A package for R that enhances htmlwidgets with client side, inter-widget interactions (currently linked brushing and filtering).
While the whole point of HTML Widgets is to allow you to easily create interactive HTML visualizations from R, the kind of interactivity they enable is generally intra-widget, meaning interactive things happen within the boundaries of the HTML element they inhabit (hover over a data point to see a tooltip, drag over a range of the time series to zoom in).
The other kind of interactivity is inter-widget, meaning that user gestures on one widget can affect other widgets. The most common example is linked brushing, where selecting points on one view of the data highlights those same observations in other views of the data. (Linked brushing is already possible in some libraries like Plotly, but again, only between sub-plots of a single composite Plotly plot, so still intra-widget.)
Crosstalk can be thought of as an add-on to the htmlwidgets package. It provides widget authors with a set of classes, functions, and conventions for implementing standard interactions (currently, linked brushing and filtering) in a standard way.
By standardizing on a set of interactions, we can let widget users coordinate widgets with surprisingly little code.
Here's an example using R Markdown:
---
title: Fiji earthquakes
output: html_document
---
```{r}
library(htmltools)
library(crosstalk)
library(leaflet)
library(DT)
# Wrap data frame in SharedData
sd <- SharedData$new(quakes)
# Create a filter input
filter_slider("mag", "Magnitude", sd, column = ~mag, step = 0.1)
# Use SharedData like a dataframe with Crosstalk-enabled widgets
leaflet(sd) %>% addTiles() %>% addMarkers()
datatable(sd)
```
We wrap the data frame quakes in a SharedData R6 object to signal to widgets and filter inputs that they should link up to other Crosstalk-enabled widgets/inputs that use the same SharedData object.
Here are the results:
There are inherent limitations in the current version of Crosstalk that widget authors and users need to be aware of.
- Crosstalk currently only works for linked brushing and filtering of views that show individual data points, not aggregate or summary views (where "observations" is defined as a single row in a data frame). For example, histograms are not supported since each bar represents multiple data points; but scatter plot points each represent a single data point, so they are supported.
- Because all data must be loaded into the browser, Crosstalk is not appropriate for large data sets. (There's no hard limit, since HTML widgets require varying amounts of CPU cycles and memory for each data point.)
Crosstalk is still experimental (i.e. APIs may change) and is not yet available on CRAN. Therefore, work on widgets is taking place on GitHub branches until Crosstalk is stable.
Plotly supports linked brushing and filtering for scatterplots for plot_ly plots. ggplotly is not supported.
Note: To make a selection on a Plotly plot, make sure to switch to the "Box Select" or "Lasso Select" tool, as the default is to zoom. You can change the default by adding %>% layout(dragmode = "select") (or "lasso") to your plotly object, as in the example below.
devtools::install_github("ropensci/plotly@joe/feature/crosstalk")library(crosstalk)
library(plotly)
set.seed(100)
sd <- SharedData$new(diamonds[sample(nrow(diamonds), 1000), ])
plot_ly(d, x = carat, y = price, text = paste("Clarity: ", clarity),
mode = "markers", color = carat, size = carat) %>%
layout(dragmode = "select")Leaflet supports linked brushing and filtering for markers, circle markers, and shapes. You can't currently make a selection on a Leaflet map, but selections on other widgets will be reflected with increased/decreased opacity on the map.
devtools::install_github("rstudio/leaflet@joe/feature/crosstalk-filter")library(crosstalk)
library(leaflet)
sd <- SharedData$new(quakes)
leaflet(sd) %>% addMarkers()DT supports filtering.
devtools::install_github("rstudio/DT@joe/feature/crosstalk")library(crosstalk)
library(DT)
sd <- SharedData$new(mtcars)
datatable(sd)Currently, the following filter controls are available. These are built-in to the Crosstalk package.
filter_slider- A range slider that filters on a single (numeric, date, or date/time) variable.filter_checkbox- List of checkboxes to filter based on a categorical variable.filter_select- Selectize-based control to filter based on a categorical variable.
These widgets work almost anywhere HTML widgets themselves do: R Markdown documents, Shiny apps, from the R console.
See the function reference in the Crosstalk package for more details.
TODO
TODO
Yes! TODO