DOC: linalg: don't recommend np.linalg.norm #2987

Merged
merged 1 commit into from Oct 14, 2013

Projects

None yet

5 participants

@larsmans
Collaborator

The scipy.linalg.norm docstring is quite confusing because it seems to recommend np.linalg.norm instead, even though that function can be several times slower. It also has an axis keyword that scipy.linalg.norm doesn't have, adding to the confusion.

@coveralls

Coverage Status

Coverage remained the same when pulling ef66145 on larsmans:scipy-not-numpy-norm into 57ac922 on scipy:master.

@rgommers rgommers merged commit 7479249 into scipy:master Oct 14, 2013

1 check passed

Details default The Travis CI build passed
@rgommers
Member

Agreed, this is clearer. Thanks @larsmans

@larsmans larsmans deleted the larsmans:scipy-not-numpy-norm branch Oct 14, 2013
@WarrenWeckesser
Collaborator

FYI: In numpy 1.8, numpy.linalg.norm has an axis argument.

@ewmoore
Member
ewmoore commented Oct 15, 2013

Warren Weckesser wrote:

FYI: In numpy 1.8, |numpy.linalg.norm| has an |axis| argument.


Reply to this email directly or view it on GitHub
#2987 (comment).

Truthfully, there is no reason that scipy and numpy should be different
here. Can we open an issue (say labeled as a Task) on the numpy tracker
(or here if necessary) to normalize these two? I understand that it may
require wrapping (s,d,sc,dz)nrm2 and adding it to blas_lite.

@larsmans
Collaborator

I have a PR over at NumPy that uses dot in norm.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment