Tag ScikitLearn.jl v0.5.0

[attobot]

Repository: cstjean/ScikitLearn.jl
Release: v0.5.0
Travis:
Diff: vs v0.4.0
requires vs v0.4.0:

--- 0.4.0/requires
+++ 0.5.0/requires
@@ -1,4 +1,4 @@
-julia 0.6
+julia 0.7
 PyCall 1.6.2
 Parameters 0.3.0
 MacroTools 0.3.0

cc: @cstjean

Please make sure that:

• CI passes for supported Julia versions (if applicable).
• Version bounds reflect minimum requirements.

[andreasnoack]

The CIBot tests currently fail because Mocking includes a project file. Later today there might be a new release of Mocking without a project file so I'll restart the tests once it has been released.

[andreasnoack]

Please add an upper bound if Julia 1.0 isn't supported.

[cstjean]

I have a new version for 1.0 already (barring a few problems with PyCall and RDatasets not being quite 1.0-compatible). I've just been waiting for this PR to be pulled into metadata to release 0.5.1.

[cstjean]

Should I just update the attobot PR? I thought that it would be simpler to review the much smaller diff from the next tag, than to restart the review process.

[andreasnoack]

I'd just move (delete and reapply) the tag here instead. I'm not sure what you mean by "restart the review process". The only thing we do here (almost) is to check that tests pass.

[attobot]

Repository: cstjean/ScikitLearn.jl
Release: v0.5.0
Travis:
Diff: vs v0.4.0
requires vs v0.4.0:

--- 0.4.0/requires
+++ 0.5.0/requires
@@ -1,10 +1,10 @@
-julia 0.6
+julia 0.7
 PyCall 1.6.2
 Parameters 0.3.0
 MacroTools 0.3.0
 Conda
 ScikitLearnBase 0.0.6
 StatsBase 0.8.0
-Iterators 0.1.9
+IterTools
 Compat 0.18.0
 DataFrames 0.11

cc: @cstjean

Please make sure that:

• CI passes for supported Julia versions (if applicable).
• Version bounds reflect minimum requirements.

[andreasnoack]

There still isn't an upper bound on Julia

[cstjean]

As I've said above, the remaining tests failures are, to the best of my knowledge, PyCall and RDatasets failures on 1.0. Otherwise the package seems to work perfectly fine on 1.0.

[andreasnoack]

I misunderstood. I thought you had a version different from this one for 1.0. RDatasets was release in a 1.0 compatible version a while ago and PyCall for 1.0 was merged earlier today so I'll try to restart tests.

[cstjean]

There's still an error, AFAICT because of PyCall/Base. JuliaPy/PyCall.jl#555

I know you're not the only person involved in setting the rules for METADATA, but I find the policy to insist on 100% test success a bit frustrating. I haven't added functionality to this package in forever. I'm just trying to fight code rot, but because I'm "just" fixing 80% of the new test failures, we can't merge into METADATA, even though the new tag is a clear improvement on all fronts. Last time, I updated ScikitLearn for the new DataFrames release in February, but it took me four months to find the time to deal with 100% of the issues. During those four months, all new users of ScikitLearn saw their DataFrames.jl install downgraded, even though the master was perfectly good. It's hardly ideal.

[andreasnoack]

Hm. For some reason, CIBot seems to get stuck when trying to process this package so I'll go ahead an merge. Regarding your comment then, generally, I think passing test is a reasonable minimum requirement for a release. Hopefully, things will be more stable now and in the future, it should also be easier to avoid that a single package keeps other packages on lower versions for all projects.

[cstjean]

Thanks!