Geom.bar - Colorkey does not adhere to the order specified in Scale.x_discrete

[dstoeckel]

To group related classes together in a Geom.bar plot I used Scale.x_discrete.
The (simplified) code reads as follows:

df = DataFrame(value=[5,1,6,3], class=["a", "b", "c", "d"]);
plot(
    df, x=:class, y=:value, color=:class,
    Geom.bar, Scale.x_discrete(levels=["b","a","c","d"])
);

The problem here is, that the colors and the position of the levels in the colorkey are not chosen consistently with the order specified in levels. This becomes a problem when multiple plots are generated where new classes are added. In this case I would like to keep a consistent ordering and coloring scheme between the plots by appending the new class to the levels. Currently, however, the classnames seem to be sorted lexicographically which will result in shift of all colors.

[Mattriks]

In Gadfly (and ggplot2), each aesthetic (e.g. x,y,color) has its own scale, and the scales operate independently.
In Gadfly, the levels argument controls the order for each scale:

using Gadfly, RDatasets
mtcars = dataset("datasets","mtcars")
p = plot(mtcars, x=:Cyl, color=:Cyl, Geom.histogram,
  Scale.x_discrete(levels=[8,6,4]), Scale.color_discrete(levels=[8,6,4]) )    

[Mattriks]

In ggplot2, the limits argument controls the order:

require("ggplot2")
ggplot(mtcars, aes(x=factor(cyl), fill=factor(cyl))) + geom_bar() + 
 scale_x_discrete(limits=c("8","6","4")) + scale_fill_hue(limits=c("8","6","4"))