251 lines
13 KiB
Python
251 lines
13 KiB
Python
"""Offline audit of an official Q3 run; never connects to the simulator.
|
|
|
|
Usage: python scripts/verify_q3_practice.py output/q3/practice/20260911-port2026
|
|
Preserves session.jsonl and summary.json; writes derived audit artifacts only.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import argparse
|
|
import csv
|
|
from dataclasses import asdict
|
|
from datetime import datetime
|
|
import hashlib
|
|
import json
|
|
import math
|
|
from pathlib import Path
|
|
from urllib.parse import urlsplit
|
|
|
|
from benchmark_q3_search import SevenBatch, seven_points
|
|
from mcm26b.config import ProblemConfig
|
|
from mcm26b.coverage import coverage_points
|
|
from mcm26b.evaluation import recompute_virtual_time
|
|
from mcm26b.planner import OmnidirectionalLoop
|
|
from mcm26b.protocol.client import SimulatorClient
|
|
from mcm26b.harness import PlaybackTransport
|
|
|
|
|
|
class StrictPlayback(PlaybackTransport):
|
|
"""Check complete action content, not merely the sequence of API paths."""
|
|
|
|
def post(self, url, body, headers, timeout):
|
|
if self.index >= len(self.events):
|
|
raise AssertionError("Replay attempted an extra action")
|
|
expected = self.events[self.index]
|
|
actual_request = json.loads(body)
|
|
|
|
def without_id(value):
|
|
return {k: v for k, v in value.items() if k != "request_id"}
|
|
|
|
if (urlsplit(url).path != expected["path"] or
|
|
without_id(actual_request) != without_id(expected["request"])):
|
|
raise AssertionError(f"Replay action differs at index {self.index}")
|
|
return super().post(url, body, headers, timeout)
|
|
|
|
|
|
class AuditMixin:
|
|
def __init__(self, client, config):
|
|
super().__init__(client, config)
|
|
self.certificates = []
|
|
|
|
def _attempt_clear(self, channel, target, cert):
|
|
vertices = self.belief.candidate_set(channel).polygon
|
|
self.certificates.append({**asdict(cert), "submitted_position": target,
|
|
"submitted_max_vertex_distance_m":
|
|
max(math.dist(target, v) for v in vertices)})
|
|
return super()._attempt_clear(channel, target, cert)
|
|
|
|
|
|
class AuditedLoop(AuditMixin, OmnidirectionalLoop):
|
|
pass
|
|
|
|
|
|
class AuditedSevenBatch(AuditMixin, SevenBatch):
|
|
pass
|
|
|
|
|
|
def save_json(path, value):
|
|
path.write_text(json.dumps(value, ensure_ascii=False, indent=2,
|
|
allow_nan=False) + "\n", encoding="utf-8")
|
|
|
|
|
|
def timing_rows(events):
|
|
"""按官方附件 2 第 4.1 节逐动作累计虚拟时间（微秒），返回明细与汇总。"""
|
|
position, channel, virtual_us = (0.0, 0.0), 1, 0
|
|
distance_total, switch_total, move_total_us = 0.0, 0, 0
|
|
rows = []
|
|
previous_recorded = 0.0
|
|
max_local_gap = 0.0
|
|
for i, event in enumerate(events):
|
|
request, response = event["request"], event["response"]
|
|
path = event["path"]
|
|
distance, switches, action_s = 0.0, 0, 0
|
|
if path in ("/measure", "/clear"):
|
|
target = (request["position"]["x"], request["position"]["y"])
|
|
distance = math.dist(position, target)
|
|
position = target
|
|
if path == "/measure":
|
|
switches = int(channel != request["channel"])
|
|
channel = request["channel"]
|
|
action_s = 5
|
|
else:
|
|
action_s = 5 if response["clear_result"] == "success" else 3
|
|
move_us = round(distance / 5 * 1_000_000)
|
|
virtual_us += move_us + (switches + action_s) * 1_000_000
|
|
distance_total += distance
|
|
switch_total += switches
|
|
move_total_us += move_us
|
|
recorded = response["virtual_time_s"]
|
|
gap_us = round(recorded * 1_000_000) - virtual_us
|
|
max_local_gap = max(max_local_gap,
|
|
abs(recorded - previous_recorded - distance/5 - switches - action_s))
|
|
previous_recorded = recorded
|
|
rows.append({"index": i, "path": path, "request_id": request["request_id"],
|
|
"channel": request.get("channel", ""), "x": position[0], "y": position[1],
|
|
"result": response.get("measure_result", response.get("clear_result", "")),
|
|
"svd_deg": response.get("svd_deg", ""), "distance_m": distance,
|
|
"move_us": move_us, "switch_s": switches, "action_s": action_s,
|
|
"recomputed_us": virtual_us, "recorded_s": recorded, "gap_us": gap_us})
|
|
return rows, virtual_us, distance_total, move_total_us, switch_total, previous_recorded, \
|
|
max_local_gap
|
|
|
|
|
|
def audit(folder: Path) -> dict:
|
|
records = [json.loads(line) for line in
|
|
(folder / "session.jsonl").read_text(encoding="utf-8").splitlines()]
|
|
events = [e for e in records if e.get("event") == "accepted"]
|
|
original = json.loads((folder / "summary.json").read_text(encoding="utf-8-sig"))
|
|
strategy = original.get("strategy", "grid49_immediate")
|
|
if strategy not in ("grid49_immediate", "seven_batch"):
|
|
raise ValueError(f"Unsupported strategy: {strategy}")
|
|
official_paths = list(folder.glob("*.result.json"))
|
|
if len(official_paths) != 1 or not events:
|
|
raise ValueError("Expected one official result and nonempty session evidence")
|
|
official_path = official_paths[0]
|
|
official = json.loads(official_path.read_text(encoding="utf-8"))
|
|
package = official_path.with_name(official_path.name.replace(".result.json", ".jlog"))
|
|
|
|
checks = {}
|
|
checks["official_problem_3_all_omnidirectional"] = (
|
|
official["problem_no"] == 3 and official["directional_jammer_count"] == 0
|
|
and 10 <= official["jammer_count"] <= 16
|
|
and official["omnidirectional_jammer_count"] == official["jammer_count"])
|
|
checks["official_package_sha256_matches"] = (
|
|
hashlib.sha256(package.read_bytes()).hexdigest() == official["package_sha256"])
|
|
checks["normal_enter_and_exit"] = (
|
|
events[0]["path"] == "/enter" and events[-1]["path"] == "/exit"
|
|
and events[-1]["response"].get("exit_reason") == "user_exit")
|
|
start_ms = datetime.fromisoformat(official["window_started_at_utc"]).timestamp() * 1000
|
|
end_ms = datetime.fromisoformat(official["ended_at_utc"]).timestamp() * 1000
|
|
checks["responses_within_official_run_window"] = all(
|
|
start_ms <= e["response"]["real_timestamp_ms"] <= end_ms + 1 for e in events)
|
|
ids = [e["request"]["request_id"] for e in events]
|
|
checks["unique_accepted_request_ids"] = len(ids) == len(set(ids))
|
|
checks["no_rejections_or_retries"] = (
|
|
len(records) == 2 * len(events)
|
|
and all(e.get("attempts") == 1 for e in events)
|
|
and all(records[2*i].get("event") == "attempt"
|
|
and records[2*i]["request"] == e["request"]
|
|
and records[2*i+1] == e for i, e in enumerate(events)))
|
|
clears = [e for e in events if e["path"] == "/clear"]
|
|
successes = [e for e in clears if e["response"].get("clear_result") == "success"]
|
|
cleared = sorted({e["request"]["channel"] for e in successes})
|
|
checks["all_official_sources_cleared"] = len(cleared) == official["jammer_count"]
|
|
checks["every_clear_succeeded_once"] = len(clears) == len(successes) == len(cleared)
|
|
checks["original_summary_completed"] = (
|
|
original["completed"] and original["cleared_channels"] == cleared
|
|
and original["unresolved_channels"] == [] and original["anomalies"] == [])
|
|
|
|
grid = set(seven_points() if strategy == "seven_batch" else coverage_points())
|
|
absent_coverage = {}
|
|
for channel in sorted(set(range(1, 21)) - set(cleared)):
|
|
channel_events = [e for e in events if e["path"] == "/measure"
|
|
and e["request"]["channel"] == channel]
|
|
points = {(e["request"]["position"]["x"], e["request"]["position"]["y"])
|
|
for e in channel_events if e["response"]["measure_result"] == "no_signal"}
|
|
absent_coverage[str(channel)] = {
|
|
"grid_points_measured": len(points & grid), "required_grid_points": len(grid),
|
|
"all_no_signal": all(e["response"]["measure_result"] == "no_signal"
|
|
for e in channel_events),
|
|
"complete": grid <= points,
|
|
}
|
|
checks["absent_channels_fully_covered"] = all(
|
|
item["complete"] and item["all_no_signal"] for item in absent_coverage.values())
|
|
|
|
# Official attachment 2 section 4.1: accumulate virtual time in microseconds.
|
|
# Round each move to the nearest microsecond, then add integer action costs.
|
|
rows, virtual_us, distance_total, move_total_us, switch_total, previous_recorded, \
|
|
max_local_gap = timing_rows(events)
|
|
checks["all_timestamps_match_microsecond_recomputation"] = all(r["gap_us"] == 0 for r in rows)
|
|
real_s = (events[-1]["response"]["real_timestamp_ms"] -
|
|
events[0]["response"]["real_timestamp_ms"]) / 1000
|
|
checks["within_real_and_virtual_limits"] = (
|
|
real_s < events[0]["response"]["remaining_real_duration_s"]
|
|
and virtual_us / 1_000_000 < events[0]["response"]["max_virtual_duration_s"])
|
|
|
|
# Exact action replay independently checks decisions using only recorded responses.
|
|
transport = StrictPlayback(events)
|
|
config = ProblemConfig()
|
|
client = SimulatorClient(original["robot_id"], transport=transport, config=config,
|
|
clock=lambda: 0.0)
|
|
loop = (AuditedSevenBatch if strategy == "seven_batch" else AuditedLoop)(client, config)
|
|
replay_summary = loop.run()
|
|
checks["deterministic_action_replay"] = (
|
|
transport.index == len(events) and replay_summary["completed"]
|
|
and all(replay_summary[k] == original[k] for k in
|
|
("stop_reason", "cleared_channels", "absent_channels", "measures", "clears")))
|
|
checks["all_clear_certificates_valid"] = (
|
|
len(loop.certificates) == len(clears)
|
|
and all(c["certified"] and c["radius_m"] <= c["limit_m"] < 20
|
|
and c["submitted_max_vertex_distance_m"] <= c["limit_m"]
|
|
for c in loop.certificates))
|
|
raw_timing = recompute_virtual_time(events)
|
|
report = {
|
|
"passed": all(checks.values()), "source": "official_practice", "strategy": strategy,
|
|
"case_code": official["case_code"], "checks": checks,
|
|
"official_jammer_count": official["jammer_count"], "cleared_channels": cleared,
|
|
"absent_channel_coverage": absent_coverage,
|
|
"accepted_requests_including_enter_exit": len(events),
|
|
"measures": len([e for e in events if e["path"] == "/measure"]),
|
|
"clear_attempts": len(clears), "clear_successes": len(successes),
|
|
"real_elapsed_server_s": real_s,
|
|
"real_elapsed_local_s": events[-1]["logged_at_unix_s"] - events[0]["logged_at_unix_s"],
|
|
"virtual_time_s": virtual_us / 1_000_000,
|
|
"last_clear_virtual_time_s": successes[-1]["response"]["virtual_time_s"],
|
|
"move_distance_m": distance_total, "move_time_s": move_total_us / 1_000_000,
|
|
"switch_time_s": switch_total,
|
|
"measurement_time_s": sum(r["action_s"] for r in rows if r["path"] == "/measure"),
|
|
"clear_time_s": sum(r["action_s"] for r in rows if r["path"] == "/clear"),
|
|
"timing_audit": {
|
|
"original_unquantized_mismatch_count": len(raw_timing["mismatches"]),
|
|
"original_unquantized_final_gap_s":
|
|
previous_recorded - raw_timing["final_virtual_time_s"],
|
|
"max_single_action_unquantized_gap_s": max_local_gap,
|
|
"microsecond_mismatch_count": sum(r["gap_us"] != 0 for r in rows),
|
|
"explanation": "原复算累计未量化浮点耗时；按动作移动耗时取最近整数微秒后，所有官方时刻完全一致。",
|
|
},
|
|
"clear_certificates": loop.certificates,
|
|
"limits": "仅验证当前官方演练案例；不证明所有案例或时间最优性。官方加密日志保持原样，未解密。",
|
|
}
|
|
with (folder / "actions.csv").open("w", encoding="utf-8-sig", newline="") as handle:
|
|
writer = csv.DictWriter(handle, fieldnames=list(rows[0]))
|
|
writer.writeheader()
|
|
writer.writerows(rows)
|
|
save_json(folder / "verification.json", report)
|
|
save_json(folder / "replay_summary.json", replay_summary)
|
|
manifest = {p.name: {"bytes": p.stat().st_size,
|
|
"sha256": hashlib.sha256(p.read_bytes()).hexdigest()}
|
|
for p in sorted(folder.iterdir()) if p.is_file() and p.name != "manifest.json"}
|
|
save_json(folder / "manifest.json", manifest)
|
|
return report
|
|
|
|
|
|
if __name__ == "__main__":
|
|
parser = argparse.ArgumentParser(description=__doc__)
|
|
parser.add_argument("folder", type=Path)
|
|
args = parser.parse_args()
|
|
result = audit(args.folder)
|
|
print(json.dumps({"passed": result["passed"], "case_code": result["case_code"],
|
|
"checks": result["checks"]}, ensure_ascii=False, indent=2))
|
|
raise SystemExit(0 if result["passed"] else 1)
|