Closed github-actions[bot] closed 1 year ago
This fail is triggered by a bug in MLJBase that I will fix shortly.
(However, this exposes the fact that the deprecated method @from_network
is being used in OutlierDetection.jl)
@davnn I've addressed the bug noted above at MLJBase. Can you please retrigger CI here?
This fail is triggered by a bug in MLJBase that I will fix shortly.
(However, this exposes the fact that the deprecated method
@from_network
is being used in OutlierDetection.jl)
Hey! Thanks for you support with the upgrade. The from_network API is just tested, not used in the library. I will remove the tests and remove it from the docs.
Edit: I did already observe the current CI error (glb missing) during my testing of the 0.21 preview. Is this function completely gone or renamed / moved?
| Edit: I did already observe the current CI error (glb missing) during my testing of the 0.21 preview. Is this function completely gone or renamed / moved?
Can you please point me to the stack trace? It's conceivable that some signature for glb
that I have regarded as private has changed or been removed. The public glb(nodes::AbstractNode...)
should be unaltered.
Hey, sorry for the long delay, I'm currently physically relocating, which leaves very little time. The method was renamed / made private in https://github.com/JuliaAI/MLJBase.jl/commit/123fd0c01f35d0dc7f936f6cc2ae081be46c0162. I would use the private glb method for now, but of course we could just copy the function as well. Based on the changes to reports with 0.21, I don't think it makes sense to maintain a cross-version (0.20, 0.21) compatible OutlierDetection.jl version. I have thus made the necessary adaptations for 0.21 and updated the package requirements to drop 0.20 support.
Edit: Not sure if I should publish a working 0.21 right away or fix the deprecated return!
warnings beforehand, what do you think? It doesn't look like a trivial change to migrate to prefit
.
Base: 93.99% // Head: 93.99% // No change to project coverage :thumbsup:
Coverage data is based on head (
d1fff62
) compared to base (a76cc22
). Patch coverage: 100.00% of modified lines in pull request are covered.
:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.
Edit: Not sure if I should publish a working 0.21 right away or fix the deprecated return! warnings beforehand, what do you think? It doesn't look like a trivial change to migrate to
prefit
.
The latest versions of MLJ and MLJBase no longer support return!
style learning network export or @from_network
. Ordinarily, I don't think migration to prefit
would be difficult, but I seem to remember there is a complication in this case with an existing use of non-public API (apart from glb
) which would cause non-trivial complications. Sorry, it's a while since I looked at this...
This pull request changes the compat entry for the
MLJBase
package from~0.20.12
to~0.20.12, 0.21
. This keeps the compat entries for earlier versions.Note: I have not tested your package with this new compat entry. It is your responsibility to make sure that your package tests pass before you merge this pull request.