Distinct behaviour of filter_spec in the absence/presence of select_spec

[cicdguy]

There is a distinct behaviour of filter_spec in the absence or presence of select_spec which is described in the following two scenarios:

1. Scenario 1: select_spec and filter_spec are both defined for a variable.
   In case select_spec is deselected, filter_spec becomes obsolete and whatever you change does not have influence. In the example below, filter_spec does not influence facetting.

2. Scenario 2: filter_spec is only defined defined for a variable.
   In this case, filter_spec takes the role of select_spec and in the example below facetting is controlled by filter_spec.

The example below reproduces the 2 scenarios on row_facet variable. To reproduce scenario 2, use the example below. To reproduce scenario 1, uncomment select of row_facet.

library(random.cdisc.data)
devtools::load_all("../teal/")
devtools::load_all("../teal.devel/")
devtools::load_all("../teal.modules.general/")

ADSL <- radsl(cached = TRUE)
ADSL$RACE[1:3] <- NA

dynamic_filter <- filter_spec(
  vars = choices_selected(variable_choices(ADSL), "COUNTRY"),
  multiple = TRUE
)

app <- init(
  data = cdisc_data(
    cdisc_dataset("ADSL", ADSL, code = "ADSL <- radsl(cached = TRUE)
                  ADSL$RACE[1:3] <- NA"),
    check = FALSE
  ),
  modules = root_modules(
    tm_g_scatterplot(
      label = "Scatterplot Choices",
      x = data_extract_spec(
        dataname = "ADSL",
        select = select_spec(
          label = "Select variable:",
          choices = variable_choices(ADSL, c("AGE", "BMRKR1", "BMRKR2")),
          selected = "AGE",
          multiple = FALSE,
          fixed = FALSE
        )
      ),
      y = data_extract_spec(
        dataname = "ADSL",
        select = select_spec(
          label = "Select variable:",
          choices = variable_choices(ADSL, c("AGE", "BMRKR1", "BMRKR2")),
          selected = "BMRKR1",
          multiple = FALSE,
          fixed = FALSE
        )
      ),
      color_by = data_extract_spec(
        dataname = "ADSL",
        select = select_spec(
          choices = variable_choices(ADSL),
          selected = "REGION1",
          multiple = TRUE
        )
      ),
      row_facet = data_extract_spec(
        dataname = "ADSL",
        # select = select_spec(
        #       choices = variable_choices(ADSL),
        #       selected = "SEX",
        #       multiple = FALSE
        #     ),
        filter = list(
          dynamic_filter
        )
      )
    )
  )
)
shinyApp(app$ui, app$server)

The expected behaviour would be to have like scenario 2 where filter_spec takes the role of select_spec when it is deselected or not defined.

Provenance:

Creator: mhallal1

[cicdguy]

Both scenarios are expected behavior.

1. It is actually not true that filter_spec has no effect whenever select_spec is provided but de-selected. It has no effect in the output of this particular module because the data is merged using full_join.

To have filter_spec behave like select_spec in the scenario where a select_spec is provided but de-selected would be an alternative design, and we can discuss it in a normal sprint.

Thus, I am putting this in the backlog with a discussion tag.

Provenance:

Creator: junlue

[cicdguy]

Both scenarios are expected behavior.

1. It is actually not true that filter_spec has no effect whenever select_spec is provided but de-selected. It has no effect in the output of this particular module because the data is merged using full_join.

To have filter_spec behave like select_spec in the scenario where a select_spec is provided but de-selected would be an alternative design, and we can discuss it in a normal sprint.

Thus, I am putting this in the backlog with a discussion tag.

@junlue Don't most modules use full_join? In this case, we will have a problem in all of them.

Provenance:

Creator: mhallal1

[cicdguy]

@mhallal1 good point. This shows some in constistency:

• why empty selection of included select differs from not existing selection

Forcing to include select_spec in second scenario could be also considered. I'm leaving this in backlog as it might involve significant changes

Provenance:

Creator: gogonzo

[kpagacz]

Well, I think we should consider reasons people would use select_spec and filter_spec in the same data_extract_spec. It means they want to select a different variable than used in filter_spec, which means they actually don't want to have filter_spec behave like select_spec. In that case, I think it's most likely, they don't want to filter_spec behave like select_spec if they deselect select_spec because they didn't want the functionality in the first place.

[gogonzo]

Let's wait with the decision after data_merge refactor discussion.

[donyunardi]

Linking this with insightsengineering/NEST-roadmap#36