reduce and reducedim have inconsistent signature

[wsshin]

Currently (in v0.7) reduce's signature is reduce(op, itr) and reduce(op, v0, itr), whereas reducedim's signature is reducedim(f, A, region[, v0]). It was surprising to me that the order of v0 is inconsistent between the two functions.

It is understandable that v0 comes earlier than itr in reduce(op, v0, itr), because v0 is the left argument of the binary operator op and itr supplies right arguments. To be consistent, wouldn't it be better to separate the definition reducedim(f, A, region[, v0]) into two: reducedim(f, A, region) and reducedim(f, v0, A, region)? Or, is there any particular reason for this inconsistency?

[StefanKarpinski]

They should definitely be made to match. @simonbyrne has an idea of introducing a Slices type that might help refactor this.

[nalimilan]

#27711 will fix this. Cc: @andyferris

[nalimilan]

Actually it's already fixed since reducedim has been merged with reduce.