, file_path:Path):
match = self.DAY_PATTERN.search(file_path.name)
if match:
@@ -83,7 +49,7 @@ class DatasetStats:
return None
def analyze(self):
files = self._resolve_input_files()
files = DatasetIO.resolve_input_files(self.input_files)
totals = {
"rows": 0,
@@ -99,57 +65,52 @@ class DatasetStats:
for file_path in files:
day = self._infer_day(file_path)
with file_path.open("r", encoding="utf-8") as source:
for line in source:
if not line.strip():
continue
for row in DatasetIO.iter_jsonl_rows(file_path):
game_id = row.get("game_id")
if not game_id:
continue
row = json.loads(line)
game_id = row.get("game_id")
if not game_id:
continue
turn = int(row.get("turn", 0))
safe_options = self._extract_safe_options(row)
snake_type = row.get("snake_type", "unknown")
move = row.get("move", "unknown")
turn = int(row.get("turn", 0))
safe_options = self._extract_safe_options(row)
snake_type = row.get("snake_type", "unknown")
move = row.get("move", "unknown")
game_type = row.get("game_type", {})
if isinstance(game_type, dict):
game_type_name = game_type.get("name", "unknown")
else:
game_type_name = str(game_type)
game_type = row.get("game_type", {})
if isinstance(game_type, dict):
game_type_name = game_type.get("name", "unknown")
else:
game_type_name = str(game_type)
totals["rows"] += 1
totals["games"].add(game_id)
totals["snake_types"][snake_type] += 1
totals["game_types"][game_type_name] += 1
totals["moves"][move] += 1
totals["days"][day] += 1
totals["rows"] += 1
totals["games"].add(game_id)
totals["snake_types"][snake_type] += 1
totals["game_types"][game_type_name] += 1
totals["moves"][move] += 1
totals["days"][day] += 1
if game_id not in games:
games[game_id] = {
"game_id": game_id,
"day": day,
"snake_type": snake_type,
"game_type": game_type_name,
"rows": 0,
"max_turn": -1,
"safe_options_sum": 0,
"safe_options_count": 0,
"pressure_turns": 0,
}
if game_id not in games:
games[game_id] = {
"game_id": game_id,
"day": day,
"snake_type": snake_type,
"game_type": game_type_name,
"rows": 0,
"max_turn": -1,
"safe_options_sum": 0,
"safe_options_count": 0,
"pressure_turns": 0,
}
game = games[game_id]
game["rows"] += 1
game["max_turn"] = max(game["max_turn"], turn)
if isinstance(safe_options, int):
game["safe_options_sum"] += safe_options
game["safe_options_count"] += 1
if safe_options <= 2:
game["pressure_turns"] += 1
game = games[game_id]
game["rows"] += 1
game["max_turn"] = max(game["max_turn"], turn)
if isinstance(safe_options, int):
game["safe_options_sum"] += safe_options
game["safe_options_count"] += 1
if safe_options <= 2:
game["pressure_turns"] += 1
day_games[day].add(game_id)
day_games[day].add(game_id)
game_summaries = []
for game in games.values():
@@ -225,7 +186,7 @@ if __name__ == "__main__":
"--input",
action="append",
required=True,
help="Input JSONL file, directory, or glob pattern. Repeat for multiple inputs.",
help="Input JSONL/JSONL.GZ file, directory, or glob pattern. Repeat for multiple inputs.",
)
parser.add_argument(
"--output",