from app.utils.climate_intelligence import (
    build_climate_intelligence_context,
    extract_nigeria_location,
    infer_crop_and_growth_stage,
)


def test_build_climate_intelligence_context_contains_climate_risk_and_recommendations():
    context = build_climate_intelligence_context(
        latitude=10.0,
        longitude=7.0,
        crop="maize",
        growth_stage="pre-planting",
        days=7,
    )

    assert isinstance(context, dict)
    assert "location" in context
    assert "current_weather" in context
    assert "forecast" in context
    assert "climate_risk" in context
    assert "recommendations" in context
    assert context["location"]["country"] == "Nigeria"
    assert context["climate_risk"] in {"low", "medium", "high"}
    assert isinstance(context["recommendations"], list)
    assert context["recommendations"]


def test_extract_nigeria_location_handles_lagos_and_ikeja():
    assert extract_nigeria_location("What is the weather like in Lagos?")[0] == "Lagos"
    assert extract_nigeria_location("Ikeja weather forecast")[0] == "Ikeja"


def test_infer_crop_and_growth_stage_handles_farm_questions():
    crop, stage = infer_crop_and_growth_stage("Should I plant maize in Kaduna before the rains?")
    assert crop == "maize"
    assert stage in {"pre-planting", "planting"}


def test_recommendations_are_deduped_within_risk_branch():
    """Recommendations coming out of build_climate_intelligence_context must not
    contain near-duplicate bullets, regardless of which branch produced them."""
    context = build_climate_intelligence_context(
        latitude=12.0,
        longitude=8.0,
        crop="maize",
        growth_stage="pre-planting",
        days=7,
    )
    recs = context["recommendations"]
    assert isinstance(recs, list)
    assert 1 <= len(recs) <= 3, f"expected 1–3 recs, got {len(recs)}"

    normalized = [set(rec.lower().split()) for rec in recs]
    for i, words_i in enumerate(normalized):
        for j, words_j in enumerate(normalized):
            if i >= j:
                continue
            overlap = len(words_i & words_j) / max(len(words_i), 1)
            assert overlap <= 0.5, (
                f"near-duplicate recommendations detected: "
                f"{recs[i]!r} vs {recs[j]!r} (overlap={overlap:.2f})"
            )


def test_risk_reason_is_a_single_readable_sentence():
    """risk_reason must be a single readable summary — no '|' separators and
    no obvious self-repetition."""
    context = build_climate_intelligence_context(
        latitude=12.0,
        longitude=8.0,
        crop="maize",
        growth_stage="pre-planting",
        days=7,
    )
    risk_reason = context["risk_reason"]
    assert isinstance(risk_reason, str)
    assert "|" not in risk_reason, f"unexpected pipe in risk_reason: {risk_reason!r}"
    # Defensive: the first half of the sentence must not be repeated verbatim
    # in the second half.
    if len(risk_reason) > 40:
        half = len(risk_reason) // 2
        first = risk_reason[:half].strip().lower()
        second = risk_reason[half:].strip().lower()
        # 25-char sequences from the first half should not reappear in the second half.
        seen = {first[i : i + 25] for i in range(0, max(len(first) - 25, 1), 10)}
        seen.discard("")
        for chunk in seen:
            assert chunk not in second, (
                f"risk_reason appears to repeat itself: {risk_reason!r}"
            )