Download multiple plotly plots to PDF Shiny

Question

My Shiny App displays a plotly plot for whatever input the user selects. I want a download button that saves ALL the plots inside a PDF file on the user\'s system. I\'m using R

Answer 1

I think @Stanislaus Stadlmann is on point. For some reason, plotly::export does not work inside a loop of a rmarkdown file. I suspect it is for the same reason knitr::include_graphic does not work inside a loop. A workaround is to use the markdown syntax to insert your images. Here is the rmarkdown file that works:

---
title: "Report"
output: pdf_document
always_allow_html: yes
params:
  n: NA
---

```{r,echo=FALSE,warning=FALSE, results="asis"}
library(plotly)

for (item in params$n) {
  tmpFile <- tempfile(fileext = ".png")
  export(item, file = tmpFile)
  cat("![](",tmpFile,")\n")
}
```

And this is my downloadplot1 function:

  output$downloadplot1 <-  downloadHandler(
    filename = "plots.pdf",
    content = function(file){

      tempReport <- file.path(tempdir(), "report1.Rmd")
      file.copy("download_content.Rmd", tempReport, overwrite = TRUE)

      list.of.measurands <- c("Measurand1", "Measurand2") #....all my measurands

      plots.gen <- lapply(list.of.measurands, function(msrnd){
        plot_ly(dummy.df, x = c(1:nrow(dummy.df)), y = ~get(msrnd), type = 'scatter', mode = 'markers')
      })

      params <- list(n = plots.gen)

      rmarkdown::render(tempReport, output_file = file,
                        params = params,
                        envir = new.env(parent = globalenv())
      )
    }
  )

Answer 2

The Problem

Okay, so after spending a decent amount of time playing around with plotly and knitr, I'm pretty sure that there's a problem with printing plotly graphs in a loop while inside a knitr report. I will file an issue at the plotly repository, because there must be some kind of bug. Even when exporting the graph as .png, then importing it again and displaying it in the knitr report, only one graph at a time can be shown. Weird.

The Solution

Anyhow, I found a solution without using knitr to get a pdf of all graphs that are produced in your Shiny Application. It relies on the staplr package to combine PDF files, so you have to install that package and also install the pdftk toolkit.

Afterwards, use the following code I wrote while adapting your Shiny App:

library(shiny)
library(plotly)
library(staplr)

dummy.df <- structure(list(
  Tid = structure(
    1:24, .Label = c("20180321-032-000001", 
                     "20180321-032-000003", "20180321-032-000004", "20180321-032-000005", 
                     "20180321-032-000006", "20180321-032-000007", "20180321-032-000008", 
                     "20180321-032-000009", "20180321-032-000010", "20180321-032-000011", 
                     "20180321-032-000012", "20180321-032-000013", "20180321-032-000014", 
                     "20180321-032-000015", "20180321-032-000016", "20180321-032-000017", 
                     "20180321-032-000018", "20180321-032-000020", "20180321-032-000021", 
                     "20180321-032-000022", "20180321-032-000024", "20180321-032-000025", 
                     "20180321-032-000026", "20180321-032-000027"), class = "factor"), 
  Measurand1 = c(4.1938661428, 4.2866076398, 4.2527368322, 
                 4.1653403962, 4.27242291066667, 4.16539040846667, 4.34047710253333, 
                 4.22442363773333, 4.19234076866667, 4.2468291332, 3.9844897884, 
                 4.22141039866667, 4.20227445513333, 4.33310654473333, 4.1927596214, 
                 4.15925140273333, 4.11148968806667, 4.08674611913333, 4.18821475666667, 
                 4.2206477116, 3.48470470453333, 4.2483107466, 4.209376197, 
                 4.04040350253333), 
  Measurand2 = c(240.457556634854, 248.218468503733, 
                 251.064523520989, 255.454918894609, 250.780599536337, 258.342398843477, 
                 252.343710644105, 249.881670507113, 254.937548700795, 257.252509533017, 
                 258.10699153634, 252.191362744656, 246.944795528771, 247.527116069484, 
                 261.060987461132, 257.770850218767, 259.844790397474, 243.046373553637, 
                 247.026385356368, 254.288899315579, 233.51454714355, 250.556819253509, 
                 255.8242909112, 254.938735944406), 
  Measurand3 = c(70.0613216684803, 
                 70.5004961457819, 70.8382322052776, 69.9282599322167, 68.3045749634227, 
                 71.5636835352475, 69.1173532716941, 71.3604764318073, 69.5045949393461, 
                 71.2211656142532, 72.5716638087178, 69.2085312787522, 70.7872214372161, 
                 70.7247180047809, 69.9466984209057, 71.8433220247599, 72.2055956743742, 
                 71.0348320947071, 69.3848050049961, 69.9884660785462, 73.160638501285, 
                 69.7524898841488, 71.1958302879424, 72.6060886025082)), 
  class = "data.frame", row.names = c(NA, 24L)
)

# Define UI for application
ui <- fluidPage(
  titlePanel("Download Demo"),
  sidebarLayout(
    sidebarPanel(
      selectInput(inputId = "variable",
                  label = "Plot Measurand",
                  choices = colnames(dummy.df)[2:11]
      ),
      hr(),
      downloadButton("downloadplot1", label = "Download plots")
    ),
    mainPanel(
      plotlyOutput("myplot1")
    )
  )
)

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

  # Output graph
  output$myplot1 <- renderPlotly({
    plot_ly(dummy.df, x = c(1:nrow(dummy.df)), y = ~get(input$variable), type = 'scatter',
            mode = 'markers') %>%
      layout(title = 'Values',
             xaxis = list(title = "Points", showgrid = TRUE, zeroline = FALSE),
             yaxis = list(title = input$variable, showgrid = TRUE, zeroline = FALSE))
  })

  # Creating plots individually and passing them as a list of parameters to RMD
  # Example for the first two measurands
  test.plot1 <- reactive({
    plot_ly(dummy.df, x = c(1:nrow(dummy.df)), y = ~Measurand1, type = 'scatter', mode = 'markers')
  })

  test.plot2 <- reactive({
    plot_ly(dummy.df, x = c(1:nrow(dummy.df)), y = ~Measurand2, type = 'scatter', mode = 'markers')
  }) 

  output$downloadplot1 <-  downloadHandler(
    filename = "plots.pdf",
    content = function(file){

      # Set up parameters to pass to Rmd document
      plots <- list(test.plot1(), test.plot2())

      # Plot indices
      ind_vec <- seq_along(plots)

      # Create tempfiles for all plots
      tfiles <- sapply(ind_vec, FUN = function(x)
        return(tempfile(fileext = ".pdf")))

      # create tempfiles for the plots with the second page deleted
      tfiles_repl <- sapply(ind_vec, FUN = function(x)
        return(tempfile(fileext = ".pdf")))

      # Save the objects as .pdf files
      for (i in ind_vec) {
        # Export files
        export(plots[[i]], tfiles[[i]])

        # Remove second page bc for some reason it is whitespace
        staplr::remove_pages(2, input_filepath = tfiles[[i]], 
                             output_filepath = tfiles_repl[[i]])
      }

      # Combine the plots into one pdf
      staplr::staple_pdf(input_files = tfiles_repl, output_filepath = file)

      # Remove .pdf files
      lapply(tfiles, FUN = file.remove)
      lapply(tfiles_repl, FUN = file.remove)
    }
  )
}

# Run the application 
shinyApp(ui = ui, server = server)

I only adapted the code inside the downloadHandler() function. This code basically produces .pdf files of all plots that are inside the plots list (where you later have to specify all your 25 plots, I would do this in a loop). Then, it combines all plots into one .pdf, before deleting the second page of each .pdf, which is necessary because for some reason export() produces a PDF with the second page being completely blank.

My Suggestion

If I were you, I would want to get rid of plotly at all, and replace it with ggplot2 graphs. It would be way easier to do exactly what you want (including the knitr solution). Graphs created with plotly create an extra layer of complexity, because they are web objects that first have to be converted to static files.