File size: 3,634 Bytes
f2cf1f1 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 | #!/usr/bin/env python3
"""Validate FinGovBench JSONL and graph references."""
from __future__ import annotations
import argparse
import json
from collections import defaultdict, deque
from datetime import datetime
from pathlib import Path
import jsonschema
ROOT = Path(__file__).resolve().parents[1]
def main() -> None:
parser = argparse.ArgumentParser()
parser.add_argument("dataset", type=Path)
args = parser.parse_args()
schema = json.loads((ROOT / "schema/case.schema.json").read_text())
rows = [json.loads(line) for line in args.dataset.read_text().splitlines() if line.strip()]
ids = [row["case_id"] for row in rows]
assert len(ids) == len(set(ids)), "duplicate case_id"
for row in rows:
jsonschema.validate(row, schema)
raw_node_ids = [node["node_id"] for node in row["graph"]["nodes"]]
assert len(raw_node_ids) == len(set(raw_node_ids)), (row["case_id"], "duplicate node")
node_ids = set(raw_node_ids)
adjacency: dict[str, list[str]] = defaultdict(list)
indegree = {node_id: 0 for node_id in node_ids}
for edge in row["graph"]["edges"]:
assert edge["source"] in node_ids, (row["case_id"], edge)
assert edge["target"] in node_ids, (row["case_id"], edge)
adjacency[edge["source"]].append(edge["target"])
indegree[edge["target"]] += 1
queue = deque(node for node, degree in indegree.items() if degree == 0)
visited = 0
while queue:
visited += 1
for child in adjacency[queue.popleft()]:
indegree[child] -= 1
if indegree[child] == 0:
queue.append(child)
assert visited == len(node_ids), (row["case_id"], "graph is not acyclic")
for measurement in row["telemetry"]["local_measurements"]:
assert measurement["node_id"] in node_ids, (row["case_id"], measurement)
lo, hi = measurement["defect_interval"]
assert lo <= hi, (row["case_id"], measurement)
for key in ("affected_nodes", "revalidation_scope"):
assert set(row["gold"][key]) <= node_ids, (row["case_id"], key)
failed = [m["node_id"] for m in row["telemetry"]["local_measurements"] if m["status"] == "fail"]
assert bool(failed) == bool(row["gold"]["iceberg_present"]), (row["case_id"], "telemetry/gold iceberg mismatch")
if not row["gold"]["iceberg_present"]:
assert not row["gold"]["affected_nodes"] and not row["gold"]["revalidation_scope"], row["case_id"]
assert row["gold"]["release_permitted"], row["case_id"]
if row["gold"]["iceberg_visibility"] == "visible":
assert row["telemetry"]["terminal_measurement"]["status"] == "fail", row["case_id"]
if row["case_origin"] == "genuine_revision":
metadata = row["revision_metadata"]
assert metadata["old_accession"] != metadata["new_accession"], row["case_id"]
assert metadata["old_filed"] < metadata["new_filed"], row["case_id"]
assert metadata["old_value"] != metadata["new_value"], row["case_id"]
assert datetime.fromisoformat(row["as_of"].replace("Z", "+00:00")).date().isoformat() >= metadata["new_filed"], row["case_id"]
if any(not control.get("compensable", True) and control["status"] == "failed" for control in row["red_line_controls"]):
assert not row["gold"]["release_permitted"], row["case_id"]
print(json.dumps({"valid": True, "cases": len(rows), "splits": sorted(set(row["split"] for row in rows))}))
if __name__ == "__main__":
main()
|