Naming conflicts between r-universe and CRAN packages during dependency

[robchallen]

I've worked around this but wondering if there is a solution other than "don't do anything so foolish in the future".

I created a package in my r-universe called "tableone". There is already a CRAN package of that name (my first mistake). It would be slightly painful to rename it as people in the organisation are using it.

I wanted to depend on it from a second package (modelfitter). using a

...
Suggests:
    tableone (>= 0.4.3)
Remotes: bristol-vaccine-centre/tableone
...

stanza in my DESCRIPTION file. The problem is that the CRAN tableone package is at version 0.13.2 so in some r-universe builds (Windows and MacOS - binary package builds maybe?) it looks like the CRAN package gets picked in preference to my r-universe version (example here). I guess that it can't see them as different things during dependency resolution.

I guess my options are:

1. rename my tableone (sigh)
2. bump my tableone version up past the tableone on CRAN 0.13.2 (flaky)
3. remove the dependency (N.B. this is what I did)

I did a bit of googling looking to see if you can specify a blacklist for a specific repository somehow but that does not seem to be an thing.

Any other ideas?

[jeroen]

Yes you should rename the package, otherwise it is unusable in practice.

R is very eager in updating packages from CRAN. Even if you get it to work locally, users will get in trouble because R will automatically "update" the package and replace it with the CRAN version on the user machine when it sees a chance to do so.

Also other CRAN packages may depend on the CRAN version of tableone and import functions from it, which don't exist in your tableone so it turns into a big mess...

In the r-universe dashboard we try to highlight if a package conflicts with a CRAN name with a red icon and a popover, although it is not always trivial to determine if this is a different package with the same name, or just an older version:

[robchallen]

thanks. I guessed that was going to be the answer :-)