riverplot中的节点排序

riverplot中的节点排序,r,plot,sankey-diagram,riverplot,R,Plot,Sankey Diagram,Riverplot,我目前正在用riverplot包开发一个riverplot图。然而,我很难在绘图上整理数据。让我举例说明: 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(

我目前正在用riverplot包开发一个
riverplot
图。然而,我很难在绘图上整理数据。让我举例说明:

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)
边缘应尽可能相互交叉。据我所知,
df.nodes
定义了节点的顺序,因此我应该以某种方式重新排序
df.nodes
。当然,我可以手动订购df.nodes,但是如果您有很多节点,这会变得很棘手


任何想法都值得赞赏。

这是一个棘手的问题,它需要使用
data.table进行大量连接操作。可能有更聪明的解决方案。然而,这是一个为给定的数据集工作

基本思想是从左到右对节点和边进行排序

资料
df.nodes
D   F
C   E   K
B   H   I
A   G
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)
# 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
# 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 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
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)