💾 Archived View for envs.net › ~jupy › minihacks › ufuncs.gmi captured on 2024-05-10 at 11:55:38. Gemini links have been rewritten to link to archived content

View Raw

More Information

⬅️ Previous capture (2024-03-21)

-=-=-=-=-=-=-

Ufuncs

Working with dataframes, I wanted a way to filter for string matching stuff. So, you can use usual boolean expressions such as `df["column"]==val`. But if I have my own boolean function, `boolfoo(str) => Bool`, I can't just do `boolfoo(df["column"])`! Because `df["column"]` isn't a single value, it's a series of data. Or maybe it's a dataframe itself. I can't remember, but either way, it's not going to work.

So why does the boolean == work? The series obviously doesn't match `val`, but == is a special function known as a ufunc which smartly works on elements of arrays. It's actually really easy to make your own function into a ufunc.

import numpy as np

def boolfoo(input):
  # your stuff here
  return output

uboolfoo = np.frompyfunc(boolfoo,1,1)

Now, you can use uboolfoo on arrays, dataframes, whatever! And it will perform the pattern match on the stuff inside.

What is the 1,1 about?

The first argument, `boolfoo` in the example above, is your function you want to make into a ufunc. The second argument is the number of inputs your function takes. The third argument is the number of outputs your function gives. So change as you like, but I guess 1,1 is the most common!