Don't include '...' to empty-dimensional matrices

[MichaelChirico]

Closes #141.

Finally understood the code well enough to offer this simple fix.

Crucially, we don't need to worry about omit='both' case because that implies both dimensions are non-empty (and, in fact, large). Also, passing NULL to rbind(), cbind(), just omits them, leading to the desired behavior without needing an else clause.

[flying-sheep]

this one also has a conflict now!

[flying-sheep]

Hmm, could you please add a test to check if they’re displayed correctly?

The example you gave in #141 should probably not show NAs when there’s no actual rows, if I understood correctly.

[MichaelChirico]

Appreciate your guidance on what a test should look like -- the main thing I had in mind when testing is to ensure there's no warning.

Beyond that, I added the simple test:

expect_match(repr(m), "A04[^A]*A07")

Which checks that A05 and A06 are elided and the repr() output goes from A04 to A07. I wanted to avoid issues of potentially-fragile reliance on whitespace, hence the regex instead of a full equality test on the repr() output.

[flying-sheep]

Yeah, maybe just add a check that the repr string is just one line for the zero-rows column (i.e. only the headers)

[flying-sheep]

beautiful, thank you!