How to change node and link colors in R googleVis sankey chart

Question

How can node and link colors be changed in R googleVis sankey chart? And link having the same color as its originating node?

library(googleVis)
datSK <- data.frame(From=c(rep("A",3), rep("B", 3)),
                To=c(rep(c("X", "Y", "Z"),2)),
                Weight=c(5,7,6,2,9,4))

Sankey <- gvisSankey(datSK, from="From", to="To", weight="Weight",
                 options=list(
                   sankey="{link: {color: { fill: '#d799ae' } },
                        node: { color: { fill: '#a61d4c' },
                        label: { color: '#871b47' } }}"))
plot(Sankey)

Ultimogeniture

Answer 1

As soon as you have to color links from 2 originated nodes you'll need 2 colors for links. Also you have 5 nodes in total, so you'll need 5 colors for them.

Lets create 2 arrays in JSON format with colors for nodes and links

colors_link <- c('green', 'blue')
colors_link_array <- paste0("[", paste0("'", colors_link,"'", collapse = ','), "]")

colors_node <- c('yellow', 'lightblue', 'red', 'black', 'brown')
colors_node_array <- paste0("[", paste0("'", colors_node,"'", collapse = ','), "]")

Next, insert that array into options:

opts <- paste0("{
        link: { colorMode: 'source',
                colors: ", colors_link_array ," },
        node: { colors: ", colors_node_array ," }
      }" )

And, finally plot graph:

plot( gvisSankey(datSK, from="From", to="To", weight="Weight",
                     options=list(
                       sankey=opts)))

Note, that in options colorMode is set to 'source' which means you would like to color links from originated nodes. Alternatively, set 'target' to color links for destinated nodes

EDIT: add description for multilevel sankeys

It is a bit tricky to find how to assign colors for multilevel sankeys.

We need to create other dateframe:

datSK <- data.frame(From=c(rep("A",3), rep("B", 3), rep(c("X", "Y", "Z"), 2 )),
                To=c(rep(c("X", "Y", "Z"),2), rep("M", 3), rep("N", 3)),
                Weight=c(5,7,6,2,9,4,3,4,5,6, 4,8))

Here we have to change only arrays of colors. Command to built plot is the same Let's assume we want these colors for the nodes and links :

colors_link <- c('green', 'blue', 'yellow', 'brown', 'red')
colors_link_array <- paste0("[", paste0("'", colors_link,"'", collapse = ','), "]")

colors_node <- c('yellow', 'lightblue', 'red', 'black', 'brown', 'green', 'brown')
colors_node_array <- paste0("[", paste0("'", colors_node,"'", collapse = ','), "]")

Result would be :

The most trickiest part is to understand how these colors are assigned:

1. Links are assigned in the order they appear in dataset (row_wise)

2. For the nodes colors are assigned in the order plot is built.

   • From A to X, Y, Z - green
   • From X to M, N - blue
   • From Y to M, N - yellow
   • From Z to M, N - brown
   • From B to X, Y, Z - red

More detailed information on how to format sankey diagram : https://developers.google.com/chart/interactive/docs/gallery/sankey

Brookins

Could you extend your answer for Multilevel Sankeys? – Funky

Is it possible to have two different colors coming out of one node? In other words, can colors be set without using "source" or "target" as colorMode? See: #45291840 – Canfield

Answer 2

I know this is older but in case anyone else is ever stuck on this - I figured out how to make the proper order and generate a string of color nodes so you can have custom colors for certain labels. Shout out to @vadym-b for the data and explaining about the order. Check it out:

#convert to list combined of source and target for color order
# edges is a dataframe from @vadym-b's answer above
edges <- data.frame(From=c(rep("A",3), rep("B", 3), rep(c("X", "Y", "Z"), 2 )),
                    To=c(rep(c("X", "Y", "Z"),2), rep("M", 3), rep("N", 3)),
                    Weight=c(5,7,6,2,9,4,3,4,5,6, 4,8))

#we have to make the order right - you need a list
# that is a combination of From, To, From, To, From, To
nc.df <- c()
for (i in 1:nrow(edges)) {
  nc.df <- c(nc.df, as.character(edges$From[i]), as.character(edges$To[i]))
}

#the actualy parsing - get the unique list and return
# colors based on what the source or target value is
nodeColor <- sapply(unique(nc.df), function(r) {
  if (grepl('A',r)) return('red')
  if (grepl('B',r)) return('red')
  if (grepl('Z',r)) return('green')
  if (grepl('X',r)) return('green')
  if (grepl('Y',r)) return('purple')
  if (grepl('M',r)) return('blue')
  if (grepl('N',r)) return('blue')
  #return a default color if you like
  return('black')
})

#make a sankey
library(googleVis)

# put the color list in a collapsed string
sankey <- gvisSankey(
  edges, 
  chartid = 'Sankey', 
  from="From", 
  to="To", 
  weight="Weight", 
  options=list(
    sankey = paste0("{
      iterations: 0,
      link: {
        colorMode: 'gradient'
      },
      node: {
        colors: ['",paste(nodeColor,collapse="','"),"']
      }
    }")
  )
)

plot(sankey)

Goodkin

I absolutely love this! Question though. What if you wanted A -> X to be one color and A -> Y to be another? how would the code change? – Rollin

@walker I think you could add another for loop but have it paste the edges$From and edges$To together as a single string so it's like c('AZ','BY',...), then sapply on that vector with grepl calls that look for that for -> to link.... I haven't tried this though – Goodkin

Answer 3

I have put on github a piece of code that does that.

#TOPLOTS[,1] = from ; TOPLOTS[,1] = to
names_pahtwayorder<-unlist(data.frame(t(TOPLOTs[,1:2])))
names_pahtwayorder<-names_pahtwayorder[!duplicated(names_pahtwayorder)]
names(names_pahtwayorder)<-NULL; names_pahtwayorder

https://github.com/SkanderMulder/ExtractIPA/blob/master/functionSankey.r

Internationale