11 Retro

A blast from the past.

Wherein we explore the lost art of the hatched choropleth.

I decided to give hatched choropleths a go for this one, as those seem appropriately ‘retro’ in the context of mapping packages primarily aimed at relatively simple thematic mapping.

TL;DR; it’s been a journey.

Neither tmap not ggplot2 has hatching available as an option natively. I experimented with the HatchedPolygons package with tmap, and the ggpattern package with ggplot2 but neither allowed me to get the end result I was after. Eventually I returned to base R plotting in conjunction with some code I found here from Diego Hernangómez, which I modified, in particular to get the legend to work.

Both hatching and legend key elements are spatial data added to the map (i.e. they aren’t patterns as such, they are linestrings).

Libraries

library(sf)
library(tmap)
library(dplyr)
library(ggplot2)

Data wrangling

ak <- st_read("data/ak-city-demographics-2018.gpkg")

The maps

So here is the best version, which uses base R. patterns.R contains code I modified from here to provide patternLayer and legendPattern functions.

library(classInt)
source("patterns.R")

brks <- round(classIntervals(ak$jedi, n = 5)$brks, 1)
intervals <- paste(brks[-6], "to", brks[-1])
brks[6] <- brks[6] + 0.01
spacings <- c(500, 350, 250, 175, 125)

par(mai = rep(0.1, 4))
plot(ak$geom, lwd = 0.75)
for (i in 1:5) {
  ak %>% 
    filter(pop > 0, jedi >= brks[i], jedi < brks[i + 1]) %>%
    st_union() %>%
  patternLayer("left2right", density = spacings[i], add = TRUE, lwd = 0.75)
}
bb <- st_bbox(ak)
legendPattern(pos = bb[1:2], title.txt = "% Jedi",
  categ = intervals, patterns = "left2right", density = spacings,
  frame = FALSE, lwd = 0.75)

tmap

I tried the HatchedPolygons package, since I’d successfully used it in this figure. It makes hatched shapes fine, but I couldn’t figure out how to make a useful legend.

library(HatchedPolygons)

hatched_subset <- function(data, var, range, density, angle) {
  data[data[[var]] >= range[1] & data[[var]] < range[2], ] %>%
    as("Spatial") %>%
    hatched.SpatialPolygons(density = density, angle = angle) %>%
    st_as_sf() %>%
    st_set_crs(2193)
}
densities <- seq(0.006, 0.01, 0.001)

m <- tm_shape(st_union(ak)) +
  tm_lines(lwd = 0.75)

for (i in 1:5) {
  m <- m + 
    tm_shape(
      hatched_subset(ak, "jedi", brks[i:(i+1)], densities[i], -45)) +
    tm_lines(lwd = 0.5, col = "black")
}
m + tm_shape(ak) +
  tm_borders(lwd = 0.75) +
  tm_layout(frame = FALSE)

ggplot2

ggpattern from the aptly named coolbutuseless stable is a lot of fun.

I played around with persuading it to make a ‘serious’ hatched choropleth—which to be clear, is possible—but in the end settled for this much more satisfying weird effect.

library(ggpattern)

ggplot(ak) +
  geom_sf_pattern(aes(pattern_angle = jedi),
    pattern_density = 0.001, 
    pattern_spacing = 0.01,
    pattern_fill = "black",
    fill = "white",
    colour = "white",
    linewidth = 0.5 * 25.4 / 72.27,
  ) + 
  guides(pattern_angle = guide_legend(title = "% Jedi")) +
  geom_sf(data = st_union(ak), linewidth = 0.75, fill = "#00000000") +
  theme_void()