Christopher Guess

Whoever wrote #pandas and #numpy have a special place in hell for their function naming conventions. Sure, why not use undiscoverable and completely unself-documenting function names? Why not write them like it's still 1978 and make sure you conserve that 16kb of RAM?

It's so clearly written by mathematicians and not software engineers it makes my head spin. Every modern language has moved on to verbose naming conventions sometime around 1994 but Python? Nope. You're still stuck writing like you're K&R on a goddamn PDP-11 in Bell Labs.

To all the new #ml / #ai students/programmers, no this is not normal, and no, no sane programmer in 30 years has written any software that looks like Pandas or NumPy's conventions (outside of microcontrollers).

Feb 22, 2023 at 8:27pmWeb
Show threadSmári McCarthyFeb 22, 2023
@cguess the best hypothesis I can think of that would explain some of the naming is name and calling convention compatibility with LAPACK and such. My second best guess is that mathematicians and physicists may have been involved, two groups renouned for their love of impenetrable terseness. Either way, that should be an option, not the default.
Show threadChristopher GuessFeb 22, 2023
@smari I'd guess a few statisticians as well, also a field well known for their verboseness and ease of understand.
Show threadSmári McCarthyFeb 22, 2023
@cguess I should say, I'm reasonably fluent in math, but I often find that the terseness and the notation is a hindrance to comprehension rather than a help. Just in the last hour I've been looking for a fast way to do a thing, only to find that all the math papers I can find on the subject are making a particular assumption of common notation which I'm sure makes perfect sense once I've figured out what it is.
Show threadChristopher GuessFeb 22, 2023
@smari my favorite is when notation is shared across tons of fields (how many things does ‘j’ stand for in math, engineering and physics?) but there's no key or explainer what various symbols mean for this specific branch. If you're not already practically doing a PhD in this sub-sub-subject you’ll just never know.
Show threadbenFeb 22, 2023
@cguess I'm sorry, this is just yet another iteration of "software engineer does not understand usecase". The function name being short and close to what it looks like in the formula that you just translated into code is a feature.
Show threadChristopher GuessFeb 22, 2023
@sqrt2 so the formulas being absolutely impenetrable is a feature as well? Mathematical notation and naming is not a goal I'd consider to aspire to.
Show threadbenFeb 22, 2023
@cguess It's not impenetrable, just governed by a set of conventions that you're not used to.
Show threadChristopher GuessFeb 22, 2023

@sqrt2 I've studied university-level maths and been programming for 20+ years. I did not study very specific subfields of math which also share notation (but not meaning of that notation) with other fields.

"select_rows" is to get multiple rows (and takes a Dictionary for some reason) but "iloc" is to get one row. That's just bad usability, no matter what field you're in.

Show threadbenFeb 22, 2023
@cguess That's a complaint about pandas, which is not about formula translating, unlike numpy, and err... you can get multiple rows with iloc, and there is no select_rows. (Not in numpy either.)