| """Inspect task statements, evaluation entry points and pinned environments.""" |
| from pathlib import Path |
| import argparse |
| import json |
| import re |
| import subprocess |
|
|
| from common import load_task |
| from statement_text import read_problem_statement |
|
|
| ROOT = Path(__file__).resolve().parents[1] |
|
|
|
|
| def entries(): |
| return {entry['instance_id']: entry for entry in json.loads( |
| (ROOT / 'data/evaluation-entrypoints.json').read_text())['instances']} |
|
|
|
|
| def support(entry): |
| if entry['candidate_support'] == 'explicit_static_API_available': |
| return 'candidate_api_detected' |
| return 'instance_specific' |
|
|
|
|
| def task_row(instance_id): |
| with (ROOT / 'data/crossswe-bench.jsonl').open() as stream: |
| for line in stream: |
| row = json.loads(line) |
| if row['instance_id'] == instance_id: |
| return row |
| raise ValueError('Task is absent from the dataset table') |
|
|
|
|
| def main(): |
| parser = argparse.ArgumentParser(description=__doc__) |
| parser.add_argument('command', choices=['list', 'show', 'statement', 'image', 'pull', 'shell', 'evaluator']) |
| parser.add_argument('--instance-id') |
| parser.add_argument('--candidate-support', choices=['all', 'detected', 'manual'], default='all', |
| help='Filter list by statically detected API or instance-specific instructions') |
| parser.add_argument('--format', choices=['json', 'text'], default='json', |
| help='Output format for evaluator inspection; never executes an evaluator') |
| args = parser.parse_args() |
| records = [json.loads(line) for line in (ROOT / 'data/images.jsonl').read_text().splitlines()] |
| table = {row['instance_id']: row for row in records} |
| if args.command == 'list': |
| index = entries() if args.candidate_support != 'all' else None |
| for row in records: |
| if index is not None: |
| detected = support(index[row['instance_id']]) == 'candidate_api_detected' |
| if detected != (args.candidate_support == 'detected'): |
| continue |
| print(row['instance_id']) |
| return |
| if args.instance_id not in table: |
| parser.error('Provide a canonical --instance-id from the list command') |
| row = table[args.instance_id] |
| image = row['image_ref'] |
| if not re.fullmatch(r'silinchen0617/crossrepo-bench@sha256:[0-9a-f]{64}', image): |
| raise ValueError('The image must use its pinned registry digest') |
| if args.command == 'image': |
| print(image) |
| elif args.command == 'statement': |
| print(read_problem_statement(load_task(args.instance_id)), end='') |
| elif args.command == 'evaluator': |
| entry = entries()[args.instance_id] |
| if args.format == 'json': |
| print(json.dumps(entry, indent=2)) |
| else: |
| print('Instance: ' + args.instance_id) |
| print('Evaluation support: ' + support(entry)) |
| if entry.get('candidate_entry'): |
| print('Candidate entry: ' + entry['candidate_entry']) |
| print('Interface detected by static inspection; runtime requirements still apply.') |
| else: |
| print('Use the instance contract and instructions; no verified candidate API is indexed.') |
| print('Contract: instances/' + args.instance_id + '/evaluation/contract.json') |
| for path in entry.get('documentation', []): |
| print('Documentation: ' + path) |
| if not entry.get('documentation'): |
| print('Environment metadata: ' + entry['image_json']) |
| if entry.get('candidate_required_runtime_inputs'): |
| print('Additional runtime inputs: ' + ', '.join(entry['candidate_required_runtime_inputs'])) |
| print('Directory: docs/evaluators.md') |
| elif args.command == 'show': |
| task = load_task(args.instance_id) |
| record = task_row(args.instance_id) |
| print(json.dumps({**row, 'repositories': task['repositories'], |
| 'statement_file': 'statements/' + args.instance_id + '.md', |
| 'PASS_TO_PASS_count': record['PASS_TO_PASS_count'], |
| 'FAIL_TO_PASS_count': record['FAIL_TO_PASS_count'], |
| 'evaluation_support': record['evaluation_support'], |
| 'evaluation_contract': 'instances/' + args.instance_id + '/' + task['evaluation']['path']}, indent=2)) |
| elif args.command == 'pull': |
| subprocess.run(['docker', 'pull', '--platform', 'linux/amd64', image], check=True) |
| else: |
| subprocess.run(['docker', 'run', '--rm', '-it', '--platform', 'linux/amd64', |
| '--network=none', '--entrypoint=/bin/bash', image], check=True) |
|
|
|
|
| if __name__ == '__main__': |
| main() |
|
|
