CRAN downloads R CMD check Codecov


Take a look at the ggpackets project page!

Easily build components of ggplots without sacrificing the ease of ggplot’s layer parameters and construction syntax.



or install the development version

devtools::install_github("dgkf/ggpackets", build_vignettes = TRUE)

Get Involved

There are plenty of ways to help contribute:

  1. File issues!
    Found a bug? Think the syntax looks ugly? Don’t like the name? Tell me! Issues are the best way to start the conversation.

  2. Write documentation!
    More resources always helps. Found a function unintuitive? Example code and improved function descriptors would be helpful. If you use the package and would feel comfortable writing about a topic not yet covered in a vignette, feel free to contribute a new vignette for it.

  3. Write Unit Tests!
    There’s some pretty sophisticated manipulations going on under the hood to make everything as clean as possible, because of that it’s important to make sure everything stays working the way we expect it to. Unit test contributions always welcome!

  4. Contribute Code!
    Last but not least, code contributors are welcome. Reach out and get in touch if you’re passionate about the goal of the project.

Quick Start

Define common ggplot layer sets together into a singled object. Connect all your plots with a single plot component definition and debug one central location. Build beautiful templates and save them once, reuse them easily and without having to abandon the ggplot construction framework.

ggpk_box_and_scatter <- ggpacket() +
  geom_point(position = position_jitter(width = 0.4), alpha = 0.02) + 
  geom_boxplot(outlier.shape = NA, fill = NA, color = 'black') + 
  geom_text(stat = 'summary', vjust = -1, = function(d) c(
    y = quantile(d, 0.75, names = F) + 1.5 * IQR(d),
    label = length(d)
  )) + 
  theme_linedraw() + 
  scale_color_distiller(palette = "Set1")

Now we can use that template repeatedly with a much simpler ggplot call

ggplot(diamonds, aes(x = cut, y = price, color = carat)) + 
  ggpk_box_and_scatter() + 
  ggtitle('Diamond price distribution by cut')

Handle custom arguments & parameter propegation

In addition to simply wrapping multiple ggplot2 layers, ggpackets can streamline a number of complicated plotting scenarios such as passing arguments to sets of layers, setting default argument values with scoped overrides, routing aesthetic mappings to be reused within specific layers for other aesthetics and scoping data usage over a set of layers.

ggpk_labelled_heatmap <- function(...) {
  ggpacket(...) %+%
    geom_tile(.id = 'tile', color = NA, ...) %+% 
    geom_text(.id = c("text", "text1"), color = "black", vjust = -0.3, 
      fontface = "bold", ...) %+%
    geom_text(.id = c("text", "text2"), 
      aes(label = sprintf("(%.1f)", ..fill..)), 
      color = "black", vjust = 1.1, ...) %+%

In this function we make use of a number of these specialized behaviors.

  1. .id parameters are set to tag specific layers with an identifier, which can be used to prefix arguments to route them to a subset of the ggpacket layers. Multiple IDs can be used, and arguments will filter down into that layer if they match any of the provided IDs.
  2. Ellipsis are first passed to ggpacket(...), which will pass them on as default values to all ggpacket layers.
  3. Ellipsis are also passed at the tail end of each layer call, allowing arguments to mask default values. The placement of the ellipsis determines whether arguments will override or be overridden by the existing parameters. After expanding the ellipsis, the last instance of each argument is used to build the call.
  4. Aesthetics are rerouted using the specialized ..<aesthetic>.. syntax.
  5. We use %+% instead of the commonly-used + to add layers together, which allows ggpackets to accept non-standard arguments before ggplot sends us warnings about them.
ggplot( + 
  aes(x = rowpos, y = colpos, label = treatment, fill = decrease) + 
  ggpk_labelled_heatmap(text.color = "white", text2.alpha = 0.5) + 
  ggtitle('Honeybee population decline in repellent trial grid')