"""Metadata for immutable, recognized Sudoku dataset artifacts."""

from __future__ import annotations

LEGACY_DATA_PROTOCOL = "sudoku_legacy_mnist_pool_v1"
LEGACY_DATA_SHA256 = "8d0070b00ffc75763ceb8c2b29006823c80b70f50b6ca0018310d0fdc7b3828e"


def data_protocol_metadata(sha256: str, requested_id: str | None = None) -> dict:
    """Identify by content; never assign legacy or corrected claims to another file."""
    known = sha256.lower() == LEGACY_DATA_SHA256
    actual_id = LEGACY_DATA_PROTOCOL if known else None
    if requested_id is not None and requested_id != actual_id:
        raise ValueError("Data-protocol identifier does not match the dataset SHA256")
    if not known:
        return {
            "id": None,
            "status": "unregistered",
            "source_image_split_disjointness": "not_established",
            "evidence_label": "Unregistered dataset protocol; no source-image separation claim.",
        }
    return {
        "id": LEGACY_DATA_PROTOCOL,
        "status": "legacy",
        "dataset_sha256": LEGACY_DATA_SHA256,
        "splits": {"train": 9000, "val": 1000, "test": 1000},
        "generation_seed": 42,
        "split_seed_offsets": {"train": 0, "val": 100000, "test": 200000},
        "initial_clues": 55,
        "max_successful_removals": 10,
        "observed_clues_per_board": 45,
        "board_distribution": "transformations_of_one_base_solution",
        "source_image_pool": "merged_original_mnist_train_and_test",
        "source_image_split_disjointness": "not_established",
        "evidence_label": "Legacy Sudoku dataset; source-image split disjointness not established.",
    }