Get the iso subdivision codes from wikipedia

iso_3166_2_subdivision_scrape.R

library(rvest)
library(dplyr)
library(purrr)
library(tidyr)
# https://en.wikipedia.org/wiki/ISO_3166-2

baseurl <- "https://en.wikipedia.org"
url <- "https://en.wikipedia.org/wiki/ISO_3166-2"

# get the tables from which spring links!
links <- read_html(url) %>%
  html_nodes("table") %>%
  `[`(1) %>% # first two tables
  html_nodes("tr") %>%
  html_nodes("a") %>%
  html_attr("href") %>%
  grep("ISO_3166", ., value = TRUE)

# weird duplicate col names that bind_rows does not want
# from tables where some columns have two headers
fix_dup_names <- function(adf) {
  n <- names(adf)
  if (length(unique(n)) == length(n)) {
    return(adf)
  }
  names(adf) <- paste(n, 1:length(n), sep = "_")
  names(adf)[n == "Code"] <- "Code"
  adf[-1, ]
}

parse_one_page <- function(a_link, quiet = FALSE) {
  if (!quiet) print(a_link)
  read_html(paste0(baseurl, a_link)) %>%
    html_nodes("body table.wikitable.sortable") %>%
    html_table(fill = TRUE) %>%
    map(fix_dup_names) %>%
    data.table::rbindlist(fill = TRUE) %>% # used instead of bind_rows to deal with mixed classes
    as_tibble() %>%
    mutate(country_code = gsub("\\/wiki\\/ISO_3166-2\\:", "", a_link))
}


reshape_tab <- function(adf, quiet = FALSE) {
  if (!quiet) print(adf[1, ])
  adf %>%
    mutate_all(as.character) %>% # sneaky integers sneaking in
    pivot_longer(
      cols = !matches("Code"),
      names_to = "subdivision_name_type",
      values_to = "subdivision_name"
    ) %>%
    rename(code = Code)
}

tabs <- map(links, parse_one_page)
tabs <- discard(tabs, ~ nrow(.x) == 0)
iso_df <- map_df(tabs, reshape_tab)

# fix missing codes
iso_df_filtered <- iso_df %>%
  filter(!is.na(subdivision_name)) %>%
  mutate(
    code = ifelse(is.na(code), `Former code`, code),
    code = ifelse(is.na(code), `Alternative code`, code),
    code = ifelse(is.na(code), `Netherlands ISO 3166-2 code`, code)
  ) %>%
  # get rid of dups with no code
  group_by(subdivision_name, country_code) %>%
  filter(!is.na(code)) %>%
  ungroup() %>%
  # get rid of duplicate rows
  group_by(code, subdivision_name, country_code) %>%
  slice(1L) %>%
  ungroup() %>%
  # sift down
  select(code, country_code, subdivision_name_type, subdivision_name) %>%
  filter(grepl("[n,N]ame", subdivision_name_type) | subdivision_name_type== "Local variant")
   
iso_df_filtered