Javascript is now better than Python

=programming

 

 

Every year, there are millions of people who decide to try programming, and, not knowing much about software, ask:
"What programming language should I learn?"

Python and Javascript are common answers to that question. Javascript can work on web pages, while Python has been considered a language that's easy for beginners to learn. But I've concluded that Python is now never the correct choice for programming, except in these 2 cases:

- You need a specific Python library.
- You're doing a leetcode interview.

 

 

Python is slow

Python is much slower than Javascript, because its basic design has features that are hard for interpreters to optimize. Notably, calling a function in Python is over 10x as expensive as in most other languages. There are multiple reasons, but the biggest one is: monkey-patching means the function must be looked up at every call, even in a loop.

I've seen some proposed uses for monkey-patching, such as: You wrote a bunch of code calling a function from a library, and now you want to change what that function does without changing the library (which might need automatic updates later). In other words, the purpose is making code do something other than what it says it does. For the same reasons that those are hard for interpreters to reason about, they're also hard for humans reading the code to reason about. Personally, I like to be able to read code and tell what it does.

Monkey-patching was a mistake, and Python was designed around making it easy because Guido von Rossum has bad taste.

 

 

Python is bad for complex projects

There are 2 main reasons why I consider Python unsuitable for large or complex programs:

 

variable scoping

In most modern languages, a variable is created with constructs like "let x = ..." and then changed with "x = ..." but in Python both of those are the same. That makes variable scoping unclear. It's not clear when reading code whether an existing variable is being modified, and people can accidentally modify an existing variable when they happen to choose an existing variable name.

 

types

I consider type systems to be essential for making reliable large software projects. And of course, a good type system is better than one with only primitive types like integers and floats. Python technically has some types, but they're so minimal as to be almost useless.

 

 

Python is hard to distribute

With Python, if you want to send something to users that reliably works, it's been rather difficult.

Many times, I've downloaded a Python project, installed the dependencies, and then...the dependencies didn't work. I'd go so far as to say that Python specifically is largely responsible for the rise of Docker - with a sane programming language you can just make a truly standalone binary and not have to include a whole OS around it. Apparently "uv" has recently made the situation a bit better, but it's been a problem for a long time, and meanwhile npm packages have just worked.

It's also hard to package standalone Python programs for users. Javascript, on the other hand, has many workable options for that:

- put your program online and let people use it as a web app
- start a local server and open the default browser
- use Tauri to use the installed WebView
- include a whole browser with your program (Electron)

 

 

obligatory whitespace mention

Every post about Python has to say something about indentation, but it seemed like everything about that was already discussed ad nauseum. Luckily, I recently learned why Python standardized on spaces instead of tabs!

When Guido was asked why PEP-8 says all lines of code should 79 chars or less, he replied:

 

Because an 80 char wide Emacs window starts wrapping at 79.

 

And tabs annoyed him because people using different tab widths - especially Javascript users with a tab width of 2 - would write code that, with a tab width of 4, would go past 79 characters.

This is an example of Guido van Rossum having the wrong priorities because he has bad taste. With wide monitors being cheap, always limiting line width to 79 characters is objectively wrong, as Linus Torvalds noted.

 

 

why Python succeeded

Per the above, Python is not very good for software that's performance-sensitive, complex, large, or distributed to non-developers. Since most software falls in one of those categories, it raises the question of why Python has been popular.

 

leetcode

Python is particularly suitable for leetcode interviews because:

- It's always an option, and something less-popular might not be.
- Many interviews don't allow installing extra libraries, because "you're supposed to use basic tools to show your low-level understanding". Python has a relatively large default library which is obviously an advantage in that case.
- It's garbage collected.
- Leetcode interviews don't involve working with a large existing system, in which case having type information would be more helpful.
- Leetcode interviews generally don't care about performance constant factors.

 

Since software companies have hired largely based on such interviews, their hiring has favored Python users more than it should have.

 

memes to libraries

Since Python has been considered "the language that's easy for beginners", library developers who want to make things that beginners can use have often picked Python. Then, having easy-to-use libraries helped Python maintain that reputation, more so than the language itself.

 

google

Google is one of the most valuable and successful companies to ever exist. Some early hires at Google happened to like Python. Google was also an early company to heavily hire based on leetcode interviews. Once Google became successful, its use of both Python and leetcode interviews was blindly copied by the management of many other companies.

Google, of course, makes most of its money from advertising, and collects personal data aggressively. This reminds me: I recommend using either Ungoogled Chromium or Firefox as your browser, and the Ublock Origin extension (which doesn't work on Chrome). Not for ideological reasons, but because it will make your life a little better.

 

falling behind

Javascript, C#, and Rust have all improved a lot over the past 10 years. Python has, compared to those, made little progress.

 

 

Javascript got better

Some ways Javascript improved more than Python include:

Faster compilers. Companies wanted their browser to be popular, and Javascript performance was directly linked to browser popularity, so companies put an enormous (for compilers) amount of money into making Javascript run faster.

Javascript only added let and const in 2015. Yes, that was 10 years ago, but it was 20 years after the languages was introduced. 20 years of people using var for everything, and now it's rarely seen.

Optional chaining with x?.y was introduced in 2020, and now you see it used constantly.

TypeScript now exists. It provides Javascript with a more sophisticated type system than most typed languages have. It got significant improvements in 2014 and 2016, and it's now common for Javascript libraries to provide types.

npm was first released in 2010. It made some improvements over time, like adding package-lock.json in 2017. It has a lot of libraries now.

 

npm dependencies

The existence of npm is both a blessing and a curse - a blessing because it's easy to use, and a curse because library authors use it too much. Because of that, Javascript has ~3x the average dependencies per project of Go, and ~7x the dependency count of Python.

It was obvious to me that supply-chain attacks were a security problem. Now they're happening regularly, and companies have finally started to be aware of that risk. I recommend the following mitigations:

- Avoid using libraries with many dependencies.
- Use pnpm, or disable npm install scripts with "npm config set ignore-scripts true".
- Use pnpm, and set minimumReleaseAge.

 

conclusion

Python deserves to go in the dustbin of history alongside COBOL, PHP, and Matlab. I still have to use it sometimes, but I'd prefer not to.