space-bundle/src/optos/generators.py

"""Synthetic instance generators for all registered problem types."""

from __future__ import annotations

import hashlib
import random
from typing import Any

from optos.constants import PROBLEM_TYPES, SIZE_PRESETS
from optos.models import InstanceFeatures, ProblemInstance


def _rng(seed: int) -> random.Random:
    return random.Random(seed)


def _instance_id(problem_type: str, size: str, seed: int) -> str:
    raw = f"{problem_type}_{size}_{seed}"
    return hashlib.md5(raw.encode()).hexdigest()[:12]


def _scale_n(base: int, size: str) -> int:
    return max(2, int(base * SIZE_PRESETS[size]["scale"]))


def generate_scheduling(rng: random.Random, size: str) -> dict[str, Any]:
    n_jobs = _scale_n(6, size)
    n_machines = _scale_n(4, size)
    processing_times = [
        [rng.randint(2, 12) for _ in range(n_machines)]
        for _ in range(n_jobs)
    ]
    machine_order = [
        list(range(n_machines))
        for _ in range(n_jobs)
    ]
    return {
        "n_jobs": n_jobs,
        "n_machines": n_machines,
        "processing_times": processing_times,
        "machine_order": machine_order,
    }


def generate_routing(rng: random.Random, size: str) -> dict[str, Any]:
    n_customers = _scale_n(12, size)
    n_vehicles = max(2, n_customers // 4)
    depot = (50.0, 50.0)
    customers = [
        (rng.uniform(0, 100), rng.uniform(0, 100))
        for _ in range(n_customers)
    ]
    demands = [rng.randint(1, 10) for _ in range(n_customers)]
    vehicle_capacity = max(demands) * 3
    return {
        "n_customers": n_customers,
        "n_vehicles": n_vehicles,
        "depot": depot,
        "customers": customers,
        "demands": demands,
        "vehicle_capacity": vehicle_capacity,
    }


def generate_assignment(rng: random.Random, size: str) -> dict[str, Any]:
    n = _scale_n(8, size)
    cost_matrix = [
        [rng.randint(1, 50) for _ in range(n)]
        for _ in range(n)
    ]
    return {"n_agents": n, "cost_matrix": cost_matrix}


def generate_inventory(rng: random.Random, size: str) -> dict[str, Any]:
    n_items = _scale_n(10, size)
    horizon = _scale_n(14, size)
    demand = [
        [rng.randint(5, 30) for _ in range(horizon)]
        for _ in range(n_items)
    ]
    holding_cost = [rng.uniform(0.5, 2.0) for _ in range(n_items)]
    stockout_cost = [rng.uniform(5.0, 20.0) for _ in range(n_items)]
    order_cost = [rng.uniform(10.0, 50.0) for _ in range(n_items)]
    initial_stock = [rng.randint(10, 40) for _ in range(n_items)]
    return {
        "n_items": n_items,
        "horizon": horizon,
        "demand": demand,
        "holding_cost": holding_cost,
        "stockout_cost": stockout_cost,
        "order_cost": order_cost,
        "initial_stock": initial_stock,
        "max_order": [max(d) * 2 for d in demand],
    }


def generate_facility_location(rng: random.Random, size: str) -> dict[str, Any]:
    n_facilities = _scale_n(6, size)
    n_customers = _scale_n(15, size)
    fixed_costs = [rng.randint(100, 500) for _ in range(n_facilities)]
    transport_costs = [
        [rng.randint(1, 30) for _ in range(n_facilities)]
        for _ in range(n_customers)
    ]
    return {
        "n_facilities": n_facilities,
        "n_customers": n_customers,
        "fixed_costs": fixed_costs,
        "transport_costs": transport_costs,
    }


def generate_packing(rng: random.Random, size: str) -> dict[str, Any]:
    n_items = _scale_n(20, size)
    bin_capacity = 100
    item_sizes = [rng.randint(10, 45) for _ in range(n_items)]
    return {
        "n_items": n_items,
        "bin_capacity": bin_capacity,
        "item_sizes": item_sizes,
    }


GENERATORS = {
    "scheduling": generate_scheduling,
    "routing": generate_routing,
    "assignment": generate_assignment,
    "inventory": generate_inventory,
    "facility_location": generate_facility_location,
    "packing": generate_packing,
}


def _estimate_features(problem_type: str, data: dict[str, Any]) -> InstanceFeatures:
    if problem_type == "scheduling":
        n_vars = data["n_jobs"] * data["n_machines"] * 2
        n_cons = data["n_jobs"] * (data["n_machines"] - 1) + data["n_machines"]
    elif problem_type == "routing":
        n_vars = data["n_customers"] * data["n_vehicles"]
        n_cons = data["n_customers"] + data["n_vehicles"]
    elif problem_type == "assignment":
        n = data["n_agents"]
        n_vars = n * n
        n_cons = 2 * n
    elif problem_type == "inventory":
        n_vars = data["n_items"] * data["horizon"]
        n_cons = data["n_items"] * data["horizon"]
    elif problem_type == "facility_location":
        nf, nc = data["n_facilities"], data["n_customers"]
        n_vars = nf + nf * nc
        n_cons = nc + nf * nc
    elif problem_type == "packing":
        n_vars = data["n_items"] * data["n_items"]
        n_cons = data["n_items"] + data["n_items"]
    else:
        n_vars, n_cons = 10, 10

    return InstanceFeatures(
        n_variables=n_vars,
        n_constraints=n_cons,
        density=round(min(1.0, n_cons / max(n_vars, 1)), 3),
        pct_integer=0.85,
        constraint_tightness=round(random.Random(0).uniform(0.4, 0.8), 3),
    )


def generate_instance(
    problem_type: str,
    size: str = "medium",
    seed: int = 42,
    constraints: dict[str, Any] | None = None,
    objectives: dict[str, Any] | None = None,
) -> ProblemInstance:
    if problem_type not in PROBLEM_TYPES:
        raise ValueError(f"Unknown problem type: {problem_type}")

    rng = _rng(seed)
    data = GENERATORS[problem_type](rng, size)
    features = _estimate_features(problem_type, data)
    meta = PROBLEM_TYPES[problem_type]

    label = f"{meta['label']} · {SIZE_PRESETS.get(size, {}).get('label', size)} · seed={seed}"

    return ProblemInstance(
        problem_type=problem_type,
        instance_id=_instance_id(problem_type, size, seed),
        label=label,
        size=size,
        seed=seed,
        data=data,
        features=features,
        constraints=constraints or {},
        objectives=objectives or {"primary": meta.get("objective", "minimize")},
    )