Re: New keyword name for linalg.lu

On Sun, May 21, 2023, at 17:08, Ilhan Polat wrote:
> Sorry for reviving this thread, but indeed "p_indices" sounding uglier 
> the more I look at it. Just to summarize what it does to save you from 
> the thread above; 
>
> when
>
> P, L, U = scipy.linalg.lu(A)
>
> is run, currently, P is returning a full 2D array. If A is a tall array 
> say, (25, 5) then P is necessarily (25, 25). And it is just a 
> permutaiton matrix, a row shuffled np.eye(25). Instead, you can ask 
> with this new keyword to return that shuffle pattern. as a 1D array and 
> hence P becomes (25, ) array. And there are more than rare cases that 
> only L and U is needed for other purposes. 
>
> So we would like to make that matrix vs array switch with a keyword. I 
> thought "p_indices = True/False" was a good idea but it became a bit 
> disturbing. Other matlab following software offers a string selector 
> with "outputForm='matrix'/'vector'" which forces users to remember the 
> options hence I am voting for a TrueFalse flag. However I can't come up 
> with the right keyword. 
>
> Could you please offer some alternatives even just for inspiration? 

How about reversing the meaning of the flag and naming it something like `p_full` or `full_p` with a `True` default? I also considered "squareform" instead of "full" as in `spatial.distance.squareform`, because there's a similar 1d <-> 2d square thing here, but the semantics are different enough that this name would probably be more confusing than helpful.

András

>
> Best,
> ilhan
>
>
>
> On Tue, Apr 25, 2023 at 9:34 AM Jake Bowhay <[email protected]> wrote:
>> It would be nice to add a quick note to the docs explaining when/which you should use. Currently both state "Compute pivoted LU decomposition of a matrix." which while true isn't very helpful for a user trying to decide which function to pick!
>> _______________________________________________
>> SciPy-Dev mailing list -- [email protected]
>> To unsubscribe send an email to [email protected]
>> https://mail.python.org/mailman3/lists/scipy-dev.python.org/
>> Member address: [email protected]
> _______________________________________________
> SciPy-Dev mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
> https://mail.python.org/mailman3/lists/scipy-dev.python.org/
> Member address: [email protected]
_______________________________________________
SciPy-Dev mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3/lists/scipy-dev.python.org/
Member address: [email protected]