import gzip import hashlib import json from pathlib import Path import tempfile import unittest from unittest.mock import patch from paper_atlas.collectors import mm2026 from paper_atlas.audits.source_identity import audit, load_inputs, load_mm_program_inputs class MMProgramIdentityInputTests(unittest.TestCase): @classmethod def setUpClass(cls): cls.fixture = json.loads(gzip.decompress((Path(__file__).parent/'fixtures/mm2026-source.json.gz').read_bytes())) def setUp(self): self.tmp = tempfile.TemporaryDirectory() self.addCleanup(self.tmp.cleanup) self.root = Path(self.tmp.name) src = self.root/'source' for name,body in self.fixture['files'].items(): path=src/name;path.parent.mkdir(parents=True,exist_ok=True);path.write_bytes(body.encode()) mm2026.collect(self.root/'program', snapshots=src) self.report_path = self.root/'program/collection-report.json' def change_report(self, change): report=json.loads(self.report_path.read_text());change(report) self.report_path.write_text(json.dumps(report)) def assert_rejected_before_database(self): with patch('paper_atlas.audits.source_identity.ReadConnection') as connection: with self.assertRaises(ValueError): audit(None,self.root/'audit',pg_service='unused',pg_schema='research',mm_program_report=self.report_path) connection.assert_not_called() self.assertFalse((self.root/'audit').exists()) def test_all2143_programs_are_unknown_readonly_inputs_not_import_records(self): rows,inputs=load_mm_program_inputs(self.report_path) self.assertEqual(len(rows),2143);self.assertEqual(len(inputs),28) self.assertEqual(sum(len(r['program_contexts']) for r in rows),4326) actual={r['program_id']:r for r in rows} for gold in self.fixture['independent_inventory']['inventory']: r=actual[gold['paper_id']] self.assertEqual((r['title'],r['authors'],r['track']), (gold['title'],gold['authors'],gold['track_literal'])) self.assertEqual(r['external_id'],'2026/'+gold['paper_id']) self.assertEqual(r['status'],'unknown');self.assertEqual(r['identifiers'],{}) self.assertEqual(r['source_observed_status'],'listed_in_official_program') self.assertFalse(r['publication_identity_verified']);self.assertFalse(r['final_decision_verified']) self.assertFalse(r['contribution_type_verified']) self.assertEqual([c['source_pointer'] for c in r['program_contexts']],gold['locations']) self.assertEqual({r['program_id'] for r in rows if not r['authors']},{'mfp4971','mfp7801'}) self.assertTrue(all('unknown is not a source decision' in i['manifest']['scope'] for i in inputs)) self.assertFalse(any(self.root.rglob('*.manifest.json'))) def test_untrusted_absolute_file_paths_are_not_followed(self): self.change_report(lambda r:[f.update(file='/untrusted/'+Path(f['file']).name) for f in r['files']]) rows,inputs=load_mm_program_inputs(self.report_path) self.assertEqual(len(rows),2143) self.assertTrue(all(Path(i['file']).parent==self.report_path.parent for i in inputs)) def test_program_files_are_not_silently_accepted_by_regular_ingestion_input_loader(self): report=json.loads(self.report_path.read_text()) with self.assertRaises(FileNotFoundError):load_inputs([report['files'][0]['file']]) def test_corrupt_jsonl_fails_before_connection_or_output(self): report=json.loads(self.report_path.read_text());path=Path(report['files'][0]['file']) path.write_bytes(path.read_bytes()+b' ') self.assert_rejected_before_database() def test_rehashed_forged_identifiers_decisions_and_contexts_rejected(self): original=self.report_path.read_bytes();report=json.loads(original);path=Path(report['files'][0]['file']);raw=path.read_bytes() for field,value in [('identifiers',{'openreview':'invented'}),('status','accepted'),('contexts',[])]: path.write_bytes(raw);self.report_path.write_bytes(original) lines=raw.splitlines();row=json.loads(lines[0]);row[field]=value;lines[0]=json.dumps(row).encode() changed=b'\n'.join(lines)+b'\n';path.write_bytes(changed) self.change_report(lambda r:r['files'][0].update(file_sha256=hashlib.sha256(changed).hexdigest())) with self.subTest(field=field):self.assert_rejected_before_database() def test_missing_or_repeated_track_rejected(self): original=self.report_path.read_bytes() for kind in ('missing','repeated'): self.report_path.write_bytes(original) if kind=='missing':self.change_report(lambda r:r['files'].pop()) else:self.change_report(lambda r:r['files'].__setitem__(1,r['files'][0])) with self.subTest(kind=kind):self.assert_rejected_before_database() def test_raw_source_corruption_or_symlink_rejected(self): path=self.report_path.parent/'evidence/program-data.json';raw=path.read_bytes() path.write_bytes(raw+b' ');self.assert_rejected_before_database() path.write_bytes(raw);saved=path.with_name('original.json');path.rename(saved);path.symlink_to(saved) self.assert_rejected_before_database() def test_mixed_input_modes_rejected_before_connection(self): with patch('paper_atlas.audits.source_identity.ReadConnection') as connection: for files,program in [([],None),([Path('unused')],self.report_path)]: with self.assertRaisesRegex(ValueError,'Choose source files'): audit(files,self.root/'out',pg_service='unused',pg_schema='research',mm_program_report=program) connection.assert_not_called()