Estimate the percentage of the plot taken according to the legend in ggplot2

Question

I have a ggplot function, but I want to give users a warning in case they abuse it. The problem is that if many groups and group names are long, the legend becomes so large that it takes up the whole plot:

But if I changed the names of the plots, I would get:

I have:

• ggplot / ggbuild object
• the size of the chart window (480 x 320 by default), but it can be changed.

I want to use them to estimate the total area accepted by the legend (so I can send a warning if the size of the relationship legend / plot size is too large). Here is an example of the code I used to create the numbers:

library(ggplot2) # stolen from https://ryouready.wordpress.com/2008/12/18/generate-random-string-name/ MHmakeRandomString <- function(n=1, lenght=12) { randomString <- c(1:n) for (i in 1:n) { randomString[i] <- paste(sample(c(0:9, letters, LETTERS), lenght, replace=TRUE), collapse="") } return(randomString) } makeData <- function(k, useLongNames = FALSE) { x <- c(1, 100) X <- cbind(1, x) b <- matrix(rnorm(2*k), k, 2) y <- numeric(2*k) for (i in seq_len(k)) y[1:2 + 2*(i-1)] <- X %*% b[i, ] df <- data.frame(x = c(1, n), y = y) if (useLongNames) { df$g <- factor(rep(MHmakeRandomString(k), each = 2)) } else { df$g <- factor(rep(1:k, each = 2)) } return(df) } # okayish plot df <- makeData(50) g0 <- ggplot(data = df, aes(x = x, y = y, group = g, color = g)) + geom_line() + guides(color=guide_legend(nrow=5)) # unreadable plot df <- makeData(50, useLongNames = TRUE) g1 <- ggplot(data = df, aes(x = x, y = y, group = g, color = g)) + geom_line() + guides(color=guide_legend(nrow=5)) # to plot # x11(width = 480, height = 320) # print(g0) # x11(width = 480, height = 320) # print(g1) 

I have an idea that the answer is somewhere in ggplotGrob() . However, I am not familiar with grobs (and could not find clear documentation) and ended up on

 gGrob0 <- ggplotGrob(g0) gGrob1 <- ggplotGrob(g1) gGrob0$grobs[[15]]$grobs[[1]]$grobs # all legend elements convertWidth(grobWidth(gGrob0$grobs[[15]]), unitTo = "inches") # 4.5128 inches convertWidth(grobWidth(gGrob1$grobs[[15]]), unitTo = "inches") # 12.419 inches # but this is not correct: # number of legend columns x legend width <= plot width # 10 * 12.419 <= 480 

which, it seems to me, gives me a lot of information about what interests me. How to convert this information to the total width of the legend? Thank you very much in advance.

Answer 1

Here is my solution (inspired by m-dz comment). The objects g0 and g1 come from the code in the question.

 plotAndPrintRatio <- function(g, width, height) { gGrob <- ggplotGrob(g) tmpfile <- tempfile(pattern = "png") png(tmpfile, width = width, height = height) # it is necessary to open a device plot(g) legendSize <- as.numeric(convertWidth(grobWidth(gGrob$grobs[[15]]), unitTo = "inches")) plotSize <- as.numeric(convertWidth(grobWidth(gGrob$grobs[[7]]), unitTo = "inches")) print(legendSize / plotSize) # the ratio of legend size to plot size dev.off() return(tmpfile) } # problem only in the first plot f1 <- plotAndPrintRatio(g0, width = 480, height = 320) # 0.6769345 f2 <- plotAndPrintRatio(g1, width = 480, height = 320) # 1.887872 --> too big! # larger window size fixes the problem f3 <- plotAndPrintRatio(g0, width = 1200, height = 900) # 0.2707738 f4 <- plotAndPrintRatio(g1, width = 1200, height = 900) # 0.7551488 filesList <- list(f1, f2, f3, f4) # to show the saved pngs: dev.off() # might be needed to clean up the plotting window grid::grid.raster(png::readPNG(f1)) grid::grid.raster(png::readPNG(f2)) grid::grid.raster(png::readPNG(f3)) grid::grid.raster(png::readPNG(f4)) # to remove the tempfiles created: # lapply(filesList, file.remove)