# Copyright (c) Streamlit Inc. (2018-2022) Snowflake Inc. (2022-2026) # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """@st.cache_data: pickle-based caching.""" from __future__ import annotations import pickle import threading from typing import ( TYPE_CHECKING, Any, Final, Literal, TypeAlias, TypeVar, cast, overload, ) from typing_extensions import ParamSpec import streamlit as st from streamlit import runtime from streamlit.errors import StreamlitAPIException from streamlit.logger import get_logger from streamlit.runtime.caching.cache_errors import CacheError, CacheKeyNotFoundError from streamlit.runtime.caching.cache_type import CacheType from streamlit.runtime.caching.cache_utils import ( Cache, CachedFunc, CachedFuncInfo, CacheScope, get_session_id_or_throw, make_cached_func_wrapper, ) from streamlit.runtime.caching.cached_message_replay import ( CachedMessageReplayContext, CachedResult, MsgData, ) from streamlit.runtime.caching.storage import ( CacheStorage, CacheStorageContext, CacheStorageError, CacheStorageKeyNotFoundError, CacheStorageManager, ) from streamlit.runtime.caching.storage.cache_storage_protocol import ( InvalidCacheStorageContextError, ) from streamlit.runtime.caching.storage.dummy_cache_storage import ( MemoryCacheStorageManager, ) from streamlit.runtime.metrics_util import gather_metrics from streamlit.runtime.stats import ( CACHE_MEMORY_FAMILY, CacheStat, StatsProvider, group_cache_stats, ) from streamlit.time_util import time_to_seconds if TYPE_CHECKING: from collections.abc import Callable, Sequence from datetime import timedelta from streamlit.runtime.caching.hashing import HashFuncsDict _LOGGER: Final = get_logger(__name__) CACHE_DATA_MESSAGE_REPLAY_CTX = CachedMessageReplayContext(CacheType.DATA) # The cache persistence options we support: "disk" or None CachePersistType: TypeAlias = Literal["disk"] | None P = ParamSpec("P") R = TypeVar("R") class CachedDataFuncInfo(CachedFuncInfo[P, R]): """Implements the CachedFuncInfo interface for @st.cache_data.""" persist: CachePersistType max_entries: int | None ttl: float | timedelta | str | None def __init__( self, func: Callable[P, R], persist: CachePersistType, max_entries: int | None, ttl: float | timedelta | str | None, show_spinner: bool | str, show_time: bool = False, hash_funcs: HashFuncsDict | None = None, scope: CacheScope = "global", ) -> None: super().__init__( func, hash_funcs=hash_funcs, show_spinner=show_spinner, show_time=show_time, scope=scope, ) self.persist = persist self.max_entries = max_entries self.ttl = ttl self.validate_params() @property def cache_type(self) -> CacheType: return CacheType.DATA @property def cached_message_replay_ctx(self) -> CachedMessageReplayContext: return CACHE_DATA_MESSAGE_REPLAY_CTX @property def display_name(self) -> str: """A human-readable name for the cached function.""" return f"{self.func.__module__}.{self.func.__qualname__}" def get_function_cache(self, function_key: str) -> Cache[R]: return _data_caches.get_cache( key=function_key, persist=self.persist, max_entries=self.max_entries, ttl=self.ttl, display_name=self.display_name, scope=self.scope, ) def validate_params(self) -> None: """ Validate the params passed to @st.cache_data are compatible with cache storage. When called, this method could log warnings if cache params are invalid for current storage. """ _data_caches.validate_cache_params( function_name=self.func.__name__, persist=self.persist, max_entries=self.max_entries, ttl=self.ttl, ) class DataCaches(StatsProvider): """Manages all DataCache instances.""" def __init__(self) -> None: self._caches_lock = threading.Lock() # Map of session IDs to map of function keys to caches. self._function_caches: dict[str | None, dict[str, DataCache[Any]]] = {} @property def stats_families(self) -> Sequence[str]: return (CACHE_MEMORY_FAMILY,) def get_cache( self, key: str, persist: CachePersistType, max_entries: int | None, ttl: int | float | timedelta | str | None, display_name: str, scope: CacheScope = "global", ) -> DataCache[Any]: """Return the mem cache for the given key. If it doesn't exist, create a new one with the given params. Raises ------ StreamlitAPIException Raised when ``scope`` is ``"session"`` and there is no thread-local run context. """ ttl_seconds = time_to_seconds(ttl, coerce_none_to_inf=False) # Fetch the session ID. Note that this will throw an exception if there is no # session associated with the current thread. session_id: str | None if scope == "global": session_id = None else: session_id = get_session_id_or_throw() # Get the existing cache, if it exists, and validate that its params # haven't changed. with self._caches_lock: session_caches = self._function_caches.get(session_id) if session_caches is None: session_caches = self._function_caches[session_id] = {} cache = session_caches.get(key) if ( cache is not None and cache.ttl_seconds == ttl_seconds and cache.max_entries == max_entries and cache.persist == persist ): return cache # Close the existing cache's storage, if it exists. if cache is not None: _LOGGER.debug( "Closing existing DataCache storage " "(key=%s, persist=%s, max_entries=%s, ttl=%s) " "before creating new one with different params", key, persist, max_entries, ttl, ) cache.storage.close() # Create a new cache object and put it in our dict _LOGGER.debug( "Creating new DataCache (key=%s, persist=%s, max_entries=%s, ttl=%s)", key, persist, max_entries, ttl, ) cache_context = self.create_cache_storage_context( function_key=key, function_name=display_name, ttl_seconds=ttl_seconds, max_entries=max_entries, persist=persist, ) cache_storage_manager = self.get_storage_manager() storage = cache_storage_manager.create(cache_context) cache = DataCache( key=key, storage=storage, persist=persist, max_entries=max_entries, ttl_seconds=ttl_seconds, display_name=display_name, ) self._function_caches[session_id][key] = cache return cache def clear_session(self, session_id: str) -> None: """Clears all caches for the given session ID.""" # Hold the lock while removing the cache, but release it while clearing. with self._caches_lock: session_caches = self._function_caches.get(session_id) if session_caches is not None: del self._function_caches[session_id] if session_caches is not None: for cache in session_caches.values(): cache.clear() cache.storage.close() def clear_all(self) -> None: """Clear all in-memory and on-disk caches.""" with self._caches_lock: try: # try to remove in optimal way if such ability provided by # storage manager clear_all method; # if not implemented, fallback to remove all # available storages one by one self.get_storage_manager().clear_all() except NotImplementedError: for data_caches in self._function_caches.values(): for data_cache in data_caches.values(): data_cache.clear() data_cache.storage.close() self._function_caches = {} def get_stats( self, _family_names: Sequence[str] | None = None ) -> dict[str, list[CacheStat]]: with self._caches_lock: # Shallow-clone our caches. We don't want to hold the global # lock during stats-gathering. function_caches = [ cache for caches in self._function_caches.values() for cache in caches.values() ] stats: list[CacheStat] = [] for cache in function_caches: cache_stats = cache.get_stats() for family_stats in cache_stats.values(): stats.extend(family_stats) if not stats: return {} # In general, get_stats methods need to be able to return only requested stat # families, but this method only returns a single family, and we're guaranteed # that it was one of those requested if we make it here. return {CACHE_MEMORY_FAMILY: group_cache_stats(stats)} def validate_cache_params( self, function_name: str, persist: CachePersistType, max_entries: int | None, ttl: int | float | timedelta | str | None, ) -> None: """Validate that the cache params are valid for given storage. Raises ------ InvalidCacheStorageContext Raised if the cache storage manager is not able to work with provided CacheStorageContext. """ ttl_seconds = time_to_seconds(ttl, coerce_none_to_inf=False) cache_context = self.create_cache_storage_context( function_key="DUMMY_KEY", function_name=function_name, ttl_seconds=ttl_seconds, max_entries=max_entries, persist=persist, ) try: self.get_storage_manager().check_context(cache_context) except InvalidCacheStorageContextError: _LOGGER.exception( "Cache params for function %s are incompatible with current " "cache storage manager.", function_name, ) raise def create_cache_storage_context( self, function_key: str, function_name: str, persist: CachePersistType, ttl_seconds: float | None, max_entries: int | None, ) -> CacheStorageContext: return CacheStorageContext( function_key=function_key, function_display_name=function_name, ttl_seconds=ttl_seconds, max_entries=max_entries, persist=persist, ) def get_storage_manager(self) -> CacheStorageManager: if runtime.exists(): return runtime.get_instance().cache_storage_manager # When running in "raw mode", we can't access the CacheStorageManager, # so we're falling back to InMemoryCache. _LOGGER.warning("No runtime found, using MemoryCacheStorageManager") return MemoryCacheStorageManager() # Singleton DataCaches instance _data_caches = DataCaches() def clear_session_cache(session_id: str) -> None: """Clears all caches for the given session ID.""" _data_caches.clear_session(session_id) def get_data_cache_stats_provider() -> StatsProvider: """Return the StatsProvider for all @st.cache_data functions.""" return _data_caches class CacheDataAPI: """Implements the public st.cache_data API: the @st.cache_data decorator, and st.cache_data.clear(). """ def __init__(self, decorator_metric_name: str) -> None: """Create a CacheDataAPI instance. Parameters ---------- decorator_metric_name The metric name to record for decorator usage. """ # Parameterize the decorator metric name. # (Ignore spurious mypy complaints - https://github.com/python/mypy/issues/2427) self._decorator = gather_metrics( # type: ignore decorator_metric_name, self._decorator ) # Type-annotate the decorator function. # (See https://mypy.readthedocs.io/en/stable/generics.html#decorator-factories) # Bare decorator usage @overload def __call__(self, func: Callable[P, R]) -> CachedFunc[P, R]: ... # Decorator with arguments @overload def __call__( self, *, ttl: float | timedelta | str | None = None, max_entries: int | None = None, show_spinner: bool | str = True, show_time: bool = False, persist: CachePersistType | bool = None, hash_funcs: HashFuncsDict | None = None, scope: CacheScope = "global", ) -> Callable[[Callable[P, R]], CachedFunc[P, R]]: ... def __call__( self, func: Callable[P, R] | None = None, *, ttl: float | timedelta | str | None = None, max_entries: int | None = None, show_spinner: bool | str = True, show_time: bool = False, persist: CachePersistType | bool = None, hash_funcs: HashFuncsDict | None = None, scope: CacheScope = "global", ) -> CachedFunc[P, R] | Callable[[Callable[P, R]], CachedFunc[P, R]]: return self._decorator( func, # ty: ignore[invalid-argument-type] ttl=ttl, max_entries=max_entries, persist=persist, show_spinner=show_spinner, show_time=show_time, hash_funcs=hash_funcs, scope=scope, ) def _decorator( self, func: Callable[P, R] | None = None, *, ttl: float | timedelta | str | None, max_entries: int | None, show_spinner: bool | str, show_time: bool = False, persist: CachePersistType | bool, hash_funcs: HashFuncsDict | None = None, scope: CacheScope = "global", ) -> CachedFunc[P, R] | Callable[[Callable[P, R]], CachedFunc[P, R]]: """Decorator to cache functions that return data (e.g. dataframe transforms, database queries, ML inference). Cached objects can be global or session-scoped. Global cached data is available across all users, sessions, and reruns. Session-scoped cached data is only available in the current session and removed when the session disconnects. Cached objects are stored in "pickled" form, which means that the return value of a cached function must be pickleable. Each caller of the cached function gets its own copy of the cached data. You can clear a function's cache with ``func.clear()`` or clear the entire cache with ``st.cache_data.clear()``. A function's arguments must be hashable to cache it. Streamlit makes a best effort to hash a variety of objects, but the fallback hashing method requires that the argument be pickleable, also. If you have an unhashable argument (like a database connection) or an argument you want to exclude from caching, use an underscore prefix in the argument name. In this case, Streamlit will return a cached value when all other arguments match a previous function call. Alternatively, you can declare custom hashing functions with ``hash_funcs``. Objects cached by ``st.cache_data`` are returned as copies. To cache a shared resource or something you want to mutate in place, use ``st.cache_resource`` instead. To learn more about caching, see `Caching overview `_. .. note:: Caching async functions is not supported. To upvote this feature, see GitHub issue `#8308 `_. Parameters ---------- func : callable The function to cache. Streamlit hashes the function's source code. ttl : float, timedelta, str, or None The maximum time to keep an entry in the cache. Can be one of: - ``None`` if cache entries should never expire (default). - A number specifying the time in seconds. - A string specifying the time in a format supported by `Pandas's Timedelta constructor `_, e.g. ``"1d"``, ``"1.5 days"``, or ``"1h23s"``. - A ``timedelta`` object from `Python's built-in datetime library `_, e.g. ``timedelta(days=1)``. Note that ``ttl`` will be ignored if ``persist="disk"`` or ``persist=True``. max_entries : int or None The maximum number of entries to keep in the cache, or None for an unbounded cache. When a new entry is added to a full cache, the oldest cached entry will be removed. Defaults to None. show_spinner : bool or str Enable the spinner. Default is True to show a spinner when there is a "cache miss" and the cached data is being created. If string, value of show_spinner param will be used for spinner text. show_time : bool Whether to show the elapsed time next to the spinner text. If this is ``False`` (default), no time is displayed. If this is ``True``, elapsed time is displayed with a precision of 0.1 seconds. The time format is not configurable. persist : "disk", bool, or None Optional location to persist cached data to. Passing "disk" (or True) will persist the cached data to the local disk. None (or False) will disable persistence. The default is None. hash_funcs : dict or None Mapping of types or fully qualified names to hash functions. This is used to override the behavior of the hasher inside Streamlit's caching mechanism: when the hasher encounters an object, it will first check to see if its type matches a key in this dict and, if so, will use the provided function to generate a hash for it. See below for an example of how this can be used. scope : "global" or "session" The scope for the data cache. If this is ``"global"`` (default), the data is cached globally. If this is ``"session"``, the data is removed from the cache when the session disconnects. Because a session-scoped cache is cleared when a session disconnects, an unstable network connection can cause the cache to populate multiple times in a single session. If this is a problem, you might consider adjusting the ``server.websocketPingInterval`` configuration option. Example ------- >>> import streamlit as st >>> >>> @st.cache_data ... def fetch_and_clean_data(url): ... # Fetch data from URL here, and then clean it up. ... return data >>> >>> d1 = fetch_and_clean_data(DATA_URL_1) >>> # Actually executes the function, since this is the first time it was >>> # encountered. >>> >>> d2 = fetch_and_clean_data(DATA_URL_1) >>> # Does not execute the function. Instead, returns its previously computed >>> # value. This means that now the data in d1 is the same as in d2. >>> >>> d3 = fetch_and_clean_data(DATA_URL_2) >>> # This is a different URL, so the function executes. To set the ``persist`` parameter, use this command as follows: >>> import streamlit as st >>> >>> @st.cache_data(persist="disk") ... def fetch_and_clean_data(url): ... # Fetch data from URL here, and then clean it up. ... return data By default, all parameters to a cached function must be hashable. Any parameter whose name begins with ``_`` will not be hashed. You can use this as an "escape hatch" for parameters that are not hashable: >>> import streamlit as st >>> >>> @st.cache_data ... def fetch_and_clean_data(_db_connection, num_rows): ... # Fetch data from _db_connection here, and then clean it up. ... return data >>> >>> connection = make_database_connection() >>> d1 = fetch_and_clean_data(connection, num_rows=10) >>> # Actually executes the function, since this is the first time it was >>> # encountered. >>> >>> another_connection = make_database_connection() >>> d2 = fetch_and_clean_data(another_connection, num_rows=10) >>> # Does not execute the function. Instead, returns its previously computed >>> # value - even though the _database_connection parameter was different >>> # in both calls. A cached function's cache can be procedurally cleared: >>> import streamlit as st >>> >>> @st.cache_data ... def fetch_and_clean_data(_db_connection, num_rows): ... # Fetch data from _db_connection here, and then clean it up. ... return data >>> >>> fetch_and_clean_data.clear(_db_connection, 50) >>> # Clear the cached entry for the arguments provided. >>> >>> fetch_and_clean_data.clear() >>> # Clear all cached entries for this function. To override the default hashing behavior, pass a custom hash function. You can do that by mapping a type (e.g. ``datetime.datetime``) to a hash function (``lambda dt: dt.isoformat()``) like this: >>> import streamlit as st >>> import datetime >>> >>> @st.cache_data(hash_funcs={datetime.datetime: lambda dt: dt.isoformat()}) ... def convert_to_utc(dt: datetime.datetime): ... return dt.astimezone(datetime.timezone.utc) Alternatively, you can map the type's fully-qualified name (e.g. ``"datetime.datetime"``) to the hash function instead: >>> import streamlit as st >>> import datetime >>> >>> @st.cache_data(hash_funcs={"datetime.datetime": lambda dt: dt.isoformat()}) ... def convert_to_utc(dt: datetime.datetime): ... return dt.astimezone(datetime.timezone.utc) """ # Parse our persist value into a string persist_string: CachePersistType if persist is True: persist_string = "disk" elif persist is False: persist_string = None else: persist_string = persist if persist_string not in (None, "disk"): # We'll eventually have more persist options. raise StreamlitAPIException( f"Unsupported persist option '{persist}'. Valid values are 'disk' or None." ) if scope not in ("global", "session"): raise StreamlitAPIException( f"Unsupported scope option '{scope}'. Valid values are 'global' or 'session'." ) def wrapper(f: Callable[P, R]) -> CachedFunc[P, R]: return make_cached_func_wrapper( CachedDataFuncInfo( func=f, persist=persist_string, show_spinner=show_spinner, show_time=show_time, max_entries=max_entries, ttl=ttl, hash_funcs=hash_funcs, scope=scope, ) ) if func is None: return wrapper return make_cached_func_wrapper( CachedDataFuncInfo( func=func, persist=persist_string, show_spinner=show_spinner, show_time=show_time, max_entries=max_entries, ttl=ttl, hash_funcs=hash_funcs, scope=scope, ) ) @gather_metrics("clear_data_caches") def clear(self) -> None: """Clear all in-memory and on-disk data caches.""" _data_caches.clear_all() class DataCache(Cache[R]): """Manages cached values for a single st.cache_data function.""" def __init__( self, key: str, storage: CacheStorage, persist: CachePersistType, max_entries: int | None, ttl_seconds: float | None, display_name: str, ) -> None: super().__init__() self.key = key self.display_name = display_name self.storage = storage self.ttl_seconds = ttl_seconds self.max_entries = max_entries self.persist = persist def get_stats( self, _family_names: Sequence[str] | None = None ) -> dict[str, list[CacheStat]]: # In general, get_stats methods need to be able to return only requested stat # families, but this method only returns a single family, and we're guaranteed # that it was one of those requested if we make it here. if isinstance(self.storage, StatsProvider): return cast("dict[str, list[CacheStat]]", self.storage.get_stats()) return {} def read_result(self, key: str) -> CachedResult[R]: """Read a value and messages from the cache. Raise `CacheKeyNotFoundError` if the value doesn't exist, and `CacheError` if the value exists but can't be unpickled. """ try: pickled_entry = self.storage.get(key) except CacheStorageKeyNotFoundError as e: raise CacheKeyNotFoundError(str(e)) from e except CacheStorageError as e: raise CacheError(str(e)) from e try: entry = pickle.loads(pickled_entry) # noqa: S301 if not isinstance(entry, CachedResult): # Loaded an old cache file format, remove it and let the caller # rerun the function. self.storage.delete(key) raise CacheKeyNotFoundError() return entry except pickle.UnpicklingError as exc: raise CacheError(f"Failed to unpickle {key}") from exc @gather_metrics("_cache_data_object") def write_result(self, key: str, value: R, messages: list[MsgData]) -> None: """Write a value and associated messages to the cache. The value must be pickleable. """ try: main_id = st._main.id sidebar_id = st.sidebar.id entry = CachedResult(value, messages, main_id, sidebar_id) pickled_entry = pickle.dumps(entry) except (pickle.PicklingError, TypeError) as exc: raise CacheError(f"Failed to pickle {key}") from exc self.storage.set(key, pickled_entry) def _clear(self, key: str | None = None) -> None: if not key: self.storage.clear() else: self.storage.delete(key)