You must log in or register to comment.
While I do agree with most of what is said here, I have a hangup on one of the points: Thinking that “docstrings and variable names” are a trustworthy way to indicate types. Python is not a statically typed language - never will be. You can have as much type hinting as you want, but you will never have a guarantee that some variable holds the type you think it does, short of checking the type at runtime. Also, code logic can change over time, and there is no guarantee that comments, docstrings and variable names will always be up to date.
By all means, having good docstrings, variable names, and type hinting is important, but none of them should be treated as some kind of silver bullet that gets you around the fact that I can access
__globals__at any time and change any variable to whatever I want if I’m so inclined.This doesn’t have to be a bad thing though. I use both Python and C++ daily, and think that the proper way to use Python is to fully embrace duck typing. However that also means my code should be written in such a way that it will work as long as whatever input to it conforms loosely to whatever type I’m expecting to receive.
I just type hint all my functions and that solves 99% of my type related issues.
Duck typing is the best if fully embraced. But it also means you have to worry just a little bit about clean failures once the project grows a little. I like this better than type checking relentlessly.
It also means that your test suite or doctests or whatever should throw some unexpected types around now and again to check how it handles ducks and chickens and such :)
I have to be honest in that, while I think duck typing should be embraced, I have a hard time seeing how people are actually able to deal with large-scale pure Python projects, just because of the dynamic typing. To me, it makes reading code so much more difficult when I can’t just look at a function and immediately see the types involved.
Because of this, I also have a small hangup with examples in some C++ libraries that use
auto. Like sure, I’m happy to useautowhen writing code, but when reading an example I would very much like to immediately be able to know what the return type of a function is. In general, I think the use ofautoshould be restricted to cases where it increases readability, and not used as a lazy way out of writing out the types, which I think is one of the benefits of C++ vs. Python in large projects.
The “but muh types” criticism of Python has always been completely ignorant of the main use cases of Python. Like yeah, probably shouldn’t use it in avionics or medical devices. But scientific computing where you’re basically just using the interpreter as a calculator? You really only care about whether something’s an integer or not. Float vs. double isn’t gonna kill you
