The r-deckgl package makes the open-source JavaScript library deck.gl available within R via the htmlwidgets package.
install.packages("deckgl")
You can install the latest version of r-deckgl from github with:
# install.packages("remotes")
::install_github("crazycapivara/deckgl") remotes
library(deckgl)
Create a deckgl
instance:
deckgl()
Add a basemap:
deckgl() %>%
add_basemap()
Add any kind of layers:
# Grid layer example
data("sf_bike_parking")
<- list(
props extruded = TRUE,
cellSize = 200,
elevationScale = 4,
getPosition = ~lng + lat,
tooltip = "Count: {{count}}"
)
deckgl(zoom = 11, pitch = 45) %>%
add_basemap() %>%
add_grid_layer(
data = sf_bike_parking,
properties = props
)
deckgl
InstanceThe deckgl
function creates the widget / renderer to
which you add layers and other configuration parameters:
<- deckgl(
rdeck latitude = 37.8,
longitude = -122.45,
zoom = 12
%>%
) add_grid_layer(
data = data,
properties = props
)
Due to the generic function add_layer
any kind of layer
defined in the deck.gl
Layer Catalog is supported. The layer type is chosen via the
class_name
parameter, e. g. ScatterplotLayer
or GeoJsonLayer
. Usually you will not use the generic
function but one of the add_*_layer
shortcuts instead:
# Generic function
deckgl() %>%
add_layer("ArcLayer", id, data, properties)
# Shortcut function
deckgl() %>%
add_arc_layer(id, data, properties)
The data
parameter can either be an url to fetch data
from or a data object. In most cases you will pass an object of type
data.frame
to the layers. Use the formula syntax to define
data accessors that deck.gl uses to access the properties of the data
object:
<- list(
props getPosition = ~lng + lat
# ...
)
sf
Data ObjectsAn object of class sf
is a data.frame
with a geometry list-column. Set the layer
prop that fetches the geometry to the geometry list-column of your
sf
object:
# Example: PolygonLayer
<- list(
props getPolygon = ~geometry
# ...
)
With add_source
you can add a source to the widget that
can be used accross layers:
data("bart_stations")
deckgl() %>%
add_source("bart-stations", bart_stations) %>%
add_scatterplot_layer(
source = "bart-stations",
getPosition = ~lng + lat,
# ...
%>%
) add_text_layer(
source = "bart-stations",
getPosition = ~lng + lat,
# ...
%>%
) add_basemap()
Please note that you use the parameter source
instead of
data
.
Layer properties are passed to the add_*_layer
functions
either as named list by the properties
argument or as named
parameters / keyword arguments via the ...
parameter. The
names correspond to the properties of the deck.gl counterparts.
Therefore, please see the deck.gl
Layer Catalog to determine the available parameters for the used
layer. You can also pass a props list and keyword arguments together.
Identical properties are overwritten by the latter ones.
Grid Layer example:
// JavaScript code
const layer = new GridLayer({
id: "grid-layer",
data: data,
extruded: true,
cellSize: 200,
elevationScale: 4,
getPosition: d => [d.lng, d.lat]
; })
# Corresponding R code
# using named arguments
<- deckgl() %>%
deck add_grid_layer(
id = "grid-layer",
data = data,
extruded = TRUE,
cellSize = 200,
elevationScale = 4,
getPosition = ~lng + lat
)
# ... using a named props list
<- list(
props cellSize = 200,
extruded = TRUE,
# ...
)
deckgl() %>%
add_grid_layer(
data = data,
properties = props
)
According to the style conventions in R, camelCased
parameters in deck.gl can also be passed as snake_cased
parameters in R. For example, getPosition
can be passed to
deck.gl as get_position
:
deckgl() %>%
add_grid_layer(
get_position = ~lng + lat,
cell_size = 200,
# ...
)
Use the formula syntax to define data accessors:
<- list(
props getPosition = ~lng + lat # js: d => [d.lng, d.lat]
getFillColor = ~color # js: d => d.color
# ...
)
The example above assumes that your data contains the columns
lng
, lat
and color
.
It is also possible to pass JavaScript code by using the
JS
function in R:
<- list(
props getColor = JS("d => d.capital ? [140, 10, 10] : [60, 10, 10]")
# ...
)
In deck.gl colors are represented by [r, g, b, a]
arrays. In R you can pass hex color codes or color names to all color
props of the add_*_layer
functions. They are automatically
converted to the required format:
deckgl() %>%
add_grid_layer(
colorRange = RColorBrewer::brewer.pal(6, "Blues"),
# ...
)
The tooltip for a layer can be set via the tooltip
parameter. You can either pass a single template string or a list with
the following properties (see also use_tooltip
):
html
: A template string that will be set as the
innerHTML
of the tooltip.style
: A cssText
string that will modefiy
the default style of the tooltip.The tooltip string is a so called “mustache” template in which variable names are identified by the double curly brackets that surround them. The variable names available to the template are given by deck.gl’s pickingInfo.object and vary by layer.
Arc Layer example:
data("bart_segments")
<- list(
props getWidth = 12,
getSourcePosition = ~from_lng + from_lat,
getTargetPosition = ~to_lng + to_lat,
getSourceColor = "yellow",
getTargetColor = "orange",
tooltip = use_tooltip(
html = "{{from_name}} to {{to_name}}",
style = "background: steelBlue; border-radius: 5px;"
)
)
deckgl(zoom = 9.5, pitch = 35) %>%
add_arc_layer(data = bart_segments, properties = props) %>%
add_basemap()
See mustache.js for a complete syntax overwiew.
Controls are displayed as overlays on top of the map / deck. Usually you can set the position and the style of the control. The most basic control is a simple text box:
deckgl() %>%
add_basemap() %>%
add_control(
html = "Plain Base Map",
pos = "top-right",
style = "background: steelblue; color: white"
)
You can add an instance of the ace
editor in JSON mode to the map by using
add_json_editor
:
deckgl() %>%
add_grid_layer(
# ...
%>%
) add_json_editor()
This allows you to change your layer props on the fly. You can toggle the visibility of the editor by pressing “e”.
With add_legend
you can add a custom legend to your
widget:
deckgl() %>%
add_basemap() %>%
add_legend(
colors = c("yellow", "orange"),
labels = c("Cake", "Icecream"),
title = "Sweets"
)
In most cases, you will create the legend automatically using a palette function:
<- 1:10
data_column <- scales::col_bin("Blues", data_column, bins = 5)
pal deckgl() %>%
add_basemap() %>%
add_legend_pal(pal, title = "Blues")
By default, add_basemap
adds a carto
basemap to the widget.
To use basemaps from mapbox it is recommended that
you store your API access token in an environment variable called
MAPBOX_API_TOKEN
:
# If not set globally
#Sys.setenv(MAPBOX_API_TOKEN = "xyz")
deckgl() %>%
add_mapbox_basemap("mapbox://styles/mapbox/light-v9")
You can run the API
examples from the add_*_layer
functions with
example(add_*_layer)
:
example(add_grid_layer)
With the renderDeckgl
and deckglOutput
functions you can use r-deckgl in shiny applications:
library(shiny)
library(deckgl)
<- function(input, output) {
backend $rdeck <- renderDeckgl({
outputdeckgl() %>%
add_grid_layer(
data = sf_bike_parking,
getPosition = ~lng + lat,
cellSize = 400,
pickable = TRUE
%>%
) add_basemap()
})
}
<- fluidPage(
frontend deckglOutput("rdeck")
)
shinyApp(frontend, backend)
To update a deckgl
instance use
deckgl_proxy
in combination with
update_deckgl
.
Furthermore, the onclick
event sends deck.gl’s picking
info object to your shiny application and updates the corresponding
input in the form of input$widget_id_onclick
. For example,
if the widget id is rdeck
, you can access the
pickingInfo
object with
input$rdeck_onclick
:
< -function(input, output) {
backend # ...
observeEvent(input$rdeck_onclick, {
<- input$rdeck_onclick
info print(info$object)
}) }
The JavaScript library of r-deckgl uses webpack as module bundler. Therefore,
you need node.js to build the module.
All JavaScript code is located in the javascript/src
folder
and test components go to
javascript/src/test-components
.
Install deps and build the library from inside the
javascript
folder with:
npm install
npm run build
To spin up the webpack-dev-server run:
npm run start
If the deckgl
widget is not visible in the viewer pane
of RStudio, just open it in your browser by clicking “Show in new
window” and everything will be fine.