DelayedMatrixStats no longer supports scalar `center=` arguments about delayedarray

[hpages]

@PeteHaitch Would you be open to the possibility of keeping length-1 centers valid in DelayedMatrixStats, at least for a little bit longer? Length-1 centers in DelayedArray, SparseArray and sparseMatrixStats have been supported for a while and they didn't start issuing warnings recently despite the fact that matrixStats doesn't like them anymore. So having them now cause a hard error in DelayedMatrixStats is inconsistent with DelayedArray, SparseArray and sparseMatrixStats .

Also I can't think of anything wrong with supporting length-1 centers -- they're just a convenience like recycling.

If you agree with this, I'll submit a patch to DelayedMatrixStats. The patch will simply recycle length-1 centers to nrow(x) or ncol(x).

Thanks,
H.

[PeteHaitch]

Argh sorry for this.
I was trying to follow the defunct-ification by Henrik in matrixStats (HenrikBengtsson/matrixStats#254), since we've (mostly) tried to mimic its API.
A patch would be welcome (it might just be reverting the most recent commits).

[hpages]

No worries. Working on it. Will also involve patching MatrixGenerics to make things consistent across the board.

[hpages]

PR for DelayedMatrixStats ready: PeteHaitch/DelayedMatrixStats#101

[PeteHaitch]

Thanks, @hpages
I'm in the midst of resolving OSCA build issues and I'll then turn to this (hopefully Friday).

[hpages]

Fixed by PeteHaitch/DelayedMatrixStats#101