Few missing arnoldi tests #22963
Few missing arnoldi tests #22963
Conversation
kshyatt
added
linear algebra
test/linalg/arnoldi.jl
Outdated
| @test norm(v) > testtol # eigenvectors cannot be null vectors | ||
| @test_warn "Use symbols instead of strings for specifying which eigenvalues to compute" eigs(a, which="LM") | ||
| @test a*v[:,2] ≈ d[2]*v[:,2] | ||
| @test norm(v) > testtol # eigenvectors cannot be null vectors |
There was a problem hiding this comment.
these 2 lines seem like exact duplicates, should there be a new d and v they operate on?
There was a problem hiding this comment.
You're right, derrrrp on my part, sorry about that
|
Sorry about the bad whitespace. Good to go? |
test/linalg/arnoldi.jl
Outdated
| @@ -31,6 +31,10 @@ using Base.Test | |||
| (d,v) = eigs(a, nev=3) | |||
| @test a*v[:,2] ≈ d[2]*v[:,2] | |||
| @test norm(v) > testtol # eigenvectors cannot be null vectors | |||
| (d,v) = eigs(a, I, nev=3) # test two argument eigs | |||
There was a problem hiding this comment.
Slightly confused: three argument eigs? :)
There was a problem hiding this comment.
A keyword argument is still an argument, no? :)
There was a problem hiding this comment.
A second class argument!
test/linalg/arnoldi.jl
Outdated
| (d,v) = eigs(a, I, nev=3) # test two argument eigs | ||
| @test a*v[:,2] ≈ d[2]*v[:,2] | ||
| @test norm(v) > testtol # eigenvectors cannot be null vectors | ||
| @test_warn "Use symbols instead of strings for specifying which eigenvalues to compute" eigs(a, which="LM") |
There was a problem hiding this comment.
Why is this a thing? Shouldn't we deprecate the change from String -> Symbol for the which kwarg instead of this warning?
Unrelated to this PR though, I can submit a PR for the deprecation if thats the correct thing to do.
There was a problem hiding this comment.
I suspect it's a "protecting the fragile feelings of users" thing - is there an ARPACK wrapper that uses strings? I agree it's messed up but if the code there we should test it.
There was a problem hiding this comment.
Yes, the symbol is converted to a string before sent to the ARPACK wrapper, but still...
There was a problem hiding this comment.
Look, I'm just reporting the fact that this warning exists. ¯_(ツ)_/¯ I have no particular attachment to the question of whether it deserves to survive.
There was a problem hiding this comment.
This whole file is full of this kind of stuff haha
|
@Sacha0 are you ok with this? Ready to go? |
I still find the comment confusing :). |
|
looks like this should be reverted, it keeps failing on CI |
This reverts commit 5041126. This has been frequently failing on 64-bit Linux Travis with ``` Error in testset linalg/arnoldi: Error During Test Test threw an exception of type Base.LinAlg.ARPACKException Expression: (eigs(speye(50), nev=10))[1] ≈ ones(10) Base.LinAlg.ARPACKException("unspecified ARPACK error: 3") ```
|
Sorry :( |
|
I think this might be an issue with OpenBLAS. I was able to reproduce this on an Ivybridge machine running Linux. If I set |
No description provided.