Preserving aspect ratio in R's grid graphics

To draw a "crossed" rectangle of height 2 times larger than its width using the low-level graphics package facilities I call:

xlim <- c(0, 500)
ylim <- c(0, 1000)
plot.new()
plot.window(xlim, ylim, asp=1)
rect(xlim[1], ylim[1], xlim[2], ylim[2])
lines(c(xlim[1], xlim[2]), c(ylim[1], ylim[2]))
lines(c(xlim[1], xlim[2]), c(ylim[2], ylim[1]))

The figure has a nice feature: the aspect ratio is preserved so that if I change the size of the plot window, I get the same height-to-width proportions.

How can I obtain an equivalent result with grid graphics?


You should create a viewport that uses Square Normalised Parent Coordinates, see ?unit :

"snpc" : (...) This is useful for making things which are a proportion of the viewport, but have to be square (or have a fixed aspect ratio).

Here is the code:

library('grid')
xlim <- c(0, 500)
ylim <- c(0, 1000)
grid.newpage() # like plot.new()
pushViewport(viewport( # like plot.window()
        x=0.5, y=0.5, # a centered viewport
        width=unit(min(1,diff(xlim)/diff(ylim)), "snpc"), # aspect ratio preserved
        height=unit(min(1,diff(ylim)/diff(xlim)), "snpc"),
        xscale=xlim, # cf. xlim
        yscale=ylim  # cf. ylim
))
# some drawings:
grid.rect(xlim[1], ylim[1], xlim[2], ylim[2], just=c(0, 0), default.units="native")
grid.lines(xlim, ylim, default.units="native")
grid.lines(xlim, rev(ylim), default.units="native")

The default.units argument in eg grid.rect forces the plotting functions to use the native ( xscale / yscale ) viewport coordinates. just=c(0, 0) indicates that xlim[1], ylim[1] denote the bottom-left node of the rectangle.


In ggplot2 (which is grid based) you can fix the aspect ratio using coord_fixed() :

library(ggplot2)
ggplot(mtcars, aes(x = wt, y = mpg)) + geom_point() + coord_fixed(ratio = 0.5)

This will fix the ratio, and the ratio will be constant even when changing the size of the graphics window.

I'm not sure if this is helpful, as you asked for a low-level grid based solution. But I thought it might be useful none the less.

链接地址: http://www.djcxy.com/p/78730.html

上一篇: 如何通过输入边缘的数量进行计数和排序

下一篇: 在R的网格图形中保留纵横比