Caching

Pass a ResultCache to the client to cache query results. The store only ever sees opaque hq1.… keys, never queries, tenant ids, or filter values.

from hypequery.datasets import MemoryCacheStore, ResultCache, create_dataset_client

cache = ResultCache(
    store=MemoryCacheStore(max_entries=1_000),
    project="acme",
    environment="production",
    ttl_seconds=60,
)
analytics = create_dataset_client(executor=executor, cache=cache)

result = analytics.execute(orders, query, context=context)
result.meta.cache  # "miss", then "hit" for the same request

• Tenant isolation. Entries are keyed by tenant, so two tenants never share an entry.
• Equivalent requests share an entry. Filter order, offset: 0, and empty lists do not create separate entries.
• Never fails a query. A store error runs the query uncached and reports meta.cache == "bypass". Pass use_cache=False to skip the cache for one call.
• Shared stores need a secret. For a store shared across processes, such as Redis, pass a secret of 32 or more random bytes, the same on every instance and distinct per environment. Without one, the cache generates a random secret for its own lifetime, which is all an in-memory store needs.