Ensure that we don't try and assign scale to NULL

[hadley]

Fixes #2599

@karawoo any ideas why this behaviour changed?

[karawoo]

It looks like the change happened with the port to tidyeval, so maybe @lionel- can weigh in too. This specific error first occurs with 1f39e13. From the first pass at tidyeval commit until then there's a different error:

ggplot(mtcars, aes(mpg, wt)) + 
  geom_point(aes(shape = NULL))
#> Error in if (length(ans) == 0L || as.character(ans[[1L]])[1L] == "~") { : 
#>   missing value where TRUE/FALSE needed

Before all the tidyeval changes, find_scale() was not being called for the shape aesthetic at all.

[hadley]

I wonder if it's 1f39e13#diff-e01fbacefeedb8a626f9a0ec103d0e12L277

[lionel-]

@hadley NULLs are stored the same way in both dev and CRAN versions:

str(aes(shape = NULL))
#> List of 1
#>  $ shape: NULL

So it seems like the commit you mention is not the cause.

[lionel-]

Also this gives the same results:

x <- aes(shape = NULL)
x[["shape"]] <- NULL
str(x)
#> Named list()

x <- aes(shape = NULL)
x["shape"] <- NULL
str(x)
#> Named list()

x <- aes(shape = NULL)
x$shape <- NULL
str(x)
#> Named list()

[lionel-]

The issue was that plyr::tryapply() strips NULL elements automatically. Compacting aesthetics after evaluation should fix the issue, cf last commit.

[hadley]

Ah I knew tryapply() was there for some reason!! (I'll try and tweak the unit tests if I have time)

[lock]

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/