Place nodes explicitly with visNetwork (or an Alte

问题:

How can I explicitly place nodes on a visNetwork graph?

Or: How can I recreate that graphic in R using visNetwork or an alternative?

Background: The ultimate goal is to represent Causal Loop Diagrams coming from Vensim files. Placing the nodes explicitly is just the first (crucial) step, because in Causal Loop Diagrams the visual mapping of nodes is part of the information (unlike in general graph theory). So if anybody has advice on the bigger picture aka. 'Bringing Causal Loop Diagram Modeling to R', I'll be more than happy.

What I tried:

library("visNetwork")

nodes <- data.frame(id = 1:3, label = c("one", "two", "three"))
edges <- data.frame(from = c(1,1,2), to = c(2,3,1))

visNetwork(nodes, edges, width = "100%", title = nodes$labels, stringsAsFactors = FALSE) %>% visEdges(arrows = "to")

which plots something like (exact layout will change, because of random seed):

With the Q&A from here I tried to place nodes manually by setting x and y values.

library("visNetwork")

nodes <- data.frame(id = 1:3, label = c("one", "two", "three"), x = c(0,1,2), y = c(0,1,2))
edges <- data.frame(from = c(1,1,2), to = c(2,3,1))

visNetwork(nodes, edges, width = "100%", title = nodes$labels, stringsAsFactors = FALSE) %>% visEdges(arrows = "to")

which plots:

..and I really don't understand what's the correspondance between x, y and the placing on the screen..

Also I could not find anything in the docs for visLayout.

回答1:

It somehow turns out, that the x and y args are not working. Here a solution:

library("visNetwork")

nodes <- data.frame(id = 1:3, label = c("one", "two", "three"))
edges <- data.frame(from = c(1,1,2), to = c(2,3,1))

coords <- as.matrix(data.frame(x = c(0,1,2),
                               y = c(0,1,2),
                               stringsAsFactors = FALSE))

visNetwork(nodes, edges, width = "100%", title = nodes$labels) %>%
    visNodes() %>%
    visOptions(highlightNearest = TRUE) %>%
    visInteraction(navigationButtons = TRUE,
                   dragNodes = TRUE, dragView = TRUE,
                   zoomView = FALSE) %>%
    visEdges(arrows = 'to') %>%
    visIgraphLayout(layout = "layout.norm", layoutMatrix = coords)

For history see also here. Perhaps these links might be helpful for what you want to achive: causaleffect and plot.CLD

回答2:

Using ggraph instead of visNetwork simplifies things.

library(ggraph)
library(igraph)

g <- make_graph(edges = c(1,2,2,1,1,3))
V(g)$name <- c('one', 'two', 'three')

ggraph(g, layout = 'manual', node.positions = data.frame(x = c(1,1,2), y = c(2,1,2.1))) + 
  geom_edge_arc(aes(start_cap = label_rect(node1.name),
                    end_cap = label_rect(node2.name)),
                 angle_calc = 'along',
                 label_dodge = unit(2.5, 'mm'),
                 arrow = arrow(length = unit(4, 'mm'))) + 
  geom_node_text(aes(label = name, x = x, y = y))

This plots

which is (apart from gridlines and colours) what I was searching for.