facets fail with uninformative error for some column names

[eliocamp]

If one tries to facet a plot using columns with name "COL", "ROW", "PANEL", "SCALE_X" or "SCALE_Y", ggplot2 fails with a rather obscure error.

library(ggplot2)

mtcars$ROW <- mtcars$cyl
ggplot(mtcars, aes(hp, mpg)) +
   geom_point() +
   facet_grid(ROW ~  am) 
#> Error: `scale_id` must not be `NA`

^{Created on 2020-07-17 by the reprex package (v0.3.0)}

[yutannihilation]

Thanks, confirmed. This seems a long-standing known issue. c.f. #838 We might not be able to fix this easily, but maybe we can show some informative error message.

[thomasp85]

it is rather unfortunate and I think we should look into a fix... This should not be unsurmountable as it is only a weird interaction between the layout data.frame (containing the reserved words) and the column names

Quick thought would be to keep all facet panel data values in a nested data.frame instead of as separate columns (which creates the issue due to multiple identical column names)

[kotliary]

Spend a lot of time today debugging a problem with the PANEL column in my data. Just a warning that the input data frame has column(s) with these names would be good enough.

[eliocamp]

Just a warning that the input data frame has column(s) with these names would be good enough.

How about if we just make it work? 😉

With this PR :

library(ggplot2)
mtcars$PANEL <- mtcars$cyl
ggplot(mtcars, aes(hp, mpg)) +
  geom_point() +
  facet_grid(PANEL ~  am) 

[kotliary]

Looks great. 👍 Hope the PR will be successful.