# Load required packages
library(dplyr)
library(ggplot2)
library(scales)
library(stringr)Custom Charting Functions Using ggplot2
R
Data Visualization
ggplot2
While R has a variety of options for 2D graphics and data visualization, it’s hard to beat ggplot2 in terms of features, functionality, and overall visual quality. This post demonstrates how to create customized charting functions for specific chart types using ggplot2 as the underlying visualization engine.
Required Libraries
Sample Dataset
For this demonstration, we’ll use a summarized version of the COVID-19 Data Repository hosted by Johns Hopkins University.
# Load COVID-19 data
df <- read.csv("https://bit.ly/3G8G63u")
# Get top 5 countries by death count
top_countries <- df %>%
group_by(country) %>%
summarise(count = sum(deaths_daily)) %>%
top_n(5) %>%
.$country
print(top_countries)[1] "Brazil" "India" "Mexico" "Russia" "US" Let’s prepare our data for visualization by creating a 7-day moving average of daily confirmed cases for the top five countries:
# Create a data frame with the required information
# Note that a centered 7-day moving average is used
plotdf <- df %>%
mutate(date = as.Date(date, format = "%m/%d/%Y")) %>%
filter(country %in% top_countries) %>%
group_by(country, date) %>%
summarise(count = sum(confirmed_daily)) %>%
arrange(country, date) %>%
group_by(country) %>%
mutate(MA = zoo::rollapply(count, FUN = mean, width = 7, by = 1, fill = NA, align = "center"))Building a Simple Line Chart Function
Let’s start by creating a basic line chart function. Note the use of aes_string() instead of just aes(). This allows us to supply arguments to ggplot2 as strings, making our function more flexible.
# Function definition
line_chart <- function(df,
x,
y,
group_color = NULL,
line_width = 1,
line_type = 1){
ggplot(df, aes(x = !! sym(x),
y = !! sym(y),
color = !! sym(group_color))) +
geom_line(linewidth = line_width,
linetype = line_type)
}
# Test run
line_chart(plotdf,
x = "date",
y = "MA",
group_color = "country",
line_type = 1,
line_width = 1.2)
Creating a Custom Theme
Now that we know how to encapsulate the call to ggplot2 in a more intuitive manner, we can create a customized theme for our charts. This is useful since this theme can be applied to any chart.
custom_theme <- function(plt,
base_size = 11,
base_line_size = 1,
palette = "Set1"){
# Note the use of "+" and not "%>%"
plt +
# Adjust overall font size
theme_minimal(base_size = base_size,
base_line_size = base_line_size) +
# Put legend at the bottom
theme(legend.position = "bottom") +
# Different colour scale
scale_color_brewer(palette = palette)
}
# Test run
line_chart(plotdf, "date", "MA", "country") %>% custom_theme()
Enhancing Our Functions
Let’s add more features to our line_chart() function to make it more versatile:
line_chart <- function(df,
x, y,
group_color = NULL,
line_width = 1,
line_type = 1,
xlab = NULL,
ylab = NULL,
title = NULL,
subtitle = NULL,
caption = NULL){
# Base plot
ggplot(df, aes(x = !! sym(x),
y = !! sym(y),
color = !! sym(group_color))) +
# Line chart
geom_line(size = line_width,
linetype = line_type) +
# Titles and subtitles
labs(x = xlab,
y = ylab,
title = title,
subtitle = subtitle,
caption = caption)
}We’ll also enhance our custom_theme() function to handle different axis formatting options:
custom_theme <- function(plt,
palette = "Set1",
format_x_axis_as = NULL,
format_y_axis_as = NULL,
x_axis_scale = 1,
y_axis_scale = 1,
x_axis_text_size = 10,
y_axis_text_size = 10,
base_size = 11,
base_line_size = 1,
x_angle = 45){
mappings <- names(unlist(plt$mapping))
p <- plt +
# Adjust overall font size
theme_minimal(base_size = base_size,
base_line_size = base_line_size) +
# Put legend at the bottom
theme(legend.position = "bottom",
axis.text.x = element_text(angle = x_angle)) +
# Different colour palette
{if("colour" %in% mappings) scale_color_brewer(palette = palette)}+
{if("fill" %in% mappings) scale_fill_brewer(palette = palette)}+
# Change some theme options
theme(plot.background = element_rect(fill = "#f7f7f7"),
plot.subtitle = element_text(face = "italic"),
axis.title.x = element_text(face = "bold",
size = x_axis_text_size),
axis.title.y = element_text(face = "bold",
size = y_axis_text_size)) +
# Change x-axis formatting
{if(!is.null(format_x_axis_as))
switch(format_x_axis_as,
"date" = scale_x_date(breaks = pretty_breaks(n = 12)),
"number" = scale_x_continuous(labels = number_format(accuracy = 0.1,
decimal.mark = ",",
scale = x_axis_scale)),
"percent" = scale_x_continuous(labels = percent))} +
# Change y-axis formatting
{if(!is.null(format_y_axis_as))
switch(format_y_axis_as,
"date" = scale_y_date(breaks = pretty_breaks(n = 12)),
"number" = scale_y_continuous(labels = number_format(accuracy = 0.1,
decimal.mark = ",",
scale = y_axis_scale)),
"percent" = scale_y_continuous(labels = percent))}
# Capitalise all names
vec <- lapply(p$labels, str_to_title)
names(vec) <- names(p$labels)
p$labels <- vec
return(p)
}Putting It All Together
Now let’s see how our enhanced functions work together to create a polished visualization:
line_chart(plotdf,
x = "date",
y = "MA",
group_color = "country",
xlab = "Date",
ylab = "Moving Avg. (in '000)",
title = "Daily COVID19 Case Load",
subtitle = "Top 5 countries by volume") %>%
custom_theme(format_x_axis_as = "date",
format_y_axis_as = "number",
y_axis_scale = 0.001)
Applying the Custom Theme to Other Chart Types
The beauty of our custom_theme() function is that it can be applied to any ggplot2 object. Let’s create a bar chart to demonstrate this flexibility:
p <- plotdf %>%
mutate(month = format(date, "%m-%b")) %>%
ggplot(aes(x = month, y = MA, fill = country)) +
geom_col(position = "dodge") +
labs(title = "Monthly COVID19 Case load trend",
subtitle = "Top 5 countries",
x = "Month",
y = "Moving Average ('000)")
custom_theme(p,
palette = "Set2",
format_y_axis_as = "number",
y_axis_scale = 0.001)
Benefits of Custom Charting Functions
Creating custom charting functions with ggplot2 offers several advantages:
Consistency: Ensures all charts in your reports or dashboards have a consistent look and feel.
Efficiency: Reduces the amount of code you need to write for commonly used chart types.
Maintainability: Makes it easier to update the style of all charts by modifying a single function.
Simplicity: Abstracts away the complexity of ggplot2 for team members who may not be as familiar with the package.
When to Use Custom Functions vs. Direct ggplot2
It’s worth noting that building customized charting functions using ggplot2 is most useful when you need to create the same type of chart(s) repeatedly. When doing exploratory work, using ggplot2 directly is often easier and more flexible since you can build all kinds of charts (or layer different chart types) within the same pipeline.