Ability to still show columns with 0 data even after filtering

[jenko1979]

Feature description

Hi, there is an undesired consequence of using the active filtering in that when you do this, a column that has 0 data in it disappears entirely form the table. Whilst i understand why this is happening, i feel like there should be an option to retain all levels of your chosen column variables as columns. So for example if you are creating a table of Serious AEs and you do not have any in "Treatment B" and you have 1 in each of the other arms "Treatment A" and "Treatment C", you will only see Treatment A and Treatment C in the table. If there was an option to retain all levels of the selected column variable (in this case TRT01A) in the table, then it would be much more desired to also see Treatment B in the table with 0 as its count.

Code of Conduct

• I agree to follow this project's Code of Conduct.

Contribution Guidelines

• I agree to follow this project's Contribution Guidelines.

Security Policy

• I agree to follow this project's Security Policy.

[llrs-roche]

Hi, I seem to recall that there was a similar petition for a specific module which got a button to drop empty factors or not. @kumamiao should we consider adding such option for all modules?

[kumamiao]

I recall that similar discussion was brought up during the refactoring of filter panel 2 years ago, and the decision was to handle the dropping of empty factors in the filtering step. @donyunardi I might not be with the team back then, do you recall the decision making back then? Thanks.

[donyunardi]

Hi, I seem to recall that there was a similar petition for a specific module which got a button to drop empty factors or not.

Yes, we did this in:
insightsengineering/teal.modules.general#890

the decision was to handle the dropping of empty factors in the filtering step

I don't think this is true. The dropping of empty factors, if desired, usually happens in the module server and not during filtering.

Here's an example where Species is filtered and still have 3 level factors:

Example Code

library(teal)

app <- init(
  data = teal_data(
    iris = iris    
  ),
  modules = modules(
    module = module(
      label = "Module 1",
      server = function(id, data) {
        moduleServer(
          id,
          function(input, output, session) {

            output$output_1 <- renderPrint({
              str(data()[["iris"]])
            })
            
            output$output_2 <- renderPrint({
              data()[["iris"]]
            })

          }
        )
      },
      ui = function(id) {
        ns <- NS(id)
        tagList(
          verbatimTextOutput(ns("output_1")),
          verbatimTextOutput(ns("output_2"))
        )
      }
    )
  ),
  title = "my teal app"
)

shinyApp(app$ui, app$server)

As @llrs-roche pointed out,
We could apply the same design/solution to modules in tmc, like what we did in tmg.

Hi @jenko1979
Could you please share which tmc module that you were using when running to this issue?

[jenko1979]

Hi, the 2 modules that i have seen this behavior are tm_t_events and tm_t_summary. An example can easily be created just by running the run example code on the help page for these modules. I'd imagine this is the behavior across all table modules.

For tm_t_summary, for example if you run the run example and in the app filter on RACE=NATIVE HAWAIIN OR OTHER PACIFIC ISLANDER you lose the 2 treatment columns A and C - instead it would be much nicer to have the option to retain treatment A column with 0 counts, which would then retain the combination column as well (we never produce static outputs without these columns)

For tm_t_events, for example if you run this and in the app filter on AETOXCR=4, AESEV=SEVERE and AEACN=DRUG WITHDRAWN then you lose treatment A column. Again, much better and more typical to retain this column in the table with 0 counts else its not going to be clear to the user whether you've just filtered that treatment column out.

Note - my last column about filtering out treatment columns is important as if user really does want to filter out certain treatment groups by choice then we still need the behavior of not having all levels displayed. So it would get complicated as depending on what variable is being used for filtering you might wish to see different behavior. e.g if you had a situation where you had 3 different treatment arms and someone only wanted to see treatment A and B then they would filter on these, but then if you filtered AESER=Y and suddenly you lost all from treatment A by doing this then it becomes challenging as here you would not be able to check the box to say you wanted all levels as you would then get treatment C back which would be unwanted. But if you didn't check the box you would lose treatment A from your table where perhaps it was desired to see this with 0 counts. So its almost like this is a variable level decision as to whether or not filtering on a variable should lead to removal of the treatment columns or not.