Revdep: bridging `+` methods

[teunbrand]

In some reverse dependency checks, it has become clear that some + methods no longer work as intended. I had already struggled to find a solution for this problem in GGally. It has come up in these packages, as far as I recall:

• GGally
• ggside
• feasts
• gggenomes

And is also reported for nlmixr2plot (#6498 (comment)).

I'm not sure what the optimal solution is for this case. I don't think using S7's double dispatch is the immediate solution, in particular because it has trouble dispatching on S3 classes in some cases (see also RConsortium/S7#544). One option could be to also implement the S3 class as an S7 class, in which case the S7 double dispatch becomes viable.

In GGally's case, the custom LHS S3 class included "gg" and the RHS in e.g. labs() also has an S3 "gg" class for legacy reasons. In this case, overriding +.gg worked but that may not be the case for every type of class.

[billdenney]

@teunbrand, the issue you raised in the linked S7 repo seems to be exactly the one that I'm experiencing with nlmixr2plot. If the package can use chooseOpsMethod, then it works for my ggtibble package. That may be a general solution.

[teunbrand]

Yeah I'm aware of that function ever since you mentioned it in #5536 (comment), but I can't seem to figure out how to make this work with the S7 methods.

[teunbrand]

@billdenney My list of packages to inform arrived at ggtibble. I saw that there was a recent update to CRAN with a news entry that it was compatible with ggplot2 4.0.0. Thanks for taking such a proactive stance! I'm guessing having to inform you is now a moot point?

[billdenney]

Yeah, ggtibble should be taken care of.