get_param: fix data length for constructed matrix

[kyleam]

When creating the matrix from a vector and labels, format_matrix() calculates the data size as 2 N, where N is the maximum label index (e.g., 2 3 for a 3x3 matrix). The correct size is N * N.

Fortunately that error doesn't functionally matter because matrix's ncol and nrow arguments, not the input data length, control the final shape.

R 4.2 or later flags the issue in cases where 2 N isn't a multiple of N N. Several spots in test-get-param.R now trigger a warning: "data length differs from size of matrix".

Fix the issue by dropping the rep() call entirely and passing a scalar. matrix() will repeat that as many times as needed.

[kyleam]

As an aside, you shouldnt be able to have a 2 x 3 matrix for an OMEGA matrix from my understanding though (not sure if that was just an example).

I think you misunderstood the first paragraph of the commit message. It's describing how the current state is incorrect. The whole point is that 2 * N is wrong, and the size calculation should be N * N (i.e. the size of a square matrix).

---

Given that's the core point, do you want to take another look in light of that and make sure you still approve?

Also, let me know if you have a suggestion for rephrasing the commit message in a way that you would find clearer.

[barrettk]

@kyleam No I see what youre saying by that - I agree that it's wrong. I honestly recall thinking about this when I implemented it (and knew it should be N x N in terms of dimension). I wish I could remember why I changed it to 2 x N, but now I am not able to make sense of why I did it that way. Would bet I saw some weird behavior in the console or something, but dont think it really matters at this point. Regardless, a single static value makes the most sense since the row and column specification ultimately puts it in the correct format/dimension.

But no I think your commit message is fine and still approve, I just wanted to make sure a 2 x 3 matrix wasnt a possible output from get_omega()/get_param().

[kyleam]

@barrettk Got it. Thanks for the review.

(Obviously this is just a focused fix, but I wouldn't be opposed to a rewrite of format_matrix that makes it bit clearer that this should always be a square matrix. For example, get max index for rows and columns and assert they're the same.)