Can't store values to data in reactive environment

[StefanMunnes]

Description

Hello,
I can't help myself again. For a kind of Wiki-Survey, which I could already implement successfully with surveydown, I fail to save values in a new variable. For the pairwise comparison of statements, an Excel file is loaded in the app.R file at the beginning and the different paired statements are stored randomly in a nested list object. Depending on the selection in a prior question, a set of different pairs is selected from this nested list, which are then displayed and evaluated one after the other. The display and answering works well, also the storing of the chosen statement. However, I can't get the pair of statements that was shown stored in the data with sd_store_value().

To give you an idea what this wiki-survey looks like (but in german), check: https://shiny2.wzb.eu/munnes/wikisurvey/

In the server() function in the app.Rfile I have an observer function, looping through the chosen set of pairs and creating a question for each set. To know which two randomly assigned statements the participants could choose from in this questions, I try to store this two values as a collapsed text value:

  shiny::observe({

    if (!is.null(input$question)) {

      for (pair in seq_along(pairs[[input$question]])) {

        pair_items <- paste(pairs[[input$question]][[pair]], collapse = ",")

        # this is the line that I expect to store each pair to the data, but it doesn't work
        sd_store_value(pair_items, paste0("wiki_pair_", pair))

        sd_question(
          id = paste0("wiki_", pair),
          type = "mc_buttons",
          label = NULL,
          direction = "vertical",
          option = pairs[[input$question]][[pair]]
        )
      }
    }
  })

[jhelvy]

Yes, this is currently on our todo list to fix. The 7th one down on the list says "sd_store_data() function - store any values generated after survey launches in the db". I also just tagged this issue as associated with that item.

Right now, sd_store_data() only stores the values that are static, meaning they are created when the survey initially launches. It will not work with reactive values - things that change later once the survey is up and running.

Here is a simpler example that we have been working with for testing to try and solve this problem:
survey.zip

This survey asks for two numbers then displays the product. The product is computed in the server like this:

observe({
  product <- as.numeric(input$first_number) * as.numeric(input$second_number)
  sd_store_value(product)
  output$product <- shiny::renderUI(product)
})

The last line creating the output$product object is how we're able to display the product in the survey. Ideally you wouldn't need this because if the sd_store_value() call was working you could just use sd_output('product', type = 'value') in the survey. But since the sd_store_value() doesn't actually store the value, we have to add this rendering line. So eventually when we get this fixed, that line won't be needed anymore.

To explain a little more why this isn't working, the sd_store_value() function is very simple. All it does is create a list (session$userData$stored_values <- list()) that keeps a record of stored static values. That list is only updated once right when the survey initially launches (see here in the server code). But it never updates.

So what we need to do is add something somewhere else in the server to update reactive values, which will probably involve updating the code in the sd_store_data() function as well as the server. If you'd like to give it a try, please feel free to do so! We've got several other simpler issues that I think we can more quickly knock out, so we'll come back to this one once we get those squashed.

[jhelvy]

I just created an issue from this discussion because this is an actual bug in how sd_store_data() works. I'll copy my reply above there.

[StefanMunnes]

Just to let you know: For now, I just saved the randomly assigned pairs for the comparison as a collapsed character vector in one variable in the server. This avoids the problem with the sd_store_values() function in the reactive environment when I tried to store it in the reactive loop of creating each question and store each pair separately.
For me, this is a great workaround and there are much more important things to correct/change/add than this, so I don't think I will be able to provide any code to fix this.

[jhelvy]

Okay. The bigger picture problem is that we still don't have a systematic way to store values that are reactively updated in the server into the resulting survey data, and that is something we definitely should be able to do. I already know what we need to do as well, I just need the time to sit down and work it out. We'll eventually get on it.

[jhelvy]

This is now addressed. As of v0.8.2, reactive values will be stored with sd_store_value(). You can also create reactive values that will be stored in the data using sd_reactive(), see here for an example.