scale_identity doesn't play well with guide = "legend"

[Henrik-P]

To get a legend guide when using scale_identity we need to specify guide = "legend". In addition, "you'll typically* also need to supply breaks and labels" (?scale_identity).

However, when trying to get legends for color, fill, shape, size and alpha, when using scale_identity, a correct legend is obtained only for color. All other aesthetics generate either an incorrect legend (fill) or an error (shape, size and alpha).

Some data:

df <- data.frame(x = 1:2, y = 1,
                 color = c("red", "blue"), fill = c("blue",  "red"),
                 shape = 21, size = c(20, 10), alpha = c(0.8, 0.4))

Then try to get legends, without and with specification of labels and breaks.

Legends for color and fill:

library(ggplot2)
ggplot(df, aes(x = x, y = y, color = color, fill = fill, shape = shape, size = size,  alpha = alpha)) + 
   geom_point(stroke = 5) +
   scale_color_identity(guide = "legend") + 
   scale_fill_identity(guide = "legend") +
   scale_shape_identity() +
   scale_size_identity() + 
   scale_alpha_identity()

ggplot(df, aes(x = x, y = y, color = color, fill = fill, shape = shape, size = size,  alpha = alpha)) + 
   geom_point(stroke = 5) +
   scale_color_identity(guide = "legend") + 
   scale_fill_identity(guide = "legend", labels = df$fill, breaks = df$fill) + 
   scale_shape_identity() +
   scale_size_identity() +
   scale_alpha_identity()

The plot itself looks OK for all aesthetics. The legend for color is OK, but the fill legend only displays black points.

---

Legend for shape errors (Error: Continuous value supplied to discrete scale)

ggplot(df, aes(x = x, y = y, shape = shape)) +
   geom_point() +
   scale_shape_identity(guide = "legend")
#> Error: Continuous value supplied to discrete scale

ggplot(df, aes(x = x, y = y, shape = shape)) +
   geom_point() +
   scale_shape_identity(guide = "legend", labels = df$shape, breaks = df$shape)
#> Error: Continuous value supplied to discrete scale

---

Legends for size and alpha error (Error in !self$na.translate : invalid argument type)

ggplot(df, aes(x = x, y = y, size = size)) +
   geom_point() +
   scale_size_identity(guide = "legend")
#> Error in !self$na.translate: invalid argument type

ggplot(df, aes(x = x, y = y, size = size)) +
   geom_point() +
   scale_size_identity(guide = "legend", labels = df$size, breaks = df$size)
#> Error in !self$na.translate: invalid argument type

ggplot(df, aes(x = x, y = y, alpha = alpha)) +
   geom_point() +
   scale_alpha_identity(guide = "legend")
#> Error in !self$na.translate: invalid argument type

ggplot(df, aes(x = x, y = y, alpha = alpha)) +
   geom_point() +
   scale_alpha_identity(guide = "legend", labels = df$alpha, breaks = df$alpha)
#> Error in !self$na.translate: invalid argument type

---

By accident, I also mapped alpha while specifying a size legend, resulting in an alpha legend:

ggplot(df, aes(x = x, y = y,
               alpha = alpha)) +
  geom_point() +
  scale_size_identity(guide = "legend")

---

* It's unclear what 'typically' means here. At least in my examples, including or excluding labels and breaks makes no difference.

[hadley]

Can you please generate the reprex using the reprex package so I can see the results inline?

[HughParsonage]

Output using reprex::reprex

df <- data.frame(x = 1:2, y = 1, color = c("red", "blue"), size = c(20, 10), 
  alpha = c(0.8, 0.4))
library(ggplot2)
ggplot(df, aes(x = x, y = y, color = color, size = size)) + geom_point() + scale_size_identity(guide = "legend")
#> Error in !self$na.translate: invalid argument type