Error on inconsistent object dimensions

[MetaEntropy]

Currently, 'test', 'yes' and 'no' can have inconsistent dimensions without any warning or error even with strict=TRUE

> test=rbind(alpha=c(A=TRUE, B=FALSE),beta=c(FALSE, TRUE))
> yes=matrix(nrow=1, ncol=4, 1:4)
> no=matrix(nrow=4, ncol=1, 1:4*10)
> ifelse1(test, yes, no)
       A  B
alpha  1 30
beta  20  4
> ifelse1(test, yes, no, strict=TRUE)
       A  B
alpha  1 30
beta  20  4
> yes+no
Error in yes + no : non-conformable arrays

I would expect an error (or at least a warning) by default on the ifelse1() operation.

[jaganmn]

In my opinion, ifelse should follow the behaviour of subassignment [<-, not the behaviour of binary operators. After all, it is implemented using subassignment, not arithmetic. The dimensions of subassignment values are irrelevant, hence the current behaviour.

The test for conformable arguments employed by the binary operators is actually not as simple as a test for identical dim attributes. Adding such a test to ifelse would make it slower and its implementation more complex. Can you argue clearly why it is worth it?