R Highcharter：Shiny on the fly 中的动态钻取

Posted 2023-03-24

【中文标题】R Highcharter：Shiny on the fly 中的动态钻取

【英文标题】：R Highcharter: dynamic drilldown in Shiny on the fly

【发布时间】：2019-08-05 16:48:01

【问题描述】：

我正在尝试使用highcharter 和shiny 中的动态数据创建一个多层钻取图。在 SO 社区（向@K. Rohde 大喊）的帮助下，能够通过遍历所有可能的钻取来解决这个问题。我实际的闪亮应用程序将有数百个可能的向下钻取，我不想将这些额外的时间添加到应用程序中，而是使用addSingleSeriesAsDrilldown 即时创建向下钻取。不确定如何在 R 中使用它。

下面是我的问题的工作示例，循环遍历所有向下钻取的可能性：

library (shinyjs)
library (tidyr)
library (data.table)
library (highcharter)
library (dplyr)
library (shinydashboard)
library (shiny)

x <- c("Farm","Farm","Farm","City","City","City","Ocean","Ocean")
y <- c("Sheep","Sheep","Cow","Car","Bus","Bus","Boat","Boat")
z <- c("Bill","Tracy","Sandy","Bob","Carl","Newt","Fig","Tony")
a <- c(1,1,1,1,1,1,1,1)

dat <- data.frame(x,y,z,a)

header <- dashboardHeader()
body <- dashboardBody(

  highchartOutput("Working"),
  verbatimTextOutput("trial")

)
sidebar <- dashboardSidebar()

ui <- dashboardPage(header, sidebar, body)

server <- function(input, output, session) 

  output$Working <- renderHighchart(
    #First Tier #Copied
    datSum <- dat %>%
      group_by(x) %>%
      summarize(Quantity = sum(a)
      )
    datSum <- arrange(datSum,desc(Quantity))
    Lvl1dfStatus <- tibble(name = datSum$x, y = datSum$Quantity, drilldown = tolower(name))

    #Second Tier # Generalized to not use one single input
    # Note: I am creating a list of Drilldown Definitions here.

    Level_2_Drilldowns <- lapply(unique(dat$x), function(x_level) 
      # x_level is what you called 'input' earlier.
      datSum2 <- dat[dat$x == x_level,]

      datSum2 <- datSum2 %>%
        group_by(y) %>%
        summarize(Quantity = sum(a)
        )
      datSum2 <- arrange(datSum2,desc(Quantity))

      # Note: The "drilldown" variable has to be unique, this is why we use level 1 plus level 2 names.
      Lvl2dfStatus <- tibble(name = datSum2$y,y = datSum2$Quantity, drilldown = tolower(paste(x_level, name, sep = "_")))

      list(id = tolower(x_level), type = "column", data = list_parse(Lvl2dfStatus))
    )


    #Third Tier # Generalized through all of level 2
    # Note: Again creating a list of Drilldown Definitions here.
    Level_3_Drilldowns <- lapply(unique(dat$x), function(x_level) 

      datSum2 <- dat[dat$x == x_level,]

      lapply(unique(datSum2$y), function(y_level) 

        datSum3 <- datSum2[datSum2$y == y_level,]

        datSum3 <- datSum3 %>%
          group_by(z) %>%
          summarize(Quantity = sum(a)
          )
        datSum3 <- arrange(datSum3,desc(Quantity))

        Lvl3dfStatus <- tibble(name = datSum3$z,y = datSum3$Quantity)

        # Note: The id must match the one we specified above as "drilldown"
        list(id = tolower(paste(x_level, y_level, sep = "_")), type = "column", data = list_parse2(Lvl3dfStatus))
      )
    ) %>% unlist(recursive = FALSE)

    highchart() %>%
      hc_xAxis(type = "category") %>%
      hc_add_series(Lvl1dfStatus, "column", hcaes(x = name, y = y), color = "#E4551F") %>%
      hc_plotOptions(column = list(stacking = "normal")) %>%
      hc_drilldown(
        allowPointDrilldown = TRUE,
        series = c(Level_2_Drilldowns, Level_3_Drilldowns)
      )
  )

  output$trial <- renderText(input$ClickedInput)




shinyApp(ui, server)

以下是使用addSingleSeriesAsDrilldown 的 R 代码示例，但我不确定如何应用它。我需要动态更改JS 字符串。

library(highcharter)
highchart() %>%
  hc_chart(
    events = list(
      drilldown = JS("function(e) 
        var chart = this,
        newSeries = [
          color: 'red',
          type: 'column',
          stacking: 'normal',
          data: [1, 5, 3, 4]
        , 
          type: 'column',
          stacking: 'normal',
          data: [3, 4, 5, 1]
        ]
        chart.addSingleSeriesAsDrilldown(e.point, newSeries[0]);
        chart.addSingleSeriesAsDrilldown(e.point, newSeries[1]);
        chart.applyDrilldown();
      ")
    )
  ) %>%
  hc_add_series(type = "pie", data= list(list(y = 3, drilldown = TRUE), list(y = 2, drilldown = TRUE))) %>%
  hc_drilldown(
    series = list()
  )

【参考方案1】：

这个问题你得到了双重答案。有两种基本方法可以实现您的愿望。一种是使用 Highcharts 提供的向下钻取，即使您必须从 R 后端收集子系列。另一种是简单地替换 Highcharts 向下钻取并实现 R 驱动的向下钻取，仅使用 Highcharts 进行渲染。

由于它可能更容易消化，我会从后者开始。

Shiny 的钻取功能

忘记 Highcharts 可以进行向下钻取。您已经拥有了所需的一切，因为您知道如何添加一个事件广播器，它会在图表上的某个点被点击时告诉您。

为此，您真正使用了renderHighcharts 的反应性，并使用代表当前向下钻取的不同数据集重新渲染图表。该过程如下：单击“农场”列，您现在使用“农场”子集呈现图表。单击下一列，您构建更深的嵌套子集并呈现它。 Highcharts 提供的唯一一项您必须自己做的事情就是添加一个“返回”按钮以再次向上钻取。

下面的解决方案一开始可能会让人感到困惑，因为它由一些反应式表达式组成，这些表达式汇聚成一个反应式数据集，其中包含您当前的钻取状态。请注意，我们必须将当前的钻取状态存储在后端，以便能够向上钻取并钻取到更深的层次。

library (shinyjs)
library (tidyr)
library (data.table)
library (highcharter)
library (dplyr)
library (shinydashboard)
library (shiny)

x <- c("Farm","Farm","Farm","City","City","City","Ocean","Ocean")
y <- c("Sheep","Sheep","Cow","Car","Bus","Bus","Boat","Boat")
z <- c("Bill","Tracy","Sandy","Bob","Carl","Newt","Fig","Tony")
a <- c(1,1,1,1,1,1,1,1)

dat <- data.frame(x,y,z,a)

header <- dashboardHeader()
body <- dashboardBody(
  actionButton("Back", "Back"),
  highchartOutput("Working"),
  verbatimTextOutput("trial")

)
sidebar <- dashboardSidebar()

ui <- dashboardPage(header, sidebar, body)

server <- function(input, output, session) 
  # To hold the current drilldown status as list, i.e. list("Farm", "Sheep")
  state <- reactiveValues(drills = list())

  # Reactive reacting to the above drill list, giving out a normalized data.frame (category, amount)
  filtered <- reactive(
    if (length(state$drills) == 0) 
      # Case no drills are present.
      data.frame(category = dat$x, amount = dat$a)

     else if (length(state$drills) == 1) 
      # Case only x_level drill is present.
      x_level = state$drills[[1]]
      sub <- dat[dat$x == x_level,]
      data.frame(category = sub$y, amount = sub$a)

     else if (length(state$drills) == 2) 
      # Case x_level and y_level drills are present.

      x_level = state$drills[[1]]
      y_level = state$drills[[2]]
      sub <- dat[dat$x == x_level & dat$y == y_level,]
      data.frame(category = sub$z, amount = sub$a)
    
  )

  # Since Drilldown from Highcharts is not used: Install own click handler that builds up the drill list.
  observeEvent(input$ClickedInput, 
    if (length(state$drills) < 2) 
      # Push drill name.
      state$drills <<- c(state$drills, input$ClickedInput)
    
  )

  # Since Drilldown from Highcharts is not used: Back button is manually inserted.
  observeEvent(input$Back, 
    if (length(state$drills) > 0) 
      # Pop drill name.
      state$drills <<- state$drills[-length(state$drills)]
    
  )

  output$Working <- renderHighchart(

    # Using normalized names from above.
    summarized <- filtered() %>%
      group_by(category) %>%
      summarize(Quantity = sum(amount))

    summarized <- arrange(summarized, desc(Quantity))
    tibbled <- tibble(name = summarized$category, y = summarized$Quantity)

    # This time, click handler is needed.
    pointClickFunction <- JS("function(event) Shiny.onInputChange('ClickedInput', event.point.name);")

    highchart() %>%
      hc_xAxis(type = "category") %>%
      hc_add_series(tibbled, "column", hcaes(x = name, y = y), color = "#E4551F") %>%
      hc_plotOptions(column = list(stacking = "normal", events = list(click = pointClickFunction)))
  )

  output$trial <- renderText(input$ClickedInput)


shinyApp(ui, server)

Highcharts 的下钻功能

这里的情况是，您需要将数据从后端发送到 javascript 以使用图表库中的 addSeriesAsDrilldown 方法。这以一种异步方式工作：Highcharts 提醒某些点被请求向下钻取（通过单击它）。然后后端必须计算相应的数据集，然后将数据集报告回 Highcharts 以便它可以被渲染。我们为此使用 CustomMessageHandler。

我们不会在原始 Highcharts 中添加任何向下钻取系列，但我们会告诉 Highcharts 在请求向下钻取时它必须发送什么关键字（向下钻取事件）。请注意，这不是点击事件，而是更专业的事件（仅在下钻可用时）。

我们发回的数据格式必须正确，所以在这里你需要了解一下 Highcharts 的 api（JS，不是 highcharter）。

创建向下钻取数据的方法有很多，所以我在这里编写了另一个更通用的函数。然而，最重要的是您使用的级别 ID 可用于确定我们当前所处的过滤级别。代码中有一些cmets来指出这些情况。

library (shinyjs)
library (tidyr)
library (data.table)
library (highcharter)
library (dplyr)
library (shinydashboard)
library (shiny)

x <- c("Farm","Farm","Farm","City","City","City","Ocean","Ocean")
y <- c("Sheep","Sheep","Cow","Car","Bus","Bus","Boat","Boat")
z <- c("Bill","Tracy","Sandy","Bob","Carl","Newt","Fig","Tony")
a <- c(1,1,1,1,1,1,1,1)

dat <- data.frame(x,y,z,a)

header <- dashboardHeader()
body <- dashboardBody(
  highchartOutput("Working"),
  verbatimTextOutput("trial")

)
sidebar <- dashboardSidebar()

ui <- dashboardPage(header, sidebar, body)

server <- function(input, output, session) 
  output$Working <- renderHighchart(
    # Make the initial data.
    summarized <- dat %>%
      group_by(x) %>%
      summarize(Quantity = sum(a))

    summarized <- arrange(summarized, desc(Quantity))
    tibbled <- tibble(name = summarized$x, y = summarized$Quantity)

    # This time, click handler is needed.
    drilldownHandler <- JS("function(event) Shiny.onInputChange('ClickedInput', event.point.drilldown);")

    # Also a message receiver for later async drilldown data has to be set.
    # Note in the JS: message.point is going to be the point ID. Highcharts addSeriesAsDrilldown need a point to attach
    #   the drilldown series to. This is retrieved via chart.get which takes the ID of any Highcharts Element.
    #   This means: IDs are kind of important here, so keep track of what you assign.
    installDrilldownReceiver <- JS("function() 
      var chart = this;
      Shiny.addCustomMessageHandler('drilldown', function(message) 
        var point = chart.get(message.point)
        chart.addSeriesAsDrilldown(point, message.series);
      );
    ")

    highchart() %>%
      # Both events are on the chart layer, not by series. 
      hc_chart(events = list(load = installDrilldownReceiver, drilldown = drilldownHandler)) %>%
      hc_xAxis(type = "category") %>%
      # Note: We add a drilldown directive (= name) to tell Highcharts that this has a drilldown functionality.
      hc_add_series(tibbled, "column", hcaes(x = name, y = y, drilldown = name, id = name), color = "#E4551F") %>%
      hc_plotOptions(column = list(stacking = "normal")) %>%
      hc_drilldown(allowPointDrilldown = TRUE)
  )

  # Drilldown handler to calculate the correct drilldown
  observeEvent(input$ClickedInput, 
    # We will code the drill levels to be i.e. Farm_Car. By that we calculate the next Sub-Chart.
    levels <- strsplit(input$ClickedInput, "_", fixed = TRUE)[[1]]
    # This is just for generalizing this function to work in all the levels and even be expandable to further more levels.
    resemblences <- c("x", "y", "z")

    dataSubSet <- dat

    # We subsequently narrow down the original dataset by walking through the drilled levels
    for (i in 1:length(levels)) 
      dataSubSet <- dat[dat[[resemblences[i]]] == levels[i],]
    

    # Create a common data.frame for all level names.
    normalized <- data.frame(category = dataSubSet[[resemblences[length(levels) + 1]]], amount = dataSubSet$a)

    summarized <- normalized %>%
      group_by(category) %>%
      summarize(Quantity = sum(amount))

    summarized <- arrange(summarized, desc(Quantity))

    tibbled <- tibble(name = summarized$category, y = summarized$Quantity)

    # Preparing the names and drilldown directives for the next level below.
    # If already in "Farm_Car", the name for column "Bob" will be "Farm_Car_Bob"
    nextLevelCodes = lapply(tibbled$name, function(fac) 
      paste(c(levels, as.character(fac)), collapse = "_")
    ) %>% unlist

    tibbled$id = nextLevelCodes

    # This is dynamic handling for when there is no further drilldown possible.
    # If no "drilldown" property is set in the data object, Highcharts will not let further drilldowns be triggered.
    if (length(levels) < length(resemblences) - 1) 
      tibbled$drilldown = nextLevelCodes
    

    # Sending data to the installed Drilldown Data listener.
    session$sendCustomMessage("drilldown", list(
      series = list(
        type = "column",
        name = paste(levels, sep = "_"),
        data = list_parse(tibbled)
      ),
      # Here, point is, as mentioned above, the ID of the point that triggered the drilldown.
      point = input$ClickedInput
    ))
  )

  output$trial <- renderText(input$ClickedInput)


shinyApp(ui, server)

【讨论】：

哇，好棒的答案！！我将采用您的第一个解决方案，因为它更容易理解和复制。太感谢了！在过去的 2 天里，我一直在用头撞墙，试图弄清楚该怎么做。

@Kevin 我的荣幸。

解决方案 1 的唯一缺点是您失去了一些在 highcharter 中向下钻取所带来的平滑过渡，但总体而言它很棒！

真棒答案@K.Rohde :) 这也将帮助我做捷径

@K. Rohde 谢谢。如果没有您的解决方案，我将无法解决我的问题。请参阅 SO 问题 -> ***.com/questions/55393013/…