from __future__ import annotations

import math

import numpy as np
import pytest

from services.financial import CashFlow, net_present_value
from services.simulation import (
    CashFlowSpec,
    DistributionConfigError,
    DistributionSource,
    DistributionSpec,
    DistributionType,
    SimulationConfig,
    SimulationMetric,
    run_monte_carlo,
)


def test_run_monte_carlo_deterministic_matches_financial_helpers() -> None:
    base_flows = [
        CashFlow(amount=-1000.0, period_index=0),
        CashFlow(amount=600.0, period_index=1),
        CashFlow(amount=600.0, period_index=2),
    ]
    specs = [CashFlowSpec(cash_flow=flow) for flow in base_flows]
    config = SimulationConfig(
        iterations=10,
        discount_rate=0.1,
        percentiles=(50,),
        seed=123,
    )

    result = run_monte_carlo(specs, config)
    summary = result.summaries[SimulationMetric.NPV]
    expected = net_present_value(0.1, base_flows)

    assert summary.sample_size == config.iterations
    assert summary.failed_runs == 0
    assert summary.mean == pytest.approx(expected, rel=1e-6)
    assert summary.std_dev == 0.0
    assert summary.percentiles[50] == pytest.approx(expected, rel=1e-6)


def test_run_monte_carlo_normal_distribution_uses_seed_for_reproducibility() -> None:
    base_flows = [
        CashFlow(amount=-100.0, period_index=0),
        CashFlow(amount=0.0, period_index=1),
        CashFlow(amount=0.0, period_index=2),
    ]
    revenue_flow = CashFlowSpec(
        cash_flow=CashFlow(amount=120.0, period_index=1),
        distribution=DistributionSpec(
            type=DistributionType.NORMAL,
            parameters={"mean": 120.0, "std_dev": 10.0},
        ),
    )
    specs = [CashFlowSpec(cash_flow=base_flows[0]), revenue_flow]
    config = SimulationConfig(
        iterations=1000,
        discount_rate=0.0,
        percentiles=(5.0, 50.0, 95.0),
        seed=42,
    )

    result = run_monte_carlo(specs, config)
    summary = result.summaries[SimulationMetric.NPV]

    assert summary.sample_size == config.iterations
    assert summary.failed_runs == 0
    # With zero discount rate the expected mean NPV equals mean sampled value minus investment.
    assert summary.mean == pytest.approx(20.0, abs=1.0)
    assert summary.std_dev == pytest.approx(10.0, abs=1.0)
    assert summary.percentiles[50.0] == pytest.approx(summary.mean, abs=1.0)


def test_run_monte_carlo_supports_scenario_field_source() -> None:
    base_flow = CashFlow(amount=0.0, period_index=1)
    spec = CashFlowSpec(
        cash_flow=base_flow,
        distribution=DistributionSpec(
            type=DistributionType.NORMAL,
            parameters={"std_dev": 0.0},
            source=DistributionSource.SCENARIO_FIELD,
            source_key="salvage_mean",
        ),
    )
    config = SimulationConfig(iterations=1, discount_rate=0.0, seed=7)

    result = run_monte_carlo(
        [CashFlowSpec(cash_flow=CashFlow(
            amount=-100.0, period_index=0)), spec],
        config,
        scenario_context={"salvage_mean": 150.0},
    )

    summary = result.summaries[SimulationMetric.NPV]
    assert summary.sample_size == 1
    assert summary.mean == pytest.approx(50.0)


def test_run_monte_carlo_records_failed_metrics_when_not_defined() -> None:
    base_flows = [CashFlow(amount=100.0, period_index=0)]
    specs = [CashFlowSpec(cash_flow=flow) for flow in base_flows]
    config = SimulationConfig(
        iterations=5,
        discount_rate=0.1,
        metrics=(SimulationMetric.IRR,),
        seed=5,
    )

    result = run_monte_carlo(specs, config)
    summary = result.summaries[SimulationMetric.IRR]

    assert summary.sample_size == 0
    assert summary.failed_runs == config.iterations
    assert math.isnan(summary.mean)


def test_run_monte_carlo_distribution_missing_context_raises() -> None:
    spec = DistributionSpec(
        type=DistributionType.NORMAL,
        parameters={"std_dev": 1.0},
        source=DistributionSource.SCENARIO_FIELD,
        source_key="unknown",
    )
    cash_flow_spec = CashFlowSpec(
        cash_flow=CashFlow(amount=0.0, period_index=0),
        distribution=spec,
    )
    config = SimulationConfig(iterations=1, discount_rate=0.0)

    with pytest.raises(DistributionConfigError):
        run_monte_carlo([cash_flow_spec], config, scenario_context={})


def test_run_monte_carlo_can_return_samples() -> None:
    base_flow = CashFlow(amount=50.0, period_index=1)
    specs = [
        CashFlowSpec(cash_flow=CashFlow(amount=-40.0, period_index=0)),
        CashFlowSpec(cash_flow=base_flow),
    ]
    config = SimulationConfig(
        iterations=3,
        discount_rate=0.0,
        metrics=(SimulationMetric.NPV,),
        return_samples=True,
        seed=11,
    )

    result = run_monte_carlo(specs, config)

    assert result.samples is not None
    assert SimulationMetric.NPV in result.samples
    samples = result.samples[SimulationMetric.NPV]
    assert isinstance(samples, np.ndarray)
    assert samples.shape == (config.iterations,)