Additional (unnecessary?) attached libraries in get r code

[kpagacz]

When running this example:

library(scda)
library(dplyr)
library(teal.modules.clinical)

adsl <- synthetic_cdisc_data("latest")$adsl
adex <- synthetic_cdisc_data("latest")$adex

set.seed(1, kind = "Mersenne-Twister")
adex <- adex %>%
  distinct(USUBJID, .keep_all = TRUE) %>%
  mutate(
    PARAMCD = "TDURD",
    PARAM = "Overall duration (days)",
    AVAL = sample(x = seq(1, 200), size = n(), replace = TRUE),
    AVALU = "Days"
  ) %>%
  bind_rows(adex)

app <- init(
  data = cdisc_data(
    cdisc_dataset("ADSL", adsl, code = 'ADSL <- synthetic_cdisc_data("latest")$adsl'),
    cdisc_dataset("ADEX", adex, code = 'ADEX <- synthetic_cdisc_data("latest")$adex
  set.seed(1, kind = "Mersenne-Twister")
  ADEX <- ADEX %>%
  distinct(USUBJID, .keep_all = TRUE) %>%
  mutate(PARAMCD = "TDURD",
         PARAM = "Overall duration (days)",
         AVAL = sample(x = seq(1, 200), size = n(), replace = TRUE),
         AVALU = "Days") %>%
  bind_rows(ADEX)'),
    check = TRUE
  ),
  modules = modules(
    tm_t_exposure(
      label = "Duration of Exposure Table",
      dataname = "ADEX",
      paramcd = choices_selected(
        choices = value_choices(adex, "PARAMCD", "PARAM"),
        selected = "TDURD"
      ),
      col_by_var = choices_selected(
        choices = variable_choices(adex, subset = c("ARM")),
        selected = NULL
      ),
      row_by_var = choices_selected(
        choices = variable_choices(adex, subset = c("ETHNIC", "SEX")),
        selected = "ETHNIC"
      ),
      parcat = choices_selected(
        choices = value_choices(adex, "PARCAT2"),
        selected = "Drug A"
      ),
      add_total = FALSE
    )
  ),
  filter = list(
    ADSL = list(SAFFL = "Y")
  )
)

shinyApp(app$ui, app$server)

I noticed show r code attaches all the behind-the-scenes libraries which might not be needed

[Polkas]

This comes from the design of how we recognize dependencies, utils::sessionInfo()$otherPkgs.

teal/R/get_rcode_utils.R

Line 41 in a0963ef

get_rcode_libraries <- function() {

I think we should leave it if there is no clever move to reduce the list of dependencies.
The static removal of packages which we think are not needed do not have to be stable as we could change sth in the future. teal.data could be used e.g. to rise the connection.
On the other hand I could imagine that teal.slice/shiny/testthat/... will be never used in the show R code.

dependencies <- utils::sessionInfo()$otherPkgs
# this process is dangerous
setdiff(dependencies, c("teal.X", "testthat"))

For sure there could be more advanced solutions, like specifying dependencies when chunks are built.

[gogonzo]

From @mhallal1

Please refer to the comments by @Polkas in #752 (comment)

We should not add blindly all library calls for each SHow R Code call.
We have many options to add only needed dependencies, for each module.

linked to insightsengineering/teal.code#93

[gogonzo]

Closing this one. We can't separate output-related and app-related libraries. We moving to the solution where we provide library calls by evaluating them in teal_data object.

teal_data() |> within({library(...)})