posts in the Programming category

A Geneva Convention for the Language Wars

I don’t tend to get too sniffy about the quality of discourse on the Internet. I have some appreciation for even the most pointless, uninformed flamewars. (And maybe my take on Web site comments is for another post.) But there’s an increasingly popular topic of articles and blog posts which is starting to annoy me a little. You’ve likely read them—they have titles like: “Python is Eating R’s Lunch,” “Why Python is Going to Take Over Data Science,” “Why Python is a Pain in the Ass and Will Never Beat R,” “Why Everyone Will Live on the Moon and Code in Julia in 5 Years,” etc.

This style of article obviously isn’t unique to data analysis languages. It’s a classic nerd flamewar, in the proud tradition of text editor wars and browser wars. Perhaps an added inflammatory agent here is the Data Science hype machine.

And that’s all okay. Go on the Internet and bitch about languages you don’t like, or tell everyone why your preferred one is awesome. That’s what the Internet’s here for. And Lord knows I’ve done it myself.

My only problem is that it ...

I’ve seen the best minds of my generation destroyed by Matlab …

(Note: this is very quick and not well thought out. Mostly a conversation starter as opposed to any real thesis on the subject.)

This post is a continuation of a Twitter conversation here, started when John Myles White poked the hornets’ nest. (Python’s nest? Where do Pythons live?)

The gist with John’s code is here.

This isn’t a very thoughtful post. But the conversation was becoming sort of a shootout and my thoughts (half-formed as they are) were a bit longer than a tweet. Essentially, I think the Python performance shootouts—PyPy, Numba, Cython—are missing the point.

The point is, I think, that loops are a crutch. A 3-nested for loop in Julia that increments a counter takes 8 lines of code (1 initialize counter, 3 for statements, 1 increment statement, 3 end statements). Only one of those lines tells me what the code does.

But most scientific programmers learned to code in imperative languages and that style of thinking and coding has become natural. I’ve often seen comments like this:


Which I think simply equates readability with familiarity. That isn’t wrong, but it isn’t the whole story.

Anyway, a lot of the ...

Page 1 / 1