A wrapper of pluggy to support asyncio and context managers.
This package provides a subclass of
pluggy.PluginManager
which
- allows async functions, context managers, and async context managers to be hooks
- and accepts plugin factories in addition to plugin instances for registration.
Table of Contents
You can install apluggy with pip:
pip install apluggyHere, we show a simple example of how to use apluggy.
We only describe the usage of additional features provided by apluggy. For the usage of pluggy itself, please refer to the pluggy documentation.
You can try this example in a Python interpreter.
$ python
Python 3.10.13 (...)
...
...
>>>Import necessary packages of this example.
>>> import asyncio
>>> import apluggy as pluggy
>>> from apluggy import asynccontextmanager, contextmanagerIn this example, apluggy is imported with the alias pluggy.
The decorators asynccontextmanager and contextmanager are imported from
apluggy. They are wrappers of the decorators of the same names in the
contextlib package. The
wrappers preserve the signatures of decorated functions, which are necessary for
pluggy to pass arguments to hook implementations correctly. (The decorator
contextmanger in apluggy is the same object as the decorator
contextmanager in the decorator
package. The decorator package does not
provide asynccontextmanager decorator as of version 5.1. The decorator
asynccontextmanger in apluggy is implemented in a similar way as the
decorator contextmanager in the decorator package.)
>>> hookspec = pluggy.HookspecMarker('project')
>>> hookimpl = pluggy.HookimplMarker('project')In this example, we define three hooks: async function, context manager, and async context manager.
>>> class Spec:
... """A hook specification namespace."""
...
... @hookspec
... async def afunc(self, arg1, arg2):
... pass
...
... @hookspec
... @contextmanager
... def context(self, arg1, arg2):
... pass
...
... @hookspec
... @asynccontextmanager
... async def acontext(self, arg1, arg2):
... passWe define two plugins as classes. Each plugin implements the three hooks defined above.
>>> class Plugin_1:
... """A hook implementation namespace."""
...
... @hookimpl
... async def afunc(self, arg1, arg2):
... print('inside Plugin_1.afunc()')
... return arg1 + arg2
...
... @hookimpl
... @contextmanager
... def context(self, arg1, arg2):
... print('inside Plugin_1.context(): before')
... yield arg1 + arg2
... print('inside Plugin_1.context(): after')
...
... @hookimpl
... @asynccontextmanager
... async def acontext(self, arg1, arg2):
... print('inside Plugin_1.acontext(): before')
... yield arg1 + arg2
... print('inside Plugin_1.acontext(): after')
>>> class Plugin_2:
... """A 2nd hook implementation namespace."""
...
... @hookimpl
... async def afunc(self, arg1, arg2):
... print('inside Plugin_2.afunc()')
... return arg1 - arg2
...
... @hookimpl
... @contextmanager
... def context(self, arg1, arg2):
... print('inside Plugin_2.context(): before')
... yield arg1 - arg2
... print('inside Plugin_2.context(): after')
...
... @hookimpl
... @asynccontextmanager
... async def acontext(self, arg1, arg2):
... print('inside Plugin_2.acontext(): before')
... yield arg1 - arg2
... print('inside Plugin_2.acontext(): after')Plugins can be registered as instances or factories. In the following
example, we register two plugins: Plugin_1 as an instance, and Plugin_2
as a factory.
>>> pm = pluggy.PluginManager('project')
>>> pm.add_hookspecs(Spec)
>>> _ = pm.register(Plugin_1()) # instantiation is optional.
>>> _ = pm.register(Plugin_2) # callable is considered a plugin factory.Pluggy accepts a class or
module as a
plugin. However, it actually accepts a class instance, not a class itself.
Consequently, when plugins are loaded with
load_setuptools_entrypoints(),
the entry points must be class instances or modules. Classes themselves cannot
be used as entry points (if understood correctly).
So that classes themselves can be entry points, apluggy accepts a class itself for a plugin registration. When apluggy receives a callable object, apluggy considers the object as a plugin factory.
The following example shows how to call hooks.
>>> async def call_afunc():
... results = await pm.ahook.afunc(arg1=1, arg2=2) # ahook instead of hook
... print(results)
>>> asyncio.run(call_afunc())
inside Plugin_2.afunc()
inside Plugin_1.afunc()
[-1, 3]>>> with pm.with_.context(arg1=1, arg2=2) as y: # with_ instead of hook
... print(y)
inside Plugin_2.context(): before
inside Plugin_1.context(): before
[-1, 3]
inside Plugin_1.context(): after
inside Plugin_2.context(): afterIn the reverse order:
>>> with pm.with_reverse.context(arg1=1, arg2=2) as y: # with_reverse instead of hook
... print(y)
inside Plugin_1.context(): before
inside Plugin_2.context(): before
[3, -1]
inside Plugin_2.context(): after
inside Plugin_1.context(): after>>> async def call_acontext():
... async with pm.awith.acontext(arg1=1, arg2=2) as y: # awith instead of hook
... print(y)
>>> asyncio.run(call_acontext())
inside Plugin_2.acontext(): before
inside Plugin_1.acontext(): before
[-1, 3]
inside Plugin_1.acontext(): after
inside Plugin_2.acontext(): afterIn the reverse order:
>>> async def call_acontext():
... async with pm.awith_reverse.acontext(arg1=1, arg2=2) as y: # awith_reverse instead of hook
... print(y)
>>> asyncio.run(call_acontext())
inside Plugin_1.acontext(): before
inside Plugin_2.acontext(): before
[3, -1]
inside Plugin_2.acontext(): after
inside Plugin_1.acontext(): after- apluggy is licensed under the MIT license.
![dependabot[bot]](https://avatars.githubusercontent.com/u/49699333?size=120)