I would swap Python with C++. Constantly dealing with stupid runtime errors that could’ve been easily captured during compile time.
Did you forget to rename this one use of the variable at the end of the program? Sucks for you, because I won’t tell you about it until after 30 minutes in of the execution.
Seriously though, writing a monolith of a function and not testing anything until you run it the first time isn’t the way to go. Even with a compiler you’re only going to catch syntactical and type issues. No compiler in the world is going to tell you you forgot to store your data in the correct variable, although it or a a linter may have helped you realize you weren’t using it anywhere else.
Agreed. Mypy pre-commit hooks are very useful if you’re starting a fresh project. Adding typing to an existing project which reuses variables with different types… We lost weeks to it.
It doesn’t need to have large complex scopes. I have the brain of a goldfish. I program because it’s challenging. It’s challenging because I’m bad at it.
Mostly number crunching and data exploration tasks. Just so I can make informed decisions about the data I got. I do this rarely enough so it hasn’t been worth for me to install all these extra third party support wheels.
Shouldn’t be forgetting for one off scripts either, if that’s the logic you want to go with.
The tool exists, either you do it or you don’t and end up getting an error until the interpreter hits that line. It’s just the nature of being compiled at runtime.
TDD only works well if the problem is clearly specified before the first line of code has been written, which is rarely the case when I need Python for something.
So which is then? You want a one off script to just quickly crunch some numbers on a problem you still need to understand? Because that is where it is perfectly normal to get some errors and doodle around. That is the entire point of it.
Or you have a concise concept of what you are going to do, and how and why? Because that is what you do, when you program more than a “one off”.
Either you go to the store with a shopping list and you work through that list or you go browsing and see what comes up. But don’t expect to be as fast and have everything you needed, when you dont write your shopping list at home.
Often I use Python for exploratory purposes. Like, I got a bunch of data, and I want to know if a particular algorithm might work or not. I implement the algorithm, but realize the results don’t look good enough. So I tweak the algorithm, maybe even do major refactoring. Or maybe I realize my visualizations or metrics don’t capture what I need to see. Or maybe I must settle for some compromise?
I iterate on this repeatedly until I find something I’m happy about (or until I give up). Sometimes I end up with something completely different from my initial idea.
TDD won’t help me much here because the end result is unknown. For each iteration of this idea process I might even need to rewrite all the tests because none of them are valid anymore.
Then maybe use an editor with a decent linter and check the problems tab or just red line markers?? I also have those kind of runtime errors sometimes but I take the blame.
Catching some errors is better than catching no errors. No compiler in any language can protect you from all runtime errors either way, but some are better at it than others.
I’ll take Python’s “program exited four hours into runtime” over C++''s “this memory is uninitialized so it’s less than, equal to, and greater than zero at the same time, hope you never try to do any of checks on it!” undefined behaviour any time.
At least with Python, you know something went wrong.
I guess as a C# guy I’ve never had to deal with an issue like this. Most of the time the exceptions are pretty easy to diagnose unless it’s in the UI or in some async function.
I would swap Python with C++. Constantly dealing with stupid runtime errors that could’ve been easily captured during compile time.
Did you forget to rename this one use of the variable at the end of the program? Sucks for you, because I won’t tell you about it until after 30 minutes in of the execution.
Yeesh. I mean, perl would tell you about that immediately, I’m just saying… :-P
Yeah, but then you have to use perl
Using perl is not the problem, now trying to read perl code later? That’s the challenge! :-P
You can solve this with git:
Seriously though, writing a monolith of a function and not testing anything until you run it the first time isn’t the way to go. Even with a compiler you’re only going to catch syntactical and type issues. No compiler in the world is going to tell you you forgot to store your data in the correct variable, although it or a a linter may have helped you realize you weren’t using it anywhere else.
Python was typeless. And it was common to reuse variables with different types of content.
So you at some point never knew what actually is within the variable you are using.
Using typing in python solve 95% of your problems of having runtime errors instead of compile errors
Agreed. Mypy pre-commit hooks are very useful if you’re starting a fresh project. Adding typing to an existing project which reuses variables with different types… We lost weeks to it.
My brother. That’s why you do unit tests.
And lint
👆 definitely linting first 👆
finding errors as you type is even better than finding errors at compile time
But are you even a real programmer if you don’t test in production?
I shouldn’t need to do unit tests for quick one off scripts
What kind of quick one off scripts have large complex scopes where variable renames are difficult to track?
Besides, these days Python has great LSPs and typing features that can even surpass the traditional typed langs
It doesn’t need to have large complex scopes. I have the brain of a goldfish. I program because it’s challenging. It’s challenging because I’m bad at it.
Mostly number crunching and data exploration tasks. Just so I can make informed decisions about the data I got. I do this rarely enough so it hasn’t been worth for me to install all these extra third party support wheels.
Those support wheels are for your own (and apparently systematic) errors…
Shouldn’t be forgetting for one off scripts either, if that’s the logic you want to go with.
The tool exists, either you do it or you don’t and end up getting an error until the interpreter hits that line. It’s just the nature of being compiled at runtime.
“Ohh, I got all these numbers I want to crunch using numpy or pandas and plot it using matplotlib. Hold on, I just need to write unit tests first.”
Well. Yeah. That’s test-driven development. It’s a very good practice.
TDD only works well if the problem is clearly specified before the first line of code has been written, which is rarely the case when I need Python for something.
So which is then? You want a one off script to just quickly crunch some numbers on a problem you still need to understand? Because that is where it is perfectly normal to get some errors and doodle around. That is the entire point of it.
Or you have a concise concept of what you are going to do, and how and why? Because that is what you do, when you program more than a “one off”.
Either you go to the store with a shopping list and you work through that list or you go browsing and see what comes up. But don’t expect to be as fast and have everything you needed, when you dont write your shopping list at home.
Often I use Python for exploratory purposes. Like, I got a bunch of data, and I want to know if a particular algorithm might work or not. I implement the algorithm, but realize the results don’t look good enough. So I tweak the algorithm, maybe even do major refactoring. Or maybe I realize my visualizations or metrics don’t capture what I need to see. Or maybe I must settle for some compromise?
I iterate on this repeatedly until I find something I’m happy about (or until I give up). Sometimes I end up with something completely different from my initial idea.
TDD won’t help me much here because the end result is unknown. For each iteration of this idea process I might even need to rewrite all the tests because none of them are valid anymore.
Then maybe use an editor with a decent linter and check the problems tab or just red line markers?? I also have those kind of runtime errors sometimes but I take the blame.
deleted by creator
Catching some errors is better than catching no errors. No compiler in any language can protect you from all runtime errors either way, but some are better at it than others.
I’ll take Python’s “program exited four hours into runtime” over C++''s “this memory is uninitialized so it’s less than, equal to, and greater than zero at the same time, hope you never try to do any of checks on it!” undefined behaviour any time.
At least with Python, you know something went wrong.
Seriously, in what way does the python interpreter protect you?
you need a linter, bro
when integrated into the editor it’ll highlight stupid mistakes as they’re typed
sure, but thats just outsourcing the problem.
It’s also a solution…
As if that’s a bad thing… it means you’re not locked in with a tool you don’t like and the language itself doesn’t dictate your workflow.
There’s very little benefit and a lot of potential problems in using a single tool for everything.
I guess as a C# guy I’ve never had to deal with an issue like this. Most of the time the exceptions are pretty easy to diagnose unless it’s in the UI or in some async function.