如何在 Shiny 中创建可点击的直方图？

Posted 2023-02-16

【中文标题】如何在 Shiny 中创建可点击的直方图？

【英文标题】：How to create a clickable histogram in Shiny?

【发布时间】：2022-01-04 17:52:18

【问题描述】：

我想在shiny 中创建一个可点击的直方图，但不知道是否可行。

几个月前，我看到了一个可点击的火山图，它为您提供了您点击的内容的表格。

来源：https://2-bitbio.com/2017/12/clickable-volcano-plots-in-shiny.html

我发现的关于创建可点击直方图的最接近的帖子是Click to get coordinates from multiple histogram in shiny

但是，我不想获取坐标。我想要数据框的行名。

有了这个数据框，我每次点击直方图中的一个条时可以得到行名吗？

mtcars <- mtcars %>% 
  select("hp")
mtcars <- as.matrix(mtcars)

闪亮的一个例子（但不可点击）：

library(shiny)
library(ggplot2)
library(scales)
library(dplyr)

ui <- fluidPage(
  
  titlePanel("Histogram"),
  
  sidebarLayout(
    sidebarPanel(
    ),
    
    mainPanel(
      plotOutput("hist"),
    )
  )
)

mtcars <- mtcars %>% 
  select("hp")
mtcars <- as.matrix(mtcars)

server <- function(input, output) 
  
  output$hist <- renderPlot(
    
    pp <- qplot(mtcars, geom = "histogram", bins = 10, xlab="values", 
                ylab="Frequency", main="Histogram",
                fill=I("red"), col=I("black"), alpha=I(0.4))
    
   pp + scale_x_continuous(breaks=pretty(mtcars, n=10))
  )
  
  


shinyApp(ui = ui, server = server)

有人知道怎么做吗？

提前非常感谢！

问候

【参考方案1】：

这是一个很好的问题，它的挑战在于 qplot/ggplot 图表是静态图像。下面的app.r 是我将如何做的一个例子。我很想看看其他方法。

本质上：

创建一个数字序列，这些数字既可用作直方图中的断点，也可用作数据框中的间隔。这些是基于用户输入的，但您可以对它们进行硬编码。 根据值所在的间隔为数据帧中的每一行分配一个“bin”值。 记录用户点击事件的 x 坐标，并根据同一组间隔为其分配一个“bin”值。 对您的数据框进行子集化，并仅保留数据的“bin”值与用户点击事件的 x 坐标的“bin”值匹配的记录。

否则，如果你愿意走 d3 路线，你可以探索 R Views 发布的something like this。

#Load libraries ----------------------------------------------------
library(shiny)
library(ggplot2)
library(scales)
library(dplyr)


# Prepare data -----------------------------------------------------
df <- mtcars
df <- cbind(model = rownames(df), data.frame(df, row.names = NULL)) # setting the rownames as the first column
dm <- df$hp %>% as.matrix()


# UI function ------------------------------------------------------
ui <- fluidPage(

  titlePanel("Histogram"),

  sidebarLayout(
    sidebarPanel(

      tags$h5("I added the below text output only to demonstrate shiny's way for tracking user interaction on static plots. You can click, double-click, or click & drag (i.e. brushing). These functions are AWESOME when exploring scatterplots."),

      tags$h3("Chart click and brushing"),

      verbatimTextOutput("info"),

      tags$h5("Now I'm applying the below UI inputs to the `vec` and `breaks` arguments in `findInterval()` and `qplot()` respectively; I'm using `findInterval()` to bin the values in the dataframe AND to bin the x-value of the user's click event input on the chart. Then we can return the dataframe rows with the same bin values as the x-value of the click input."),

      sliderInput("seq_from_to"
                  , label = h3("Sequence 'From' and 'To'")
                  , min = 0
                  , max = 500
                  , value = c(50, 350)
                  ),

      sliderInput("seq_by"
                  , label = h3("Sequence 'By'")
                  , min = 25
                  , max = 200
                  , value = 50
                  , step = 5)

    ),

    mainPanel(

      plotOutput("hist",
                 click = "plot_click",
                 dblclick = "plot_dblclick",
                 hover = "plot_hover",
                 brush = "plot_brush"),

      dataTableOutput("table")

    )
  )
)


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

  # Render Histogram Plot
  output$hist <- renderPlot(

    # Using the same `qplot` function but inserting the user inputs to set the breaks values in the plot
    pp <- qplot(dm
                , geom = "histogram"
                , breaks = seq(from = input$seq_from_to[1], to = input$seq_from_to[2], by = input$seq_by)
                , xlab = "values"
                , ylab = "Frequency"
                , main = "Histogram"
                , fill = I("red")
                , col = I("black")
                , alpha = I(0.4)
                )

    # Also using the user inputs to set the breaks values for the x-axis
    pp + scale_x_continuous(breaks = seq(from = input$seq_from_to[1], to = input$seq_from_to[2], by = input$seq_by))
  )

  # This is purely explanatory to help show how shiny can read user interaction on qplot/ggplot objects
  # It's taken from the Shiny docs here: https://shiny.rstudio.com/articles/plot-interaction.html
  output$info <- renderText(

    # Retain the x and y coords of the user click event data
    xy_str <- function(e) 
      if(is.null(e)) return("NULL\n")
      paste0("x=", round(e$x, 1), " y=", round(e$y, 1), "\n")
    

    # Retain the x and y range coords of click & drag (brush) data
    xy_range_str <- function(e) 
      if(is.null(e)) return("NULL\n")
      paste0("xmin=", round(e$xmin, 1), " xmax=", round(e$xmax, 1),
             " ymin=", round(e$ymin, 1), " ymax=", round(e$ymax, 1))
    

    # Paste this together so we can read it in the UI function for demo purposes
    paste0(
      "click: ", xy_str(input$plot_click),
      "dblclick: ", xy_str(input$plot_dblclick),
      "hover: ", xy_str(input$plot_hover),
      "brush: ", xy_range_str(input$plot_brush)
    )
  )

  # Back to the story. Set a listener to trigger when one of the following is updated:
  toListen <- reactive(list(
    input$plot_click    # user clicks on the plot
    , input$seq_from_to # user updates the range slider
    , input$seq_by      # user updates the number input
    )
  )

  # When one of those events are triggered, update the datatable output
  observeEvent(toListen(), 

    # Save the user click event data
    click_data <- input$plot_click
    print(click_data) # during your app preview, you can watch the R Console to see what click data is accessible

    # Assign bin values to each row using the intervals that are set by the user input
    df$bin <- findInterval(dm, vec = seq(from = input$seq_from_to[1], to = input$seq_from_to[2], by = input$seq_by))

    # Similarly assign a bin value to the click event based on what interval the x values falls within
    click_data$x_bin <- findInterval(click_data$x, vec = seq(from = input$seq_from_to[1], to = input$seq_from_to[2], by = input$seq_by))

    # Lastly, subset the df to only those records within the same interval as the click event x-value
    df_results <- subset(df, bin == click_data$x_bin)

    # Select what values to view in the table
    df_results <- df_results %>% select(model, hp)

    # And push these back out to the UI
    output$table <- renderDataTable(df_results,
                                     options = list(
                                       pageLength = 5
                                     )
    )

  )




shinyApp(ui = ui, server = server)

【讨论】：

非常感谢！你的回答很完整。以防万一，你知道我怎样才能显示刷子的桌子吗？除了“点击选项”之外，我也想拥有那个选项。我在想brushedPoints，但我需要 x 和 y，而我没有 y。 (shiny.rstudio.com/reference/shiny/0.12.0/brushedPoints.html)

【参考方案2】：

嗯，有人回答了。由于我花时间把它放在一起，这里是另一个潜在的解决方案。

library(shiny)
library(ggplot2)
library(scales)
library(dplyr)
library(DescTools)                             # added for Closest()

ui <- fluidPage(                    
  
  titlePanel("Histogram"),
  sidebarLayout(
    sidebarPanel(
    ),
    
    mainPanel(
      plotOutput("hist", click = 'plot_click'),  # added plot_click
      verbatimTextOutput("x_value"),             # added queues for interactivity
      verbatimTextOutput("selected_rows")        # added table for bin values
    )
  )
)

# this can be a dataframe or matrix for qplot or ggplot 
           # (not sure if there was another reason you had this code?)
# mtcars <- mtcars %>% 
#   select("hp")                      # if you only want hp
# mtcars <- as.matrix(mtcars)         # I suggest making row names a column
                                      # to keep 2 columns 

pp <- ggplot(mtcars) +
  geom_histogram(aes(x = hp),
                 bins = 10,
                 fill = "red",
                 color = "black",
                 alpha = .4) +
  labs(x = "values",
       y = "Frequency",
       title = "Histogram")

# extract data from plot to find where each value falls within the histogram bins
        # I kept the pkg name, function in more than one library
bd <- ggplot_build(ggplot2::last_plot())$data[[1]]   

# add the assigned bin number to the mtcars frame; used for filtering matches
mtcars$bins <- lapply(mtcars$hp,
                      function(y) 
                        which(bd$x == Closest(bd$x, y))
                      ) %>% unlist()

server <- function(input, output)  

  output$hist <- renderPlot(
    # moved the plot outside of server, so that global variables could be created
    
    # pp <- qplot(mtcars[,"hp"], geom = "histogram", bins = 10, xlab="values", 
    #             ylab = "Frequency", main = "Histogram",
    #             fill = I("red"), col = I("black"), alpha = I(0.4))
    # scale_x_continuous(breaks=pretty(mtcars, n=10)) # can't use this

    pp
  )
  # # Print the name of the x value                 # added all that's below with server()
  output$x_value <- renderPrint(
    if (is.null(input$plot_click$x)) return()

    # find the closest bin center to show where the user clicked on the histogram
    cBin <- which(bd$x == Closest(bd$x, input$plot_click$x))
    paste0("You selected bin ", cBin)   # print out selected value based on bin center
  )
  # Print the rows of the data frame which match the x value
  output$selected_rows <- renderPrint(
    if (is.null(input$plot_click$x)) return()
    
    # find the closest bin center to show where the user clicked on the histogram
    cBin <- which(bd$x == Closest(bd$x, input$plot_click$x))
    mtcars %>% filter(bins == cBin)
    # mtcars
  )


shinyApp(ui = ui, server = server)

【参考方案3】：

以防万一有人在这篇文章结束时寻找一种包含brushedPoints的方法...灵感来自post，我找到了一种方法！

代码：

#Load libraries ----------------------------------------------------
library(shiny)
library(ggplot2)
library(scales)
library(dplyr)


# Prepare data -----------------------------------------------------
df <- mtcars
df <- cbind(model = rownames(df), data.frame(df, row.names = NULL)) # setting the rownames as the first column
breaks_data = pretty(mtcars$hp, n=10)
my_breaks = seq(min(breaks_data), to=max(breaks_data), by=30)


# UI function ------------------------------------------------------
ui <- fluidPage(
  
  titlePanel("Histogram"),
  
  sidebarLayout(
    sidebarPanel(
      actionButton("draw_plot", "Draw the plot")
    ),
    
    mainPanel(
      
      plotOutput("hist",
                 brush = brushOpts("plot_brush", resetOnNew = T, direction = "x")),
      
      dataTableOutput("table"),
    )
  )
)


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

  observeEvent(input$plot_brush, 
   info_plot <- brushedPoints(df, input$plot_brush)
   output$table <- renderDataTable(info_plot)
   
  )
  
  # If the user didn't choose to see the plot, it won't appear.
  output$hist <- renderPlot(
    df %>% ggplot(aes(hp)) +
      geom_histogram(alpha=I(0.4), col = I("black"), fill = I("red"), bins=10) +
      labs(x = "values",
           y = "Frequency",
           title = "Histogram") +
      scale_x_continuous(breaks = my_breaks)
    
    
  )
  


shinyApp(ui = ui, server = server)