Reverse geocoding with Google API.R

# Call library data.table
library(data.table)

# ============================================ #
# Prepare data                                 #
# ============================================ #
# Read your data.
DT <- fread("Hawaii_sites.csv")

# Get only lat-long unique records.
# Is better to work only on unique lat-long combinations 
# because of the Google Maps Geocoding API usage limits.
# You can merge back the results later.
XY.unq <- unique(DT[, .(LAT, LON)])

# Give your Google API key
# A key can look something like xIzaAyFWWziisk2V-IwkHLoQ3flu2vN6xp7Tms0
API_key <- "your key"

# Build google maps API XML link/query for each record
XY.unq[, link := paste0("https://maps.googleapis.com/maps/api/geocode/xml?latlng=", LAT, ",", LON, "&key=", API_key)]
# Check out more at https://developers.google.com/maps/documentation/geocoding/intro#reverse-example
# Their examples are based on json and not xml, but the logics are the same.

# Get XML return for each link with readLines()
# Use lapply because readLines is not vectorized.
XY.unq[, XY_XML := lapply(link, readLines)]
# The new column XY_XML is actually a list of character vectors 
# Note that, the response from the reverse geocoder can be a bit slow 
# e.g. 17-20 sec for 250 queries (links);
# there are also limitations: https://developers.google.com/maps/documentation/geocoding/usage-limits

# Save/Load object
# save(XY.unq, file = "XY.unq.rda")
# load(file = "XY.unq.rda")

# ============================================ #
# Function to get desired data from XML format #
# ============================================ #
Get_from_XML <- function(xml, type){
    # The "type" meanings can be found here: 
    # https://developers.google.com/maps/documentation/geocoding/intro#Types
    # Get index of first match in the xml string
    idx <- grep(pattern = type, x = xml, fixed = TRUE)[1]
    # Replace everything between <> with empty string ''
    # Also pay attention to where the needed data is located in the xml string,
    # therefore, pay attention to the corresponding index.
    # For example, the name of the country is 2 lines up in the xml structure
    # from where “country” searching word is actually located.
    my.gsub <- function(x) gsub(pattern = ' *<.*?> *', replacement = '', x)
    switch(type,
           status            = my.gsub(x = xml[idx]),
           formatted_address = my.gsub(x = xml[idx]),
           administrative_area_level_1 = ,
           administrative_area_level_2 = ,
           administrative_area_level_3 = ,
           administrative_area_level_4 = ,
           administrative_area_level_5 = ,
           country = my.gsub(x = xml[idx-2]),
           stop("Provide a type!")
    )
}

# ============================================ #
# Extract desired address data in new columns  #
# ============================================ #
# Use sapply to vectorize Get_from_XML() function built above
XY.unq[, ":=" 
       (Status  = sapply(XY_XML, function(i) Get_from_XML(xml = i, type = 'status')),
        # status codes possible in a reverse geocoding response at:
        # https://developers.google.com/maps/documentation/geocoding/intro#reverse-response
        Country = sapply(XY_XML, function(i) Get_from_XML(xml = i, type = 'country')),
        Admin1  = sapply(XY_XML, function(i) Get_from_XML(xml = i, type = 'administrative_area_level_1')),
        Admin2  = sapply(XY_XML, function(i) Get_from_XML(xml = i, type = 'administrative_area_level_2')),
        Admin3  = sapply(XY_XML, function(i) Get_from_XML(xml = i, type = 'administrative_area_level_3')),
        Admin4  = sapply(XY_XML, function(i) Get_from_XML(xml = i, type = 'administrative_area_level_4')),
        Admin5  = sapply(XY_XML, function(i) Get_from_XML(xml = i, type = 'administrative_area_level_5')),
        Address = sapply(XY_XML, function(i) Get_from_XML(xml = i, type = 'formatted_address'))
        # Check Address Types and Address Component Types at:
        # https://developers.google.com/maps/documentation/geocoding/intro#Types
)]

# Remove unwanted columns
colnames(XY.unq)
XY.unq[, c("link", "XY_XML") := NULL]

# Left Outer Join DT with XY.unq (using data.table::merge() syntax)
RevGeo <- merge(x = DT, y = XY.unq, by = c("LAT", "LON"), all.x = TRUE)
# or using data.table syntax
# RevGeo <- XY.unq[DT, on = c("LAT", "LON")]

# Save/Load object
# save(RevGeo, file = "RevGeo.rda")
# load(file = "RevGeo.rda")