Default 'useNames' argument - take 2

[HenrikBengtsson]

Alright, I've listened, and I thought about it, and then I thought about it even more. I think it's okay to make:

• useNames = TRUE

the new default.

Advantages:

• Name attributes are processed by default, which makes life easier for most people (less typing)
• It is consistent with colMeans(), rowMeans(), colSums(), rowSums(), and apply()

Disadvantages:

• It adds performance and memory overhead
• Not everyone will know about useNames = FALSE (compared to everyone would know about useNames = TRUE if they need names)
• It introduces a new overhead to existing packages (n=443; 2023-05-24)
• Package maintainers who suffer from the new overhead might not notice for a long time, so it might be a while before they add useNames = FALSE

But, as others tried to convince me in the past, the advantages outweight the disadvantages.

So, the new plan is to go with useNames = TRUE.

FWIW, there are currently a few packages that will break when we make useNames = TRUE the default, cf. Issue #227. They have been informed several times already.

[const-ae]

Hi Henrik, I started making sparseMatrixStats compatible with useNames = TRUE (https://github.com/const-ae/sparseMatrixStats/commit/5441917d2e8e158e86ff8dbb75f58d061440b2bc). There is one edge case which I don't cover yet. matrixStats::colColsums output differs for matrices with no column name and matrices with a dimnames attribute set to NULL. ...

Moved to Issue #234.

[HenrikBengtsson]

Thanks for spotting this. Can you please repost this as a separate issue?

[HenrikBengtsson]

UPDATE: I'm doing the last validations before submitting this new version to CRAN.

[HenrikBengtsson]

matrixStats 1.0.0 is now on CRAN.

[const-ae]

Congrats :)