#!/usr/bin/env python3
"""Local structural checks for the BehaviorBench NeurIPS artifact."""
import json, re, sys
from pathlib import Path
import pyarrow as pa
import pyarrow.parquet as pq

ROOT=Path(__file__).resolve().parents[1]
errors=[]
required=["README.md","DATASET_CARD.md","LICENSE","CHANGELOG.md","croissant.json","docs/RESPONSIBLE_AI.md"]
for x in required:
    if not (ROOT/x).exists(): errors.append(f"missing {x}")

try: meta=json.loads((ROOT/"croissant.json").read_text())
except Exception as e: errors.append(f"invalid croissant JSON: {e}"); meta={}

rai=["rai:dataLimitations","rai:dataBiases","rai:personalSensitiveInformation","rai:dataUseCases","rai:dataSocialImpact","rai:hasSyntheticData","rai:dataCollection"]
for k in rai:
    if k not in meta: errors.append(f"missing Croissant RAI field {k}")

files={p.relative_to(ROOT).as_posix():p for top in (ROOT/"dataset",ROOT/"benchmark") for p in top.rglob("*.parquet") if "evaluation/sources" not in p.as_posix()}
dists={x.get("@id"):x for x in meta.get("distribution",[])}
records={x.get("name"):x for x in meta.get("recordSet",[])}
for rel,p in files.items():
    if rel not in dists: errors.append(f"no distribution for {rel}")
    if rel not in records: errors.append(f"no recordSet for {rel}")
    else:
        schema=pq.ParquetFile(p).schema_arrow
        actual=set(schema.names)
        described={x.get("name") for x in records[rel].get("field",[])}
        if not described.issubset(actual):
            errors.append(f"schema contains unknown fields for {rel}")
        omitted=actual-described
        for name in omitted:
            t=schema.field(name).type
            # Current mlcroissant record generation cannot reliably cast
            # nested Parquet structs, including list<struct<...>> columns.
            nested_struct = pa.types.is_struct(t) or (
                (pa.types.is_list(t) or pa.types.is_large_list(t))
                and pa.types.is_struct(t.value_type)
            )
            if not nested_struct:
                errors.append(f"unexplained omitted field {name} for {rel}")
for rel in dists:
    if rel not in files: errors.append(f"distribution path not found: {rel}")

# Relative Markdown links.
for p in [ROOT/"README.md",ROOT/"DATASET_CARD.md",*sorted((ROOT/"docs").glob("*.md"))]:
    for link in re.findall(r"\[[^]]+\]\(([^)]+)\)",p.read_text()):
        if "://" not in link and not link.startswith("#") and not (p.parent/link).resolve().exists():
            errors.append(f"broken link {p.relative_to(ROOT)} -> {link}")

if errors:
    print("FAIL")
    print("\n".join(f"- {x}" for x in errors));sys.exit(1)
print(f"PASS: {len(files)} Parquet distributions/record sets; required RAI fields and local links present.")