How To Write Unit Tests For Existing Python Code // Part 1 of 2
ฝัง
- เผยแพร่เมื่อ 8 ก.ค. 2024
- In this video, I show you a practical example of adding unit tests to existing code. This first part focuses on adding tests while not changing the original code. In the second part (coming out next week), I'll show how refactoring the code simplifies test writing while also improving the design.
NOTES:
1) There are a few things in the test setup that are not ideal, like how dates are used in the test code, using a real API key, and doing actual card charges. I'll address these things in part 2.
2) patching means that you're replacing a dependency in a function or method with something else. The thing that you replace it with is called a mock (apologies, my use of the terms in the video is a bit messy).
The code I worked on in this episode is available here: github.com/ArjanCodes/2022-te....
💡 Get my FREE 7-step guide to help you consistently design great software: arjancodes.com/designguide.
💻 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.
👍 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
👀 Code reviewers:
- Yoriz
- Ryan Laursen
- Sybren A. Stüvel
- Dale Hagglund
🎥 Video edited by Mark Bacskai: / bacskaimark
🔖 Chapters:
0:00 Intro
1:05 Explaining the example
4:10 Analysis & Setup
6:05 Adding LineItem tests
8:40 Adding Order tests
10:56 Adding PaymentProcessor tests
15:50 Analysis of the test setup so far
17:05 Adding pay_order tests
#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!
💡 Get my FREE 7-step guide to help you consistently design great software: arjancodes.com/designguide.
btw Arjan just wanted to mention that I literally got a job offer with twice my current salary by remembering things from your videos in the technical interview. most probably wouldn't have gotten it without you. thanks so much for what you do!
That’s awesome, congratulations!
That basically means that you owe him 20 percent of that, right? :p
Grats on the job man!
Was that a course you took?
Well done and congratulations!!
That’s amazing. Congrats.
I ll be Frank, design patterns, unit tests for existing code. You are my mentor. And you put up videos which are useful in a production environment.
Thank you so much, glad you like the videos!
Hi Frank!
9:23 there are two functions called `test_order_total`. Python only sees the second one, and the first one doesn't run. This is a great example of how you can have 100% test coverage and still be missing tests!
I have my vscode configured to run flake8 and mypy. I make this particular mistake all the time when writing tests, but I get a helpfully annoying red underline under the name of the first duplicated function name.
Came here to report the same thing! Nice spot 👏
Actually pytest shows the coverage of 82% for `test_order.py`, which gives a hint. So it's not a 100% coverage.
When writing code with pytest.raises, I recommend only putting the line that you expect to raise in the block, otherwise it might accidentally catch a failure that is happening elsewhere.
Thank you, this is exactly the nightmare I have to deal with. Instead of payment processors it’s expensive computations on gigabytes of data that need to be tested. Sigh. Can’t wait for part 2
In my experience, old code that doesn't have tests is often nearly untestable. I recently rewrote a (very old) program that dealt with databases. The UI and the rest of the code are so interwoven that you can't even let it generate an SQL string without first creating a UI component that's responsible for it. That's why I didn't even bother with it. I just rewrote it from scratch.
You're quickly becoming one of my favourite programming tutorial channels. TH-cam has so much content which, on one end is all beginner programming concepts, and at the other end it's all frameworks and fancy programming features.
I think you fill that gap in between where we take basic programming features and compose them in ways that make software development scalable and maintainable.
I'm so glad you've made this video, even though I don't need it anymore.
I got a bit annoyed that there were so many Unit Test introductions, but nothing about what to do if you already have existing code, especially if said code contains a state, because all introductions have a nice, pure (meaning the function only has input and output - no sideeffects like printing text or saving files), function, which simply doesn't reflect reality.
Love your intermediate tutorials with examples + refactoring! Would love more TDD and software design patterns guides (for dummies + for non-dummies) as well as I can write code in a pretty straightforward manner but not so much with the mindset of a production setting with best practices. Also really nice setup :D
I've been programming for a while using Python, and I'd plateaued quite badly.
Your videos are fantastic, and they're constantly teaching me something new.
Thank you!
Thank you Chadd, glad you like the videos!
Arjan, you always exceed the expectations. Thank you once more for the content. The quality of the tips and the video itself are amazing!
Nice timing! I'm currently doing something like an internship where one of the things I'll be doing is to help betterify code written by my supervisor. One of the first things I'm going to be doing will be to set up tests and automate the various "did you write it appropriately".
His code is quite good in general, but he has no tests, and when I asked if he used PEP-8 or another styleguide ... the answer was "I try to make sure to keep consistent tabs".
I might not be able to help make the code itself better/faster (don't know yet!), but i will for sure be able to help make sure it's easier to refactor it and show that it's *working as it should*
you an mCoding are some of the most valuable resources I've seen on youtube in a looong time....
After 3 hours of pain googling and youtubing, finally only your video made it clear how to run test from "tests" folder and make import work. FINALLY! THANK YOU SO MUCH
Glad it helped!
Same here. I find it quite complicated to make import work correctly. I would like to ask, is it right I can't run test_line_item.py as separate file getting import error "No module named 'pay'? So far when I wanted to test my code I just run separate parts of it, but with import in siblings folders it's not possible.
Now we're all leveling up guys, great 👏🏻😊
When you know the content is going to be worth the watch, so you like the video even before you start watching. Vreeslik baie dankie.
I learned all these techniques hard way and hours of hard work. I wish I had come across your channel before. If you are a python developer, you can’t appreciate Arjans content enough this content is literally the abstract of years of hard work.
phenomenal editing job. Thanks for creating this. All content, no fluff, right to the point, no hemming/hawing, quick pace but calm delivery.
I tip my hat to y'all. Subscribed and liked. Thank you for a top tier tutorial.
Thank you so much for the kind words, Beau!
Excited for the Part2!! Arjan please post it asap...
Your videos are fantastic, and each time I am learning something new.
Great one. Good point that you can realize where the code needs refactoring from writing unit tests. The tests that check raise value error can actually Raise it from different calls (one the init of the object and the other from the call of the method) so those tests are actually not unit and are not testing in the way they are intended.
Long time watcher. This was the video i was waiting for. I'm learning how to do testing and didn't know where to start.
I just started delving into unit testing this week, so this video came at a perfect time!! Thank you so much, looking forward for part 2 :)
Coming up on Friday 😉.
You're a legend, Arjan 👍
Honestly im so greatful for your videos.
Im learning so much, enjoying my time and gertting alot of motivation to improve as a datascientist/developer.
Thank you so much!!!!
Youre awesome!
I'm glad to hear the videos are helpful!
This tutorial is very complete, thanks man!
This is fantastic, eagerly waiting for the next video!!
This was super helpful! Thanks for sharing.
Glad the video was useful!
Great videos mate. Always learning something new from them.
Thank you for this valuable content. I have watched nearly all of your videos because I like the way you explain things. I'll start leaving comments on your videos to help the TH-cam algorithm even more 🙂
Glad you like the content! And thanks for liking/commenting under the video. It’s indeed good to help the TH-cam algorithm, but I also find the feedback really helpful in improving the videos I post.
Just in ...Thanks Arjan for picking up this topic.
Arjan you are a true gem, hope to learn more about python from you! Keep it going!
Hey Samir thank you, and will do! 😉
Thank you, Arjan, for your brilliant work and helping!
Thanks so much, glad it was helpful!
Excellent tutorial.
Glad you liked it!
#11 here. Can't wait until next week. Great video again!
Thanks a lot! Very much helpful!
Glad you enjoyed the video :)
Awesome video, I really enjoyed it. It would be very interesting to see mocking applied to a database connection. I work with a 25 year old project in Perl, and a relatively younger Python code, and both have no tests at all. Both are inmense codebases and is usually seen as "imposible" to add unit testing to it. Everything is based on data stored in a mysql database, but I have no clue on how to even start testing when I have to mock so much data
Found this to be simple and effective guide!!!
I'm glad to hear you enjoyed the video!
Fantastic stuff!
Thanks, glad you liked it!
super useful video, thanks again
Amazing I love it, thanks for doing what u do
Thanks Arjan for this good video.
Glad you liked it!
Thanks Arjan, thisvideo was super util for me.
Glad to hear it was helpful!
I hope you bring Mock into the conversation! Easier to use than writing lambdas and mock functions. Mock(side_effect=inputs) would have been a simpler way to monkeypatch the input calls.
Mock is a class not well understood, I've had to teach other engineers about it many times. Well worth learning about if you're serious about writing test code! It's very powerful to use in conjunction with monkeypatch.
Hey, @Joshua
That is indeed an interesting functionality. Do you have any material to share?
I have been using patch and magicMock, but feels so unnatural.
Every time I have to Google/read up on some examples again
great video thanks!
Awesome stuff man, as always!
Thank you!
This is a super mega important topic, thanks for the video! Test order total is defined twice, I think, so only one will be exercised
Also monekypatch is a rather exotic path. Why not mock.patch with side-effect for the 3 different return values?
What a great discovery your channel is! Thanks a lot for all videos you are doing, fabulous job! Many greetings from Ukraine 💛💙
Thank you, happy you’re enjoying the content!
Thank you Arjan, really glad that somehow TH-cam algorithms showed your channel) your content is very helpful! Nice
Thank you Jury, glad the content is helpful!
These videos are great.
Thank you Jacob, glad you like them!
Great video!
Thanks!
etymology suggestion for monkey patch: 1) guerillia -> gorilla -> monkey; 2) monkeying about with code (en.wikipedia.org/wiki/Monkey_patch)
Arjan, I always enjoy your videos. You really do a good job of pacing the complexity from simple to advanced that makes learning here so appealing. If I could request that you follow up these videos with a "how to write unit tests for existing python code using unittest and mock/patch" that would be great. my organization recently switched over from the simpler pytest to now using the more complex unittest lib as standard, and the abstractions for mock and mocked classes for testing really make it hard to grasp exactly whats going on sometimes. Great video anyways
Thanks!
Best python content here on YT. Also really like that shirt you're wearing. I would like to know where I can get one if you don't mind.
Sir, You're a genius!
Thanks, happy you’re enjoying the content!
My professor assigned us to make unit test without ever telling us what a unit test is. Thank you for teaching me. I regret paying for college now lol
I swear you have a hidden video camera in my office. You keep making videos on exactly what I am working on.
Excellent video, I've been following it for some time and it's adding a lot of knowledge to me. A video on how to apply Mocks in situations involving API requests and database connection would be interesting.
Great suggestion Igor, Thank you!
I must say that I'm a little disappointed that he starts by understanding the code, and the method isn't a thousand lines long.
That said I did learn a lot and I love this channel thank you for all the amazing content!!! Thanks for testing videos!! Please more!!!
very informative ! , I hope you'll cover also the type checking with mypy
Very good video - as always :-). Just one remark: If you don't want to have some tests suddenly fail on 1/1/2025, you probably want to use either something like the freezegun module or set the card info dynamically :-).
Hi Marc, I’ll address this issue in part 2 this Friday.
really good video. thx!
Thanks, glad you liked it!
I'd like to know more about your choice of pytest over unittest. Are there particular factors that lead to a preference, or is it just what's more familiar? I've been working on some projects that mix bits from both pytest and unittest, and it's kind of a mess but I don't know which one to go with. There's things that annoy me about both. :D
Thanks for the video, wasn't aware of monkeypatch, looks much easier to use than the other patches available to pytest and unittests.
Had just one remark, be careful with only checking the exception type, especially if the function being tested has multiple of the same exception types, you had 2x ValueError in the same function, but could've triggered the wrong if statement without knowing.
I prefer to also check the actual error message within an assert, to make sure that I'm triggering the correct ValueError.
That's a great reason to use custom exceptions, and create a different one for each fail case!
The monkeypatch example was particularly interesting. I've only used it some to test a method that called input function. However, I modified sys.stdin to place my input strings rather than replacing the input function. Is there a preferred method or are they considered relatively equal?
I hope the code quality of tests in part2 will be much better, for example many many cases in here should use fixtures to not repeat code or create classes itself inside every test.
Yes - I’m covering fixtures in part 2.
Super interesting content. I have a question (open to anyone):
I see that in testing with a "valid" card the expiry date is being hard coded. Would this not be a problem in the future, when the hard coded values become a "past date" and the tets are run again (and fail)? Thank you!
PS (after watching part 2/2) My question is answered there!
I'm not doing this professionally, so I really doubt I will do many tests... But knowing how to is nice
Depending on the size of your growing project tests will help you keep your own code stable while growing and extending it. If your feature set is very volatile tests will feel like a burden and overhead.
Cool video topic!
Thanks!
It seems that monkeypatch is even more powerful than Java's mockito in some aspects. Nice video.
Thanks!
Didn't know I could actually write a comment with the purchase 😅 I Just wanted to give you a huge shout-out! I've been learning and growing in my engineering career for some time now and your videos have been a fantastic resource for me! Thank you for sharing your incredible expertise!
Thank you so much! I'm happy to hear the videos are helpful to you!
I'd love to know where you got that cool geometric led lights behind you. Also, big fan of all your content!
Really great video! I love to see things about testing :)
Wouldn't it be better to put only the line which raises an error into pytest.raises (like in try/catch) or why is all the code inside the block?
Thanks Nico! Yes, good point, it would be better to only put the line that actually raises the error in the with statement, as this focuses the error detection better.
Great video! I'm catching up with your videos, so maybe you've done that already, but it would be a nice follow up to show how to create the same payment processing system and its tests if we were writing it using TDD from scratch, probably using dependency injection for the input and the payment service and therefore not need to monkeypatch
Thanks and love that suggestion!
Very interesting video, can't wait for part 2!
You mentioned writing tests before the actual code. I wonder, how would this work for functions, that are computationally heavy, for example because they do some complex image processing, or perhaps create a 3D scene? How would you provide the correct result for the assertions? It's hard for me to imagine how this scenario would work.
I'm only a beginner when it comes to programming, so I hope my question makes sense. :)
Hi Dávid, if a function performs a very complex task, you would normally split up the function into several smaller functions that you can then test separately. For parts of your application that produce a complex result like a rendered image, a GUI, etc, a common technique used is snapshot testing. You generate the correct complex representation once and indicate this is the correct ‘snapshot’. When you change things in the code or add new features, the snapshot test verifies that the resulting output is still the same (or if it’s supposed to be different, you generate a new snapshot).
@@ArjanCodes Got it, thanks for the explanation Arjan!
Will the branch coverage (additional to line coverage) be in part 2?
Can the coverage metric also be shown directly in the editor?
You're all dressed up these days. I need hacker hoodie back once in a while
When should I use assert vs self.assertEqual?
Hello thanks for this video really helped... but is it possible to do a real payout system tutorial with stripe putting all these tests together ?
Can you please suggest what is the right way to put constants (like error msgs, log messages, other strings used)?
What is the best practice to use constants in python :
(1) a separate .py file for storing strings
(2) Enums
(3) Dictionary
(4) Named Tuples
(5) Data classes
(6) anything else
I got my first software testing job were we use pytest to automate the API testing.
I am thankful that youtube recommended your video because i would rather learn this than watch Netflix :D
Thanks for putting this out for free 🙏🏽
Glad to hear it’s helpful!
Hi Arjan, Are these tests possible when my application built on streamlit?
That's what I need now 🐱 I wrote the code, but not unit test for it: (
Awesome. I am from mexico Your channel is fucking great. And thanks for sharing your knowledge. Keep the good work.
Hi . Would be awesome if you could make a video about agile principles. Thanks in advance :)
I would have also "asserted" that when the order item is created, its created with Status == OPEN
Good suggestion!
Hey man - moved house and I need to redecorate. Where did you get that thing on your wall? I love it!
Pytest is easily my favourite testing framework! And my favourite feature are fixtures! So easy to create and you can do all sorts of cool things with them too, parametrize, etc... What I am still struggling with is to do true TDD. I most often find myself having to do some exploration before being able to think what the tests will look like. Any ideas for that?
I often start by thinking about the program flow and writing tests for the "intended flow" first. Then I think about what can go wrong and write tests for that. It's also important to remember that you can't test for every case, so do not feel too bad for missing something. Also, restricting per module helps a lot.
Like in this video, the flow is "You have an item, order it, and get charged for the order.". So you first write a set of tests for the items, then the order, and finally the payment processor.
When writing tests, the most essential part is not thinking of the implementation code but instead, "How would I interact with this object/function/...?".
At least, those are my opinions/tips.
@@veni_vidi_victorian good tips, emphasis on the last part. Thinking about how you would interact with it is more important than the implementation, good tests should also be able to tell a story of how you interact with the code base
This is a great question and often asked. I would suggest exploring an idea completely separately (like for example in a Jupyter Notebook) and learning about API’s or error habdling, and then returning to TDD. There are many ways … 🖖
Hi Arjan, would be nice to know how to test very complex clases like the ones that usually appears in Deep Learning. For example, pytorch_lightning.LightningDataModule... things like the shape of tensors in several configurations, training loops, etc. I don't really know if it's worth to unit test somehow these kind of things. Currently, I'm placing asserts in my the code itself but not unit testing systematically, would be nice to know at least your point of view about this topic!
Waiting for it!
You are quite old, but you are not ancient. You have adapted and conquered so many changes, that's very impressive. Beside that the information and the enterprise ready information is just on another level. I no longer believe that programming has an age limit, you are so skilled at teaching and so skilled at technically coding. FUCKING Brilliant.
Hi Arjan. As always, it was a great video! Thank you for it. I noticed there were 2 test functions with the same name "test_order_total()". Is it ok to do that when writing tests?
No, it's not OK. It's a bug caused by copy-paste, and it prevents one of the tests from running.
Nice video, thanks. I do have a question, though:
14:17 I don't understand what "test_card_valid_date()" is actually testing, given that it has neither an assert statement nor a "with".
In this case it is testing whether an exception will be raised when calling the charge method. Since the date is valid, the method should not throw an exception. The test passes if it doesn't.
Excellent example software -- too often I see horrible example code when talking about testing (which leads me to conclusions I will not say). Question: what are your thoughts about starting with writing acceptance tests for brownfield projects?
Hey Arjan, love your videos :) any way you could comment on writing testable code for data science? I find it hard to write unit tests for, for example, the result of a more complex data processing function or the result of a training job.
Thanks! A general tip to write code that's easier to test: whenever you can, treat data as immutable. So: read data, process it to create new data, but don't directly modify the data. By keeping data immutable, your processing functions are going to be easier to test since they are going to be mostly pure functions. Hope that makes sense. It's an interesting topic, I might do a separate video about this.
does anyone know which VSCode appearance theme is being used? :O
how do i test functions that returns unknown data? like fetching from an api? or functions that do really complex calculations and you cannot know the result before running it?
Mocking for non-deterministic responses or validating the HTTP response code
Could you please tell
(1) how to do Unit Testing for decorators - both methods and class based?
(2) Flask's request module throws lot of context related errors while patching how to resolve those?
Hello Arjan, how does it look like if you've used "unittest" built-in lib? And why do you prefer pytest? Thnks