data_studio/operations.py

from __future__ import annotations

import hashlib
from datetime import datetime, timezone
from typing import Any

from data_studio.identity import canonical_json


def backup_manifest(environment: str, collections: dict[str, str], hf_artifacts: dict[str, str]) -> dict[str, Any]:
    payload = {
        "schema_version": "1.0",
        "environment": environment,
        "collections": collections,
        "hf_artifacts": hf_artifacts,
    }
    return {
        **payload,
        "manifest_checksum": hashlib.sha256(canonical_json(payload).encode("utf-8")).hexdigest(),
        "created_at": datetime.now(timezone.utc).isoformat(),
    }


def impact_report(
    revoked_entity_id: str,
    dependency_edges: list[dict[str, str]],
) -> dict[str, Any]:
    affected: set[str] = set()
    frontier = [revoked_entity_id]
    while frontier:
        current = frontier.pop()
        for edge in dependency_edges:
            if edge["from"] == current and edge["to"] not in affected:
                affected.add(edge["to"])
                frontier.append(edge["to"])
    return {"revoked_entity_id": revoked_entity_id, "affected_entities": sorted(affected)}


def quality_drift(
    baseline_failure_rate: float,
    current_failure_rate: float,
    investigation_delta: float,
) -> dict[str, Any]:
    delta = current_failure_rate - baseline_failure_rate
    return {
        "baseline_failure_rate": baseline_failure_rate,
        "current_failure_rate": current_failure_rate,
        "delta": delta,
        "investigation_required": delta >= investigation_delta,
    }