"""Tests for sber_index service — pure unit tests, no network/DB.

Coverage:
  (a) build_sber_route: base64 route builder produces correct encoded string, per dashboard.
  (b) decode_sber_response: decodes a sample base64 payload into correct rows.
  (c) pull_sber_indices upsert SQL is idempotent (ON CONFLICT shape assertion via
      fake-db roundtrip — verifies upsert is called with correct SQL and params).
  (d) Real-fixture decode: exercises actual sberindex.ru response format end-to-end.
  (e) Period parsing edge-cases.
"""

from __future__ import annotations

import base64
import json
import os
import sys
from datetime import date
from pathlib import Path
from unittest.mock import MagicMock, patch
from urllib.parse import unquote

# DATABASE_URL required by config before any app import.
os.environ.setdefault("DATABASE_URL", "postgresql+psycopg://test:test@localhost:5432/test")

# WeasyPrint stub — not installed in CI without GTK.
_wp_mock = MagicMock()
sys.modules.setdefault("weasyprint", _wp_mock)

import pytest  # noqa: E402

from app.services.sber_index import (  # noqa: E402
    SBER_DASHBOARDS,
    SberDashboard,
    _parse_period_month,
    build_sber_route,
    decode_sber_response,
)

# Lookup helper: slug → SberDashboard
_DASH: dict[str, SberDashboard] = {d.slug: d for d in SBER_DASHBOARDS}

# ---------------------------------------------------------------------------
# Fixture path (real sberindex.ru response)
# ---------------------------------------------------------------------------

FIXTURE_PATH = Path(__file__).parent / "fixtures" / "sber_real_estate_deals_66.json"


# ---------------------------------------------------------------------------
# (a) Route builder — per-dashboard filter assertions
# ---------------------------------------------------------------------------


def test_build_sber_route_rф_real_estate_deals() -> None:
    """Route for РФ (643) + real_estate_deals should decode to expected path."""
    d = _DASH["real_estate_deals"]
    encoded = build_sber_route(d.slug, "643", d.extra_filter)
    decoded = base64.b64decode(encoded).decode("utf-8")

    assert decoded.startswith("/dataset/v1/real_estate_deals?filter=")
    assert "limit=1000" in decoded
    assert "offset=0" in decoded

    # Extract filter JSON from decoded path
    filter_part = decoded.split("filter=")[1].split("&")[0]
    filter_json = json.loads(unquote(filter_part))

    assert filter_json["REF_AREA"] == "643"
    assert filter_json["FREQ"] == "M"
    assert filter_json["SOURCE"] == "SI"
    assert filter_json["REALTY"] == "2"


def test_build_sber_route_different_dashboard() -> None:
    """Route slug is correctly embedded in the decoded path."""
    d = _DASH["dinamika-tsen-obyavlenii"]
    encoded = build_sber_route(d.slug, "643", d.extra_filter)
    decoded = base64.b64decode(encoded).decode("utf-8")
    assert "/dataset/v1/dinamika-tsen-obyavlenii?" in decoded


def test_build_sber_route_is_valid_base64() -> None:
    """Encoded route must be decodeable ASCII base64."""
    d = _DASH["residential_real_estate_prices"]
    encoded = build_sber_route(d.slug, "77", d.extra_filter)
    # Should not raise
    decoded_bytes = base64.b64decode(encoded)
    assert len(decoded_bytes) > 0


def test_build_sber_route_ref_area_embedded() -> None:
    """REF_AREA value is preserved exactly in the filter JSON."""
    d = _DASH["real_estate_deals"]
    encoded = build_sber_route(d.slug, "66", d.extra_filter)
    decoded = base64.b64decode(encoded).decode("utf-8")
    filter_part = decoded.split("filter=")[1].split("&")[0]
    filter_json = json.loads(unquote(filter_part))
    assert filter_json["REF_AREA"] == "66"


@pytest.mark.parametrize("dashboard", SBER_DASHBOARDS)
def test_build_sber_route_per_dashboard(dashboard: SberDashboard) -> None:
    """Each dashboard must produce the correct per-dashboard filter keys (locks #794 fix)."""
    route_b64 = build_sber_route(dashboard.slug, "66", dashboard.extra_filter)
    path = base64.b64decode(route_b64).decode("utf-8")

    # Extract the filter JSON from the path: ?filter=<urlencoded>&limit=...
    filter_part = path.split("filter=")[1].split("&")[0]
    filter_dict = json.loads(unquote(filter_part))

    # All routes must have REF_AREA + FREQ=M
    assert filter_dict.get("REF_AREA") == "66"
    assert filter_dict.get("FREQ") == "M"

    if dashboard.slug == "real_estate_deals":
        assert filter_dict.get("SOURCE") == "SI"
        assert filter_dict.get("REALTY") == "2"

    elif dashboard.slug == "residential_real_estate_prices":
        # #902: вторичка hedonic — re-captured live 2026-06-17 (was TYPE=1/NOVELTY=3 → 404).
        assert filter_dict.get("SOURCE") == "SI"
        assert filter_dict.get("REAL_ESTATE_TYPE") == "2"
        assert filter_dict.get("REAL_ESTATE_NOVELTY") == "1"

    elif dashboard.slug == "dinamika-tsen-obyavlenii":
        assert filter_dict.get("SOURCE") == "DK"
        assert filter_dict.get("PRICE_TYPE") == "1"

    else:
        pytest.fail(f"Unknown dashboard slug: {dashboard.slug!r}")


# ---------------------------------------------------------------------------
# (a2) #902 — exact base64 lock for the live-captured /dataset/v1/<slug> routes
# ---------------------------------------------------------------------------

# REF_AREA=643 (РФ) routes captured live 2026-06-17 via Playwright route-interception
# on the sber dashboards (all 200 OK). Each entry: (slug, filter_dict, expected_base64).
# The expected base64 is base64(/dataset/v1/<slug>?filter=<urlencode(json)>&limit=1000&offset=0)
# with json.dumps(separators=(",", ":")) field order = REF_AREA, FREQ, SOURCE, <dims...>.
_CAPTURED_ROUTES_643: list[tuple[str, dict[str, str], str]] = [
    (
        "residential_real_estate_prices",
        {"SOURCE": "SI", "REAL_ESTATE_TYPE": "1", "REAL_ESTATE_NOVELTY": "1"},
        "L2RhdGFzZXQvdjEvcmVzaWRlbnRpYWxfcmVhbF9lc3RhdGVfcHJpY2VzP2ZpbHRlcj0lN0IlMjJSRUZf"
        "QVJFQSUyMiUzQSUyMjY0MyUyMiUyQyUyMkZSRVElMjIlM0ElMjJNJTIyJTJDJTIyU09VUkNFJTIyJTNB"
        "JTIyU0klMjIlMkMlMjJSRUFMX0VTVEFURV9UWVBFJTIyJTNBJTIyMSUyMiUyQyUyMlJFQUxfRVNUQVRF"
        "X05PVkVMVFklMjIlM0ElMjIxJTIyJTdEJmxpbWl0PTEwMDAmb2Zmc2V0PTA=",
    ),
    (
        "residential_real_estate_prices",
        {"SOURCE": "SI", "REAL_ESTATE_TYPE": "2", "REAL_ESTATE_NOVELTY": "1"},
        "L2RhdGFzZXQvdjEvcmVzaWRlbnRpYWxfcmVhbF9lc3RhdGVfcHJpY2VzP2ZpbHRlcj0lN0IlMjJSRUZf"
        "QVJFQSUyMiUzQSUyMjY0MyUyMiUyQyUyMkZSRVElMjIlM0ElMjJNJTIyJTJDJTIyU09VUkNFJTIyJTNB"
        "JTIyU0klMjIlMkMlMjJSRUFMX0VTVEFURV9UWVBFJTIyJTNBJTIyMiUyMiUyQyUyMlJFQUxfRVNUQVRF"
        "X05PVkVMVFklMjIlM0ElMjIxJTIyJTdEJmxpbWl0PTEwMDAmb2Zmc2V0PTA=",
    ),
    (
        "dinamika-tsen-obyavlenii",
        {"SOURCE": "DK", "PRICE_TYPE": "1"},
        "L2RhdGFzZXQvdjEvZGluYW1pa2EtdHNlbi1vYnlhdmxlbmlpP2ZpbHRlcj0lN0IlMjJSRUZfQVJFQSUy"
        "MiUzQSUyMjY0MyUyMiUyQyUyMkZSRVElMjIlM0ElMjJNJTIyJTJDJTIyU09VUkNFJTIyJTNBJTIyREsl"
        "MjIlMkMlMjJQUklDRV9UWVBFJTIyJTNBJTIyMSUyMiU3RCZsaW1pdD0xMDAwJm9mZnNldD0w",
    ),
    (
        "dinamika-tsen-obyavlenii",
        {"SOURCE": "DK", "PRICE_TYPE": "2"},
        "L2RhdGFzZXQvdjEvZGluYW1pa2EtdHNlbi1vYnlhdmxlbmlpP2ZpbHRlcj0lN0IlMjJSRUZfQVJFQSUy"
        "MiUzQSUyMjY0MyUyMiUyQyUyMkZSRVElMjIlM0ElMjJNJTIyJTJDJTIyU09VUkNFJTIyJTNBJTIyREsl"
        "MjIlMkMlMjJQUklDRV9UWVBFJTIyJTNBJTIyMiUyMiU3RCZsaW1pdD0xMDAwJm9mZnNldD0w",
    ),
]


@pytest.mark.parametrize(("slug", "extra_filter", "expected_b64"), _CAPTURED_ROUTES_643)
def test_build_sber_route_matches_captured_b64(
    slug: str, extra_filter: dict[str, str], expected_b64: str
) -> None:
    """#902: build_sber_route must reproduce the live-captured base64 route byte-for-byte.

    Locks the /dataset/v1/<slug> path (NOT /dataset/v1/data/<slug>) and the exact filter
    dimension set + JSON key order. Guards against silent dataset-path regressions.
    """
    encoded = build_sber_route(slug, "643", extra_filter)
    assert encoded == expected_b64

    # And the decoded path must use the bare /dataset/v1/<slug> form (not /data/).
    decoded = base64.b64decode(encoded).decode("utf-8")
    assert decoded.startswith(f"/dataset/v1/{slug}?filter=")
    assert "/dataset/v1/data/" not in decoded


def test_active_dashboards_match_captured_secondary_series() -> None:
    """#902: the two previously-404'ing dashboards now use a captured-200 filter combo.

    residential_real_estate_prices → вторичка hedonic (TYPE=2, NOVELTY=1).
    dinamika-tsen-obyavlenii       → ДомКлик asking (PRICE_TYPE=1).
    Each active SBER_DASHBOARDS entry's base64 must equal one of the captured 643 routes.
    """
    captured_b64 = {b for _, _, b in _CAPTURED_ROUTES_643}
    for slug in ("residential_real_estate_prices", "dinamika-tsen-obyavlenii"):
        d = _DASH[slug]
        encoded = build_sber_route(d.slug, "643", d.extra_filter)
        assert encoded in captured_b64, (
            f"active dashboard {slug!r} route not in captured-live set — "
            f"filter={d.extra_filter}"
        )


# ---------------------------------------------------------------------------
# (b) Response decoder
# ---------------------------------------------------------------------------


def _make_b64_cell(value: str) -> str:
    return "__string__" + base64.b64encode(value.encode("utf-8")).decode("ascii")


def _make_num_cell(value: float) -> str:
    return f"__number__{value}"


def _make_sowa_response(rows: list[list[str]]) -> dict:  # type: ignore[type-arg]
    """Build a minimal SOWA response dict with a fixed field list."""
    fields = [
        "indicator_id",
        "kpi_id",
        "period",
        "value",
        "obs_status",
        "source",
        "ref_area",
        "realty",
        "freq",
        "unit_measure",
        "unit_mult",
    ]
    encoded_fields = [_make_b64_cell(f) for f in fields]
    return {
        "SOWA": {
            "method": "GET",
            "route": "test_route",
            "data": {
                "type": "object",
                "value": [
                    {"key": "fields", "value": encoded_fields},
                    {"key": "data", "value": rows},
                ],
            },
        }
    }


def test_decode_sber_response_basic() -> None:
    """Decode a two-row response and check period + value extraction."""
    row1 = [
        _make_b64_cell("1"),  # indicator_id
        _make_num_cell(643032001),  # kpi_id
        _make_b64_cell("2017-01-30T21:00:00Z"),  # period
        _make_num_cell(48549),  # value
        _make_b64_cell("A"),  # obs_status
        _make_b64_cell("Данные СберИндекса"),  # source
        _make_b64_cell("Россия"),  # ref_area
        _make_b64_cell("Вторичный"),  # realty
        _make_b64_cell("Месяц"),  # freq
        _make_b64_cell("руб"),  # unit_measure
        _make_num_cell(0),  # unit_mult
    ]
    row2 = [
        _make_b64_cell("1"),
        _make_num_cell(643032001),
        _make_b64_cell("2024-06-30T21:00:00Z"),
        _make_num_cell(120000),
        _make_b64_cell("A"),
        _make_b64_cell("Данные СберИндекса"),
        _make_b64_cell("Россия"),
        _make_b64_cell("Вторичный"),
        _make_b64_cell("Месяц"),
        _make_b64_cell("руб"),
        _make_num_cell(0),
    ]
    payload = _make_sowa_response([row1, row2])
    decoded = decode_sber_response(payload)

    assert len(decoded) == 2
    assert decoded[0]["ref_area"] == "Россия"
    assert decoded[0]["value"] == 48549.0
    assert decoded[0]["period"] == "2017-01-30T21:00:00Z"
    assert decoded[1]["value"] == 120000.0


def test_decode_sber_response_period_normalises_to_first_of_month() -> None:
    """_parse_period_month always returns the first day of the month."""
    result = _parse_period_month("2017-01-30T21:00:00Z")
    assert result == date(2017, 1, 1)

    result2 = _parse_period_month("2024-06-30T21:00:00Z")
    assert result2 == date(2024, 6, 1)


def test_decode_sber_response_empty_data() -> None:
    """Empty data section returns empty list."""
    payload = _make_sowa_response([])
    decoded = decode_sber_response(payload)
    assert decoded == []


def test_decode_sber_response_missing_fields_key_raises() -> None:
    """Response without 'fields' key raises ValueError."""
    bad_payload = {
        "SOWA": {
            "method": "GET",
            "route": "x",
            "data": {"type": "object", "value": [{"key": "data", "value": []}]},
        }
    }
    with pytest.raises(ValueError, match="fields"):
        decode_sber_response(bad_payload)


# ---------------------------------------------------------------------------
# (d) Real-fixture decode — exercises actual sberindex.ru response format
# ---------------------------------------------------------------------------


def test_decode_sber_response_real_fixture() -> None:
    with FIXTURE_PATH.open(encoding="utf-8") as f:
        raw = json.load(f)

    rows = decode_sber_response(raw)

    assert len(rows) >= 1, "Expected at least 1 decoded row from real fixture"

    # All rows should have the expected fields present
    first = rows[0]
    assert "ref_area" in first
    assert "realty" in first
    assert "value" in first
    assert "period" in first

    # Values must be typed correctly
    assert isinstance(first["value"], float)
    assert isinstance(first["period"], str) and first["period"]

    # Verify region label decoded correctly
    regions = {r["ref_area"] for r in rows}
    assert (
        "Свердловская область" in regions
    ), f"Expected 'Свердловская область' in ref_area values, got: {regions}"

    # Verify secondary-market segment present
    realty_vals = {r["realty"] for r in rows}
    assert any(
        "тори" in str(v) or "Вторичн" in str(v) for v in realty_vals
    ), f"Expected secondary-market label in realty field, got: {realty_vals}"


# ---------------------------------------------------------------------------
# (e) Period parsing edge-cases
# ---------------------------------------------------------------------------


def test_parse_period_month() -> None:
    result = _parse_period_month("2026-04-29T21:00:00.000Z")
    assert result == date(2026, 4, 1)


def test_parse_period_month_z_suffix() -> None:
    """Original format from live API."""
    result = _parse_period_month("2017-01-30T21:00:00Z")
    assert result == date(2017, 1, 1)


# ---------------------------------------------------------------------------
# (c) Upsert SQL idempotency — fake-db roundtrip
# ---------------------------------------------------------------------------


class _ExecuteCapture:
    """Captures execute() calls and their params; simulate no-op DB."""

    def __init__(self) -> None:
        self.calls: list[tuple[str, dict]] = []  # type: ignore[type-arg]

    def execute(self, stmt: object, params: dict | None = None) -> MagicMock:  # type: ignore[return-value]
        sql_str = str(stmt)
        self.calls.append((sql_str, params or {}))
        return MagicMock(fetchone=lambda: None, fetchall=lambda: [])

    def commit(self) -> None:
        pass

    def rollback(self) -> None:
        pass


@pytest.mark.asyncio
async def test_pull_sber_indices_upsert_on_conflict_idempotent() -> None:
    """pull_sber_indices emits ON CONFLICT ... DO UPDATE upsert SQL (idempotent)."""
    from app.services.sber_index import pull_sber_indices

    fake_db = _ExecuteCapture()
    fake_row = ("Россия", date(2024, 1, 1), "Вторичный", "real_estate_deals", 95000.0)

    # Patch fetch_sber_index to yield one deterministic row per call
    async def _mock_fetch(  # type: ignore[override]
        client: object, *, dashboard: SberDashboard, ref_area: str
    ):
        yield fake_row

    with patch("app.services.sber_index.fetch_sber_index", side_effect=_mock_fetch):
        # Only pull one dashboard × one city to keep assertion simple
        result = await pull_sber_indices(
            fake_db,  # type: ignore[arg-type]
            cities={"643": "Россия"},
            dashboards=[_DASH["real_estate_deals"]],
        )

    assert result["upserted"] == 1
    assert result["errors"] == 0

    # Find the INSERT call
    insert_calls = [sql for sql, _ in fake_db.calls if "INSERT INTO sber_price_index" in sql]
    assert insert_calls, "Expected at least one INSERT INTO sber_price_index call"

    upsert_sql = insert_calls[0]
    # Must contain ON CONFLICT clause with DO UPDATE SET
    assert "ON CONFLICT" in upsert_sql
    assert "DO UPDATE SET" in upsert_sql
    assert "index_value_rub_m2" in upsert_sql
    assert "fetched_at" in upsert_sql

    # Must NOT contain :: type casts (psycopg v3 rule)
    import re

    assert not re.search(
        r":[a-z_]+::[a-z]", upsert_sql
    ), "SQL must not contain ::type casts — use CAST(... AS type) instead"


@pytest.mark.asyncio
async def test_pull_sber_indices_error_per_series_continues() -> None:
    """A failing series increments errors but does not abort other series."""
    from app.services.sber_index import pull_sber_indices

    fake_db = _ExecuteCapture()
    call_count = 0

    async def _mock_fetch_with_error(  # type: ignore[override]
        client: object, *, dashboard: SberDashboard, ref_area: str
    ):
        nonlocal call_count
        call_count += 1
        if dashboard.slug == "real_estate_deals":
            raise Exception("Simulated network failure")
        yield ("Россия", date(2024, 1, 1), "Вторичный", dashboard.slug, 90000.0)

    with patch("app.services.sber_index.fetch_sber_index", side_effect=_mock_fetch_with_error):
        result = await pull_sber_indices(
            fake_db,  # type: ignore[arg-type]
            cities={"643": "Россия"},
            dashboards=[_DASH["real_estate_deals"], _DASH["residential_real_estate_prices"]],
        )

    # First series failed, second succeeded
    assert result["errors"] == 1
    assert result["upserted"] == 1


@pytest.mark.asyncio
async def test_pull_sber_indices_404_logs_warning_with_path_hint(
    caplog: pytest.LogCaptureFixture,
) -> None:
    """#902: a 404 logs a WARNING naming the slug + 'dataset-path invalid?' hint."""
    import logging

    import httpx

    from app.services.sber_index import pull_sber_indices

    fake_db = _ExecuteCapture()

    async def _mock_fetch_404(  # type: ignore[override]
        client: object, *, dashboard: SberDashboard, ref_area: str
    ):
        request = httpx.Request("POST", "https://sberindex.ru/api/sowa")
        response = httpx.Response(404, request=request)
        raise httpx.HTTPStatusError("Not Found", request=request, response=response)
        yield  # pragma: no cover — makes this an async generator

    with patch("app.services.sber_index.fetch_sber_index", side_effect=_mock_fetch_404):
        with caplog.at_level(logging.WARNING, logger="app.services.sber_index"):
            result = await pull_sber_indices(
                fake_db,  # type: ignore[arg-type]
                cities={"643": "Россия"},
                dashboards=[_DASH["residential_real_estate_prices"]],
            )

    assert result["errors"] == 1
    warnings = [r for r in caplog.records if r.levelno == logging.WARNING]
    assert any(
        "residential_real_estate_prices" in r.message and "dataset-path invalid" in r.message
        for r in warnings
    ), f"Expected a 404 dataset-path warning, got: {[r.message for r in warnings]}"


@pytest.mark.asyncio
async def test_pull_sber_indices_asking_benchmark_logged(caplog: pytest.LogCaptureFixture) -> None:
    """After upserting dinamika-tsen-obyavlenii, a benchmark INFO line is logged."""
    import logging

    from app.services.sber_index import pull_sber_indices

    fake_db = _ExecuteCapture()

    async def _mock_fetch(  # type: ignore[override]
        client: object, *, dashboard: SberDashboard, ref_area: str
    ):
        yield ("Россия", date(2024, 6, 1), "Вторичный", dashboard.slug, 115000.0)

    with patch("app.services.sber_index.fetch_sber_index", side_effect=_mock_fetch):
        with caplog.at_level(logging.INFO, logger="app.services.sber_index"):
            await pull_sber_indices(
                fake_db,  # type: ignore[arg-type]
                cities={"643": "Россия"},
                dashboards=[_DASH["dinamika-tsen-obyavlenii"]],
            )

    assert any(
        "benchmark" in record.message for record in caplog.records
    ), "Expected a benchmark log line after upserting dinamika-tsen-obyavlenii"
    assert any(
        "115000" in record.message for record in caplog.records
    ), "Benchmark log should include the latest asking index value"