New Show Hacker News story: Show HN: Generic dual-paradigm hooking mechanism
Show HN: Generic dual-paradigm hooking mechanism
3 by alexrustic | 0 comments on Hacker News.
Hi HN ! I am Alex, a tech enthusiast, I'm excited to show you a major iteration of my library for performing hooking in Python. I redesigned the whole project because it didn't not cover all my needs. I'm happy with the current iteration that I've written tests for and look forward to spending weeks and months using it in my projects. Python has a concept called Decorator [1] which is a function that takes another function and extends the behavior. In the following script, the timeit decorator is used to measure the execution time of the heavy_computation function: import time from functools import wraps def timeit(text): def deco(target): @wraps(target) def wrapper(*args, **kwargs): # execute and measure the target run time start_time = time.perf_counter() result = target(*args, **kwargs) total_time = time.perf_counter() - start_time # print elapsed time print(text.format(total=total_time)) return result return wrapper return deco @timeit(text="Done in {total:.3f} seconds !") def heavy_computation(a, b): time.sleep(2) # doing some heavy computation ! return a*b if __name__ == "__main__": result = heavy_computation(6, 9) print("Result:", result) Output: $ python -m test Done in 2.001 seconds ! Result: 54 Besides benchmarking, there are many other cool things that can be done with the Python decorator. For example, the Flask [2] and Bottle [3] web frameworks implement routing with decorators. While decorators are cool, it's worth mentioning that using a decorator is much more intuitive than writing its code. The code is entirely different depending on whether the decorator takes arguments or not. The following code performs the same task as the previous one, except it is more clear and intuitive: import time from hooking import on_enter def timeit(context, *args, **kwargs): # execute and measure the target run time start_time = time.perf_counter() context.result = context.target(*args, **kwargs) total_time = time.perf_counter() - start_time # print elapsed time text = context.config.get("text") # get 'text' from config data print(text.format(total=total_time)) context.target = None @on_enter(timeit, text="Done in {total:.3f} seconds !") def heavy_computation(a, b): time.sleep(2) # doing some heavy computation ! return a*b if __name__ == "__main__": result = heavy_computation(6, 9) print("Result:", result) Output: $ python -m test Done in 2.001 seconds ! Result: 54 The Hooking library used in the code above uses Python decorators to wrap, augment, and override functions and methods. It is a generic hooking [4] mechanism which is perfect for creating a plug-in mechanism for a project, performing benchmarking and debugging, implementing routing in a web framework, et cetera. Also, it is a dual paradigm hooking mechanism since it supports tight and loose coupling [5]. The previous code uses the tight coupling paradigm, that's why the timeit hook is directly tied to the target function. In loose coupling paradigm, targets functions and methods are tagged using a decorator, and hooks are bound to these tags. So when a target is called, the bound hooks are executed upstream or downstream. This paradigm is served by a class designed for pragmatic access via class methods [6]. This class can be easily subclassed to group tags by theme for example. Here is an example of the loose coupling paradigm: import time from hooking import H @H.tag def heavy_computation(a, b): print("heavy computation...") time.sleep(2) # doing some heavy computation ! return a*b def upstream_hook(context, *args, **kwargs): print("upstream hook...") def downstream_hook(context, *args, **kwargs): print("downstream hook...") # bind upstream_hook and downstream_hook to the "heavy_computation" tag H.wrap("heavy_computation", upstream_hook, downstream_hook) if __name__ == "__main__": result = heavy_computation(6, 9) print("Result:", result) Output: $ python -m test upstream hook... heavy computation... downstream hook... Result: 54 This library is available on PyPI and you can play with the examples [7] which are on the project's README. I would like to know what you think [8] of this project. Your questions, suggestions and criticisms are welcome ! [1] https://ift.tt/YXqReyC [2] https://ift.tt/TgtH1Bn [3] https://ift.tt/14BDsWt [4] https://ift.tt/W3CHlud [5] https://ift.tt/rwm5fqT... [6] https://ift.tt/4ecjzFh [7] https://ift.tt/ROhWYVP [8] https://ift.tt/RDJouZl
3 by alexrustic | 0 comments on Hacker News.
Hi HN ! I am Alex, a tech enthusiast, I'm excited to show you a major iteration of my library for performing hooking in Python. I redesigned the whole project because it didn't not cover all my needs. I'm happy with the current iteration that I've written tests for and look forward to spending weeks and months using it in my projects. Python has a concept called Decorator [1] which is a function that takes another function and extends the behavior. In the following script, the timeit decorator is used to measure the execution time of the heavy_computation function: import time from functools import wraps def timeit(text): def deco(target): @wraps(target) def wrapper(*args, **kwargs): # execute and measure the target run time start_time = time.perf_counter() result = target(*args, **kwargs) total_time = time.perf_counter() - start_time # print elapsed time print(text.format(total=total_time)) return result return wrapper return deco @timeit(text="Done in {total:.3f} seconds !") def heavy_computation(a, b): time.sleep(2) # doing some heavy computation ! return a*b if __name__ == "__main__": result = heavy_computation(6, 9) print("Result:", result) Output: $ python -m test Done in 2.001 seconds ! Result: 54 Besides benchmarking, there are many other cool things that can be done with the Python decorator. For example, the Flask [2] and Bottle [3] web frameworks implement routing with decorators. While decorators are cool, it's worth mentioning that using a decorator is much more intuitive than writing its code. The code is entirely different depending on whether the decorator takes arguments or not. The following code performs the same task as the previous one, except it is more clear and intuitive: import time from hooking import on_enter def timeit(context, *args, **kwargs): # execute and measure the target run time start_time = time.perf_counter() context.result = context.target(*args, **kwargs) total_time = time.perf_counter() - start_time # print elapsed time text = context.config.get("text") # get 'text' from config data print(text.format(total=total_time)) context.target = None @on_enter(timeit, text="Done in {total:.3f} seconds !") def heavy_computation(a, b): time.sleep(2) # doing some heavy computation ! return a*b if __name__ == "__main__": result = heavy_computation(6, 9) print("Result:", result) Output: $ python -m test Done in 2.001 seconds ! Result: 54 The Hooking library used in the code above uses Python decorators to wrap, augment, and override functions and methods. It is a generic hooking [4] mechanism which is perfect for creating a plug-in mechanism for a project, performing benchmarking and debugging, implementing routing in a web framework, et cetera. Also, it is a dual paradigm hooking mechanism since it supports tight and loose coupling [5]. The previous code uses the tight coupling paradigm, that's why the timeit hook is directly tied to the target function. In loose coupling paradigm, targets functions and methods are tagged using a decorator, and hooks are bound to these tags. So when a target is called, the bound hooks are executed upstream or downstream. This paradigm is served by a class designed for pragmatic access via class methods [6]. This class can be easily subclassed to group tags by theme for example. Here is an example of the loose coupling paradigm: import time from hooking import H @H.tag def heavy_computation(a, b): print("heavy computation...") time.sleep(2) # doing some heavy computation ! return a*b def upstream_hook(context, *args, **kwargs): print("upstream hook...") def downstream_hook(context, *args, **kwargs): print("downstream hook...") # bind upstream_hook and downstream_hook to the "heavy_computation" tag H.wrap("heavy_computation", upstream_hook, downstream_hook) if __name__ == "__main__": result = heavy_computation(6, 9) print("Result:", result) Output: $ python -m test upstream hook... heavy computation... downstream hook... Result: 54 This library is available on PyPI and you can play with the examples [7] which are on the project's README. I would like to know what you think [8] of this project. Your questions, suggestions and criticisms are welcome ! [1] https://ift.tt/YXqReyC [2] https://ift.tt/TgtH1Bn [3] https://ift.tt/14BDsWt [4] https://ift.tt/W3CHlud [5] https://ift.tt/rwm5fqT... [6] https://ift.tt/4ecjzFh [7] https://ift.tt/ROhWYVP [8] https://ift.tt/RDJouZl
Comments
Post a Comment