verify_release.py

"""Verify local release hashes, retained originals, and derived data. No network."""
from pathlib import Path, PurePosixPath
import hashlib
import json
import sys

ROOT = Path(__file__).resolve().parent


def safe_path(name):
    p = PurePosixPath(name)
    if p.is_absolute() or not p.parts or any(x in ('..', '.') for x in p.parts) or '\\' in name or ':' in name:
        raise ValueError(f'Unsafe release path: {name!r}')
    local = ROOT.joinpath(*p.parts)
    if local.is_symlink() or not local.resolve().is_relative_to(ROOT.resolve()):
        raise ValueError(f'Path escapes release: {name!r}')
    return local


def verify():
    manifest = json.loads((ROOT/'release_manifest.json').read_text(encoding='utf-8'))
    files = manifest['files']
    if not files:
        raise ValueError('Empty manifest')
    for name, entry in files.items():
        raw = safe_path(name).read_bytes()
        if len(raw) != entry['bytes'] or hashlib.sha256(raw).hexdigest() != entry['sha256']:
            raise ValueError(f'Release checksum mismatch: {name}')
    inventory = json.loads((ROOT/'provenance/source_inventory.json').read_text(encoding='utf-8'))
    for entry in inventory['files']:
        raw = safe_path(entry['preserved_path']).read_bytes()
        if hashlib.sha256(raw).hexdigest() != entry['sha256'] or len(raw) != entry['bytes']:
            raise ValueError(f'Original source changed: {entry["archive_path"]}')
    counts = {}
    for name in ('research_corpus','claims','review_tasks'):
        rows = [json.loads(line) for line in (ROOT/f'data/{name}.jsonl').read_text(encoding='utf-8').splitlines()]
        if len({row['id'] for row in rows}) != len(rows):
            raise ValueError(f'Duplicate record IDs: {name}')
        schema = json.loads((ROOT/f'schemas/{name}.schema.json').read_text(encoding='utf-8'))
        for row in rows:
            if set(row) != set(schema['required']):
                raise ValueError(f'Record fields differ from schema: {name}')
            for key, spec in schema['properties'].items():
                value = row[key]
                expected = spec['type']
                if expected == 'string' and not isinstance(value,str):
                    raise ValueError(f'Expected string in {name}.{key}')
                if expected == 'integer' and (not isinstance(value,int) or isinstance(value,bool)):
                    raise ValueError(f'Expected integer in {name}.{key}')
            if name == 'research_corpus':
                raw = safe_path(row['source_path']).read_bytes()
                text = raw.decode('utf-8')
                start,end = row['char_start'],row['char_end']
                if not 0 <= start < end <= len(text) or row['text'] != text[start:end]:
                    raise ValueError(f'Corpus slice mismatch: {row["id"]}')
                if hashlib.sha256(raw).hexdigest() != row['source_sha256']:
                    raise ValueError(f'Corpus source mismatch: {row["id"]}')
            elif name == 'claims':
                raw = safe_path(row['source_path']).read_bytes()
                if hashlib.sha256(raw).hexdigest() != row['source_sha256']:
                    raise ValueError(f'Claim source mismatch: {row["id"]}')
            elif name == 'review_tasks':
                for source in json.loads(row['required_sources']):
                    if not safe_path(source).is_file():
                        raise ValueError(f'Missing task source: {source}')
        counts[name] = len(rows)
    import csv
    from tools.build_agent_data import typed_csv
    for path in sorted((ROOT/'results').glob('*.csv')):
        with path.open(newline='',encoding='utf-8') as f:
            source_rows = [{key:typed_csv(value) for key,value in row.items()} for row in csv.DictReader(f)]
        derived_rows = [json.loads(line) for line in (ROOT/f'data/{path.stem}.jsonl').read_text(encoding='utf-8').splitlines()]
        if source_rows != derived_rows:
            raise ValueError(f'Derived experiment table differs: {path.stem}')
        counts[path.stem] = len(source_rows)
    if counts != json.loads((ROOT/'data/counts.json').read_text(encoding='utf-8')):
        raise ValueError('Dataset count mismatch')
    print(f'PASS: {len(files)} release files; {len(inventory["files"])} original files; dataset source slices and counts verified.')
    print('Checksums establish local integrity, not authorship, mathematical truth, or publication priority.')
    return manifest


if __name__ == '__main__':
    try:
        verify()
    except (OSError, ValueError, KeyError, TypeError) as exc:
        print(f'FAIL: {exc}',file=sys.stderr)
        raise SystemExit(1)