Scripts for scraping divorce demographics by country from Wikipedia and plotting it in R with ggplot2 with the respective country flags in place of points.

data.R

## Script for scraping Wikipedia for data to use with the geom_flag() prototype
## CONTACT: Mikhail Popov (@bearloga // mikhail[at]mpopov[dot]com)
## URL: https://gist.github.com/bearloga/519a701a6a9bc7c3ba9f

# install.packages("import")
library(rvest) # install.packages("rvest")
library(magrittr)

import::from(dplyr, mutate, select, keep_where = filter, left_join, distinct)

population <- read_html("https://en.wikipedia.org/wiki/List_of_countries_and_dependencies_by_population") %>%
  html_nodes("table") %>%
  { .[[1]] } %>%
  html_table() %>%
  mutate(Population = as.integer(gsub(",", "", Population)),
         Country = sub("\\(.*\\)", "", gsub("\\[.*\\]", "", `Country (or dependent territory)`))) %>%
  select(c(Country, Population)) %>%
  distinct()

marriage_divorce <- read_html("https://en.wikipedia.org/wiki/Divorce_demography") %>%
  html_nodes("table") %>%
  { .[[1]] } %>%
  html_table() %>%
  mutate(Country = stringr::str_trim(Country)) %>%
  distinct() %>%
  keep_where(!is.na(`Crude marriage rate`) & !is.na(`Crude divorce rate`))

geom_flag prototype.R

## Prototype for what might someday become geom_flag() for use with ggplot2
## CONTACT: Mikhail Popov (@bearloga // mikhail[at]mpopov[dot]com)
## URL: https://gist.github.com/bearloga/519a701a6a9bc7c3ba9f
## LICENSE: MIT
## SVG Flag Collection by Jakob Flierl

## Prerequisites:
## 1. Install the R packages we'll be using in this script:
install.packages(c("magrittr", "dplyr", "rvest", "import",
                   "ISOcodes", "ggplot2", "grid", "png"))
## 2. Download the MIT-licensed flag-icon-css repository from GitHub:
git2r::clone("https://github.com/lipis/flag-icon-css", "flag-icon-css")
## 3. Convert SVGs to PNGs:
##   Unfortunately, we can only use PNGs and not SVGs, so we need to convert the flags to PNGs,
##   which requires inkscape (for example), either as an Application or as a MacPort.
##   e.g. $> sudo port install inkscape
##   To convert SVGs to PNGs, run the following commands in Terminal:
##     $> cd /path/to/flag-icon-css/flags/4x3/
##     $> for f in *.svg; do g=$(echo $f | sed 's=svg=png='); inkscape -z -w 200 -h 150 -e $g $f; done
##     $> mkdir ../pngs && mv *.png ../pngs/
## ...this will take some time...

library(magrittr)
import::from(dplyr, mutate, left_join, keep_where = filter)
library(ggplot2)

## The flag PNGs are named following the ISO 3166-1 standard. Let's get the country names:
data("ISO_3166_1", package = "ISOcodes")

flags <- data.frame(filename = list.files("flag-icon-css/flags/pngs", "\\.png"),
                    stringsAsFactors = FALSE) %>%
  mutate(filepath = file.path("flag-icon-css/flags/pngs", filename),
         code = toupper(stringr::str_extract(filename, "^[a-z]{2}"))) %>%
  left_join(ISO_3166_1[, c("Name", "Alpha_2")], by = c("code" = "Alpha_2")) %>%
  set_names(c("filename", "filepath", "code", "country")) %>%
  keep_where(!is.na(country))

## Scrape data from Wikipedia (requires rvest & dplyr packages)
# source("data.R")
## To avoid problems later, let's restrict the scraped datasets
##   to countries for which we have the flags for.
population %<>% keep_where(Country %in% flags$country)
marriage_divorce %<>% keep_where(Country %in% population$Country)
marriage_divorce %<>% left_join(population, by = "Country") 

## Let's make the base ggplot that we'll add the flags to later:
gg <- ggplot(data = marriage_divorce,
             aes(x = `Crude marriage rate`,
                 y = `Crude divorce rate`)) +
  geom_point(alpha = 0.1) +
  labs(title = "Divorce demography of countries",
       x = "Number of marriages per 1,000 population",
       y = "Number of divorces per 1,000 population")
# print(gg) # To test that everything so far has worked as it should.

## This next step is going to take a while. We do this to load the flags into the memory
##   as grobs so we can have an easier and faster time adding them to any ggplots.
flag_grobs <- lapply(flags$country, function(country) {
  return(grid::rasterGrob(png::readPNG(flags$filepath[flags$country == country])))
})
names(flag_grobs) <- flags$country

## The following chunk will use the data to inform where the flags should be added:
pseudogeom_flag <- mapply(function(x, y, country, flag_scale = 1/4) {
  return(annotation_custom(flag_grobs[[country]],
                           x - 2 * flag_scale, x + 2 * flag_scale,
                           y - 1.5 * flag_scale, y + 1.5 * flag_scale))
  ## The final annotation is a flag with an aspect ratio of 4:3, centered at the (x, y).
},
x = marriage_divorce$`Crude marriage rate`,
y = marriage_divorce$`Crude divorce rate`,
country = marriage_divorce$Country)

## Plot it!
(gg <- gg + pseudogeom_flag)
## We can also make it look a little better by applying the following theme:
(gg <- gg + ggthemes::theme_tufte(base_family = "Gill Sans") +
  theme(panel.grid = element_line(color = "gray80", size = 1)))

plot.png