mirror of
https://github.com/ets-labs/python-dependency-injector.git
synced 2024-11-29 21:14:00 +03:00
99 lines
3.5 KiB
ReStructuredText
99 lines
3.5 KiB
ReStructuredText
Singleton providers
|
|
-------------------
|
|
|
|
.. currentmodule:: dependency_injector.providers
|
|
|
|
:py:class:`Singleton` provider creates new instance of specified class on
|
|
first call and returns same instance on every next call.
|
|
|
|
Example:
|
|
|
|
.. image:: /images/providers/singleton.png
|
|
:width: 80%
|
|
:align: center
|
|
|
|
.. literalinclude:: ../../examples/providers/singleton.py
|
|
:language: python
|
|
:linenos:
|
|
|
|
Singleton providers resetting
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
Created and memorized by :py:class:`Singleton` instance can be reset. Reset of
|
|
:py:class:`Singleton`'s memorized instance is done by clearing reference to
|
|
it. Further lifecycle of memorized instance is out of :py:class:`Singleton`
|
|
provider's control and dependes on garbage collection strategy.
|
|
|
|
Example:
|
|
|
|
.. literalinclude:: ../../examples/providers/singleton_reseting.py
|
|
:language: python
|
|
:linenos:
|
|
|
|
Singleton providers and injections
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
:py:class:`Singleton` provider extends :py:class:`Factory` provider, so, all
|
|
of the rules about injections are the same, as for :py:class:`Factory`
|
|
provider.
|
|
|
|
.. note::
|
|
|
|
Due that :py:class:`Singleton` provider creates specified class instance
|
|
only on the first call, all injections are done once, during the first
|
|
call, also. Every next call, while instance has been already created
|
|
and memorized, no injections are done, :py:class:`Singleton` provider just
|
|
returns memorized earlier instance.
|
|
|
|
This may cause some problems, for example, in case of trying to bind
|
|
:py:class:`Factory` provider with :py:class:`Singleton` provider (provided
|
|
by dependent :py:class:`Factory` instance will be injected only once,
|
|
during the first call). Be aware that such behaviour was made with opened
|
|
eyes and is not a bug.
|
|
|
|
By the way, in such case, :py:class:`Delegate` or
|
|
:py:class:`DelegatedSingleton` provider can be useful
|
|
. It makes possible to inject providers *as is*. Please check out
|
|
`Singleton providers delegation`_ section.
|
|
|
|
Singleton providers delegation
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
:py:class:`Singleton` provider could be delegated to any other provider via
|
|
any kind of injection.
|
|
|
|
Delegation of :py:class:`Singleton` providers is the same as
|
|
:py:class:`Factory` providers delegation, please follow
|
|
:ref:`factory_providers_delegation` section for examples (with exception
|
|
about using :py:class:`DelegatedSingleton` instead of
|
|
:py:class:`DelegatedFactory`).
|
|
|
|
Singleton providers specialization
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
:py:class:`Singleton` provider could be specialized for any kind of needs via
|
|
declaring its subclasses.
|
|
|
|
Specialization of :py:class:`Singleton` providers is the same as
|
|
:py:class:`Factory` providers specialization, please follow
|
|
:ref:`factory_providers_specialization` section for examples.
|
|
|
|
Singleton providers and multi-threading
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
:py:class:`Singleton` provider is thread-safe and could be used in
|
|
multi-threading applications without any negative impact. Race condition on
|
|
singleton's initialization is escaped by using a global reentrant mutex -
|
|
:py:obj:`dependency_injector.utils.GLOBAL_LOCK`.
|
|
|
|
Also there could be a need to use thread-scoped singletons and there is a
|
|
special provider for such case - :py:class:`ThreadLocalSingleton`.
|
|
:py:class:`ThreadLocalSingleton` provider creates instance once for each
|
|
thread and returns it on every call.
|
|
|
|
Example:
|
|
|
|
.. literalinclude:: ../../examples/providers/singleton_thread_locals.py
|
|
:language: python
|
|
:linenos:
|