"""Quality-gate tests locking the 5 P0 audit fixes (issue #781).

One test per merged defect -- anti-regression guards so that future refactors
cannot silently re-introduce the same bugs.

PRs verified merged into main before writing any assert:
  #877 / commit d1034bb -- anchor safe-guard (#755)
  #761 / commit 79a7900 -- stable dedup_hash (#753)
  #784 / commit ec84637 -- expected_sold guard (#773)
  #779 / commit 489666c -- avito block-page HTTP-200 (#754)
  #743 / commit 798fb12 -- insufficient_data flag (#740)
"""

from __future__ import annotations

import os
from datetime import UTC, datetime
from pathlib import Path
from typing import Any
from unittest.mock import AsyncMock, MagicMock, patch

import anyio
import pytest

# Settings requires DATABASE_URL at init time.
os.environ.setdefault("DATABASE_URL", "postgresql+psycopg://test:test@localhost/test_db")


# ---------------------------------------------------------------------------
# Helpers shared across tests
# ---------------------------------------------------------------------------


def _make_listing_qa(*, price_per_m2: float, area_m2: float = 60.0) -> dict[str, Any]:
    return {
        "source": "cian",
        "source_url": "https://cian.ru/offer/1",
        "address": "EKB, ul. Khokryakova, 48",
        "lat": 56.830,
        "lon": 60.592,
        "rooms": 2,
        "area_m2": area_m2,
        "floor": 5,
        "total_floors": 14,
        "price_rub": int(price_per_m2 * area_m2),
        "price_per_m2": price_per_m2,
        "listing_date": datetime(2026, 5, 1),
        "days_on_market": 10,
        "photo_urls": [],
        "scraped_at": datetime(2026, 5, 20, tzinfo=UTC),
        "distance_m": 0.0,
        "relevance_score": 0.0,
        "listing_segment": "vtorichka",
    }


_RADIUS_ANALOGS_QA: list[dict[str, Any]] = [
    _make_listing_qa(price_per_m2=200_000.0),
    _make_listing_qa(price_per_m2=210_000.0),
    _make_listing_qa(price_per_m2=220_000.0),
]

# 2 comps -- below min_comps=4 threshold introduced by #755.
_TWO_COMPS: list[dict[str, Any]] = [
    {"price_per_m2": 399_478, "area_m2": 153.2, "rooms": 3},
    {"price_per_m2": 472_298, "area_m2": 110.1, "rooms": 3},
]

# 4 comps satisfying min_comps=4 (#755).
_SB_COMPS_QG: list[dict[str, Any]] = [
    {"price_per_m2": 399_478, "area_m2": 153.2, "rooms": 3},
    {"price_per_m2": 472_298, "area_m2": 110.1, "rooms": 3},
    {"price_per_m2": 683_995, "area_m2": 146.2, "rooms": 4},
    {"price_per_m2": 510_000, "area_m2": 125.0, "rooms": 3},
]


def _run_qa_estimate(
    *,
    anchor_comps: list[dict[str, Any]],
    anchor_tier: str | None,
    ratio_tuple: tuple[float | None, str | None] = (0.92, "per_rooms"),
    radius_analogs: list[dict[str, Any]] | None = None,
) -> Any:
    from app.core.config import settings
    from app.schemas.trade_in import TradeInEstimateInput
    from app.services.estimator import estimate_quality
    from app.services.geocoder import GeocodeResult

    db = MagicMock()
    payload = TradeInEstimateInput(
        address="Ekaterinburg, ul. Khokryakova, 48",
        area_m2=60.0,
        rooms=2,
        floor=5,
        total_floors=14,
    )
    geo = GeocodeResult(
        lat=56.830,
        lon=60.592,
        full_address="Sverdlovskaya obl., Ekaterinburg, ul. Khokryakova, 48",
        provider="nominatim",
    )
    radius = list(_RADIUS_ANALOGS_QA if radius_analogs is None else radius_analogs)

    async def _run() -> Any:
        with (
            patch.object(settings, "estimate_same_building_anchor_enabled", True),
            patch("app.services.estimator.geocode", new=AsyncMock(return_value=geo)),
            patch("app.services.estimator.dadata_clean_address", new=AsyncMock(return_value=None)),
            patch("app.services.estimator.match_house_readonly", return_value=None),
            patch("app.services.estimator.get_house_metadata", new=AsyncMock(return_value=None)),
            patch("app.services.estimator._fetch_analogs", return_value=(radius, False, "W")),
            patch("app.services.estimator._fetch_deals", return_value=[]),
            patch("app.services.estimator._fetch_dkp_corridor", return_value=None),
            patch(
                "app.services.estimator._fetch_anchor_comps",
                return_value=(list(anchor_comps), anchor_tier),
            ),
            patch("app.services.estimator._fetch_house_imv_anchor", return_value=None),
            patch(
                "app.services.estimator._get_or_fetch_imv_cached",
                new=AsyncMock(return_value=None),
            ),
            patch(
                "app.services.estimator._get_or_fetch_yandex_valuation_cached",
                new=AsyncMock(return_value=None),
            ),
            patch(
                "app.services.estimator.estimate_via_cian_valuation",
                new=AsyncMock(return_value=None),
            ),
            patch("app.services.estimator._get_asking_sold_ratio", return_value=ratio_tuple),
        ):
            return await estimate_quality(payload, db)

    return anyio.run(_run)


# ---------------------------------------------------------------------------
# Test 1 -- Anchor confidence n=2 (#755)
#
# Same-building anchor with n_comps < min_comps threshold (2 < 4) must NOT
# fire at all in the full estimate path: headline stays at the radius median
# (210 000 rub/m2), and confidence is never "high" from an under-populated
# anchor.  The DoD of #755 is: min_comps 2->4; confidence cap n<5->medium;
# MAD-clip.
# ---------------------------------------------------------------------------


def test_755_anchor_n2_does_not_fire_headline_stays_radius() -> None:
    """#755 quality-gate: 2 same-building comps < min_comps=4 -> anchor does NOT fire.

    Headline stays at the radius median (210 000 rub/m2) and is NOT inflated by
    an under-populated uplift. Before #755, min_comps was 2, so this anchor
    would have fired and replaced the headline with the premium anchor value.
    Refs: PR #877 / commit d1034bb.
    """
    est = _run_qa_estimate(anchor_comps=_TWO_COMPS, anchor_tier="A")
    # Anchor did not fire -> headline is the radius-path median (210 000).
    assert est.median_price_per_m2 == 210_000, (
        f"Expected radius median 210_000, got {est.median_price_per_m2} -- "
        "anchor with n=2 comps must NOT fire (min_comps=4 post-#755)"
    )
    # Confidence from 3 radius analogs must not be "high" (n_analogs=3 < threshold).
    assert (
        est.confidence != "high"
    ), f"Confidence should not be 'high' with 3 radius analogs, got {est.confidence!r}"


def test_755_anchor_n2_pure_unit_confidence_never_high() -> None:
    """#755 quality-gate (pure unit): _compute_same_building_anchor with n=2 comps,
    min_comps=1 override, verifies confidence is capped below 'high' (n < 5).

    Even if a caller bypasses the min_comps threshold, the confidence for n<5
    must remain 'medium' or 'low' -- the cap introduced by #755.
    Refs: PR #877 / commit d1034bb.
    """
    from app.services.estimator import _compute_same_building_anchor

    comps = [
        {"price_per_m2": 399_478, "area_m2": 153.2, "rooms": 3},
        {"price_per_m2": 472_298, "area_m2": 110.1, "rooms": 3},
    ]
    res = _compute_same_building_anchor(
        comps,
        area_target=110.0,
        rooms_target=3,
        tier="A",
        sigma=0.18,
        rooms_boost=1.6,
        min_comps=1,  # bypass threshold to test confidence cap independently
    )
    assert res is not None
    assert res["n"] == 2
    # n < 5 -> confidence cap: must NOT be 'high'.
    assert res["confidence"] != "high", (
        f"confidence must not be 'high' for n=2 anchor comps (post-#755 cap), "
        f"got {res['confidence']!r}"
    )
    assert res["confidence"] in ("medium", "low")


# ---------------------------------------------------------------------------
# Test 2 -- dedup_hash stable across price change and ?context= query (#753)
#
# Before #753 the hash included price_rub, so a reprice of the same Avito
# listing produced a duplicate active listing.  After #753 the hash is based
# on source_id (if present) or source_url stripped of query-string -- so
# same-listing reprices and re-scrapes with different ?context= tokens give
# the SAME hash.
# ---------------------------------------------------------------------------


def _make_scraped_lot(
    *,
    source: str = "avito",
    source_url: str,
    source_id: str | None = None,
    price_rub: int = 5_000_000,
) -> Any:
    from scraper_kit.base import ScrapedLot

    return ScrapedLot(
        source=source,
        source_url=source_url,
        source_id=source_id,
        address="EKB, ul. Tkachei, 13",
        rooms=2,
        area_m2=52.0,
        price_rub=price_rub,
    )


def test_753_dedup_hash_stable_across_reprice_with_source_id() -> None:
    """#753 quality-gate: same Avito item (same source_id) with different price_rub
    produces identical compute_dedup_hash.

    Before #753, price_rub was part of the hash key, so a seller repricing
    from 5M to 4.9M created a second active listing -- poisoning
    asking_to_sold_ratio and inventory counts.
    Refs: PR #761 / commit 79a7900.
    """
    url = "https://www.avito.ru/ekaterinburg/kvartiry/2k_52m_13_1234567890"
    sid = "1234567890"

    lot_original = _make_scraped_lot(source_url=url, source_id=sid, price_rub=5_000_000)
    lot_repriced = _make_scraped_lot(source_url=url, source_id=sid, price_rub=4_900_000)

    assert (
        lot_original.compute_dedup_hash() == lot_repriced.compute_dedup_hash()
    ), "dedup_hash must be stable across reprice when source_id is present (#753)"


def test_753_dedup_hash_stable_across_context_query_no_source_id() -> None:
    """#753 quality-gate: same Avito URL with different ?context= tokens -> same hash.

    Avito appends a volatile ?context=... tracking token to every listing URL.
    When source_id is absent, dedup_hash falls back to source_url stripped of
    the query-string. Two re-scrapes with different tokens must yield the same
    hash (no phantom duplicates).
    Refs: PR #761 / commit 79a7900.
    """
    base_url = "https://www.avito.ru/ekaterinburg/kvartiry/2k_52m_13_1234567890"
    lot_v1 = _make_scraped_lot(
        source_url=f"{base_url}?context=H4sIAAAAA",
        source_id=None,
        price_rub=5_000_000,
    )
    lot_v2 = _make_scraped_lot(
        source_url=f"{base_url}?context=X9zKBQAA",
        source_id=None,
        price_rub=5_000_000,
    )
    assert (
        lot_v1.compute_dedup_hash() == lot_v2.compute_dedup_hash()
    ), "dedup_hash must strip ?context= query and match for same listing URL (#753)"


def test_753_dedup_hash_differs_for_distinct_listings() -> None:
    """#753 sanity: two genuinely different Avito listings produce different hashes.

    Guards against the hash function becoming trivially constant.
    Refs: PR #761 / commit 79a7900.
    """
    lot_a = _make_scraped_lot(
        source_url="https://www.avito.ru/ekaterinburg/kvartiry/aaa",
        source_id="111",
    )
    lot_b = _make_scraped_lot(
        source_url="https://www.avito.ru/ekaterinburg/kvartiry/bbb",
        source_id="222",
    )
    assert lot_a.compute_dedup_hash() != lot_b.compute_dedup_hash()


def test_753_dedup_hash_source_id_takes_priority_over_url() -> None:
    """#753: when source_id is set, URL is ignored for the hash key.

    Two lots with the same source_id but different URLs produce the same
    hash -- only source_id matters.
    Refs: PR #761 / commit 79a7900.
    """
    lot_canonical = _make_scraped_lot(
        source_url="https://www.avito.ru/ekaterinburg/kvartiry/listing_9999",
        source_id="9999",
    )
    lot_redirect = _make_scraped_lot(
        source_url="https://www.avito.ru/ru/offer/9999?some=other",
        source_id="9999",
    )
    assert (
        lot_canonical.compute_dedup_hash() == lot_redirect.compute_dedup_hash()
    ), "source_id takes priority: same source_id -> same hash regardless of URL (#753)"


# ---------------------------------------------------------------------------
# Test 3 -- expected_sold > 0 on anchor-only path with no radius comps (#773)
#
# Before #773, the expected_sold guard was `and listings_clean` (truthy check
# on the radius listing list).  When radius_analogs=[] the list is empty ->
# falsy -> expected_sold was skipped even when the anchor gave median_price>0.
# After #773 the guard is `and median_price > 0`, so anchor-only estimates
# with a valid ratio also populate expected_sold.
# ---------------------------------------------------------------------------


def test_773_expected_sold_positive_on_anchor_only_path(
    monkeypatch: pytest.MonkeyPatch,
) -> None:
    """#773 quality-gate: anchor-only path (radius_analogs=[]) with ratio present
    -> expected_sold_price_rub > 0 (not NULL).

    Before #773 fix, the guard was `and listings_clean` (empty list = falsy),
    so expected_sold was skipped on the anchor-only path even though the anchor
    produced a valid median_price. The fix changed the guard to `and median_price > 0`.
    Refs: PR #784 / commit ec84637.
    """
    # #2002: asserts expected_sold == headline × ratio; hold the orthogonal hedonic
    # year+area correction OFF (OFF ⇒ exact legacy expected_sold).
    from app.core.config import settings as _settings

    monkeypatch.setattr(_settings, "estimate_hedonic_correction_enabled", False)
    est = _run_qa_estimate(
        anchor_comps=_SB_COMPS_QG,
        anchor_tier="A",
        radius_analogs=[],  # no radius comps -- anchor-only path
        ratio_tuple=(0.92, "per_rooms"),
    )
    # Anchor gave a valid headline.
    assert est.median_price_rub > 0, "anchor must produce non-zero headline"
    # expected_sold must also be computed (not skipped by empty listings_clean).
    assert (
        est.expected_sold_price_rub is not None
    ), "expected_sold_price_rub must not be None on anchor-only path when ratio present (#773)"
    assert (
        est.expected_sold_price_rub > 0
    ), f"expected_sold_price_rub must be > 0, got {est.expected_sold_price_rub} (#773)"
    # Consistency: sold = asking * ratio.
    assert est.expected_sold_price_rub == round(est.median_price_rub * 0.92)


def test_773_expected_sold_null_when_no_ratio_anchor_only() -> None:
    """#773 control: no ratio (None) -> expected_sold stays None even with anchor median>0.

    The fix must not fabricate an expected_sold when no ratio row exists.
    Refs: PR #784 / commit ec84637.
    """
    est = _run_qa_estimate(
        anchor_comps=_SB_COMPS_QG,
        anchor_tier="A",
        radius_analogs=[],
        ratio_tuple=(None, None),
    )
    assert est.median_price_rub > 0  # anchor headline is valid
    assert (
        est.expected_sold_price_rub is None
    ), "expected_sold must be None when ratio is None, even on anchor-only path (#773)"


# ---------------------------------------------------------------------------
# Test 4 -- Avito block-page HTTP-200 behavior (#754)
#
# Before #754, an HTTP-200 response with 0 parsed cards was treated as a
# silent end-of-pagination.  After #754:
#   - page=1 with 0 cards -> raises AvitoContentBlockedError (no silent empty)
#   - page>1 with 0 cards -> graceful break (end of pagination), no exception
# ---------------------------------------------------------------------------

_FIXTURES = Path(__file__).parent / "fixtures"
_BLOCKPAGE_HTML = (_FIXTURES / "avito_blockpage_sample.html").read_text(encoding="utf-8")
_SERP_HTML = (_FIXTURES / "avito_serp_sample.html").read_text(encoding="utf-8")


@pytest.mark.asyncio
async def test_754_block_page_http200_page1_raises() -> None:
    """#754 quality-gate: HTTP-200 with 0 parsed cards on page=1 -> AvitoContentBlockedError.

    Before #754 this scenario caused a silent empty return ([]) masking a content block.
    Refs: PR #779 / commit 489666c.
    """
    from scraper_kit.avito_exceptions import AvitoContentBlockedError
    from scraper_kit.providers.avito.serp import AvitoScraper

    from app.services.scraper_adapters import RealScraperConfig

    scraper = AvitoScraper(RealScraperConfig())
    with patch.object(scraper, "_fetch_serp_html", new=AsyncMock(return_value=_BLOCKPAGE_HTML)):
        with pytest.raises(AvitoContentBlockedError):
            await scraper.fetch_around(56.838, 60.605, radius_m=1000, pages=1)


@pytest.mark.asyncio
async def test_754_block_page_http200_page_gt1_graceful_return() -> None:
    """#754 quality-gate: HTTP-200 with 0 parsed cards on page>1 -> graceful return.

    page=1 succeeds (real SERP HTML), page=2 returns block-page (0 cards ->
    end-of-pagination break). The call must NOT raise.
    Refs: PR #779 / commit 489666c.
    """
    from scraper_kit.providers.avito.serp import AvitoScraper

    from app.services.scraper_adapters import RealScraperConfig

    scraper = AvitoScraper(RealScraperConfig())

    async def _serp_html_side_effect(url: str, page: int) -> str:
        return _SERP_HTML if page == 1 else _BLOCKPAGE_HTML

    with patch.object(
        scraper, "_fetch_serp_html", new=AsyncMock(side_effect=_serp_html_side_effect)
    ):
        with patch.object(scraper, "sleep_between_requests", new=AsyncMock(return_value=None)):
            result = await scraper.fetch_around(56.838, 60.605, radius_m=1000, pages=2)

    # The important invariant: no exception raised, result is a list.
    assert isinstance(
        result, list
    ), "fetch_around must return a list, not raise, when page>1 returns 0 cards (#754)"


# ---------------------------------------------------------------------------
# Test 5 -- median=0 -> insufficient_data=True (#740 / #697)
#
# When the estimate has no analogs (radius empty, no anchor, no IMV) the
# median is 0.  AggregatedEstimate.insufficient_data must be True so the
# frontend shows the empty-state UI instead of "0 rub".
# ---------------------------------------------------------------------------


def test_740_median_zero_gives_insufficient_data_true() -> None:
    """#740 quality-gate: estimate with no analogs -> insufficient_data=True, median=0.

    Before #740/#697, the flag did not exist; the frontend showed "0,00 mln rub".
    The fix added AggregatedEstimate.insufficient_data as a computed property
    that is True when median_price_rub <= 0.
    Refs: PR #743 / commit 798fb12.
    """
    from uuid import uuid4

    from app.schemas.trade_in import AggregatedEstimate

    est = AggregatedEstimate(
        estimate_id=uuid4(),
        median_price_rub=0,
        range_low_rub=0,
        range_high_rub=0,
        median_price_per_m2=0,
        confidence="low",
        n_analogs=0,
        period_months=12,
        analogs=[],
        actual_deals=[],
        expires_at=datetime(2026, 6, 1, tzinfo=UTC),
    )
    assert (
        est.insufficient_data is True
    ), "insufficient_data must be True when median_price_rub=0 (#740)"
    assert est.median_price_rub == 0


def test_740_positive_median_gives_insufficient_data_false() -> None:
    """#740 control: estimate with analogs (median>0) -> insufficient_data=False.

    Guards against the flag becoming trivially True.
    Refs: PR #743 / commit 798fb12.
    """
    from uuid import uuid4

    from app.schemas.trade_in import AggregatedEstimate

    est = AggregatedEstimate(
        estimate_id=uuid4(),
        median_price_rub=6_900_000,
        range_low_rub=6_000_000,
        range_high_rub=7_800_000,
        median_price_per_m2=132_000,
        confidence="medium",
        n_analogs=8,
        period_months=12,
        analogs=[],
        actual_deals=[],
        expires_at=datetime(2026, 6, 1, tzinfo=UTC),
    )
    assert (
        est.insufficient_data is False
    ), "insufficient_data must be False when median_price_rub>0 (#740)"


def test_740_insufficient_data_serialized_in_model_dump() -> None:
    """#740: insufficient_data appears in model_dump() output (API response shape).

    The frontend reads this field via JSON; it must survive Pydantic serialization.
    Refs: PR #743 / commit 798fb12.
    """
    from uuid import uuid4

    from app.schemas.trade_in import AggregatedEstimate

    zero_est = AggregatedEstimate(
        estimate_id=uuid4(),
        median_price_rub=0,
        range_low_rub=0,
        range_high_rub=0,
        median_price_per_m2=0,
        confidence="low",
        n_analogs=0,
        period_months=12,
        analogs=[],
        actual_deals=[],
        expires_at=datetime(2026, 6, 1, tzinfo=UTC),
    )
    dump = zero_est.model_dump()
    assert "insufficient_data" in dump, "insufficient_data must be present in model_dump()"
    assert dump["insufficient_data"] is True