fiftyone.operators.cache.decorator

Decorator for caching function results in an ExecutionStore.

The function must:

• accept a ctx argument as the first parameter
• return a serializable value
• should produce the same output for the same input
• should have no side effects
• have serializable parameters and return values

note:

When using ``link_to_dataset=True``:
    - the associated store is deleted the dataset is deleted
    - the cache entry is namespaced to the dataset

note:

Return values will be coerced from JSON unsafe types to safe types.
This may yield unexpected return values if the cached function returns
non-serializable types (e.g., NumPy arrays), since they are converted
to a JSON-compatible format.

This behavior can be overridden by providing custom ``serialize`` and/or
``deserialize`` functions.

Example Usage:

# Standalone function with default caching @execution_cache def expensive_query(ctx, path):

return ctx.dataset.count_values(path)

# Instance method with dataset-scoped caching class Processor:

@execution_cache(ttl=60, store_name="processor_cache") def expensive_query(self, ctx, path):

return ctx.dataset.count_values(path)

# Using a custom key function def custom_key_fn(ctx, path):

return [path, get_day_of_week()]

# Combines the custom key function with user-scoped caching @execution_cache(ttl=90, key_fn=custom_key_fn, jwt_scoped=True) def user_specific_query(ctx, path):

return ctx.dataset.match(

F("creator_id") == ctx.user_id

).count_values(path)

# Bypass the cache result = expensive_query.uncached(ctx, path)

# Set the cache for the given arguments expensive_query.set_cache(ctx, path, value_to_cache)

# Clear the cache for the given arguments expensive_query.clear_cache(ctx, path)

# Remove all cache entries for the function expensive_query.clear_all_caches()

# Clear all cache entries for the function expensive_query.clear_all_caches()

# NOTE: dataset_id is required if link_to_dataset=True expensive_query.clear_all_caches(dataset_id=dataset._doc.id)