This project is an Aspect Oriented Programming library for python with reflective concerns.

• Homepage
• PyPI
• Documentation

pip install b3j0f.aop

1. Free and unlimited access: no limits to sharing of ideas and knowledges with the license MIT.

2. Performance:

   • less memory consumption in using the __slots__ class property.
   • less time on (un-)weaving and advice application improvement with binary python encoding and in using constants var in code.
   • (dis/en)abling advices without remove them in using dedicated Advice class.

3. Easy to use:

   • joinpoint matching with function or regex.

   • distributed programming:

     • interception context sharing in order to ease behaviour sharing between advices.
     • uuid for advice identification in order to ease its use in a distributed context.

   • maintenable with well named variables and functions, comments and few lines.

   • extensible through pythonic code (PEP8), same logic to function code interception and concern modularisation with one module by joinpoint or advice.

   • respect of aspects vocabulary in order to ease its use among AOP users.

   • close to callable python objects in weaving all types of callable elements such as (built-in) functions, (built-in) class, (built-in) methods, callable objects, etc.

   • advices are callable objects.

   • Unit tests for all functions such as examples.

4. Benchmark:

   • speed execution

• Do not weave advices on readonly instance methods (where class use __slots__ attribute).

>>> from b3j0f.aop import weave, is_intercepted
>>> double_advice = lambda joinpoint: joinpoint.proceed() * 2
>>> weave(target=min, advices=double_advice)
>>> min(6, 7)
12

>>> from b3j0f.aop import is_intercepted
>>> is_intercepted(min)
True

Ok, let's get back its previous behaviour ...

>>> from b3j0f.aop import unweave
>>> unweave(min)
>>> min(6, 7)
6
>>> is_intercepted(min)
False

>>> from b3j0f.aop import weave_on
>>> weave_on(advices=double_advice)(min)
>>> min(6, 7)
12
>>> is_intercepted(min)
True
>>> unweave(min)  # do not forget to unweave if weaving is useless ;)

Remember it is not possible to weave on readonly methods such as str.split.

>>> @weave_on(advices=double_advice, pointcut='simple.*')
>>> class Test(object):
>>>     def simple_one(self):
>>>         return 1
>>>     def simple_x(self, x):
>>>         return x
>>>     def three(self):
>>>         return 3
>>> assert Test().simple_one() == 1 * 2  # result is doubled
>>> assert Test().simple_x(2) == 2 * 2  # result is doubled
>>> assert Test().three() == 3  # normal result
>>> unweave(Test, pointcut='simple_one')  # unweave simple_one only
>>> assert Test().simple_one() == 1  # back to normal
>>> assert Test().simple_x(2) == 2 * 2  # result is still doubled

Enjoy ...

Related to improving criteria points (1. Free and unlimited access, etc.), a state of the art is provided here.

• Not functional approach: Aspect class definition.
• Side effects: Not close to python API.
• Not optimized Weaving and Time execution: use classes and generic methods.
• Not maintenable: poor comments.
• open-source and use limitations: GPL 2.
• limited in weave filtering.

• open-source and use limitations: LGPL 2.1.
• more difficulties to understand code with no respect of the AOP vocabulary, packaged into one module.
• limited in weave filtering.

• invert the AOP in decorating advices with joinpoint instead of weaving advices on joinpoint.
• open-source and no use limitations: BSD.

• Simple and functional approach with use of python tools.
• maintenable: commented in respect of the PEP8.
• limited in weave filtering.

• a very powerful library dedicated to develop strong systems based on component based software engineering.
• unittests.
• huge community.

• require to understand a lot of concepts and install an heavy library before doing a simple interception with AOP concerns.

• Very complex and full library for doing aspects and other things.

• open-source and use limitations: GPL 3.
• not maintenable: missing documentations and not respect of the PEP8.
• Executon time is not optimized with several classes used with generic getters without using __slots__. The only one optimization comes from the yield which requires from users to use it in their own advices (which must be a class).

• wait feedbacks during 6 months before passing it to a stable version.
• Cython implementation.