Brushing options with igraph object in Shiny

Question

What would be the best way to integrate Shiny brushing option with igraph library? Now, I know that usage of brushedPoints works best with ggplot2 but is there a way to be used with igraph? Is there a way to transforme network into ggplot2 object so that panelvar1 and panelvar=NONE can be used?

I have created Shiny application to visualize network. Sorry for the slightly bigger data frames (they are from https://kateto.net/network-visualization tutorial). Please create data frames first, because Shiny app is dependent upon them. App includes brushing option.

# nodes for network
id <- c('s01', 's02', 's03', 's04', 's05', 's06', 's07', 's08', 's09', 's10', 's11', 's12', 's13', 's14', 's15', 's16', 's17')
media <- c('NY Times', 'Washington Post', 'Wall Street Journal', 'USA Today', 'LA Times', 'New York Post', 'CNN', 'MSNBC', 'FOX News', 'ABC', 'BBC', 'Yahoo News', 'Google News', 'Reuters.com', 'NYTimes.com', 'WashingtonPost.com', 'AOL.com')
media.type <- c(1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3)
type.label <- c('Newspaper', 'Newspaper', 'Newspaper', 'Newspaper', 'Newspaper', 'Newspaper', 'TV', 'TV', 'TV', 'TV', 'TV', 'Online', 'Online', 'Online', 'Online', 'Online', 'Online')
audience.size <- c(20, 25, 30, 32, 20, 50, 56, 34, 60, 23, 34, 33, 23, 12, 24, 28, 33)
nodes <- data.frame(id, media, media.type, type.label, audience.size)
str(nodes)

# edges for network
from <- c('s01', 's01', 's01', 's01', 's02', 's02', 's02', 's02', 's03', 's03', 's03', 's03', 's03', 's03', 's03', 's04', 's04', 's04', 's04', 's04', 's05', 's05', 's05', 's05', 's06', 's06', 's06', 's07', 's07', 's07', 's07', 's08', 's08', 's08', 's09', 's10', 's12', 's12', 's12', 's13', 's13', 's14', 's14', 's15', 's15', 's15', 's16', 's16', 's17')
to <- c('s02', 's03', 's04', 's15', 's01', 's03', 's09', 's10', 's01', 's04', 's05', 's08', 's10', 's11', 's12', 's03', 's06', 's11', 's12', 's17', 's01', 's02', 's09', 's15', 's06', 's16', 's17', 's03', 's08', 's10', 's14', 's03', 's07', 's09', 's10', 's03', 's06', 's13', 's14', 's12', 's17', 's11', 's13', 's01', 's04', 's06', 's06', 's17', 's04')
type <- c('hyperlink', 'hyperlink', 'hyperlink', 'mention', 'hyperlink', 'hyperlink', 'hyperlink', 'hyperlink', 'hyperlink', 'hyperlink', 'hyperlink', 'hyperlink', 'mention', 'hyperlink', 'hyperlink', 'hyperlink', 'mention', 'mention', 'hyperlink', 'mention', 'mention', 'hyperlink', 'hyperlink', 'mention', 'hyperlink', 'hyperlink', 'mention', 'mention', 'mention', 'hyperlink', 'mention', 'hyperlink', 'mention', 'mention', 'mention', 'hyperlink', 'mention', 'hyperlink', 'mention', 'hyperlink', 'mention', 'mention', 'mention', 'hyperlink', 'hyperlink', 'hyperlink', 'hyperlink', 'mention', 'hyperlink')
weight <- c(22, 22, 21, 20, 23, 21, 1, 5, 21, 22, 1, 4, 2, 1, 1, 23, 1, 22, 3, 2, 1, 21, 2, 21, 1, 21, 21, 1, 22, 21, 4, 2, 21, 23, 21, 2, 2, 22, 22, 21, 1, 1, 21, 22, 1, 4, 23, 21, 4)
links <- data.frame(from, to, type, weight)
str(links)

# Shiny App

library(shiny)
library(igraph)
# ui----
ui <- fluidPage(
  tabsetPanel(
    tabPanel("tab", #tab ----
             sidebarLayout(
               sidebarPanel("text",
                            helpText("Submit for Analysis"),
                            actionButton("button1", "Submit")
               ),
               mainPanel(
                 verbatimTextOutput("info_n_head"),
                 verbatimTextOutput("info_e_head"),
                 plotOutput("graph_1", brush = "plot_brush", width = "100%", height = '600px'),
                 verbatimTextOutput("info")
               )
             )

    )
  )
)
# server----
server <- function(input, output) {

  observeEvent(input$generate, {
    net <- graph_from_data_frame(d=links(), vertices=nodes(), directed=F)
  })

  output$info_n_head <- renderPrint({
    # Shows peview of Data.
    head(nodes)
  })
  output$info_e_head <- renderPrint({
    # Shows peview of Data.
    head(links)
  })

  observeEvent(input$button1, {
    output$graph_1 <- renderPlot({
      #---For Graph creation----
      net <- graph_from_data_frame(d=links, vertices=nodes, directed=F)

      plot(net, edge.arrow.size=.2, edge.curved=0,
           vertex.color="orange", vertex.frame.color="#555555",
           vertex.label=V(net)$media, vertex.label.color="black",
           vertex.label.cex=.7)
      # The second way to set attributes is to add them to the igraph object.

      # Generate colors based on media type:
      colrs <- c("gray50", "tomato", "gold")
      V(net)$color <- colrs[V(net)$media.type]


      # Compute node degrees (#links) and use that to set node size:
      deg <- degree(net, mode="all")
      V(net)$size <- deg*3
      # Alternatively, we can set node size based on audience size:
      V(net)$size <- V(net)$audience.size*0.7

      # The labels are currently node IDs.
      # Setting them to NA will render no labels:
      V(net)$label.color <- "black"
      V(net)$label <- NA

      # Set edge width based on weight:
      E(net)$width <- E(net)$weight/6

      #change arrow size and edge color:
      E(net)$arrow.size <- .2
      E(net)$edge.color <- "gray80"

      # We can even set the network layout: layout_with_lgl, layout_with_kk
      graph_attr(net, "layout") <- layout_with_lgl
      plot(net)
    })

  })

  output$info <- renderPrint({
    brushedPoints(nodes, input$plot_brush, panelvar1 = "media", panelvar2 = NULL)
  })

}

shinyApp(ui, server)

My goal is to produce data insights for selected nodes by brushing tool. In the current example, when the brushing square is placed over the nodes following error is produced: Error in brushedPoints: brushedPoints: not able to automatically infer xvar from brush.

Have you considered the `ggraph` package? It can plot a `igraph` graph as a `ggplot`. — Stéphane Laurent, May 04 '19 at 14:21
Yes, there is good example here: (https://stackoverflow.com/questions/46408841/retrieving-node-coordinates-from-ggraph-network-chart/46409095#46409095). It is able to get the coordinates. I am looking to establish link to the nodes file and all other attributes. Maybe something is possible with crosstalk library? I am exploring different options at the moment but none of them is giving me that filtering from the nodes file. — vrizma, May 06 '19 at 01:25

missuse · Accepted Answer · 2019-05-09T08:08:53.523

Here is an attempt using ggraph as suggested by Stéphane Laurent in the comments. I omitted graph aesthetics such as node color, size etc.

library(shiny)
library(igraph)
library(ggraph)

ui is the same as yours:

ui <- fluidPage(
  tabsetPanel(
    tabPanel("tab", #tab ----
             sidebarLayout(
               sidebarPanel("text",
                            helpText("Submit for Analysis"),
                            actionButton("button1", "Submit")
               ),
               mainPanel(
                 verbatimTextOutput("info_n_head"),
                 verbatimTextOutput("info_e_head"),
                 plotOutput("graph_1", brush = "plot_brush", width = "100%", height = '600px'),
                 verbatimTextOutput("info")
               )
             )

    )
  )
)

server <- function(input, output) {
 output$info_n_head <- renderPrint({
    # Shows peview of Data.
    head(nodes)
  })
  output$info_e_head <- renderPrint({
    # Shows peview of Data.
    head(links)
  })

  net <- graph_from_data_frame(d = links,
                               vertices = nodes,
                               directed = F)
  set.seed(1)
  net_lay <- ggraph::create_layout(net,
                                   layout = "nicely")

  observeEvent(input$button1, {
    output$graph_1 <- renderPlot({
      ggraph(net_lay) +
        geom_edge_link() +
        geom_node_point() 
    })
  })

  output$info <- renderPrint({
    brush_net <- brushedPoints(net_lay,
                               input$plot_brush)
      nodes[as.character(nodes$id) %in%  as.character(brush_net$name),]
  })

}

shinyApp(ui, server)

First you create a ggraph layout data.frame (using create_layout) which contains the original node data as well as some additional plotting variables, and then you subset this data frame using brushedPoints. Then subset the original nodes data frame based on the variables present in the brushed data frame.

EDIT: updated the server code since I found it did not work in a fresh R session due to misplaced variables.

Thank you, it was great approach to solve this problem. – vrizma May 11 '19 at 00:16 — vrizma, May 11 '19 at 00:16

Brushing options with igraph object in Shiny

1 Answers1