[Numpy-discussion] ndim argsort question

Vincent Schut schut@sarvision...
Tue Nov 20 02:30:01 CST 2007


Hi,

I'm getting a bit lost in numpy's fancy indexing, using argsort on a
slice of a 4d arrays, and then using the resulting indices to sort other
slices of that same array, preferably in one go.

I have a data[d, b, y, x] array (for the curious: date, band, y, x:
satellite image)
Then I create a sorting index from band 0:
idx = data[:, 0, :, :].argsort(axis=0)
Now I want to sort all 7 bands accoring to this index, without looping,
or at least avoid looping as far as possible. Something like sortedData
= data[idx], but then for each 'b'. However, I cant get multidim
indexing using the argsort result to work. Any hints?

Hard to explain. I hope someone understands it, otherwise please let me
know and I'll try to refrase :)

Cheers,
Vincent Schut.



More information about the Numpy-discussion mailing list