#!/usr/bin/env python3 """Verify both track exports, physical separation, long-paper proof and provenance.""" import csv import gzip import hashlib import io import json import os import re import subprocess import sys import xml.etree.ElementTree as ET from pathlib import Path from urllib.parse import unquote,urlsplit from paper_length_policy import excluded_publication_reason ROOT=Path(__file__).resolve().parents[1] AUDIT=ROOT/'checks/track_screening' def rows(path): with path.open(encoding='utf-8-sig',newline='') as stream:return list(csv.DictReader(stream)) def main(): for track in ('main','findings'): subprocess.run([sys.executable,str(ROOT/'scripts/verify_collection.py')], env={**os.environ,'AI4FINANCE_TRACK':track},check=True) catalogs={t:{p['id']:p for p in rows(ROOT/t/'finance_catalog.csv')} for t in ('main','findings')} assert not (set(catalogs['main']) & set(catalogs['findings'])) assert not ({p['paper_url'] for p in catalogs['main'].values()} & {p['paper_url'] for p in catalogs['findings'].values()}) assert set(p['venue'] for p in catalogs['main'].values())=={'ICML','ICLR','ACL','EMNLP','NeurIPS','AAAI'} assert set(p['venue'] for p in catalogs['findings'].values())<={'ACL','EMNLP'} for forbidden in ('finance_catalog.csv','finance_catalog.jsonl','pdfs','indexes'): assert not (ROOT/forbidden).exists(),('Mixed root artifact',forbidden) original={p['id']:p for p in csv.DictReader(io.StringIO(gzip.decompress((AUDIT/'finance_catalog.csv.baseline.gz').read_bytes()).decode('utf-8-sig')))} original_records={p['id']:p for p in map(json.loads,gzip.decompress((AUDIT/'finance_catalog.jsonl.baseline.gz').read_bytes()).decode().splitlines())} main_records={p['id']:p for p in map(json.loads,(ROOT/'main/finance_catalog.jsonl').read_text().splitlines())} length={p['id']:p for p in json.loads((AUDIT/'main_length_review.json').read_text())} assert set(catalogs['main'])=={k for k,v in length.items() if v['paper_length']=='long' and not excluded_publication_reason(original_records[k])} for pid,row in catalogs['main'].items(): assert {k:row[k] for k in original[pid]}==original[pid],(pid,'Original financial catalog fields changed') assert {k:v for k,v in main_records[pid].items() if k not in ('paper_length','paper_length_evidence')}==original_records[pid],pid findings_records={p['id']:p for p in map(json.loads,(ROOT/'findings/finance_catalog.jsonl').read_text().splitlines())} decisions=json.loads((AUDIT/'findings_review_decisions.json').read_text()) pdfs={p['id']:p for p in json.loads((AUDIT/'findings_fulltext_results.json').read_text())} candidate_records=json.loads((AUDIT/'findings_records.json').read_text()) assert set(pdfs)=={p['id'] for p in candidate_records if p['fulltext_candidate']} assert all(p['status']=='inspected' for p in pdfs.values()) assert set(catalogs['findings'])=={k for k in decisions if pdfs[k]['content_exceeds_short_limit']} for track,records in [('main',main_records),('findings',findings_records)]: for pid,p in records.items(): e=p['paper_length_evidence'] assert e['paper_length']=='long' if e['evidence_type']=='publisher_content_boundary_inspection': proof=e['inspection'];assert proof['inspected_content_page']>=6 assert proof['sixth_content_page_words']>=160 assert proof['first_backmatter_page']>=proof['inspected_content_page'] assert e['pdf_sha256']==p['download']['sha256'] if proof['inspected_content_page']>9: visual=json.loads((AUDIT/'long_only_recheck.json').read_text())['visual_checks'][pid] assert visual['pdf_sha256']==p['download']['sha256'] assert visual['inspected_pdf_page']==6 and visual['main_body_confirmed'] if track=='findings': assert pdfs[pid]['title_token_overlap']>=0.8 assert p['finance']['scope']==decisions[pid]['scope'] official_ids=set() for source in json.loads((AUDIT/'findings_sources.json').read_text()): path=ROOT/source['file'] data=gzip.decompress(path.read_bytes()) assert hashlib.sha256(data).hexdigest()==source['uncompressed_sha256'] tree=ET.fromstring(data);year=tree.get('id').split('.')[0] for volume in tree.findall('volume'): if volume.get('id') in ('acl','emnlp'): official_ids.update(f"{year}.findings-{volume.get('id')}.{p.get('id')}" for p in volume.findall('paper')) inventory=list(csv.DictReader(io.StringIO(gzip.decompress((AUDIT/'screening_inventory.csv.gz').read_bytes()).decode('utf-8-sig')))) assert official_ids=={p['id'] for p in inventory if p['track']=='findings'} assert {p['id'] for p in inventory if p['disposition']=='admitted_long_finance'}==set(main_records)|set(findings_records) main_source=json.loads((AUDIT/'main_sources.json').read_text()) assert hashlib.sha256(gzip.decompress((ROOT/main_source['file']).read_bytes())).hexdigest()==main_source['uncompressed_sha256'] links=0 for path in ROOT.rglob('*.md'): if '.git' in path.parts:continue for target in re.findall(r'\]\(([^)]+)\)',path.read_text()): url=urlsplit(target.strip('<>')) if url.scheme or url.netloc or not url.path:continue resolved=(path.parent/unquote(url.path)).resolve() assert resolved.is_relative_to(ROOT) and (resolved.exists() or resolved==AUDIT/'validation.json'),(str(path.relative_to(ROOT)),target) links+=1 report=dict(main_long_papers=len(main_records),findings_long_papers=len(findings_records), all_track_ids_disjoint=True,all_root_and_track_links_resolve=True, local_markdown_links_checked=links,main_original_financial_fields_preserved=True, official_findings_inventory_complete=True,findings_source_records=len(official_ids), screening_inventory_records=len(inventory),long_paper_evidence_validated=True) report['explicit_short_paper_records']=0 recheck=json.loads((AUDIT/'long_only_recheck.json').read_text()) assert recheck['admitted_main']==len(main_records) and recheck['admitted_findings']==len(findings_records) (AUDIT/'validation.json').write_text(json.dumps(report,indent=2)+'\n') print(json.dumps(report,indent=2)) if __name__=='__main__':main()