jamesotto852/ggrrr

ggrrr

ggrrr provides purrr-style functions for creating patchwork objects. This is primarily achieved with the map_pwk() family of functions which wrap around patchwork::wrap_plots(), although there are also the map_aln() functions implemented for patchwork::align_patches().

Installation

You can install the development version of ggrrr from GitHub with:

# install.packages("devtools")
devtools::install_github("jamesotto852/ggrrr")

map_pwk()

First, we’ll use map_pwk() to compare different regression models with ggplot2::geom_smooth(). Let’s start by generating some data:

library("tidyverse"); theme_set(theme_minimal())

set.seed(1)

df <- 
  tibble(
    x = runif(100, -10, 10),
    y = 2 + x + .2 * x^2  - .05 * x^3+ rnorm(100, sd = 5),
  )

ggplot(df, aes(x, y)) +
  geom_point()

Now all we need is a function that generates plots and an object to map over. We define f which creates a plot with an estimated trend line according to the specified formula. Next, we apply f over formulas to produce the desired graphic!

library("ggrrr")
#> Loading required package: patchwork

# f: formula -> ggplot object
f <- function(formula) {
  ggplot(df, aes(x, y)) +
    geom_smooth(formula = formula, method = "lm")
}

# list of formulas to apply f over
formulas <- list(
  y ~ x,
  y ~ poly(x, 2),
  y ~ poly(x, 3)
)

map_pwk(formulas, f)

The map_pwk() family of functions (map_pwk(), map2_pwk(), pmap_pwk(), etc.) all have a .patchwork_options argument. This optional argument is a named list, and is passed to patchwork::align_plots(). This allows for strict control over the layout of the resulting plots. Here we show a common use case—setting the number of columns in the grid of plots.

map_pwk(formulas, f, .patchwork_options = list(ncol = 1))

Finally, note that it is often convenient to use patchwork’s & operator for adding ggplot elements to plots in a patchwork:

map_pwk(formulas, f, .patchwork_options = list(ncol = 1)) &
  theme(
    axis.title = element_blank(),
    axis.text = element_blank(),
    axis.ticks = element_blank()
  )

A more advanced example with pmap_pwk()

Just like purrr, there are different functions for different types of inputs. For mapping over a data.frame (or lists of lists), there is pmap_pwk(). Here is an example in which pmap_pwk() is used to compare histogram and kernel density estimators:

# plot a geom + data combination
f <- function(geom, data) {
  ggplot(data, aes(x, y)) +
    geom
}

geoms <- list(
  geom_point(size = .5),
  geom_bin2d(),
  geom_density2d_filled()
)

datasets <- list(

  # Unimodal
  data.frame(
    x = rnorm(4e3),
    y = rnorm(4e3)
  ),

  # Bimodal
  data.frame(
    x = rnorm(4e3) + rep(c(-2, 2), length.out = 1e3),
    y = rnorm(4e3)
  ),

  # Constrained Support
  data.frame(
    x = rexp(4e3),
    y = rexp(4e3)
  )

)

# cross product between geoms and datasets,
# saved as a tibble
df_cross <-
  tidyr::expand_grid(
    geom = geoms,
    data = datasets
  )

# use pmap_pwk to apply across rows of tibble or data.frame
pmap_pwk(df_cross, f, .patchwork_options = list(ncol = 3)) &
  theme_void() &
  theme(
    legend.position = "none",
    axis.ticks = element_blank(),
    axis.title = element_blank(),
    axis.text = element_blank()
  )