tests/test_validate.py

"""Preprocess validate stage + pipeline — integrity/graded checks, flag-never-drop, contract columns."""

from __future__ import annotations

import copy
from pathlib import Path

import orjson

from preprocess import DATASET_AS_OF, FLAG_COLUMNS, load_candidates, run_preprocess, validate


def _clean_record() -> dict:
    """A coherent candidate that should trip no checks."""
    return {
        "candidate_id": "CAND_0000001",
        "profile": {"years_of_experience": 5.0, "current_title": "AI Engineer"},
        "career_history": [
            {"start_date": "2018-01-01", "end_date": "2023-01-01", "duration_months": 60,
             "is_current": False, "company": "X", "title": "Eng", "industry": "Tech",
             "company_size": "51-200", "description": "built things"}
        ],
        "skills": [{"name": "Python", "proficiency": "advanced", "endorsements": 5, "duration_months": 36}],
        "redrob_signals": {
            "signup_date": "2023-01-01", "last_active_date": "2024-01-01",
            "expected_salary_range_inr_lpa": {"min": 10, "max": 20},
            "verified_email": True, "verified_phone": True, "linkedin_connected": True,
        },
    }


def _validate_records(tmp_path: Path, records: list[dict]):
    path = tmp_path / "c.jsonl"
    path.write_bytes(b"\n".join(orjson.dumps(r) for r in records))
    return validate(load_candidates(path), DATASET_AS_OF)


def test_clean_record_trips_no_checks(tmp_path: Path) -> None:
    row = _validate_records(tmp_path, [_clean_record()]).row(0, named=True)
    assert row["validation_failed"] is False
    assert row["incoherence_score"] == 0.0
    assert row["trust_score"] == 3
    assert row["active_validated"] == 1 and row["experience_validated"] == 1


def test_expert_zero_is_integrity_failure(tmp_path: Path) -> None:
    record = _clean_record()
    record["skills"] = [{"name": "Rust", "proficiency": "expert", "endorsements": 1, "duration_months": 0}]
    row = _validate_records(tmp_path, [record]).row(0, named=True)
    assert row["chk_expert_zero"] is True
    assert row["validation_failed"] is True  # any integrity hit hard-fails


def test_career_dates_inverted_is_integrity_failure(tmp_path: Path) -> None:
    record = _clean_record()
    record["career_history"][0]["start_date"] = "2023-01-01"
    record["career_history"][0]["end_date"] = "2020-01-01"
    row = _validate_records(tmp_path, [record]).row(0, named=True)
    assert row["chk_career_dates_invalid"] is True and row["validation_failed"] is True


def test_impossible_yoe_is_integrity_failure(tmp_path: Path) -> None:
    record = _clean_record()
    record["profile"]["years_of_experience"] = 50.0  # vs a ~8yr career span
    row = _validate_records(tmp_path, [record]).row(0, named=True)
    assert row["chk_yoe_impossible"] is True and row["validation_failed"] is True


def test_salary_incoherence_is_graded_not_hard_fail(tmp_path: Path) -> None:
    record = _clean_record()
    record["redrob_signals"]["expected_salary_range_inr_lpa"] = {"min": 30, "max": 10}
    row = _validate_records(tmp_path, [record]).row(0, named=True)
    assert row["chk_salary_incoherent"] is True
    assert row["validation_failed"] is False  # graded → penalty, not a hard fail
    assert abs(row["incoherence_score"] - 0.10) < 1e-6
    assert row["salary_range_ok"] == 0


def test_date_order_is_graded(tmp_path: Path) -> None:
    record = _clean_record()
    record["redrob_signals"]["signup_date"] = "2024-06-01"
    record["redrob_signals"]["last_active_date"] = "2024-01-01"
    row = _validate_records(tmp_path, [record]).row(0, named=True)
    assert row["chk_date_order_invalid"] is True
    assert abs(row["incoherence_score"] - 0.15) < 1e-6
    assert row["date_order_ok"] == 0 and row["active_validated"] == 0


def test_validate_never_drops_and_emits_every_flag_column(tmp_path: Path) -> None:
    validated = _validate_records(tmp_path, [_clean_record(), _clean_record()])
    assert validated.height == 2  # flag, never drop
    for flag_column in FLAG_COLUMNS:
        assert flag_column in validated.columns, f"missing flag column {flag_column}"


def test_run_preprocess_returns_frame_and_reference_date(tmp_path: Path) -> None:
    path = tmp_path / "c.jsonl"
    records = [_clean_record(), copy.deepcopy(_clean_record())]
    records[1]["candidate_id"] = "CAND_0000002"
    path.write_bytes(b"\n".join(orjson.dumps(r) for r in records))
    frame, as_of = run_preprocess(path)
    assert as_of == DATASET_AS_OF
    assert frame.height == 2
    assert "validation_failed" in frame.columns