On Sat, 12 Jan 2013, Andreas Noack Jensen wrote: > I have a function which takes lists as argument and then tranform them to > matrices. However, sometimes the lists are null and I therefore get a 0x0 > matrix which is okay but when I concatenate with a nx0 matrix I get a 0x0 > matrix. I think a nx0 matrix is more reasonable but I agree that things are > fishy here. An example is > > ? eval ones(5,1)~{null} > 1 > 1 > 1 > 1 > 1 > > ? eval ones(5,0)~{null} > matrix is null (0 x 0) > > What do you say? Would it cause problems elsewhere if the latter was 5x0? I'm inclined to say that following the pattern of some other recent changes, producing a 5x0 result here would be consistent. If nobody raises an objection of principle, I'll make that change.