Conversation
Search for `baselearner` in this package, not in the caller's namespace.
When `baselearner` is given as a string, it's limited (by `match.arg`) to the pre-defined choices `c("bbs", "bols", "btree", "bss", "bns")`. It doesn't make sense to require the caller to import that symbol when we know it's defined in the `mboost` package anyway.
|
Thanks a lot for the patch and sorry for my rather slow response. Time is always (too) scarce. Unfortunately I cannot check the PR on windows as Appveyor is causing problems. Hope it is fixed soon. |
|
Hi @hofnerb , it looks like the Appveyor failure doesn't really have anything to do with this patch, it looks like some configuration problems: I think it's trying to install |
|
@hofnerb about the previous comment - what happens if you move the |
|
Still not working, now due to another issue related to BioC... I was hoping it is working out by the time by itself. Very annoying. |
|
So does this mean no fixes can happen on |
Search for
baselearnerin this package, not in the caller's namespace.When
baselearneris given as a string, it's limited (bymatch.arg) to the pre-defined choicesc("bbs", "bols", "btree", "bss", "bns"). It doesn't make sense to require the caller to import that symbol when we know it's defined in themboostpackage anyway.In my case, I'm invoking
mboostthrough a third-party package that does cross-validation, so it's impossible for me to importbtreeinto that namespace. The only solution is to do a shotgunlibrary(mboost)to export thebtreesymbol globally, but I'd like to avoid that.I also noticed that the
bnamevariable is never used, so I got rid of it.