Installation

You can install from GitHub with:

# install.packages("devtools")
devtools::install_github("coolbutuseless/lofi")      # Colour encoding
devtools::install_github("coolbutuseless/minisvg")   # SVG creation
devtools::install_github("coolbutuseless/devout")    # Device interface
devtools::install_github("coolbutuseless/devoutsvg") # This package

Basic usage of the svgout device

Use this device in the same way you would use pdf(), png() any of the other graphics output devices in R.

This results is thankfully uninteresting! That is, if the svgout device works properly then the output should like the standard ggplot output you’ve seen hundreds of times before.

devoutsvg::svgout(filename = "img/devoutsvg/example-basic.svg")
ggplot(mtcars) + 
  geom_point(aes(mpg, wt, colour = as.factor(cyl))) + 
  labs(title = basename("Example - Basic")) 
invisible(dev.off())
Warning in file(con, "w"): cannot open file 'img/devoutsvg/example-
basic.svg': No such file or directory
Warning in dev.off(): rdevice_close: svg_callback() Evaluation error:
cannot open the connection.

Filling with patterns

Because we now have control over the graphics device at quite a low level, we get to misbehave!

The svgout device can be instructed to use patterns instead of the actual RGB colour - this is achieved by

• instructing the device to use a particular pattern package (pattern_pkg), and
• encoding particular patterns from that package as their RGB representations using {pattern_pkg}::encode_pattern_params_as_hex_colour()

When svgout is asked to render a colour, it instead asks pattern_pkg to **decode**hex_colour_as_pattern_params and returns the desired SVG pattern snippet.

Some examples of pattern packages are:

• svgpatternsimple
• svgpatternusgs

library(svgpatternsimple)

#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Encode the parameters for 3 different patterns into 3 different colours
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
gear4_colour <- svgpatternsimple::encode_pattern_params_as_hex_colour(
  pattern_name = 'null', 
  colour       = '#123456'
)

gear6_colour <- svgpatternsimple::encode_pattern_params_as_hex_colour(
  pattern_name = 'stipple', 
  colour       = '#ff4455', 
  spacing      = 10
)

gear8_colour <- svgpatternsimple::encode_pattern_params_as_hex_colour(
  pattern_name  = 'hex', 
  angle         = 0, 
  spacing       = 20, 
  fill_fraction = 0.1,
  colour        = '#125634'
)

c(gear4_colour, gear6_colour, gear8_colour)


svgout(filename = "img/devoutsvg/example-manual.svg", pattern_pkg = 'svgpatternsimple')
ggplot(mtcars) + 
  geom_bar(aes(as.factor(cyl), fill = as.factor(cyl)), colour = 'black') + 
  labs(title = basename("Example - manual pattern specification")) + 
  theme_bw() +
  theme(legend.key.size = unit(1.5, "cm")) + 
  scale_fill_manual(
    values = c(
      '4' = gear4_colour,
      '6' = gear6_colour,
      '8' = gear8_colour
    )
  )
invisible(dev.off())

Using scale_fill_pattern_simple()

Rather than specifying individual patterns to map to colours, you can use a scale_fill. Since RGB colours don’t intuitively map to patterns, if you let ggplot choose the patterns using the default scale_fill_discrete() you will get a pretty ugly combination of patterns in your plot.

scale_fill_pattern_simple() allows the user to specify the envelope of desired outputs and ggplot will map values-to-patterns automatically within this envelope.

By default this will give an OK spread of pattern styles, but it probably won’t be great.
See the next section for how to specify the desired envelope of possible patterns.

svgout(pattern_pkg = 'svgpatternsimple', filename = "img/devoutsvg/example-scale-fill-2.svg")
ggplot(mtcars) + 
  geom_bar(aes(as.factor(cyl), fill = as.factor(cyl)), colour = 'black') + 
  labs(title = "scale_fill_pattern_simple() - defaults") + 
  theme_bw() +
  theme(legend.key.size = unit(1.5, "cm")) +
  svgpatternsimple::scale_fill_pattern_simple()
invisible(dev.off())

Using scale_fill_pattern_simple() with a defined pattern set.

By default scale_fill_pattern_simple() will give an OK spread of pattern styles, but it probably won’t be great.

The call to scale_fill_pattern_simple() can be customised to limit the possible choices of pattern which will appear in the plot. The default options are as follows:

• pattern_name = c('stripe', 'dot', 'hatch', 'check', 'stipple', 'hex')
• angle = c(22., 45, 67.5)
• spacing = seq(5, 50, length.out = 7)
• fill_fraction = seq(0.1, 0.9, length.out = 3)

In this following example, the choice of possible patterns is limited to ‘stripe’ and ‘hatch’, all at 45 degrees, with a range of fill_fraction and spacing

svgout(pattern_pkg = 'svgpatternsimple', filename = "img/devoutsvg/example-patternsimple-scale-fill.svg",
       width = 8, height = 6)
ggplot(mtcars) + 
  geom_density(aes(mpg, fill = interaction(cyl, am)), alpha = 1) +
  theme_bw() +
  theme(legend.key.size = unit(1.2, "cm")) +
  labs(title = "scale_fill_pattern_simple() - custom") +
  svgpatternsimple::scale_fill_pattern_simple(
    pattern_name  = c('stripe', 'hatch'),
    fill_fraction = seq(0.1, 0.4, length.out = 5), 
    angle         = c(45), 
    spacing       = c(10, 20)) 
invisible(dev.off())

Pattern fills with base plots

colours <- c(
  svgpatternsimple::encode_pattern_params_as_hex_colour(
    pattern_name = 'null', 
    colour       = '#123456'
  ),
  
  svgpatternsimple::encode_pattern_params_as_hex_colour(
    pattern_name = 'stipple', 
    colour       = '#ff4455', 
    spacing      = 10
  ),
  
  svgpatternsimple::encode_pattern_params_as_hex_colour(
    pattern_name = 'hex', 
    colour       = '#ddff55', 
    spacing      = 8
  ),
  
  svgpatternsimple::encode_pattern_params_as_hex_colour(
    pattern_name = 'check', 
    colour       = '#ee55ff', 
    spacing      = 10
  )
)


devoutsvg::svgout(pattern_pkg = 'svgpatternsimple', filename = "img/devoutsvg/example-pie.svg")
pie(c(cool = 4, but = 2, use = 1, less = 8), col = colours)
invisible(dev.off())

U.S. Geological Survey patterns on geom_sf() plots

The following plot uses the SVG pattern library made available in svgpatternusgs.

The 6 areas are manually assigned a colour which corresponds to a set of pattern params. The parameters are encoded to a colour by the encode_pattern_params_as_hex_colour() function in the svgpatternusgs package.

library(sf)
library(svgpatternusgs)

#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Select some data
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
nc <- sf::st_read(system.file("shape/nc.shp", package = "sf"), quiet = TRUE)
nc$mid <- sf::st_centroid(nc$geometry)
nc <- nc[nc$NAME %in% c('Surry', 'Stokes', 'Rockingham', 'Yadkin', 'Forsyth', 'Guilford'), ]

#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Encode specific USGS pattern numbers into colours
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
colours <- c(
  Surry      = svgpatternusgs::encode_pattern_params_as_hex_colour(usgs_code = 601, spacing = 100, fill='#77ff99'),
  Stokes     = svgpatternusgs::encode_pattern_params_as_hex_colour(usgs_code = 606, spacing = 100),
  Rockingham = svgpatternusgs::encode_pattern_params_as_hex_colour(usgs_code = 629, spacing = 100),
  Yadkin     = svgpatternusgs::encode_pattern_params_as_hex_colour(usgs_code = 632, spacing = 100),
  Forsyth    = svgpatternusgs::encode_pattern_params_as_hex_colour(usgs_code = 706, spacing = 100),
  Guilford   = svgpatternusgs::encode_pattern_params_as_hex_colour(usgs_code = 717, spacing = 100)
)

devoutsvg::svgout(filename = "img/devoutsvg/example-usgs.svg", pattern_pkg = 'svgpatternusgs')
ggplot(nc) +
  geom_sf(aes(fill = NAME)) +
  scale_fill_manual(values = colours) + 
  theme(legend.key.size = unit(0.6, "cm")) + 
  labs(title = "U.S. Geological Survey Patterns with `geom_sf()`") +
  theme_bw()
invisible(dev.off())

SVG to PDF

If you need a PDF version of an SVG file, there are a number of options.

1. Inkscape
2. rsvg on the command line
   • rsvg-convert -f pdf -o t.pdf t.svg
3. CairoSVG on the command line (python based)
   • cairosvg in.svg -o out.pdf
4. Imagemagick (not 100% sure)
   • convert file.svg file.pdf
5. Chrome headless (maybe?)
   • chrome --headless --disable-gpu --print-to-pdf="output.pdf" "input.svg"
6. Web-based. There are lots of these e.g.
   • https://cloudconvert.com/svg-to-pdf