ordering nodes in riverplot

Question

I'm currently developing a riverplot diagram with the riverplot package. However, I struggle to get my data ordered on the plot. Let me explain with an example:

library(riverplot)
df.nodes <- cbind.data.frame(c("A", "B", "C", "D", "E", "F", "G", "H","I","K"), c(1,1,1,1,2,2,2,2,3,3))
colnames(df.nodes) <- c("ID", "x")
df.edges <- cbind.data.frame(c("A", "B", "C", "D", "E", "F", "G", "H"), c("G", "H", "E", "F", "K", "K","I","I"), rep(1, 8))
colnames(df.edges) <- c("N1", "N2", "Value")
ex.river <- makeRiver(df.nodes, df.edges)
plot(ex.river)

As you can see, the edges in this example cross each other. This becomes very confusing when you have more nodes, than in my example. My question: do you have any creative idea how to order the nodes that I get something like this: (this is not a code, but the output order on the graph that I would like)

D   F
C   E   K
B   H   I
A   G

The edges should cross each other the least possible. As I understand the df.nodes defines the order of the nodes, so I should reorder df.nodes somehow. Of course I could order the df.nodes manually, but if you have a lots of nodes that gets tricky.

Any ideas are much appreciated.

Uwe · Accepted Answer · 2017-06-09T08:32:33.230

This was a tricky one, it require a lot of join operations using data.table. There are possibly smarter solutions. However, this one is working for the given data set.

The basic idea is to sort the nodes and edges from left to right.

Data

df.nodes <- data.frame(ID = c("A", "B", "C", "D", "E", "F", "G", "H","I","K"),
                       x  = c(rep(1:2, each = 4L), 3L, 3L), 
                       stringsAsFactors = FALSE)
df.edges <- data.frame(N1 = c("A", "B", "C", "D", "E", "F", "G", "H"), 
                       N2 = c("G", "H", "E", "F", "K", "K","I","I"), 
                       Value = rep(1L, 8),
                       stringsAsFactors = FALSE)

library(data.table)   # CRAN version 1.10.4 used
# coerce to data.table and use abbreviated object names
edt <- setDT(df.edges)
ndt <- setDT(df.nodes)

Get edge positions

# add x positions of nodes to edges
# two joins required for each of the two nodes of an edge 
edt2 <- ndt[ndt[edt, on = c(ID = "N2")], on = c(ID = "N1")][
  , setnames(.SD, c("N1", "x1", "N2", "x2", "Value"))]
# add unique id number for edge x-positions from left to right
# id reflects order of x pos 1-2, 2-3, ..., 10-11 
edt2[order(x1, x2), e.pos := rleid(x1, x2)]
edt2
#   N1 x1 N2 x2 Value e.pos
#1:  A  1  G  2     1     1
#2:  B  1  H  2     1     1
#3:  C  1  E  2     1     1
#4:  D  1  F  2     1     1
#5:  E  2  K  3     1     2
#6:  F  2  K  3     1     2
#7:  G  2  I  3     1     2
#8:  H  2  I  3     1     2

Sort from left to right

# initialize: get order of nodes in leftmost x position
# update edt2 with row number 
edt2 <- ndt[x == 1L, .(N1 = ID, rn1 = .I)][edt2, on = "N1"]
# loop over edge positions
# determine row numbers (sort order) for nodes from left to right
for (p in edt2[, head(unique(e.pos), -1L)]) {
  edt2[p == e.pos, rn2 := rn1]
  edt2 <- edt2[p == e.pos, .(N1 = N2, rn1 = rn2)][edt2, on = "N1"]
  edt2[, rn1 := dplyr::coalesce(rn1, i.rn1)][, i.rn1 := NULL]
}
edt2[e.pos == last(e.pos), rn2 := rn1]
edt2
#   N1 rn1 x1 N2 x2 Value e.pos rn2
#1:  A   1  1  G  2     1     1   1
#2:  B   2  1  H  2     1     1   2
#3:  C   3  1  E  2     1     1   3
#4:  D   4  1  F  2     1     1   4
#5:  E   3  2  K  3     1     2   3
#6:  F   4  2  K  3     1     2   4
#7:  G   1  2  I  3     1     2   1
#8:  H   2  2  I  3     1     2   2

Extract sort order of nodes from edge table

# extract sort order of all nodes from edge table,
# update node table
ndt <- unique(edt2[, .(ID = c(N1, N2), rn = c(rn1, rn2))], by = "ID")[ndt, on = "ID"]
ndt
#    ID rn x
# 1:  A  1 1
# 2:  B  2 1
# 3:  C  3 1
# 4:  D  4 1
# 5:  E  3 2
# 6:  F  4 2
# 7:  G  1 2
# 8:  H  2 2
# 9:  I  1 3
#10:  K  3 3

Create riverplot

library(riverplot)
# pass sorted node table
# coercion back to data.frame required due to type check in `makeRiver()`
ex.river <- makeRiver(setDF(ndt[order(x, rn), .(ID, x)]), setDF(edt))
plot(ex.river)