Specify position of geom_text by keywords like "top", "bottom", "left", "right", "center"

Question

I wish to position text in a ggplot without specifying x and y positions, but instead using keywords, like e.g. in graphics::legend ("The location may also be specified by setting x to a single keyword from the list "bottomright", "bottom", "bottomleft", "left", "topleft", "top", "topright", "right" and "center").

Lets say I'm making a graph like this.

sp <- ggplot(mpg, aes(hwy, cty, label = "sometext")) +
       geom_point()

I want to add the label to be printed in the same way in every graph. Calling the following simply prints text at every x, y value supplied to aes.

sp + geom_text()

I could manipulate the x and y data supplied to geom_text() to ensure that the text remained in the same relative position between graphs, but is there not a simply way to call position by default positions like "top", "bottom" etc? I.e. sp + geom_text(position = "top").

Daina

Answer 1

geom_text wants to plot labels based on your data set. It sounds like you're looking to add a single piece of text to your plot, in which case, annotate is the better option. To force the label to appear in the same position regardless of the units in the plot, you can take advantage of Inf values:

sp <- ggplot(mpg, aes(hwy, cty, label = "sometext"))+
  geom_point() +
  annotate(geom = 'text', label = 'sometext', x = -Inf, y = Inf, hjust = 0, vjust = 1)
print(sp)

Adactylous

Answer 2

I avoid annotate like the plague and just use an empty data frame data argument for geom_text:

ggplot(mpg, aes(hwy, cty, label = "sometext"))+
  geom_point() +
  geom_text(data=data.frame(), aes(label = 'sometext', x = -Inf, y = Inf),
            hjust = 0, vjust = 1)

Bodywork

Clever idea to supply an empty data.frame. What is your reason for avoiding annotate? – Daina

it's focused on text elements; it's easier to post-plot munge them with grid ops (esp targeting them) — but, keep in mind, i do alot with ggplot2 under the covers in my spare time, so not everyone does that; in the long run, i usually end up having to annotate differently across facets and i can just add vs remove without forgetting i used annotate. just a personal pref. – Bodywork

I agree that when facets come into play, annotate becomes hard to work with and I usually resort to a data frame and geom_text, but for simpler cases like this, annotate is the more obvious choice. – Adactylous

Many thanks! I just stumbled upon the possibilities to supply hjust = "inward", vjust = "inward", which worked great in my case. – Portie

You can also add extra padding above your text with vjust=1.1 – Phosgenite

Answer 3

In ggpp::geom_text_npc the x and y positions are given in npc units (0-1). However, the positions can also be specified as "words":

d = data.frame(x = rep(c("left", "center", "right"), each = 3),
               y = rep(c("bottom", "middle", "top"), 3))
d$lab = with(d, paste0(x, "-", y))
d
#        x      y           lab
# 1   left bottom   left-bottom
# 2   left middle   left-middle
# 3   left    top      left-top
# 4 center bottom center-bottom
# 5 center middle center-middle
# 6 center    top    center-top
# 7  right bottom  right-bottom
# 8  right middle  right-middle
# 9  right    top     right-top

ggplot(d) +
  geom_text_npc(aes(npcx = x, npcy = y, label = lab))

Tish

Answer 4

The solution using infinity is good and is definitely the easiest option.

However, if you want more control over the position of your labels (for example, if you want them centered, or if you want more space between the axis line and annotation), you can use some math with min() and max() of your plot titles to create centered titles at top, bottom, right, or left. The code below is a bit lengthy, but will still place labels correctly if the values in your plot change. Also, to copy to other plots, you won't need to manually calculate values, just change the names of the x and y variables.

sp <- ggplot(mpg, aes(hwy, cty)) +
  geom_point() +
  theme_classic() +
  annotate("text", label = "top", 
           x = 0.5*(min(mpg$hwy) + max(mpg$hwy)), y = max(mpg$cty), vjust = 1) +
  annotate("text", label = "bottom", 
           x = 0.5*(min(mpg$hwy) + max(mpg$hwy)), y = min(mpg$cty), vjust = 0) +
  annotate("text", label = "right", 
           x =  max(mpg$hwy), y = 0.5*(min(mpg$cty) + max(mpg$cty)), hjust = 1) +
  annotate("text", label = "left", 
            x =  min(mpg$hwy), y = 0.5*(min(mpg$cty) + max(mpg$cty)), hjust = 0)

sp   

Impassable

Thanks for the answer. I may not have been very clear but this is what I meant by "manipulate the x y data". Seeing you do it though makes me realize its not really all that much work. Could be cool if you wrapped it in a function that responds to the "top" bottom" etc arguments. – Daina

Wrapping it in a function is definitely possible. However, there is one drawback with this workflow: changing the length of the axes (let's say manually set limits for the axes) will result in changes of the annotations. It would be so nice if 'annotate' had an option to define relative text placement. – Acridine

Answer 5

It's certainly possible to write a wrapper but the way units and justification are defined makes it rather verbose,

library(ggplot2)

qplot(1,1) + 
  annotation_compass('testN') + 
  annotation_compass('testE','E') + 
  annotation_compass('testSW','SW') + 
  annotation_compass('testW','W')

annotation_compass <- function(label,
                               position = c('N','NE','E','SE','S','SW','W','NW'),
                               padding = grid::unit(c(0.5,0.5),"line"), ...){
  position <- match.arg(position)
  x <- switch (position,
    N = 0.5,
    NE = 1,
    E = 1,
    SE = 1,
    S = 0.5, 
    SW = 0,
    W = 0, 
    NW = 0
  )
  y <- switch (position,
               N = 1,
               NE = 1,
               E = 0.5,
               SE = 0,
               S = 0, 
               SW = 0,
               W = 0.5, 
               NW = 1
  )
  hjust <- switch (position,
               N = 0.5,
               NE = 1,
               E = 1,
               SE = 1,
               S = 0.5, 
               SW = 0,
               W = 0, 
               NW = 0
  )
  vjust <- switch (position,
               N = 1,
               NE = 1,
               E = 0.5,
               SE = 0,
               S = 0, 
               SW = 0,
               W = 0.5, 
               NW = 1
  )
  f1 <- switch (position,
                   N = 0,
                   NE = -1,
                   E = -1,
                   SE = -1,
                   S = 0, 
                   SW = 1,
                   W = 1, 
                   NW = 1
  )
  f2 <- switch (position,
                   N = -1,
                   NE = -1,
                   E = 0,
                   SE = 1,
                   S = 1, 
                   SW = 1,
                   W = 0, 
                   NW = -1
  )
  annotation_custom(grid::textGrob(label, 
                                   x=grid::unit(x,"npc") + f1*padding[1] , 
                                   y=grid::unit(y,"npc") + f2*padding[2],
                                   hjust=hjust,vjust=vjust, ...))
}

Woodenware

change annotation_custom(grid::textGrob to annotation_custom(ggpubr::text_grob(. in order to add size argument to text. qplot(1,1) + annotation_compass('testN',size=30) – Landgraviate