CODEWITHSUNDEEP
rggplot2
Richard
Richard

Reputation: 61289

Overlay grid rather than draw on top of it

I have a ggplot2 graph which appears as follows:

ggplot2 graph without axis lines visible

Notice that the grid or axis lines do not show through the ribbons. One way of dealing with this is to alter the alpha property of the ribbons; however, this can make the lighter colours too light.

An alternative idea would be to draw the grid/axis lines on top of the ribbons rather than beneath them. How can I achieve this render ordering?

Naturally, the question could be asked of any plot generated by ggplot2. But a copy-and-paste command that illustrates the issue is as follows:

 ggplot(data.frame(x=sample(1:100),y=sample(1:100)),aes(x=x,y=y))+geom_point(size=20)

Upvotes: 20

Views: 10881

Answers (3)

AF7
AF7

Reputation: 3242

ggplot introduced an option in theme() that lets you do just that with pull number 993, on June 18, 2015.

Just add to your plot:

+ theme(
  panel.background = element_rect(fill = NA),
  panel.ontop = TRUE
)

There is an example in the ggplot docs.

Upvotes: 30

narendra-choudhary
narendra-choudhary

Reputation: 4818

Here's a workaround using geom_hline and geom_vline.

f <- ggplot(mpg, aes(cty, hwy))
f + geom_smooth(color="red")

It generates this plot.

enter image description here

To add horizontal and vertical lines manually:

f + geom_smooth(color="red") 
  + geom_vline(xintercept = c(10,15,20,25,30,35), color="white", size=1.25) 
  + geom_hline(yintercept = c(20,30,40), color="white", size=1.25)

To automatically add xintercept and yintercept:

f <- ggplot(mpg, aes(cty, hwy)) + geom_smooth(color="red")
x_intercept <- ggplot_build(f)$panel$ranges[[1]]$x.major_source
## x_intercept
## [1] 10 15 20 25 30 35
y_intercept <- ggplot_build(f)$panel$ranges[[1]]$y.major_source
## y_intercept
## [1] 20 30 40
f + geom_vline(xintercept=x_intercept, color="white", size=1.25)
  + geom_hline(yintercept=y_intercept, color="white", size=1.25)

Now any changes in axis-ticks introduced by scale-* functions will reflect in the final plot.
Here we have horizontal and vertical lines (similar to grid) on top of the plot. You can vary size to make lines more thick. enter image description here

But it's just a workaround. Given the flexibility of ggplot2 package, I think something similar can be achieved using theme. But I do not know how.

Edit1 : We can try following but it won't put grids on top. This way we can change size, color, linetype but that's all.

f + geom_smooth(color="red") 
  + theme(panel.grid.major=element_line(color="white", size=2))

Edit2 : Added automatically insertion of xintercept and yintercept using ggplot_build(f) as explained here.

Upvotes: 7

user20650
user20650

Reputation: 25854

You could use grid package functionality to extract the grid lines from the plot, and then redraw them, which would avoid some of the manual specification when adding horizontal or vertical lines.

library(ggplot2)
library(grid)

# Draw your plot
ggplot(data.frame(x=sample(1:100),y=sample(1:100)), aes(x=x,y=y))+
   geom_point(size=20)

# This extracts the panel including major and minor gridlines
lines <- grid.get("grill.gTree", grep=TRUE)

# Redraw plot without the gridlines
# This is done, as otherwise when the lines are added again they look thicker
last_plot() + 
  theme(panel.grid.minor = element_blank(),
        panel.grid.major = element_blank())

# Navigate to relevant viewport
# To see these use grid.ls(viewports=TRUE)
seekViewport("panel.3-4-3-4")

# Redraw lines
grid.draw(lines$children[-1])

Which produces

enter image description here

Alternatively, if you wanted to automate the adding of the vertical and horizontal lines within ggplot (as in Narendra's answer), but without specifying the breaks manually, you can access their positions using ggplot_build(p), where p is your plot.

It may be worth showing this for a graph with facets. Same procedure, except you select multiple lines and panels, and then just loop through them to draw.

# New plot with facets
ggplot(mtcars, aes(mpg, wt)) + geom_point(size=10) + facet_grid(am~cyl)

gr <- grid.ls(print=FALSE)
# Get the gTree for each of the panels, as before    
lines <- lapply(gr$name[grep("grill.gTree", gr$name)], grid.get)

last_plot() + 
  theme(panel.grid.minor = element_blank(),
        panel.grid.major = element_blank())

# Get the names from each of the panels
panels <- gr$name[grep("panel.\\d", gr$name)]

# Loop through the panels redrawing the gridlines
for(i in 1:length(panels)) {
             seekViewport(panels[i])
             grid.draw(lines[[i]]$children[-1])
             }

This will also work for the plots without facts.

Upvotes: 8

Related Questions