limits are not used for the domain of manual scales with named values

[mikmart]

Currently, the limits argument to manual scales is not used to define the domain of the scale when a named values vector is provided. Take the example from ?scale_colour_manual:

library(ggplot2)

p <- ggplot(mtcars, aes(mpg, wt)) +
  geom_point(aes(colour = factor(cyl)))

cols <- c("8" = "red", "4" = "blue", "6" = "darkgreen", "10" = "orange")
p + scale_colour_manual(values = cols)

Specifying limits will not change the domain of the scale---below, cyl == 6 values are still mapped to a colour and plotted---but will alter the legend (as the breaks used to draw the guide are assumed to be a subset of limits):

p + scale_colour_manual(values = cols, limits = c("4", "8"))

With more exotic limits this can lead to very confusing legends:

p + scale_colour_manual(values = cols, limits = 2:5)

This could be fixed by ensuring that with named values, only elements with names within the limits are used in the scale mapping.

Created on 2018-05-27 by the reprex package (v0.2.0).

PS. I'd be happy to contribute a PR if this is something that should be fixed.

[paleolimbot]

I think the issue is that ScaleDiscrete$map() does not censor out-of-bounds values in the way that one might expect. I don't imagine this causes problems in many places, but in the above example I would expect values outside the limits to have the na.value colour, rather than the colour defined by values.

[teunbrand]

I think this might have been fixed already in #4093. With current master branch:

library(ggplot2)

p <- ggplot(mtcars, aes(mpg, wt)) +
  geom_point(aes(colour = factor(cyl)))

cols <- c("8" = "red", "4" = "blue", "6" = "darkgreen", "10" = "orange")

p + scale_colour_manual(values = cols, limits = as.character(2:5),
                        na.value = "grey50")

^{Created on 2021-04-28 by the reprex package (v1.0.0)}

[clauswilke]

Can we close this issue then?

[paleolimbot]

This looks good to me!