---
title: "Using infoelectoral to make electoral maps"
vignette: >
  %\VignetteIndexEntry{Using infoelectoral to make electoral maps}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

 
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = T, results = T, include = T, warning = F, message = F)
```


## Install and load the packages

Install [infoelectoral](https://ropenspain.github.io/infoelectoral/) and load the packages needed.

```{r}
# install_packages("infoelectoral")
library(infoelectoral)
# Cargo el resto de librerías
library(dplyr)
library(tidyr)
```

## Download the results

Download some results. In this case we download the election for Congress of December 2015.
```{r}
results <- municipios("congreso", "2015", "12") # Descargo los datos
```


## Import the geometries
Import the geometry shapes for the municipalities using [mapSpain](https://github.com/rOpenSpain/mapSpain). 

```{r}
library(mapSpain)
shp <- esp_get_munic_siane(year = "2016") %>% select(LAU_CODE)
shp_ccaa <- mapSpain::esp_get_ccaa_siane()
```


## Recode party names
Since most parties have different names throughout the country, you will need to recode them to group their results. You can use the column `codigo_partido_nacional` included in the resulting `data.frame` that indicates the grouping party code at the national level. After that you'll have to create the complete municipality code (`LAU_CODE`) for the merge with the `sf` object and transform the data from long to wide format.

First, let's group the parties by `codigo_partido_nacional` and `siglas` and sum the votes to see which party codes correspond to the main parties.
```{r}
results %>%
  group_by(codigo_partido_nacional) %>%
  summarise(
    siglas_r = paste(unique(siglas)[1], collapse = ", "),
    votos = sum(votos)
  ) %>%
  arrange(-votos)
```

Then, we'll have to recode the party names and calculate the percentage of votes.

```{r}
results <-
  results %>%
  mutate(
    siglas_r = case_when(
      codigo_partido_nacional == "903316" ~ "PP",
      codigo_partido_nacional == "903484" ~ "PSOE",
      codigo_partido_nacional == "901079" ~ "Cs",
      codigo_partido_nacional %in% c("903736", "905033", "905008", "905041") ~ "Podemos",
      codigo_partido_nacional == "904850" ~ "IU"
    ),
    # Construyo la columna que identifica al municipio (LAU_CODE)
    LAU_CODE = paste0(codigo_provincia, codigo_municipio),
    # Calculo el % sobre censo
    pct = round((votos / censo_ine) * 100, 2)
  ) %>%
  filter(!is.na(siglas_r)) %>%
  # Selecciono las columnas necesarias
  select(codigo_ccaa, LAU_CODE, siglas_r, censo_ine, votos_candidaturas, pct)
```


## Join data and sf object
With the `LAU_CODE` column merge the data with the geometries of the municipalities.

```{r}
shp <- left_join(shp, results, by = "LAU_CODE")
```

## Visualize

At last, we may use [ggplot2](https://ggplot2.tidyverse.org/) to visualize the data. In this case we use `purrr::map` to create a list of plots each of them with their own color gradient scale and `patchwork` to show them together.

```{r fig.align="center", fig.height = 12, fig.width=8}
library(ggplot2)
library(purrr)
library(patchwork)

colores <- c("#0cb2ff", "#E01021", "#612d62", "#E85B2D", "#E01021")
names(colores) <- c("PP", "PSOE", "Podemos", "Cs", "IU")

# Creo una lista de plots
maps <-
  map(names(colores), function(p) {
    shp %>%
      filter(siglas_r == p) %>%
      ggplot() +
      geom_sf(
        aes(fill = pct, color = pct),
        linewidth = 0, show.legend = F
      ) +
      geom_sf(
        data = shp_ccaa, fill = NA, color = "black",
        linewidth = 0.1
      ) +
      facet_wrap(~siglas_r) +
      scale_fill_gradient(
        low = "white", high = colores[p],
        na.value = "grey90", aesthetics = c("fill", "color")
      ) +
      theme_void()
  })


# Uso patchworks para mostrar los plots
wrap_plots(maps, ncol = 2)
```