File size: 19,763 Bytes
49665ce c8b91c2 49665ce c8b91c2 49665ce c8b91c2 49665ce c8b91c2 49665ce c8b91c2 49665ce | 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 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 | #!/usr/bin/env python3
"""Run CodeQL on the pinned public cohort and compare method-level alerts.
The caller supplies the downloaded public benchmark.jsonl files. GitHub's
CodeQL CLI and query packs are installed separately; this script does not
download or redistribute them.
"""
import argparse
import concurrent.futures
import json
import re
import shutil
import subprocess
import time
from collections import defaultdict
from functools import lru_cache
from pathlib import Path
from zipfile import ZipFile
LANGUAGES = {"python": "python", "javascript": "javascript", "typescript": "javascript",
"c": "cpp", "c++": "cpp", "cpp": "cpp"}
PACKS = {"python": "1.8.11", "javascript": "2.4.6", "cpp": "1.9.0"}
def rows(path):
with open(path, encoding="utf-8") as stream:
for line in stream:
if line.strip():
yield json.loads(line)
def selected_pairs(args):
cohort = None if args.all_supported else json.loads(Path(args.cohort).read_text())
selected = {}
for domain in ("bug", "security"):
repos = set(cohort[domain]) if cohort else None
selected[domain] = [
pair for pair in rows(Path(args.data) / domain / "benchmark.jsonl")
if (repos is None or pair["repository"] in repos)
and pair["language"].lower() in LANGUAGES
and all(isinstance(pair[role].get("revision"), str)
and re.fullmatch(r"[0-9a-f]{40}", pair[role]["revision"])
for role in ("before", "after"))
]
return selected
def run(command, log_path, timeout=None, container_name=None):
log_path.parent.mkdir(parents=True, exist_ok=True)
start = time.monotonic()
with log_path.open("w", encoding="utf-8") as log:
try:
result = subprocess.run(command, stdout=log, stderr=subprocess.STDOUT,
timeout=timeout)
code = result.returncode
except subprocess.TimeoutExpired:
if container_name:
subprocess.run(["docker", "stop", "--time", "2", container_name],
stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
log.write(f"CodeQL stage exceeded {timeout} seconds.\n")
code = 124
return code, round(time.monotonic() - start, 2)
def prepare_rebench_patches(pairs, root, source_parquet):
"""Recreate SWE-rebench repaired snapshots from its published gold patches."""
if not source_parquet:
return
instances = {}
for pair in pairs["bug"]:
if not pair["pair_id"].startswith("swe-rebench:"):
continue
if "+" in pair["after"]["revision"]:
continue
instance = pair["pair_id"].split(":", 2)[1]
if not re.fullmatch(r"[A-Za-z0-9_.-]+", instance):
raise ValueError(f"Invalid instance ID: {instance}")
instances[instance] = pair
if not instances:
return
quoted = ",".join("'" + instance + "'" for instance in sorted(instances))
parquet = Path(source_parquet).resolve().as_posix().replace("'", "''")
query = ("SELECT instance_id,base_commit,patch FROM read_parquet('"
+ parquet + "') WHERE instance_id IN (" + quoted + ")")
response = subprocess.run(["duckdb", "-json", "-c", query], check=True,
capture_output=True, text=True)
source_rows = {row["instance_id"]: row for row in json.loads(response.stdout)}
errors = []
for instance, pair in instances.items():
source = source_rows.get(instance)
if not source:
errors.append({"instance_id": instance, "reason": "missing source patch"})
continue
if pair["before"]["revision"] != source["base_commit"]:
errors.append({"instance_id": instance, "reason": "base commit mismatch"})
continue
repo = pair["repository"]
slug = repo.replace("/", "__")
git_dir = root / "git" / slug
checkout = root / "checkouts" / slug / pair["after"]["revision"]
patch_file = root / "patches" / f"{instance}.patch"
patch_file.parent.mkdir(parents=True, exist_ok=True)
patch_file.write_text(source["patch"], encoding="utf-8")
try:
if not git_dir.exists():
git_dir.parent.mkdir(parents=True, exist_ok=True)
subprocess.run(["git", "clone", "--filter=blob:none", "--no-checkout",
f"https://github.com/{repo}.git", str(git_dir)], check=True,
stdout=subprocess.DEVNULL, stderr=subprocess.PIPE)
have = subprocess.run(
["git", "-C", str(git_dir), "cat-file", "-e",
f"{source['base_commit']}^{{commit}}"],
stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
).returncode == 0
if not have:
subprocess.run(["git", "-C", str(git_dir), "fetch", "origin",
source["base_commit"]], check=True,
stdout=subprocess.DEVNULL, stderr=subprocess.PIPE)
if not checkout.exists():
checkout.parent.mkdir(parents=True, exist_ok=True)
subprocess.run(["git", "-C", str(git_dir), "worktree", "add", "--detach",
str(checkout), source["base_commit"]], check=True,
stdout=subprocess.DEVNULL, stderr=subprocess.PIPE)
subprocess.run(["git", "-C", str(checkout), "apply", "--whitespace=nowarn",
str(patch_file)], check=True,
stdout=subprocess.DEVNULL, stderr=subprocess.PIPE)
except subprocess.CalledProcessError as error:
errors.append({"instance_id": instance,
"reason": f"{error.cmd[0]} exited {error.returncode}"})
(root / "patch-errors.json").write_text(json.dumps(errors, indent=2) + "\n")
def prepare_sources(pairs, root):
tasks = defaultdict(set)
for domain, group in pairs.items():
for pair in group:
repo = pair["repository"]
language = LANGUAGES[pair["language"].lower()]
for role in ("before", "after"):
tasks[(repo, pair[role]["revision"], language)].add(domain)
sources = {}
errors = []
for repo, revision, language in sorted(tasks):
if not re.fullmatch(r"[0-9a-f]{40}", revision):
errors.append({"repository": repo, "revision": revision,
"reason": "not a Git commit"})
continue
slug = repo.replace("/", "__")
git_dir = root / "git" / slug
checkout = root / "checkouts" / slug / revision
try:
if not git_dir.exists():
git_dir.parent.mkdir(parents=True, exist_ok=True)
subprocess.run(
["git", "clone", "--filter=blob:none", "--no-checkout",
f"https://github.com/{repo}.git", str(git_dir)],
check=True, stdout=subprocess.DEVNULL, stderr=subprocess.PIPE,
)
if not checkout.exists():
checkout.parent.mkdir(parents=True, exist_ok=True)
have = subprocess.run(
["git", "-C", str(git_dir), "cat-file", "-e", f"{revision}^{{commit}}"],
stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
).returncode == 0
if not have:
subprocess.run(["git", "-C", str(git_dir), "fetch", "origin", revision],
check=True, stdout=subprocess.DEVNULL, stderr=subprocess.PIPE)
subprocess.run(["git", "-C", str(git_dir), "worktree", "add", "--detach",
str(checkout), revision], check=True,
stdout=subprocess.DEVNULL, stderr=subprocess.PIPE)
sources[(repo, revision, language)] = (checkout, tasks[(repo, revision, language)])
except subprocess.CalledProcessError as error:
errors.append({"repository": repo, "revision": revision,
"reason": f"{error.cmd[0]} exited {error.returncode}"})
(root / "source-errors.json").write_text(json.dumps(errors, indent=2) + "\n")
return sources
def docker(args, name, *command):
return ["docker", "run", "--rm", "--name", name,
"-v", f"{Path(args.codeql).resolve()}:/opt/codeql:ro",
"-v", f"{Path(args.root).resolve()}:/work", args.image,
"/opt/codeql/codeql", *command]
def scan_one(args, task):
(repo, revision, language), (checkout, domains) = task
root = Path(args.root).resolve()
slug = repo.replace("/", "__")
stem = f"{slug}-{revision[:12]}-{language}"
db = root / "dbs" / stem
log_dir = root / "logs"
result = {"repository": repo, "revision": revision, "language": language,
"domains": sorted(domains)}
if not (db / f"db-{language}").exists():
if db.exists():
shutil.rmtree(db)
container = f"codeql-create-{stem}".lower()
code, seconds = run(docker(args, container, "database", "create", f"/work/dbs/{stem}",
f"--language={language}", "--build-mode=none",
f"--source-root=/work/checkouts/{slug}/{revision}",
"--threads=4", "--ram=4096"),
log_dir / f"{stem}-create.log", args.scan_timeout, container)
result["create_seconds"] = seconds
if code:
result["error"] = f"database create exit {code}"
return result
for domain in sorted(domains):
sarif = root / "sarif" / domain / f"{stem}.sarif"
if sarif.exists():
continue
sarif.parent.mkdir(parents=True, exist_ok=True)
suite = "security-extended" if domain == "security" else "security-and-quality"
query = (f"/work/packs/codeql/{language}-queries/{PACKS[language]}/"
f"codeql-suites/{language}-{suite}.qls")
container = f"codeql-{domain}-{stem}".lower()
code, seconds = run(docker(args, container, "database", "analyze", f"/work/dbs/{stem}",
query, "--format=sarif-latest",
f"--output=/work/sarif/{domain}/{stem}.sarif",
"--threads=4", "--ram=4096"),
log_dir / f"{stem}-{domain}-analyze.log", args.scan_timeout,
container)
result[f"{domain}_analyze_seconds"] = seconds
if code:
result["error"] = f"{domain} analyze exit {code}"
return result
result["ok"] = True
return result
def scan(args):
root = Path(args.root).resolve()
(root / "dbs").mkdir(parents=True, exist_ok=True)
pairs = selected_pairs(args)
prepare_rebench_patches(pairs, root, args.rebench_parquet)
sources = prepare_sources(pairs, root)
print(f"Fixed cohort: {len(pairs['bug'])} bug pairs, "
f"{len(pairs['security'])} security pairs, {len(sources)} databases", flush=True)
manifest = {domain: [p["pair_id"] for p in group] for domain, group in pairs.items()}
(root / "cohort-pair-ids.json").write_text(json.dumps(manifest, indent=2) + "\n")
with concurrent.futures.ThreadPoolExecutor(max_workers=args.jobs) as executor:
for result in concurrent.futures.as_completed(
[executor.submit(scan_one, args, task) for task in sources.items()]
):
item = result.result()
with (root / "scan-status.jsonl").open("a", encoding="utf-8") as stream:
stream.write(json.dumps(item, sort_keys=True) + "\n")
print(f"{'OK' if item.get('ok') else 'FAILED'} "
f"{item['repository']} {item['revision'][:12]} "
f"{','.join(item['domains'])}", flush=True)
def method_span(side):
lines = [int(match) for match in re.findall(r"(?m)^L(\d+)\|",
side["state"]["method"]["source"])]
return (min(lines), max(lines)) if lines else None
def source_matches(side, source_path):
content = source_path.read_text(errors="replace").replace("\r\n", "\n")
needle = side["source"].replace("\r\n", "\n").strip()
start = content.find(needle)
return start >= 0 and content.count("\n", 0, start) + 1 == method_span(side)[0]
@lru_cache(maxsize=128)
def archived_sources(archive):
with ZipFile(archive) as zipped:
return set(zipped.namelist())
def codeql_alert(sarif, side, domain):
path = side["path"].lstrip("/")
span = method_span(side)
if not span:
raise ValueError(f"Missing line-numbered method: {path}")
data = json.loads(sarif.read_text())
alerts = []
for run_data in data["runs"]:
rule_tags = {
rule["id"]: set(rule.get("properties", {}).get("tags", []))
for rule in run_data["tool"]["driver"].get("rules", [])
}
for finding in run_data.get("results", []):
tags = rule_tags.get(finding["ruleId"], set())
if domain == "security" and "security" not in tags:
continue
if domain == "bug" and not tags.intersection(
("correctness", "reliability", "security")
):
continue
for location in finding.get("locations", [])[:1]:
physical = location.get("physicalLocation", {})
uri = physical.get("artifactLocation", {}).get("uri", "")
region = physical.get("region", {})
if uri.lstrip("/") != path or "startLine" not in region:
continue
start = region["startLine"]
end = region.get("endLine", start)
if start <= span[1] and end >= span[0]:
alerts.append(finding["ruleId"])
return sorted(set(alerts))
def metrics(items):
items = list(items)
tp = sum(label and alert for label, alert in items)
fp = sum(not label and alert for label, alert in items)
fn = sum(label and not alert for label, alert in items)
tn = sum(not label and not alert for label, alert in items)
precision = tp / (tp + fp) if tp + fp else None
recall = tp / (tp + fn) if tp + fn else 0.0
fpr = fp / (fp + tn) if fp + tn else 0.0
f1 = (2 * precision * recall / (precision + recall)
if precision is not None and precision + recall else 0.0)
return {"tp": tp, "fp": fp, "fn": fn, "tn": tn, "precision": precision,
"recall": recall, "fpr": fpr, "f1": f1, "recall_minus_fpr": recall - fpr}
def score(args):
root = Path(args.root).resolve()
selected = selected_pairs(args)
all_results = {}
for domain, group in selected.items():
jev = {(r["pair_id"], r["role"]): r for r in
rows(Path(args.predictions) / domain / "jev.jsonl")}
output = []
unavailable = []
for pair in group:
decisions = []
for role in ("before", "after"):
side = pair[role]
lang = LANGUAGES[pair["language"].lower()]
stem = f"{pair['repository'].replace('/', '__')}-{side['revision'][:12]}-{lang}"
sarif = root / "sarif" / domain / f"{stem}.sarif"
source_path = (root / "checkouts" / pair["repository"].replace("/", "__")
/ side["revision"] / side["path"])
if not sarif.exists() or (pair["pair_id"], role) not in jev:
reason = ("source snapshot unavailable" if not source_path.exists()
else "CodeQL analysis unavailable" if not sarif.exists()
else "Jev prediction unavailable")
unavailable.append({"pair_id": pair["pair_id"], "role": role,
"reason": reason})
decisions = []
break
archive = root / "dbs" / stem / "src.zip"
archived_name = (f"work/checkouts/{pair['repository'].replace('/', '__')}/"
f"{side['revision']}/{side['path']}")
if not source_path.exists():
unavailable.append({"pair_id": pair["pair_id"], "role": role,
"reason": "source snapshot unavailable"})
decisions = []
break
if (not archive.exists()
or archived_name not in archived_sources(archive)
or not source_matches(side, source_path)):
unavailable.append({"pair_id": pair["pair_id"], "role": role,
"reason": "target source not verified in CodeQL database"})
decisions = []
break
rules = codeql_alert(sarif, side, domain)
label = int(side["labels"][domain])
decisions.append({"role": role, "label": label, "codeql_alert": bool(rules),
"codeql_rules": rules,
"jev_alert": bool(jev[(pair["pair_id"], role)]["alert"]),
"jev_score": jev[(pair["pair_id"], role)]["score"]})
if decisions:
output.append({"pair_id": pair["pair_id"], "repository": pair["repository"],
"language": pair["language"], "decisions": decisions})
model_scores = {}
for model in ("codeql", "jev"):
model_scores[model] = metrics(
(d["label"], d[f"{model}_alert"]) for row in output for d in row["decisions"]
)
all_results[domain] = {"pairs": len(output), "requested_pairs": len(group),
"missing_sides": unavailable, "models": model_scores}
out = root / f"{domain}-codeql-comparison.jsonl"
out.write_text("".join(json.dumps(r, sort_keys=True) + "\n" for r in output))
(root / "codeql-comparison-results.json").write_text(
json.dumps(all_results, indent=2, sort_keys=True) + "\n"
)
print(json.dumps(all_results, indent=2))
def main():
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("operation", choices=("scan", "score"))
parser.add_argument("--root", required=True)
parser.add_argument("--data", required=True, help="directory with bug/security/benchmark.jsonl")
parser.add_argument("--predictions", required=True, help="directory with bug/security/jev.jsonl")
parser.add_argument("--cohort", default=str(Path(__file__).with_name("codeql-cohort.json")))
parser.add_argument("--all-supported", action="store_true",
help="scan every pair in supported languages with two Git revisions")
parser.add_argument("--codeql", help="unpacked CodeQL CLI directory")
parser.add_argument("--rebench-parquet", help="SWE-rebench-V2 source Parquet with gold patches")
parser.add_argument("--image", default="node:22-bookworm")
parser.add_argument("--jobs", type=int, default=4)
parser.add_argument("--scan-timeout", type=int, default=600,
help="seconds allowed per CodeQL database or analysis stage")
args = parser.parse_args()
if args.operation == "scan":
if not args.codeql:
parser.error("--codeql is required for scanning")
scan(args)
else:
score(args)
if __name__ == "__main__":
main()
|