tests/test_validators.py

"""Deterministic-validator tests.

These pin the behaviour judges care about: the rules engine catches what it
should and stays quiet on a clean report. No LLM, no network — pure and fast.

Run from the project root:  pytest -q
"""

import os
import sys

sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))

import schema                                        # noqa: E402
from validators import run_all                      # noqa: E402
from sample_reports import CLEAN, BROKEN            # noqa: E402


def _by_category(findings):
    return {f.category for f in findings}


def _with(**overrides):
    """A copy of the clean report with specific fields overridden."""
    return {**CLEAN, **overrides}


def test_clean_report_has_no_findings():
    findings = run_all(CLEAN)
    assert findings == [], f"clean report should be silent, got: {findings}"


def test_broken_report_flags_all_three():
    findings = run_all(BROKEN)
    cats = _by_category(findings)
    assert "INCONSISTENT_DATA" in cats   # speed mismatch + FO ROB drift
    assert "OUT_OF_RANGE" in cats        # wind force 14 Bf


def test_speed_mismatch_detected():
    findings = run_all(BROKEN)
    msgs = " ".join(f.message.lower() for f in findings)
    assert "speed" in msgs


def test_fo_rob_drift_is_critical():
    findings = run_all(BROKEN)
    rob = [f for f in findings if "ROB" in f.field]
    assert rob, "expected an FO ROB reconciliation finding"
    assert any(f.severity == "CRITICAL" for f in rob)


def test_wind_force_out_of_range():
    findings = run_all(BROKEN)
    assert any(f.category == "OUT_OF_RANGE" and "Wind" in f.field for f in findings)


def test_missing_required_field_is_critical():
    incomplete = {k: v for k, v in CLEAN.items() if k != "fo_rob"}
    findings = run_all(incomplete)
    assert any(f.category == "MISSING_DATA" and f.severity == "CRITICAL"
               for f in findings)


def test_findings_sorted_critical_first():
    findings = run_all(BROKEN)
    order = {"CRITICAL": 0, "ERROR": 1, "WARNING": 2}
    sev = [order[f.severity] for f in findings]
    assert sev == sorted(sev), "findings should be sorted CRITICAL-first"


# --- advanced checks -------------------------------------------------------

def test_me_fo_rate_out_of_band():
    # 240 mt over 24 h = 10 mt/h, well above the band -> WARNING.
    findings = run_all(_with(me_fo_cons=240.0, fo_rob_prev=None))
    assert any("ME FO" in f.field and "mt/h" in f.message for f in findings)


def test_me_fo_rate_in_band_is_silent():
    # CLEAN burns 38.5 over 24 h = 1.6 mt/h -> inside band, no rate finding.
    findings = run_all(CLEAN)
    assert not any("mt/h" in f.message for f in findings)


def test_high_slip_in_calm_water():
    findings = run_all(_with(slip_pct=25.0, wind_force=2.0))
    assert any(f.field == "Slip" and "calm" in f.message.lower()
               for f in findings)


def test_high_speed_in_heavy_weather():
    # keep distance/hours consistent with the higher speed so only the
    # weather rule fires, not the speed/distance mismatch.
    findings = run_all(_with(avg_speed=15.0, distance_run=360.0, wind_force=9.0))
    assert any(f.field == "Average speed" and "heavy weather" in f.message
               for f in findings)


def test_slip_reconciliation_disabled_without_pitch():
    # Default pitch is None -> the RPM x pitch reconciliation must not fire,
    # even with an absurd reported slip.
    findings = run_all(_with(slip_pct=20.0))
    assert not any("computed from" in f.message for f in findings)


def test_slip_reconciliation_fires_with_pitch(monkeypatch):
    monkeypatch.setattr(schema, "PROP_PITCH_M", 5.6)
    # geometry implies ~1% slip; reporting 20% is a clear disagreement.
    findings = run_all(_with(slip_pct=20.0))
    assert any(f.field == "Slip" and "computed from" in f.message
               for f in findings)