←back to thread

I don't like NumPy

(dynomight.net)
480 points MinimalAction | 1 comments | 15 May 25 16:05 UTC | HN request time: 0.203s | source
Show context
harpiaharpyja ◴[15 May 25 20:03 UTC] No.43998760[source]▶
I don't use numpy much, but based on the documentation for that function and the fact that you can index by None to add another dimension, it seems like the correct call is:

  y = linalg.solve(A,x[:,:,None])
The documentation given says that A should be (..., M, M), and x should be (..., M, K). So if A is 100x5x5 and x is 100x5, then all you need to do is convert x to 100x5x1.

Is that right? It doesn't seem that bad.

replies(1): >>43999735 #
1. geysersam ◴[15 May 25 21:56 UTC] No.43999735[source]▶
You're right but if you want `y` to have the same shape as `x` I think you also need to slice it after

    y = linalg.solve(A, x[..., None])[..., 0]
I don't really mind numpy syntax much, it gets the job done in most scenarios. Numba complements it really well when the code is easier to express like a couple of nested loops.

I used Julia for a while, but found it easier to predict the performance of python+numpy+numba, Julia has a few footguns and the python ecosystem is just insanely polished.