import json, pandas as pd
from jof250 import inventory, provenance

def test_sha256_and_duplicates(tmp_path, synthetic_pdf):
    dup = tmp_path / "copy.pdf"; dup.write_bytes(synthetic_pdf.read_bytes())
    (tmp_path / "notes.txt").write_text("not a pdf")
    df = inventory.scan(tmp_path)
    assert len(df) == 3 and df.sha256.nunique() == 2
    assert provenance.sha256_file(synthetic_pdf) == provenance.sha256_file(dup)
    assert df[df.ext == ".pdf"].pdf_valid.all()

def test_pdf_probe_extracts_doi(synthetic_pdf):
    r = inventory.pdf_probe(synthetic_pdf); assert r["pages"] == 1 and r["doi_in_pdf"] == "10.1111/jofi.99999"

def test_classify_excludes_non_pdf_and_duplicates(tmp_path, synthetic_pdf, monkeypatch):
    dup = tmp_path / "copy.pdf"; dup.write_bytes(synthetic_pdf.read_bytes()); (tmp_path / "x.csv").write_text("a")
    df = inventory.scan(tmp_path); df["bib_doi"] = "10.1111/jofi.99999"; df["bib_year"] = 2018; df["bib_section"] = "ARTICLES"; df["bib_title"] = "A Synthetic Test Article"; df["bib_joined"] = True
    out = inventory.classify(df)
    assert (out.decision == "exclude").sum() == 3  # csv + 2 identical PDFs (both flagged as duplicate content)
    assert "non-PDF file" in set(out.decision_reason)

def test_frozen_manifest_exists_and_counts_match():
    from jof250 import config
    p = config.manifests_dir() / "corpus_v1_frozen.json"
    if not p.exists(): import pytest; pytest.skip("real corpus not audited on this machine")
    fz = json.loads(p.read_text()); man = pd.read_parquet(config.manifests_dir() / "corpus_manifest.parquet")
    assert fz["n_included"] == int((man.decision == "include").sum()) == len(fz["documents"])