Etienne Bacher

> It also suffers from the problem that aggregate() doesn't keep NA values so the following wouldn't work since the NA group level would be removed. Related to this, in...

Right, it would make a lot of sense given how much `dplyr` evolves (e.g the future `.by` argument)

Yes it's in the upcoming 1.1.0

Coming back to the slowness issue, I can gain ~30% speed by replacing: ```r for (i in seq_len(n_comb)) { rows[[i]]

I can make a PR but I think it shouldn't be implemented before the behavior of `NA` in groups is fixed (or clarified). Currently there's only one test in `test_group_by.R`...

> What I'm unsure about is whether I can write a script that modifies an intermediary generated when going from Rd to Qmd or md. Is this possible? Hi, I...

> Ultimately it'll be @etiennebacher 's call, of course. The thing I want to avoid is adding code that could have side effects on other websites (I'm already spending way...

The problem probably comes from `.grouped_list()`, and more particularly from `split()` which automatically reorders the groups. One solution (taken on [SO](https://stackoverflow.com/questions/17611734/r-split-preserving-natural-order-of-factors)) is to manually specified the levels. ``` r test...

Same issue here. Note that this is also the case when I use `bioc = FALSE`

In the meantime, I have a fork where I disabled most of the things that were related to Bioconductor: ```r pak::pak("etiennebacher/revdepcheck") ``` It seems to work fine, but of course...