2-columns-ggplot.Rmd

---
title: "Two column ggplot2 charts"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(
	echo = TRUE,
	message = FALSE,
	warning = FALSE
)
library(tidyverse)
library(patchwork)
library(cowplot)
library(palmerpenguins)
```

```{r original_charts, include=FALSE}
gg_penguin_scatter <- penguins %>% 
  ggplot(aes(x = bill_length_mm,
             y = bill_depth_mm,
             color = species)) +
  geom_point()

gg_penguin_bar_chart <- penguins %>% 
  count(island, species) %>% 
  ggplot(aes(x = n,
             y = island,
             fill = species)) +
  geom_col()

gg_penguins_timeline <- penguins %>% 
  count(year, species) %>% 
  ggplot(aes(x = year,
             y = n,
             color = species)) +
  geom_line() +
  scale_x_continuous(n.breaks = 3)
```


# Intro

Often you'll want to arrange multiple {ggplot2} charts together with tags and titles, eg:

```{r echo=FALSE}
ptchw_chart <- ( ( gg_penguin_scatter | gg_penguin_bar_chart ) + plot_layout(tag_level = 'new') ) / gg_penguins_timeline + plot_layout(guides = 'collect') 

gg_ptch_chart <- ptchw_chart &
  guides(color = guide_none()) &
  plot_annotation(tag_levels = c('1', 'a'), tag_sep = ".") &
  plot_annotation(title = "Arranging ggplot2 charts") &
  theme_minimal()

ggsave("gg_ptch_chart.png",
       gg_ptch_chart)

gg_ptch_chart
```

There are two different packages you can choose from, {cowplot} and {patchwork}. They are both very popular in the R community and it's almost down to personal choice which one you prefer. I've tried to differentiate them a little bit:

- {cowplot}

  - Charts are explicitly built within a grid using `plot_grid()`. The layout is controlled by specifying the number of rows and columns in the chart
  
  - Nested `plot_grid()` are required to get a single chart to span multiple rows or columns.
  
  - Themes need to be applied to individual charts.
  
  - Legends need to be extracted from charts and manually placed within a `plot_grid()`.
  
> {cowplot} allows extreme precision over your charts. Complex collections of charts with inset charts and custom looking legends can be created.
  
  
- {patchwork}

  - Charts are built using `(p1 + p2) / p3` syntax, `p3` will be placed under `p1` and `p2`.
  
  - Because there is no grid system `p3` will automatically span the entire width of the chart.
  
  - Themes can be applied to the entire patchwork chart.
  
  - Legends can be automatically collected.
  
> {patchwork} feels and behaves like a ggplot2 extension, a lot of things are automated. It can become painful to create extremely customised charts.

## cowplot

We need to create a nested plot_grid() for the timeline chart to span the width of the chart:


```{r}
plot_grid(plot_grid(gg_penguin_scatter, gg_penguin_bar_chart),
          plot_grid(gg_penguins_timeline),
          nrow = 2)
```

In the chart below our goal is to collect together the legends and change the theme:

- The legend is extracted from the bar chart with `get_legend()`

- The legends for all charts are disabled with `theme(legend.position = "none")`

- The legend is attached to the chart using another `plot_grid()`

- The theme has to be changed for all individual charts.

```{r}
cwp_legend <- get_legend(gg_penguin_bar_chart)

cwp_collected <-
  plot_grid(
    plot_grid(
      gg_penguin_scatter + theme_minimal() + theme(legend.position = "none"),
      gg_penguin_bar_chart + theme_minimal() + theme(legend.position = "none")
    ),
    plot_grid(gg_penguins_timeline + theme_minimal() + theme(legend.position = "none")),
    nrow = 2
  )

plot_grid(
  cwp_collected,
  cwp_legend,
  ncol = 2,
  rel_widths = c(8, 1)
)
```

Automatic labelling only works within an individual `plot_grid()`. We therefore need to add manual labels into the chart:

```{r}
cwp_labelled <-
  plot_grid(
    plot_grid(
      gg_penguin_scatter + theme_minimal() + theme(legend.position = "none"),
      gg_penguin_bar_chart + theme_minimal() + theme(legend.position = "none"),
      labels = c("1.a", "1.b")
    ),
    plot_grid(gg_penguins_timeline + theme_minimal() + theme(legend.position = "none"),
              labels = "2"),
    nrow = 2
  )

plot_grid(
  cwp_labelled,
  cwp_legend,
  ncol = 2,
  rel_widths = c(8, 1)
)
```


## Patchwork

Our basic layout is achieved as follows

```{r}
(gg_penguin_scatter + gg_penguin_bar_chart ) / gg_penguins_timeline
```

To change the theme of all subplots we use `&`

```{r}
ptwc_basic <- (gg_penguin_scatter + gg_penguin_bar_chart ) / gg_penguins_timeline

ptwc_basic & theme_minimal()
```

To collect together the legends we go through two steps:

- Remove the guides for the `color` aesthetic, leaving only the fill guide.

- Use `plot_layout(guides = "collect")` to collect the remaining guides together

```{r}
ptchw_collected <- ( gg_penguin_scatter | gg_penguin_bar_chart ) / gg_penguins_timeline + plot_layout(guides = "collect")

ptchw_collected &
  guides(color = guide_none()) &
  theme_minimal()
```

{patchwork} can do automatic tagging, the documentation shows the different systems of counting available.

```{r}
ptchw_auto_tagging <- ( ( gg_penguin_scatter | gg_penguin_bar_chart ) + plot_layout(tag_level = 'new') ) / gg_penguins_timeline + plot_layout(guides = "collect")

ptchw_auto_tagging &
  guides(color = guide_none()) &
  plot_annotation(tag_levels = c('1', 'a'), tag_sep = ".") &
  theme_minimal()
```

Manual tagging

```{r}
ptchw_manual_tagging <- ( gg_penguin_scatter + labs(tag = "scatter") | gg_penguin_bar_chart + labs(tag = "bar") ) / gg_penguins_timeline + labs(tag = "line") + plot_layout(guides = "collect")


ptchw_manual_tagging &
  guides(color = guide_none()) &
  theme_minimal()
```