Python Decorators: The Complete Guide
Vložit
- čas přidán 11. 06. 2024
- Python decorators are a great way to add functionality to your Python functions. In this video, I'll show you what they are, how they work, and some of the most useful decorators you can use in your code.
Git repository: github.com/ArjanCodes/2023-de...
✍🏻 Take a quiz on this topic: www.learntail.com/quiz/iufnfs
👷 Join the FREE Code Diagnosis Workshop to help you review code more effectively using my 3-Factor Diagnosis Framework: www.arjancodes.com/diagnosis
💻 ArjanCodes Blog: www.arjancodes.com/blog
🎓 Courses:
The Software Designer Mindset: www.arjancodes.com/mindset
The Software Designer Mindset Team Packages: www.arjancodes.com/sas
The Software Architect Mindset: Pre-register now! www.arjancodes.com/architect
Next Level Python: Become a Python Expert: www.arjancodes.com/next-level...
The 30-Day Design Challenge: www.arjancodes.com/30ddc
🛒 GEAR & RECOMMENDED BOOKS: kit.co/arjancodes.
🎓 Sign up to Brilliant now and get 20% off the annual subscription: brilliant.sjv.io/arjancodes.
👍 If you enjoyed this content, give this video a like. If you want to watch more of my upcoming videos, consider subscribing to my channel!
💬 Discord: discord.arjan.codes
🐦Twitter: / arjancodes
🌍LinkedIn: / arjancodes
🕵Facebook: / arjancodes
📱Instagram: / arjancodes
👀 Code reviewers:
- Yoriz
- Ryan Laursen
- James Dooley
- Dale Hagglund
🎥 Video edited by Mark Bacskai: / bacskaimark
🔖 Chapters:
0:00 Intro
2:20 Object-Oriented decorator
9:40 Functional decorator
17:17 Functools wraps
21:20 Decorators with arguments
24:03 Should you use decorators?
26:21 Outro
#arjancodes #softwaredesign #python
DISCLAIMER - The links in this description might be affiliate links. If you purchase a product or service through one of those links, I may receive a small commission. There is no additional charge to you. Thanks for supporting my channel so I can continue to provide you with free content each week!
The way you progressively take the viewer through the level of abstraction is one of the best parts of this, and many of your videos.
Yep, exactly!
Agreed. While I really enjoy James Powell videos like the one below, @ArjanCodes just presents things in a very fluid manner.
czcams.com/video/cKPlPJyQrt4/video.html
My colleagues were lost with the content presented in that James Powell video. But I hope that no one was lost with what @ArjanCodes presented in this one.
Good job @ArjanCodes! Thank you.
I've consumed a lot of programming content recently and this channel is one of the best I've seen. Well produced, professional and perfect pacing. Even the occasional joke!
Thanks so much!
".. a pain in the class.." 🤣 Added to my sayings... Thanks Arjan!
Great stuff. I have never understood decorators... Until now. It was so nice how you showed step by step how it works. I thought decorators were only python thing but now I see its a general concept doable in all languages
Glad you liked it!!
"@ArjanCodes"
It even works on CZcams. I'm just trying to wrap Arjan's Glad-you-liked-it to decorate my comment.
This is just gold on youtube. So many junior devs don't understand decorators in Python, and think of them only as this "Magic" that you put on top without understanding anything. Thank you, Arjan.
Thank you!
Hi, Arjan! Can you make a video about "module/package structure" in Python? For instance, using a project as an example, what classes should be grouped together in which modules/packages (nomenclature, subjects, etc). Regards!
I totally support this idea!
For me it is still a bit of a mystery how to BEST create a library/package from your own files...
^
Yes x1000. While the syntax isn’t too hard to get my head around, best practices is always difficult to get a handle on. And most info on web is for open source projects. It’s actually quite difficult to get an idea of how to robustly share internal packages with the rest of your team in a non-open way.
+999 to this :)
agreed. i'd like to see that too
This dude i can tell came from Java/C++ world he mastered OOP - this is a great for a seasoned developer. Thanks bro.
One of the best channel for learning python best practices. Thanks alot Arjan!
Thank you, Tim, glad you find the content helpful!
THANK YOU 😊 Wanted something on this topic and just in time for lunch break today ! Gonna use it immediately
Turtles all the way down. Glad you mentioned the "harder to read" downside. Excellent presentation of a fun tool.
Thank you, glad you enjoyed it!
I know the decorator pattern from top of my head, but I use python decorators so rarely I have to re-learn them every time :)
Another good use case for decorators is error handling. We catch different exceptions in a web app on different layers and turn them into http errors.
But overall decorators are a niche thing.
It's literally like you're spying on my current challenges at work! There's a heavily nested decorators function that is being used and this has cleared things up a lot! Thanks so much, keep going with the great work!
Thank you and will do!
Welp. I guess I now know exactly what I need to rewrite part of a work project. There's some complicated and annoying interdependencies that I am not a fan of, but I hadn't figured out how to solve it.
Also, thanks for explaining something that as far as I could tell were just magic!
You're welcome, Marie! :)
Omg, thought how to combine music with your videos - and here it comes. Much easier to switch from coding to learning now, nice feature:)
Huge pain in the 'class' haha! Very nice!
Really nice video! I discovered Decorators in my new job (C++), but we use the traditional pattern because we can change the order and list of them by the DI configuration tool or inside factories. Amazing how things like anticipation, sync or cached operations are decoupled from the model object.
One of your best videos yet
cette chaine est une mine d'or ! 🧡
Decorators are truly a double edged sword. I have used them in the past to pass on a database session through multiple layers of controllers in a rest application. The decorator took ownership of the session and decided when to open / flush / commit / and close the session. It also added some parameters to every controller function in order to implement some security features. It worked wonders for our project as even after countless controller calls a session could seamlessly revert to an old state if an error occurred and controller functions were automatically secured, but to this day it confuses some developers when working with it, because not everyone understands python decorators well. Though I still stand to that design choice today.
I came for the high quality and well thought out content. I stayed for it too. But I loved it for all of the glorious puns. :D Thank you for the education and the puns. Both are much appreciated!
Glad you enjoyed it!
This channel has the most useful python tutorials.
Easy to follow and very helpfull!
Great to hear!
Just way awesome than any video. Thank you loads Arjan ❤
Thank you so much 😀
Amazing content! Thanks so much for sharing that.
Thanks for the excellent and easy-to-follow explanation! Brilliant teaching :)
I'm glad to hear that the video was helpful :)
Thanks for the lesson Arjan. Very well done.
Glad you liked it!
I definitely agree that wrapping logic in decorators makes code harder to read, but I love using them to register handlers, like how Flask defines routes.
I use it like that all the time for event handlers, and I just think it makes the code super easy to read and understand.
Love your videos, so happy to see someone else who loves Python as much as I do 😁
BTW, flask's decorator functions also return the original function unchanged. So even though a function is a flask route, you can call it just like a regular function if needed.
Flask's types were broken regarding this fact, and I had to fix them sometime last year.
JIC: we also can implement decorators using class with `__init__` and `__call__` method (`__init__` for parameterizing and `__call__` for applying deco to func of class)
And in addition we can decorate not only functions but classes as well
I have watched a few videos on this topic and you explained the best. It's so easy to understand with some background (classic decorator pattern in OOP and then how Python gets it done). Thank you.
Glad it was helpful!
Easily the best presentation on decorators I’ve ever seen. 👏🏻
Wow! Thanks so much 😊
I started watching recently and I have been pleasantly surprised by the quality of your content. There is one caveat, I think, to using the @decorator syntax of python and that is if you need to provide different parameters to the wraper function in different moments. Since the decorator is set at the beginning of the definition of a function, you would either define a function multiple times with the decorator with updated parameters, or use one of the other ways to use a decorator as you showed in the video.
Realized now that Arjan codes is one of the only channels that when you ask for a like i actually stop and like the video. Thought goes like: "that's so useful, it really deserves a like" hehe
btw: i would like to see your deep dive on the import system ...
Thanks for your support! I noted your suggestion :)
Exactly the video I was looking for. 🎉🎉
Glad I could help!
Finally understand how decorators work in detail
This is the most practical advise on decorators i've ever seen or heard.
Thanks!
I'm learning from you a lot. Many thanks!
My pleasure!
Great video. For anyone wanting to use the with_default_logging decorator, be aware of a potential issue. Most likely, the logger object will have the name from the module where it was instantiated. If you import it via decorator into another module, it might get confusing where the logging call actually came from because the logger name can be mistaken for the module name. I definitely had the issue in the past and it took me a while to figure out what was going on.
Liked the video when I saw the like video button light up as soon as you talked about it. Nice!
thanks this super insightful guide! much appreciated
Thank you!
An amazing tutorial!
Thanks
You're welcome!
In my previous work, VW group, decorators were used to decorate classes and decorators created instances of classes that we wanted to use in the class. So they were some kind of Factories for other classes. I had no idea how that worked at the time :)
I liked the first part of your tutorial. It reminded me how I used to decorator before java 1.5, and instantly boosted my learning speed.
Thanks for sharing!
Absolutely great video
Thank you so much!
My python code always looks like it's straight out of the 90s and yours reminds me of something from another time completely. Definitely should brush up more :)
As a functional programmer I love the functional decorator, combinators are good.
Simply an amazing video 😌 Thank you.
Glad you enjoyed it, Diego!
Thanks. It is a great video explaining how decorator works. It is a good feature of dynamic programming language.
Thank you for the kind words! Glad you liked the video.
Thanks Arjan!
You’re welcome!
I had a hard time understanding decorators in my high school days. Never created my own, but I think I might have to reconsider now
great video, thx a million, making me revisit those ideas, i tend to forget ;)
Thank you, glad you liked it!
Superb, like always.
Thanks so much!
OH MY GOD IT FINALLY MAKES SENSE
Very good!
Thank you!
powerfully awesome
Thanks, i like the path from classic patter to python’s decorator style.
Glad you like it!
I wrote a decorator for a web application once that enforced authorization. It would get the user session object from the coroutine context and verify that the user possessed the permissions defined in the decorator call. e.g. @secure('write_permission')
Great videos btw! I would appreciate a video on decorator classes. I've always found them very hard to understand how they work and what the pros/cons are. Oh, and a video on the coroutine context! Thanks!
Great video as always, I use the built in decorators but Ive never reach the point of making my own. It looks awesome but in practice I feel its a bit convoluted as an end user. I will definitely will try them out now. Thank you for the great video.
Glad you liked the video! 😊
Thank you so much for the videos!
Could you use more visuals (drawings) before writing the code? I guess if you draw what you are trying to build before building it, this will deliver the message perfectly
I just started getting into decorators, and this is such a comprehensive guide on what the hell a decorator is, thank you so much
You're so welcome!
For me the option that makes most sense is to use functions in the decorator design pattern. That way, if you're benchmarking code (say, in a jupyter environment) you can called the decorated functions of methods, but when you're deploying the code in a script that needs to run for hours and no one will look at, you can call the function without the decorator, thus leaving out a bunch of logging and calls to perf_time()
Love it!
Thanks!!
22:38 You shouldn’t have bothered with functools.partial. Just do
with_default_logging = with_logging(logger)
Danke!
Happy you enjoyed the video, Andreas!
He is back!!!
I recently implemented a version elevator system using decorators. The version elevators were just simple functions that take in a config object and modify it in-place updating it to the next highest version. The decorators were needed because some versions needed multiple mending functions due to inconsistend versioning plus we also had some functions that were just there to provide some convenience options in the configs. As such, the decorator would take the version for which the mending function should be applied as an argument or None it is a convenience thing. The decorator would then take care of adding logging and registering the mending function. Now, outdated configs can be updated by just running all the version escalators on it in the given order, e.g. if the config is version 3 and the current version is 5 it would be 3->4->5 and then run all of the convenience stuff. I think the decorators make the code quite tidy since all you see is really only the migration logic. I would recommend using the class-based approach with `__call__` when you want decorators that take arguments though. I find them easier to read than three nested functions and it makes dealing with defaults easier as well.
I wanted to ask a question but writing my question answered it, thanks
I was trying to accomplish a decorator in a class and couldn't make any sense of it until i saw this video.
Arjan explains the best and most generic way of achieving a decorator.
Thank you so much!
you are gods gift for me
thank you very much.
are not you planning for doing some projects in python.
One of the best python video makers ever.. Keep going sir.. 😁 I have been following since so long, I'm a machine learning engineer and I come here once in a while, and always watch the latest ones to be sure my python skills are state of the art 😎
Thank you so much for your support!
6 mins in and im completely lost🥲
We’ll try again
One more fantastic python guide. ❤️
You make it easy to learn.
Are you also going to tackle another programming language, like javascript?
Or will that hurt your CZcams algorithm too much?
I expect mixing programming languages will hurt the channel. But I might start another channel in the future that focuses on a different programming language.
Great video, I use decorators all the time, but only the cases as you show in the video benchmarking and logging start of end of a method call.
Never use multiple decorators, if i need both timing and logging, i create a single decorator that does both
this is not always the case. there are quite well known frameworks which use stacked decorators frequently as middleware.
also order matters. if you have a number of decorators, do you want to write a decorator for every combination of them?
One notable case where I built my own decorators was to build a mini-ETL framework. I was using decorators to inject data validation checks pre and post transformation steps.
Also reused the @nested_dataclass trick someone posted online to be able to create dataclasses with nested dataclasses members off of unpacked dictionaries.
Do you have a link for that by chance?
I really like functional programming and specially haskell. I often try to see how to reproduce haskell behavior in python for fun (and sometimes profit). Decorators were the obvious choice to implement currying. The not so obvious choice used by functools.partial (class) is much better, but requires more work to work properly.
I use decorators mostly for testing and debugging. Having a dedicated decorators for benchmark and logging produces less clutter and I can simply remove the decorator if I no longer need it. I also use my own @NOT_IMPLEMENTED decorator to automatically raise a NotImplemented exception and print message to the console, so I don't have to write the code manually in every function over and over.
Otherwise I try to stay away from decorators (unless it's property, dataclass or pytest fixture).
Thank you so much for your hard work and the effort you put into your channel. Great video!!!
Quick question/ idea for another video
Could you describe the difference between Decorator and Proxy Design patterns?
Thanks in advance!
Great suggestion!
Great Video Arjan. How would you implement an async decorator for async functions or a decorator that can be used on both sync and async. Still found a solution that "feels" right
haha djeez 'pain in the class', you outdid yourself. Angry upvote.
Everyone else is giving proper accolades for the video, so I just came in to say well done for "Pain in the Class" 😂👏
Haha, thank you Chris!
Amazingg
Several years ago the term "aspect oriented programming" was in vogue. Decorators seem to me _the_ way to implement AOP.
wow real magic
I was hoping that you would talk about ParamSpec near the end to be able to preserve the argument types of decorated functions for better autocomplete :)
Note that if you want decorators to support both sync and async functions, you can switch based on the value of inspect.iscoroutinefunction()
I usually use the decorators from packages, I didn't had the need to create stuff for myself.. but watching this video, I see there is a lot of small stuff I do a lot in my code that could be an decorator.. just low level stuff like error handling and logging..
The most useful tip with decorators especially in-case where a decorator handles input, for example in authentication the best
possible way to handle invalid input is to raise an Error : raise UnAuthenticatedError(message) then handle the error else where on your code, any other means of handling invalid arguments or invalid credentials is bound to cause problems with Flow.
for example if you handle invalid arguments by passing along a variable or credentials , then if they are invalid and you forget to handle the case where the credentials where invalid in the method or function in which the decorator is being used, you may not find this bug... straight away,
Yes, good points!
You have my decoration.
Thank you, that brightens my day! 😎
For logging I use loguru - a neat little library that saves me some time
Tak!
Thank you for your support, Cassidy!
I've only used a handful of decorators in python, but what I'm curious about is what @abstractmethod brings to the table here, in the first section of the video. I'm familiar with the idea that declaring something abstract (in languages that allow it) prevents you from creating a subclass without providing your own definition of that method. Is that all that's being done here? I admit that not usually being a library programmer, I tend to just document what things need to be done rather than build out a hierarchy to make the compiler enforce it.
Hey, Can you make a video about the difference between decorator and proxy design patterns
Lovely vidoe as always
here is a video idea for you: How to Code inefficiently.
Haha, good suggestion, with some reverse psychology :)
Brilliant video. May I ask which IDE is that - that run button and auto switching to terminal is quite nifty.
Thanks! I'm using VS Code.
I use decorators in fastapi for user's token authorization.
Is it common/good practice to allow exceptions to be raised inside a wrapper function?
This is extremely interesting, I am working on a Flask 2.0 project and would like to know if you have any videos on Flask. Search did not find anything apart from GarphQL VS Rest. Thank you.
Haven't done a video on Flask, but soon a video will come out on FastAPI.
@@ArjanCodes I look forward to it, thank you very much.
I notice you only talked about function decorators, not class decorators. They have their uses, too.
I think the most extensive use I have made of custom decorators so far has been in my DBussy package, a pure-Python wrapper for libdbus. The information defining the D-Bus interface to a Python class is specified via a function decorator on each Python method implementing a D-Bus interface method, with a class decorator to tie it all together. Then when you register your class to receive messages, the dispatch mechanism knows how to decode incoming D-Bus method calls, convert the arguments and dispatch to the appropriate Python method.
As a bonus, this same information is used to automatically generate the D-Bus XML introspection format, when another D-Bus peer sends you the introspection message. And conversely, when you introspect another D-Bus peer, this info is used to dynamically generate a “proxy interface class” -- a Python class whose methods simply send the appropriate D-Bus messages and return the responses.
What if I want to decorate a function that I don't have access to? E.g. there is a lib_function from some library and I want to add a decorator to it?
Decorators are a wonderful way to challenge people to explore the more abstract tools in Python and their also wonderful for making DRY-er code.
I recently used decorators to support a test structure, where the decorator would display information, e.g. pass/fail the return of all the test, given the test methods only returned bools.
hey Arjan.
can you share your thoughts on using result tuples such as the one below in python? good idea or bad idea?
```
def result(func):
def safecall(*args, **kwargs):
try:
return (None, func(*args, **kwargs))
except Exception as e:
return (e, None)
return safecall
@result
def div(x, y):
return x / y
X = [1, 1, 1, -1]
Y = [2, 0, [], 2]
for x, y in zip(X, Y):
err, res = div(x, y)
if err:
print(repr(err))
continue
print(res)
>>> 0.5
>>> ZeroDivisionError('division by zero')
>>> TypeError("unsupported operand type(s) for /: 'int' and 'list'")
>>> -0.5
```