add function scale_label_identity function

[Yunuuuu]

Through these commits, we are enabled to effortlessly exhibit legend labeling using the guide parameter within the scale_label_identity function, aligning it with analogous aesthetic settings for other geoms. Furthermore, the assignment of the label key can be readily accomplished through the intrinsic data within the draw_key_ function (#2004). This bring scale_label_identity function and catch following concers:

1. label as a geom aesthetic should appear in the scaled data from draw_key_text function. As we can see with following code

library(ggplot2)
# Create sample data
data <- data.frame(
    x = c(1, 2, 3, 4, 5),
    y = c(1, 2, 3, 4, 5),
    label = c("A", "B", "C", "D", "E")
)
# Create the plot with geom_text
ggplot(data, aes(x, y)) +
    geom_text(aes(label = label, color = label),
        key_glyph = function(data, params, size) {
            browser()
        }
    )

2. users are inclined to modify the parameter show.legend to exhibit textual aesthetics (only to discover that it cannot fulfill their intended purpose), while the help document means the default value of show.legend shoule have include the mapping aesthetic:

show.legend: logical. Should this layer be included in the legends? NA, the default, includes if any aesthetics are mapped. FALSE never includes, and TRUE always includes. It can also be a named logical vector to finely select the aesthetics to display.

3. The hard-coded label in the draw_key_text function appears superfluous.

[teunbrand]

I'm a little bit worried with this line.
I fear that the following code wouldn't draw labels in the legend:

ggplot(mtcars, aes(mpg, disp)) +
  geom_text(aes(
    label = rownames(mtcars), 
    colour = factor(cyl)
  ))

Could you confirm and amend if necessary?

[teunbrand]

Overall I'm a little bit cautious about the need for such function, which might have been fruitful to discuss in an issue before posting a PR. Labels typically aren't mapped to a scale and introducing scale_label_identity() might prompt the expectation of various other label scales. I don't think the identity scale is a problem, but the other ones are tricky.

[Yunuuuu]

Thanks for your comment, I wanted bring it into ggplot2 package for these issues: #2004, https://stackoverflow.com/questions/49965758/change-geom-texts-default-a-legend-to-label-string-itself?noredirect=1&lq=1, https://stackoverflow.com/questions/18337653/remove-a-from-legend-when-using-aesthetics-and-geom-text?noredirect=1&lq=1, https://stackoverflow.com/questions/59091627/add-new-legend-for-geom-text-with-text-labels-as-legend-key, without scale_label function, the data in draw_key_text function will not contain the label informations, which has prompted users to modify the internal grob object or GEOM object which shouldn't be touched in usual.

[Yunuuuu]

I have noticed that the existence of the scale_label function leads to the incorporation of the label data within the draw_key_text function. Evidently, many users, including myself, have encountered difficulties with the hard-coded "a" in the textual legend. I have conducted a search on Stack Overflow and found several questions that are similar to this issue.

[Yunuuuu]

Perhaps it might be advisable not to alter the draw_key_text function. Instead, could we explore the possibility of integrating the scale_label_identify function? Alternatively, are there any superior approaches to address the hardcoded "a" issue?

[Yunuuuu]

It has come to my attention that the current draw_key_text function is expected to include the scaled label information as outlined in the documentation

ggplot2/R/legend-draw.R

Line 10 in c76b9ae

#' @param data A single row data frame containing the scaled aesthetics to

[teunbrand]

I'll admit that the current way of dealing with this is not ideal. I'd say the recommended route without this PR would be to use the override.aes argument of the guide and manually relabelling it.

library(ggplot2)
#> Warning: package 'ggplot2' was built under R version 4.3.1

data <- data.frame(
  x = c(1, 2, 3, 4, 5),
  y = c(1, 2, 3, 4, 5),
  label = c("A", "B", "C", "D", "E")
)

ggplot(data, aes(x, y)) +
  geom_text(aes(label = label, color = label)) +
  guides(colour = guide_legend(override.aes = list(label = LETTERS[1:5])))

You could still have an identity scale, but you'd need to set the guide for it to work as expected.

scale_label_identity <- function(..., guide = "none") {
  sc <- discrete_scale("label", "identity", force, ...,
                       guide = guide,
                       super = ScaleDiscreteIdentity
  )
  
  sc
}

ggplot(data, aes(x, y)) +
  geom_text(aes(label = label, color = label)) +
  scale_label_identity(guide = "legend")

^{Created on 2023-08-17 with reprex v2.0.2}

It has come to my attention that the current draw_key_text function is expected to include the scaled label information as outlined in the documentation

There is a subtle nuance in what is 'mapped' and what is 'scaled'. The label aesthetic is typically mapped but not scaled.

[Yunuuuu]

Thank you for your explanations. I now realize the differences between "map" and "scale". Given this understanding, it might not be advisable to include such a function in this context. For me, It'll be okay to close the pull request.

[yutannihilation]

As it seems closing this pull request is the conclusion here, I'm closing now.