import pandas as pd

from isora_hf.panel import COMPOSITE_COLUMNS, G16, G18, G20, PANEL_COLUMNS, build_panel


def _obs(rows):
    frame = pd.DataFrame(
        rows,
        columns=[
            "jurisdiction_code",
            "fiscal_year",
            "questionnaire_generation",
            "indicator_code",
            "value_numeric",
        ],
    )
    frame["jurisdiction_name"] = "X"
    frame["survey_round"] = frame["fiscal_year"].map(
        {2015: "ISORA 2016", 2017: "ISORA 2018", 2023: "ISORA 2024"}
    )
    frame["indicator_label"] = "lbl " + frame["indicator_code"]
    frame["value_status"] = "value"
    frame["monetary_unit"] = None
    frame["value_local_currency_units"] = frame["value_numeric"] * 1000
    return frame


def _jur():
    return pd.DataFrame(
        {
            "jurisdiction_code": ["AAA"],
            "imf_region": ["Europe"],
            "world_bank_region_current": ["Europe & Central Asia"],
            "world_bank_income_group_current": ["High income"],
            "member_oecd": [True],
            "member_eu": [False],
            "member_iota": [True],
            "member_ciat": [False],
            "member_adb": [False],
        }
    )


def test_column_definitions_are_unique_and_reference_generations():
    names = [c.name for c in PANEL_COLUMNS]
    assert len(names) == len(set(names))
    for c in PANEL_COLUMNS:
        assert set(c.codes) <= {G16, G18, G20}
        assert c.codes or c.name in COMPOSITE_COLUMNS


def test_panel_uses_generation_specific_codes_and_drops_implausible_zeros():
    obs = _obs(
        [
            ("AAA", 2015, G16, "10790", 0.0),  # 2016 revenue-to-GDP published as 0 -> missing
            ("AAA", 2017, G18, "337_001", 17.5),
            (
                "AAA",
                2017,
                G18,
                "337_015",
                12.0,
            ),  # 2018 meaning differs -> must NOT feed capex_to_opex_pct
            ("AAA", 2023, G20, "337_015", 3.0),
            ("AAA", 2023, G20, "337_031", 10.0),
            ("AAA", 2023, G20, "337_032", 20.0),
            ("AAA", 2023, G20, "80040_3", 5.0),
        ]
    )
    income = pd.Series(
        ["High income", "High income", "High income"],
        index=pd.MultiIndex.from_tuples(
            [("AAA", 2015), ("AAA", 2017), ("AAA", 2023)],
            names=["jurisdiction_code", "fiscal_year"],
        ),
    )
    panel, dictionary = build_panel(obs, _jur(), income)
    p = panel.set_index("fiscal_year")
    assert pd.isna(p.at[2015, "revenue_to_gdp_pct"])
    assert p.at[2017, "revenue_to_gdp_pct"] == 17.5
    assert pd.isna(p.at[2017, "capex_to_opex_pct"]) and p.at[2023, "capex_to_opex_pct"] == 3.0
    assert p.at[2023, "staff_under_35_pct"] == 30.0
    assert p.at[2023, "net_revenue_lcu"] == 5000.0  # money columns come from harmonised units
    assert p.at[2023, "income_group_wb_classification_fy"] == "FY25"
    assert dictionary.loc[
        dictionary["column"] == "revenue_to_gdp_pct", "zeros_treated_as_missing"
    ].all()
    assert set(
        dictionary.loc[dictionary["column"] == "capex_to_opex_pct", "questionnaire_generation"]
    ) == {G20}