# =====================================================================================================================
# EThekwini Ward Maps                                                             Andrew Collier (andrew@exegetic.biz)
#                                            Exegetic Analytics / www.exegetic.biz / info@exegetic.biz / @exegeticdata
# =====================================================================================================================

# https://edge.durban/dataset/ethekwini-ward-maps

library(sf)
library(ggplot2)
library(ggmap)
library(dplyr)
library(readr)
library(readxl)
library(janitor)

# DOWNLOAD KML --------------------------------------------------------------------------------------------------------

KMLFILE = "eth.kml"

URL = "http://edge.durban/dataset/64ae3b1f-4ce9-48a3-8136-8c74a139de0f/resource/b6336b5c-49af-4b3a-89fe-e72c1db3a3ea/download/"
#
URL = file.path(URL, KMLFILE)

download.file(URL, KMLFILE)

# LOAD DATA -----------------------------------------------------------------------------------------------------------

# Use ogrinfo to find out the name of the layer.
#
# $ ogrinfo eth.kml
# INFO: Open of `eth.kml'
#       using driver `LIBKML' successful.
# 1: ETH

wards <- st_read(KMLFILE)

# How many records?
#
nrow(wards)
#
# Is this the number of wards in Ethekwini?

st_geometry(wards) %>% head()
st_geometry(wards) %>% tail()
#
# Aha! There are two types of data: POINT (centroid) and MULTIPOLYGON (boundary).

# Let's keep only the polygons.
#
wards <- st_collection_extract(wards, "POLYGON")
#
nrow(wards)

# QUICK MAPS ----------------------------------------------------------------------------------------------------------

# Draw the ward outlines.
#
ggplot(wards) +
  geom_sf()

# Add labels.
#
ggplot(wards) +
  geom_sf() +
  geom_sf_label(aes(label = Name))

# BETTER MAPS ---------------------------------------------------------------------------------------------------------

# Get bounding box.
#
bbox <- st_bbox(wards)
#
names(bbox)
#
# These names differ from what's required by get_stamenmap(), so remove.
#
bbox <- unname(bbox)
#
# Expand the bounding box.
#
bbox <- bbox + c(-1, -1, +1, +1) / 16

map <- get_stamenmap(bbox, zoom = 10)

# Plot map.
#
ggmap(map)

# Plot map and overlay ward boundaries.
#
ggmap(map) +
  geom_sf(data = wards, alpha = 0, col = "#3498db", inherit.aes = FALSE)

# LOAD SKILLS DATA ----------------------------------------------------------------------------------------------------

XLSXFILE = "ethekwini-skills-development-per-ward.xls"

URL = "https://edge.durban/dataset/c5ea99ec-ba22-4b37-973d-9dc30de26d03/resource/1463ecd0-e742-42df-8665-b9c9307d2770/download/"
#
URL = file.path(URL, XLSXFILE)

download.file(URL, XLSXFILE)

skills <- read_excel(XLSXFILE) %>%
  clean_names()

# PREPARE SKILLS DATA -------------------------------------------------------------------------------------------------

skills <- skills %>%
  mutate(
    drivers_licence = ifelse(drivers_licence == "NOT SPECIFIED", NA, drivers_licence),
    drivers_licence = factor(drivers_licence),
    ward_no = as.integer(ward_no),
    ward_no = ifelse(between(ward_no, 1, 110), ward_no, NA),
    ward_no = factor(ward_no, levels = 1:110)
  )

skills_summary <- skills %>%
  group_by(ward_no) %>%
  summarise(
    drivers_licence_fraction = mean(drivers_licence == "YES", na.rm = TRUE),
    age_mean = mean(age, na.rm = TRUE)
  )

wards_augmented <- inner_join(wards, skills_summary, by = c("Name" = "ward_no"))

ggmap(map) +
  geom_sf(
    data = wards_augmented,
    aes(fill = drivers_licence_fraction),
    col = "black",
    alpha = 0.8,
    inherit.aes = FALSE
  ) +
  scale_fill_gradient("% Driver License", limits = c(0, 1), low = "white", high = "#3498db") +
  labs(x = NULL, y = NULL)

ggsave("drivers-license-spatial.png", width = 6, height = 6, dpi = 150)

# OTHER DATA ----------------------------------------------------------------------------------------------------------

# gis.durban.gov.za/DataDownloads/showdocs.asp
# datafirst.uct.ac.za
# scoda.co.za
# odza.opendata.durban