3 min read

Week40 - networkD3 0.2.1 -> 0.2.4

2015-10-08
Week40 - networkD3 0.2.1 -> 0.2.4

This Week’s Widget - networkD3

networkD3 from Christopher Gandrud is the second most popular htmlwidget based on Github stars, so I think it is important that it stays in elite shape. Last week networkD3 hit CRAN with version 0.2.4, so I’ll claim networkD3 as the htmlwidget of the week. The main improvements are a cycle-supporting sankeyNetwork (no more infinite loops) and I hope a much better fitting diagonalNetwork and radialNetwork.

For those of you who don’t know or forgot, I am using Github issues on the buildingwidgets repo to list ideas for future htmlwidgets. We only have 12 more in the year long adventure, so I would love to hear from you if you have any particular widgets that intrigue you.

I also wanted to highlight scatterD3 which has turned into a very nice and powerful htmlwidget.

Installation

networkD3 is on CRAN, so installation is simple.

install.packages("networkD3")

For the latest and greatest, you can use devtools::install_github.

devtools::install_github("christophergandrud/networkD3")

Examples

diagonal/radialNetwork

diagonalNetwork and radialNetwork needed a little help automatically sizing themselves, so the user didn’t need to endlessly fiddle with margins. Also, the margins argument only supported a right margin. Now, margins can be a single number if you only want right, but now margins also can be a list with any of top, right, bottom, or left.

Let’s take a look at what happens if we make a tiny plot of a big network. Change the size of your screen and see that we also get a little bit of “responsive”-ness.

library(networkD3)

URL <- paste0("https://cdn.rawgit.com/christophergandrud/networkD3/",
              "master/JSONdata//flare.json")

## Convert to list format
Flare <- jsonlite::fromJSON(URL, simplifyDataFrame = FALSE)

## Recreate Bostock example from http://bl.ocks.org/mbostock/4063550
diagonalNetwork(
  List = Flare, fontSize = 10, opacity = 0.9,
  height = "100%", width = "60%"
)

sankeyNetwork

sankeyNetwork now needs a lot fewer arguments to work. The function tries to be “smart” about guessing the structure of your data, such as in Links, it assumes Source will be column 1 and Target will be column 2 if not explicitly told where to look.

sankeyNetwork would result in a blank screen caused by cycles in the network. These cycles would induce an infinite loop in the sankey plugin. Fortunately, Stefaan Lippens @soxofaan fixed and improved Sankey with d3-plugin-captain-sankey. We were able to just plug it right in. I screwed up the CRAN sankeyNetwork, so to get the same output as below please install from Github.

library(networkD3)

# simple network with cycles
net_cycles <- list(
  links = data.frame(
    source = c(0,0,0,1,1,5),
    target = c(1,2,3,4,5,0),
    value = 10
  ),
  nodes = data.frame(
    name = letters[1:6]
  )
)

# notice how few arguments we need now
# some output but not the nice output I expect
sankeyNetwork(
  net_cycles$links,
  net_cycles$nodes,
  Value = "value"
)

Thanks

Thanks Christopher Gandrud for spearheading networkD3. Thanks Mike Bostock and Jason Davies for what seems like everything, but most especially d3.js.

As always, thanks to

  • Ramnath Vaidyanathan and RStudio for htmlwidgets
  • all the contributors to R and JavaScript