| """Structural validation for SWE-DockerLess.jsonl. |
| |
| Checks that every row is SWE-bench shaped, that the two patches apply to the |
| embedded snapshot, that they touch disjoint files, and that the gold patch stays |
| inside a budget a 2048-token completion can produce. |
| |
| Run with: python tools/validate.py |
| """ |
|
|
| import json |
| import subprocess |
| import sys |
| import tempfile |
| from pathlib import Path |
|
|
| ROOT = Path(__file__).resolve().parent.parent |
| DATASET = ROOT / "SWE-DockerLess.jsonl" |
|
|
| SWE_BENCH_FIELDS = [ |
| "instance_id", |
| "repo", |
| "base_commit", |
| "patch", |
| "test_patch", |
| "problem_statement", |
| "hints_text", |
| "created_at", |
| "version", |
| "environment_setup_commit", |
| "FAIL_TO_PASS", |
| "PASS_TO_PASS", |
| ] |
|
|
| MAX_PROBLEM_CHARS = 4000 |
| MAX_GOLD_ADDED_LINES = 120 |
|
|
| failures = [] |
|
|
|
|
| def check(condition, message): |
| if not condition: |
| failures.append(message) |
|
|
|
|
| def write_tree(files, dest): |
| for rel, content in files.items(): |
| path = dest / rel |
| path.parent.mkdir(parents=True, exist_ok=True) |
| with open(path, "w", encoding="utf-8", newline="\n") as handle: |
| handle.write(content) |
|
|
|
|
| def git(args, cwd): |
| return subprocess.run( |
| [ |
| "git", |
| "-c", "core.autocrlf=false", |
| "-c", "user.name=v", |
| "-c", "user.email=v@example.invalid", |
| ] |
| + args, |
| cwd=str(cwd), |
| capture_output=True, |
| text=True, |
| encoding="utf-8", |
| ) |
|
|
|
|
| def touched_paths(patch): |
| paths = set() |
| for line in patch.splitlines(): |
| if line.startswith("+++ b/") or line.startswith("--- a/"): |
| paths.add(line[6:]) |
| elif line.startswith("+++ ") or line.startswith("--- "): |
| value = line[4:].strip() |
| if value != "/dev/null": |
| paths.add(value.split("/", 1)[-1]) |
| return paths |
|
|
|
|
| def added_lines(patch): |
| return sum(1 for line in patch.splitlines() if line.startswith("+") and not line.startswith("+++")) |
|
|
|
|
| def main(): |
| rows = [json.loads(line) for line in open(DATASET, encoding="utf-8") if line.strip()] |
| check(len(rows) == 4, "expected 4 rows, found %d" % len(rows)) |
|
|
| seen = set() |
| for row in rows: |
| iid = row.get("instance_id", "<missing>") |
| for field in SWE_BENCH_FIELDS: |
| check(field in row, "%s: missing field %s" % (iid, field)) |
| check(iid not in seen, "%s: duplicate instance_id" % iid) |
| seen.add(iid) |
|
|
| check(isinstance(row["FAIL_TO_PASS"], str), "%s: FAIL_TO_PASS must be a JSON string" % iid) |
| check(isinstance(row["PASS_TO_PASS"], str), "%s: PASS_TO_PASS must be a JSON string" % iid) |
| f2p = json.loads(row["FAIL_TO_PASS"]) |
| p2p = json.loads(row["PASS_TO_PASS"]) |
| check(len(f2p) >= 1, "%s: empty FAIL_TO_PASS" % iid) |
| check(len(p2p) >= 1, "%s: empty PASS_TO_PASS" % iid) |
| check(not (set(f2p) & set(p2p)), "%s: F2P/P2P overlap" % iid) |
| check(len(row["base_commit"]) == 40, "%s: base_commit is not a 40-char sha" % iid) |
|
|
| problem = row["problem_statement"] |
| check(len(problem) <= MAX_PROBLEM_CHARS, "%s: problem statement too long" % iid) |
| for test_id in f2p + p2p: |
| check("::" not in problem, "%s: problem statement leaks a test id" % iid) |
|
|
| gold_paths = touched_paths(row["patch"]) |
| test_paths = touched_paths(row["test_patch"]) |
| check(not (gold_paths & test_paths), "%s: patch and test_patch touch the same files" % iid) |
| check( |
| all(not path.startswith("tests/") for path in gold_paths), |
| "%s: gold patch edits test files" % iid, |
| ) |
| check( |
| all(path.startswith("tests/") for path in test_paths), |
| "%s: test_patch edits non-test files" % iid, |
| ) |
| check( |
| added_lines(row["patch"]) <= MAX_GOLD_ADDED_LINES, |
| "%s: gold patch adds %d lines (budget %d)" |
| % (iid, added_lines(row["patch"]), MAX_GOLD_ADDED_LINES), |
| ) |
|
|
| files = row["dockerless"]["files"] |
| check(bool(files), "%s: empty snapshot" % iid) |
|
|
| with tempfile.TemporaryDirectory(prefix="swedl-val-") as tmp: |
| work = Path(tmp) / "w" |
| work.mkdir() |
| write_tree(files, work) |
| git(["init", "-q", "-b", "main"], work) |
| git(["add", "-A"], work) |
| git(["commit", "-q", "-m", "base"], work) |
|
|
| for label, patch_text in (("test_patch", row["test_patch"]), ("patch", row["patch"])): |
| patch_file = Path(tmp) / (label + ".patch") |
| patch_file.write_text(patch_text, encoding="utf-8", newline="\n") |
| result = git(["apply", "--check", str(patch_file)], work) |
| check( |
| result.returncode == 0, |
| "%s: %s does not apply (%s)" % (iid, label, result.stderr.strip()[:200]), |
| ) |
| if label == "test_patch": |
| git(["apply", str(patch_file)], work) |
|
|
| print( |
| "%-22s fields=ok F2P=%-2d P2P=%-2d +%-3d lines problem=%d chars files=%d" |
| % ( |
| iid, |
| len(f2p), |
| len(p2p), |
| added_lines(row["patch"]), |
| len(problem), |
| len(files), |
| ) |
| ) |
|
|
| total_tests = sum(len(json.loads(row["FAIL_TO_PASS"])) for row in rows) |
| print("\n%d instances, %d FAIL_TO_PASS tests total" % (len(rows), total_tests)) |
| if failures: |
| print("\nFAILED %d checks:" % len(failures)) |
| for message in failures: |
| print(" - %s" % message) |
| return 1 |
| print("all checks passed") |
| return 0 |
|
|
|
|
| if __name__ == "__main__": |
| sys.exit(main()) |
|
|
