ggplot2: Bring one line to the front, but save the colors

Question

Consider the following code:

library(ggplot2)
foo <- data.frame(x=1:10,A=1:10,B=10:1)
ggplot(melt(foo,id.vars=\"x\"),aes(x,value,color=variable))+geom_lin         


        

Answer 1

The solution with replotting is fine unless you have reasons to avoid it. I can think of at least two: using alpha (transparency) or performance issues (you need to do it in one run, big data).

Here's what I propose:

require(scales)
# give the desired order here, I just use reverse
# separate function since we apply it over both levels & colors
shuffle <- function(x) rev(x)
foo <- data.frame(x=1:10, A=1:10, B=10:1, C=2.5, D=7.5)
melt_foo <- melt(foo, id.vars="x") 
# original plot
ggplot(melt_foo, aes(x, value, color=variable)) + 
  geom_line(size=5)

orig_order <- levels(melt_foo$variable)
new_order <- shuffle(orig_order)
# grabbing default colours
orig_colors <- hue_pal()(length(new_order))
new_colors <- shuffle(orig_colors)
melt_foo$variable <- factor(melt_foo$variable, levels=new_order)
# levels _and_ colours reversed, legend appearance stays the same
ggplot(melt_foo, aes(x, value, color=variable)) + 
  geom_line(size=5) + 
  scale_colour_manual(values=new_colors, labels=orig_order, breaks=orig_order) 

Answer 2

The order of your A,B inside data.frame(x=1:10,A=1:10,B=10:1) is causing this layering issue because it is drawing A (1:10) first.

If you use data.frame(x=1:10,A=10:1,B=1:10) is will put the ascending line on top (since B is drawn last).

Answer 3

Try this,

last_plot() + aes(group=rev(variable))

Answer 4

Replotting the red line using a subsetted dataframe does the trick.

library(ggplot2)
foo <- data.frame(x=1:10,A=1:10,B=10:1)
require(reshape2)
fooM <- melt(foo,id.vars="x")
p<-ggplot()
p<-p+geom_line(data=fooM[fooM$variable!="A",],aes(x,value,color=variable),size=5)
p<-p+geom_line(data=fooM[fooM$variable=="A",],aes(x,value,color=variable),size=5)
p

EDIT: Note that ggplot applies layers sequentially on top of each other - this can be best leveraged when constructing the plot line by line.

EDIT2: @tonytonov is right that one may want to avoid plotting identical things twice. Modified my answer to plot everything but A the first time, then only A. Result remains the same and is now also compatible with transparency or big data ;)