R ggplot2 facetting keep ratio but override/define output plot size

Question

I'm currently using ggplot2 to compare statistics from different groups, each belonging to a different region. This is done via a web application (tikiwiki CMS + plugin R), who runs the R script. Per region I can have 2 to 30 or more groups. A same R script runs for all the data in a unique web page, the result adapting depending on the chosen region - as parameter of the page.

Currently, I have this :

region 1: 12 groups = 12 medium facets
region 2: 3 groups = 3 **HUGE** facets
region 3: 24 groups = 24 **TINY** facets
region 4: 16 groups = 16 medium facets
...

Region 2

Region 3

What kind of result I would like to have is :

Region 2

Region 3 Stays the same

I'm facetting per group with facet_wrap(~data, ncol=4), so I can have either 2 or 30+ facets. My problem is the output: either the 30 facets are stuffed into a tiny box, or the two ones are oversized in the same sized box -> See pictures above... I can't find how to fix a ratio (or image output max width) but keep the final picture size free.

Is it possible in ggplot2 / R to fix a default size for the facets and to have the size of the resulting picture adapting to the numbers of facets ?

If this is not possible in ggplot2 / R is there any javascript/jquery library that can pick up the code and show the results correctly (d3, ...) ?

Answer 1

you can edit the gtable setting the heights to physical units (e.g. cm) instead of relative ("null")

require(ggplot2)
p = qplot(Sepal.Width, Sepal.Length, data=iris) + facet_wrap(~Species, ncol=1)
g = ggplotGrob(p)

panels = which(sapply(g[["heights"]], "attr", "unit") == "null")
g[["heights"]][panels] = list(unit(4, "cm"), unit(8, "cm"), unit(2, "cm"))

device.height = convertHeight(sum(g[["heights"]]), "in", valueOnly=TRUE)

pdf("test.pdf", height = device.height)
grid.draw(g)
dev.off()

Edit As a follow-up, here's a function to set the height and width of all panels to a fixed value,

freeze_panels <- function(p, draw=TRUE,
                          width=unit(5,"cm"),
                          height=unit(1,"in")){
  require(grid)
  g  <-  ggplotGrob(p)

  vertical_panels <-which(sapply(g[["heights"]], "attr", "unit") == "null")
  horizontal_panels <-which(sapply(g[["widths"]], "attr", "unit") == "null")

  g[["heights"]][vertical_panels] <- replicate(length(vertical_panels), 
                                               height, simplify=FALSE)

  g[["widths"]][horizontal_panels] <- replicate(length(horizontal_panels), 
                                               width, simplify=FALSE)

  device.height  <-  convertHeight(sum(g[["heights"]]), "in", valueOnly=TRUE)
  device.width  <-  convertWidth(sum(g[["widths"]]), "in", valueOnly=TRUE)
  if(draw){
    dev.new(height=device.height, width=device.width)
    grid.newpage()
    grid.draw(g)
  }
  invisible(g)
}



require(ggplot2)
d1 <- subset(mtcars, carb != 8)
d2 <- subset(mtcars, carb %in% c(1,2,3))
p = qplot(vs, am, data=d1) + facet_wrap(~carb)

freeze_panels(p)
freeze_panels(p %+% d2)
freeze_panels(p %+% d2 + facet_wrap(~carb, ncol=1))