Dynamically assign pathGrob in .onLoad() #3331
Conversation
…class GeomPolygon from throwing an install warning. Fixes tidyverse#3312.
|
If I understand correctly what is happening, the install warnings should only occur if you're installing with R version 3.5 or earlier. Do you have an R 3.5 environment with which you could test this? |
|
I think I can figure something out with RStudio cloud...stay tuned. |
|
With ggspatial in RStudio Cloud running R 3.5, this PR does appear to fix the warning. I used: remotes::install_github("tidyverse/[email protected]") # no warning
remotes::install_github("tidyverse/ggplot2") # warning
remotes::install_github("paleolimbot/ggplot2@13ff581") # no warning! |
| # do not install with error `possible error in 'pathGrob(munched$x, munched$y, ': | ||
| # unused argument (pathId = munched$group)` despite the fact that this is correct | ||
| # usage | ||
| pathGrob <- NULL |
There was a problem hiding this comment.
Just for my own understanding: Why do we need this assignment outside of .onLoad()?
There was a problem hiding this comment.
I copied what was in sf:::.onLoad() that @thomasp85 suggested we use as a reference. My guess is that it prevents the superassign from accidentally assigning in the global environment (or prevents a warning to that effect)
|
Thanks! |
|
This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/ |
This is a PR to fix #3313, where packages that subclass
GeomPolygonwere giving install warnings with the new version of ggplot2 in the revdep checks (#3303).I was hesitant to prepare this PR, because I don't understand why this works, and because I can't replicate the install warning that occurred in the revdep checks, I can't verify that this PR fixes the problem. I can verify that it doesn't cause any CMD check failures for either ggplot2 or ggspatial, which was one of the packages affected by the revdep check issue.