incorrect labels in the legend for scale_color_manual()

[alyst]

Consider the following example:

test.df <- data.frame( a = rnorm(100), b = rnorm(100), c = sample( c('1','2','3'), 100, replace = TRUE ) )
ggplot( test.df ) + geom_text( aes( x = a, y = b, label = c, color = c ) ) + scale_color_manual(
        values = c( "1" = 'red', "2" = 'green', "3" ='blue', "4" = 'orange' ),
        labels = c(  "4" = '4', "2" = '2', "1" = '1',  "3" = '3' ) )

It's expected that in the legend you would see 1(red), 2(green), 3(blue).
Instead, labels are assigned to the colors by their order in which they are listed in labels: 4(red) 2(green) 1(blue)

What is worse, in this example:

test.df <- data.frame( a = rnorm(100), b = rnorm(100), c = sample( c('1','2','4'), 100, replace = TRUE ) )
ggplot( test.df ) + geom_text( aes( x = a, y = b, label = c, color = c ) ) + scale_color_manual(
        values = c( "1" = 'red', "2" = 'green', "3" ='blue', "4" = 'orange' ),
        labels = c(  "1" = '1', "2" = '2', "3" = '3',  "4" = '4' ) )

the order of the values and the labels is perfectly matched, but since value 3 is missing in the data, instead of
1(red), 2(green), 4(orange) we have 1(red), 2(green), 3(orange).

ggplot2 0.9, R2.14.1

[hadley]

Regardless of the other issues, do you think the legend should include all 4 values?

[hadley]

In the second example, the only problem is really the labelling of the legend, right? You asked that 4 be mapped to orange, and that's what you got.

[hadley]

And the real issue is that the names of labels are not used have no impact - labels are matched purely on position. Did using names ever work?

[alyst]

1. In my usecases it's fine to hide those labels that never appear in the data (like it is now), but it would be cool to have them shown as an option. :)
2. Yes, the matching of the colors in the plot itself is ok.
3. I'm not sure anymore and have no system to test, but I think in 0.8.x series the legend was working correctly.