Видео

I'd guess if you have a significant conftest setup that xdist would also slow down significantly, since it needs to run conftest for each process

GREAT video with super useful tips; thank you!

2 things that could have been mentioned: - pytest --cov=my_package --cov-report=html : it outputs an html file which color codes coverage and is quite nice visual representation of cov - might not be strictly pytest related, but configuring the testing extension in VS Code enables the tests to be run individually by just clicking on the green "play" button to the left of the functions (next to line numbers), which is super nice, especially when you are writing the test and don't know if it will work, you can just click that button over and over again instead of going to terminal and starting the whole unit test session with the commands.

2:31: one question: does the test cover the free space between the bananas? 😉You test everytime against your own assumptions... not against the universe.

When I write pytest tests I try to pull as much setup/teardown code as possible out into fixtures so the tests themselves don't get obscured by loads of setup/teardown. One slightly sneaky trick is using fixtures to set defaults for other fixtures and overriding them with parametrize. So maybe I have a fixture 'user' that creates a user object used by lots of the tests and that fixture takes another fixture 'permissions'. Most of the tests expect a default set of permissions maybe 'read-only' so the permissions fixture sets up that default. Then tests that need a different set of permissions in the user object use `@pytest.mark.parametrize("permissions", [("read-write",)]` and pull in the `user` fixture with elevated permissions (you don't have to reference the overridden fixture directly in the test, pytest will complain if you parametrize a fixture that isn't used by the test but it's fine if it is only used indirectly). That works well if you only need to override the default in one or two places: if you need to do it more often then you wrap the relevant tests in a class and either redefine the fixture inside the class or use parametrize on the entire class or use multiple files with the defaults in conftest.py and override them as needed in individual files.

This is awesome haha, congratulations on 5 years mate!

What vsc theme is this? Looks nice

Just use Taichi.

i tried Pylance, mypy and pylint. But the Ruff the best what i used

Ooh, not sure yet but we use some matrix multiplication functions at work to optimize insurance policies. I bet there's at least a few places numba could significantly speed things up! Right now some larger scenarios with larger input sets take multiple MINUTES to run, if some of those could be brought down to seconds it would not only make our web application more user friendly, it will save a bunch of money in Google cloud function costs. I'll have to do some experimentation and see if they're willing to pay me for the improvements haha!

Better way to do this: numbers = [1,2,3,4,5] total = sum(numbers)

I used numba on my Mandlebrot code - 45x times speedup per core for a total of 180x speedup

"source.organizeImports": true gives error, saying it expects string

*Sometimes 😂 Thank you, dude😊

have you heard about sum?

you are a great lad

Simple and good, thanks!

While it’s neat, it was removed from the builtins in python 3 because it’s not very performant, and not that readable (therefore not very pythonic). This example you’ve written is proof of that, because you could just call sum on the list.

That was a great explanation. I can see that this a private repo. Could you please share the source code in a public repo.Thanks..!!

This description is wildly confusing. It would be better to discuss the third optional parameter which is the `initializer` which defaults to the first value of the iterable

If you use sum it is not so much ease?

"So many functions are under this hood" should be the title. I think you made a typo.

this isnt quite how filter would realistically should used, its if you're going to loop through something and want to filter your iterator to avoid some logic inside your loop, this is exactly where list comprehensions is supposed to be used though, apples to oranges cprofile is a good thing to use for actually finding speed issues as it sees which functions is adding time to your program, this is just benchmarking (which should include variable length lists)

Hey. For benchmarking in python, I highly recommend the Foss project pyquickbench

Thanks for explaining with the application along with the Syntax 🙂

Thanks for the video. Why do you use the minimum instad of the mean or median to compare both approaches?

Thank you 😊

If you do it again you might try creaming your butter and sugar together first. Then the dry and liquid ingredients get mixed separately before combined. The cream doesn't have to be hot before going into the tempered chocolate if you add it in very slowly and mix the whole time you are adding it. If you have too much egg I would mix it first before pouring any out so it's homogeneous first. - source: guy who watches a lot of cooking shows but never bakes himself

You can measure liquids more accurately using the scales. Density of double cream is apparently about 1.01g/ml so 54ml would be 54.5g. Beaten eggs have a density between 1.03g/ml and 1.1 (weight required left as an exercise for the reader). Nice little Denby bowl btw, we have a full set of those and it's the first time I've seen anyone else with that design.

thank u so much mate, really helped me out

Does it include Michaelmas?

So this is a tradition now right? I will mark my calendar 🎉

Congrats Carberra! Keep up the good work!

A short worth while saving🎉

Congrats on reaching 5 years, here's to another 5 more!

Happy birthday!