Added support for sparse data matrices.

[slundberg]

Needed when fitting to large sparse data sets. Addresses #9

[slundberg]

The problems with travis seem to be an unrelated 0.6 issue. Any chance we can merge this? ...it is a pain keeping my own copy of GLMNet in sync since I need to use sparse data.

[jeffwong]

Would love to have this functionality too. @slundberg I looked at the travis-ci log and saw this

while loading /home/travis/.julia/v0.6/GLMNet/test/runtests.jl, in expression starting on line 143
ERROR: LoadError: MethodError: convert(::Type{Array{Float64,2}}, ::GLMNet.CompressedPredictorMatrix) is ambiguous. Candidates:
  convert(::Type{T}, X::GLMNet.CompressedPredictorMatrix) where T<:Array{Float64,2} in GLMNet at /home/travis/.julia/v0.6/GLMNet/src/GLMNet.jl:80
  convert(::Type{Array{T,n}}, x::AbstractArray{S,n}) where {T, n, S} in Base at array.jl:328
Possible fix, define
  convert(::Type{Array{Float64,2}}, ::GLMNet.CompressedPredictorMatrix)

Does this mean anything to you?

[jeffwong]

I'm still new to Julia and don't know what happened between v 0.5 and v 0.6, though through some digging I believe the issue is in this test:

https://github.com/simonster/GLMNet.jl/blob/master/test/runtests.jl#L121

At this point path.betas is a GLMNet.CompressedPredictorMatrix. I think the fix for travis might be a one line change here

https://github.com/simonster/GLMNet.jl/blob/master/src/GLMNet.jl#L79

Does the function signature just need to change from

X::CompressedPredictorMatrix

to

X::GLMNet.CompressedPredictorMatrix

[slundberg]

Hey! It does seem like a 0.6 change. Things are busy here for me right now (just had a kid) so I won't be able to look into it but I think you are on the right track.

…

On Tue, Jul 4, 2017 at 8:51 PM Jeffrey Wong ***@***.***> wrote: I'm still new to Julia and don't know what happened between v 0.5 and v 0.6, though through some digging I believe the issue is in this test: https://github.com/simonster/GLMNet.jl/blob/master/test/runtests.jl#L121 At this point path.betas is a GLMNet.CompressedPredictorMatrix. I think the fix for travis might be a one line change here https://github.com/simonster/GLMNet.jl/blob/master/src/GLMNet.jl#L79 Does the function signature just need to change from X::CompressedPredictorMatrix to X::GLMNet.CompressedPredictorMatrix — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#12 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADkTxXZT50Txdt_Yt6C90EUN_BVAZZuwks5sKwhDgaJpZM4JJ6fT> .

[jeffwong]

Congratulations! Maybe @simonster can help me to verify the issue and then it can get merged?

[jeffwong]

bump. @simonster would you be able to take a look? Thanks!

[ararslan]

I'm going to close and reopen this so that (hopefully) the Travis status hook will update to point to JuliaStats instead of Simon's account. That will make it a bit easier to look at the failures in depth.

[ararslan]

The issue causing the failure has been fixed. This PR now just needs some syntactic updates for deprecated syntax and it should be good to go. I can push some commits to this branch soon to fix it up if you're still busy, @slundberg.

[slundberg]

@ararslan That would be great. Thanks!

[ararslan]

Continued in #25 with Scott's authorship preserved.