"""Very small solver prototype.

Provides a deterministic greedy warm-start optimizer for harvesting losses.
For Phase-0 we avoid external MILP dependencies and implement a clear,
deterministic algorithm suitable for unit testing. Future versions will
replace or augment this with MILP (e.g., using pulp or ortools).
"""
from datetime import date
from typing import List, Tuple
from .ir import TaxLot, HarvestAction, PlanDelta, AuditLog


def _lot_loss(lot: TaxLot) -> float:
    """Return (market_value - basis) * quantity. Positive=gain, negative=loss.

    Note: if market_value is None treat as 0 to keep deterministic behavior.
    """
    mv = lot.market_value if lot.market_value is not None else 0.0
    return (mv - lot.basis) * lot.quantity


def optimize_harvest(lots: List[TaxLot], target_loss: float) -> Tuple[PlanDelta, AuditLog]:
    """Greedy harvest: choose lots with largest losses first until reaching
    target_loss (in absolute terms). target_loss should be positive number
    indicating total loss to realize (e.g., harvest $10,000 of losses).

    Returns a PlanDelta and AuditLog.
    """
    # Work only with lots that currently show a loss
    losses = []
    for l in lots:
        loss = -_lot_loss(l) if _lot_loss(l) < 0 else 0.0
        if loss > 0:
            losses.append((loss, l))

    # Sort descending by loss magnitude (largest losses first)
    losses.sort(key=lambda x: x[0], reverse=True)

    accumulated = 0.0
    actions = []
    audit = AuditLog()

    for loss_amt, lot in losses:
        if accumulated >= target_loss:
            break
        # sell entire lot for simplicity in this prototype
        sell_qty = lot.quantity
        expected_gain = _lot_loss(lot)
        action = HarvestAction(lot_id=lot.id, sell_qty=sell_qty, date=date.today(), expected_gain=expected_gain)
        actions.append(action)
        accumulated += loss_amt
        audit.entries.append({
            "lot_id": lot.id,
            "realized_loss": loss_amt,
            "basis": lot.basis,
            "market_value": lot.market_value,
        })

    metadata = {"target_loss": target_loss, "realized_loss": accumulated}
    plan = PlanDelta(actions=actions, metadata=metadata)
    return plan, audit