Subset issues when using piggyback and renv

[PMassicotte]

I have found multiple issues when using piggyback with renv enabled project that are likely linked to how subset works for tibble vs data.frame.

For example, the function pb_releases() do not work when renv is enabled. This is occurring here:

piggyback/R/utils.R

Line 69 in 4b7cdb9

addr <- remotes[remotes[["name"]] == remote, "url"][["url"]]

With renv enabled, remotes[remotes[["name"]] == remote, "url"] is a character. Hence, using [["url"]] does not work. When renv is not enabled, the same is a tibble and can be subsetted using [["url"]].

I have made an animated GIF that hopefully show the problem clearly. You can see that pb_releases() stops working when renv is activated. When renv is deactivated, pb_releases() works again.

Note that the same problem occurs for some other functions of the packages.

[tanho63]

the problem is specifically guess_repo() right? not pb_releases()?

[PMassicotte]

Well, as I understand, pb_releases() calls guess_repo(), but yes is it occurring specifically in guess_repo().

[tanho63]

Interesting! Can you try using gh::gh_tree_remote() as the argument to repo? Might be an easy drop-in fix

[PMassicotte]

I have tried this but no luck:

> pb_releases(repo = gh::gh_tree_remote())
Error in strsplit(repo, "/") : non-character argument

[tanho63]

Oh. Duh, it returns a list. But at least it returns something! I suspect the problem is that renv only caches the import dependencies so when guess_repo tries to use gert it fails somehow.

[PMassicotte]

If I install tibble in the project (which is afterwards cached by renv) it works. So it looks like that

piggyback/R/utils.R

Line 51 in 4b7cdb9

remotes <- gert::git_remote_list(repo)

returns a tibble. Since it was not needed in my project, it simply returns a dataframe.

Then, subsetting a data.frame here:

remotes[remotes[["name"]] == remote, "url"][["url"]]

does not work since remotes[remotes[["name"]] == remote, "url"] returns a character rather than a tibble if tibble is loaded.

[tanho63]

Oh so it just needs to be remotes$url[remotes$name == remote] then…

I wonder if that’s better or worse than shortening dependencies and using gh here for that too

[PMassicotte]

Oh so it just needs to be remotes$url[remotes$name == remote]] then…

Exactly!

I wonder if that’s better or worse than shortening dependencies and using gh here for that too

Good question :)