What is serializable session state?
Serializable Session State
Serialization refers to the process of converting an object or data structure into a format that can be persisted and shared, and allowing you to recover the data’s original structure. Python’s built-in pickle module serializes Python objects to a byte stream ("pickling") and deserializes the stream into an object ("unpickling").
By default, Streamlit’s Session State allows you to persist any Python object for the duration of the session, irrespective of the object’s pickle-serializability. This property lets you store Python primitives such as integers, floating-point numbers, complex numbers and booleans, dataframes, and even lambdas returned by functions. However, some execution environments may require serializing all data in Session State, so it may be useful to detect incompatibility during development, or when the execution environment will stop supporting it in the future.
To that end, Streamlit provides a
runner.enforceSerializableSessionState configuration option that, when set to
true, only allows pickle-serializable objects in Session State. To enable the option, either create a global or project config file with the following or use it as a command-line flag:
# .streamlit/config.toml [runner] enforceSerializableSessionState = true
By "pickle-serializable", we mean calling
pickle.dumps(obj) should not raise a
PicklingError exception. When the config option is enabled, adding unserializable data to session state should result in an exception. E.g.,
import streamlit as st def unserializable_data(): return lambda x: x #👇 results in an exception when enforceSerializableSessionState is on st.session_state.unserializable = unserializable_data()