Code chunk with shared expansionContext disappears when input changes

[willgearty]

I'm splitting parts of my shinymeta output into separate chunks; however, to avoid duplicate code, I'm having all of the chunks share the same expansionContext. Here's a small reproducible example cobbled together based on existing shinymeta examples:

library(shiny)
library(shinymeta)
library(dplyr)

ui <- fluidPage(
  selectInput("dataset", "Dataset", c("iris", "pressure")),
  varSelectInput("col_x", "column for x", NULL),
  varSelectInput("col_y", "column for y", NULL),
  verbatimTextOutput("code_1"),
  verbatimTextOutput("code_2"),
  verbatimTextOutput("code_3"),
  plotOutput("plot")
)

server <- function(input, output, session) {
  df <- metaReactive({
    get(..(input$dataset), "package:datasets")
  })
  
  observeEvent(df(), {
    updateVarSelectInput(session, "col_x", data = df())
    updateVarSelectInput(session, "col_y", data = df())
  })
  
  output$plot <- metaRender(renderPlot, {
    "# This is a helpful comment"
    ggplot(..(filtered())) +
      geom_point(aes(x = !!..(input$col_x), y = !!..(input$col_y)))
  })
  
  filtered <- metaReactive({
    ..(df()) %>% distinct()
  })
  
  exp_ctx <- newExpansionContext()
  output$code_1 <- renderPrint({
    expandChain(.expansionContext = exp_ctx,
                quote(library(magrittr)),
                quote(library(ggplot2))
    )
  })
  output$code_2 <- renderPrint({
    expandChain(.expansionContext = exp_ctx,
                invisible(filtered())
    )
  })
  output$code_3 <- renderPrint({
    expandChain(.expansionContext = exp_ctx,
                output$plot()
    )
  })
}

shinyApp(ui, server)

This app loads fine and the chunks look great:

However, when you then change the selected dataset, the middle chunk just disappears:

Any thoughts? Am I missing something in the way I'm implementing the expansionContext?

[willgearty]

@cpsievert any thoughts about this?

[cpsievert]

I think the invisible() is at least part of the issue -- since this is a renderPrint(), I think that would be expected behavior (that an invisible value won't print)

[willgearty]

Thanks for taking a peak! If I remove the invisible() call it results in this before/after:

So the part of the second codeblock that we care about still disappears.

[willgearty]

@cpsievert Just checking in to see if you have any insight about this issue?