When two columns have the same name: data_merge_module renames them in an unpredictable way

[cicdguy]

to reproduce:

library(random.cdisc.data)
library(teal.module.clinical)
ADSL <- radsl(cached = TRUE)
ADAE <- radae(cached = TRUE)
ADCM <- radcm(cached = TRUE)

#' Modify ADCM
ADCM$CMINDC <- paste0("Indication_", as.numeric(ADCM$CMDECOD))
ADCM$CMDOSE <- 1
ADCM$CMTRT <- ADCM$CMCAT
ADCM$CMDOSU <- "U"
ADCM$CMROUTE <- "CMROUTE"
ADCM$CMDOSFRQ <- "CMDOSFRQ"
ADCM$CMSTDY <- 1
ADCM[ADCM$CMCAT == "medcl B", ]$CMSTDY <- 20
ADCM[ADCM$CMCAT == "medcl C", ]$CMSTDY <- 150
ADCM$CMENDY <- 500
ADCM[ADCM$CMCAT == "medcl B", ]$CMENDY <- 700
ADCM[ADCM$CMCAT == "medcl C", ]$CMENDY <- 1000
ADCM$CMASTDTM <- ADCM$ASTDTM
ADCM$CMAENDTM <- ADCM$AENDTM
rtables::var_labels(
  ADCM[c("CMINDC", "CMTRT", "CMSTDY", "CMENDY")]
) <- c(
  "Indication",
  "Reported Name of Drug, Med, or Therapy",
  "Study Day of Start of Medication",
  "Study Day of End of Medication"
)
adcm_keys <- c("STUDYID", "USUBJID", "ASTDTM", "CMSEQ", "ATC1", "ATC2", "ATC3", "ATC4")

app <- init(
  data = cdisc_data(
    cdisc_dataset("ADSL", ADSL, code = "ADSL <- radsl(cached = TRUE)"),
    cdisc_dataset("ADAE", ADAE, code = "ADAE <- radae(cached = TRUE)"),
    cdisc_dataset("ADCM", ADCM,
                  code = 'ADCM <- radcm(cached = TRUE)
      ADCM$CMINDC <- paste0("Indication_", as.numeric(ADCM$CMDECOD))
      ADCM$CMDOSE <- 1
      ADCM$CMTRT <- ADCM$CMCAT
      ADCM$CMDOSU <- "U"
      ADCM$CMROUTE <- "CMROUTE"
      ADCM$CMDOSFRQ <- "CMDOSFRQ"
      ADCM$CMSTDY <- 1
      ADCM[ADCM$CMCAT == "medcl B", ]$CMSTDY <- 20
      ADCM[ADCM$CMCAT == "medcl C", ]$CMSTDY <- 150
      ADCM$CMENDY <- 500
      ADCM[ADCM$CMCAT == "medcl B", ]$CMENDY <- 700
      ADCM[ADCM$CMCAT == "medcl C", ]$CMENDY <- 1000
      ADCM$CMASTDTM <- ADCM$ASTDTM
      ADCM$CMAENDTM <- ADCM$AENDTM
      rtables::var_labels(
        ADCM[c("CMINDC", "CMTRT", "CMSTDY", "CMENDY")]) <- c(
          "Indication",
          "Reported Name of Drug, Med, or Therapy",
          "Study Day of Start of Medication",
          "Study Day of End of Medication")',
                  keys = adcm_keys
    ),
    check = TRUE
  ),
  modules = root_modules(
    tm_g_pp_patient_timeline(
      label = "Vitals",
      dataname_adae = "ADAE",
      dataname_adcm = "ADCM",
      parentname = "ADSL",
      patient_col = "USUBJID",
      plot_height = c(600L, 200L, 2000L),
      cmtrt = choices_selected(
        choices = variable_choices(ADCM, "CMTRT"),
        selected = "CMTRT",
      ),
      aeterm = choices_selected(
        choices = variable_choices(ADAE, "AETERM"),
        selected = c("AETERM")
      ),
      aetime_start = choices_selected(
        choices = variable_choices(ADAE, "ASTDTM"),
        selected = c("ASTDTM")
      ),
      aetime_end = choices_selected(
        choices = variable_choices(ADAE, "AENDTM"),
        selected = c("AENDTM")
      ),
      dstime_start = choices_selected(
        choices = variable_choices(ADCM, "CMASTDTM"),
        selected = c("CMASTDTM")
      ),
      dstime_end = choices_selected(
        choices = variable_choices(ADCM, "CMAENDTM"),
        selected = c("CMAENDTM")
      ),
      aerelday_start = choices_selected(
        choices = variable_choices(ADAE, "ASTDY"),
        selected = c("ASTDY")
      ),
      aerelday_end = choices_selected(
        choices = variable_choices(ADAE, "AENDY"),
        selected = c("AENDY")
      ),
      dsrelday_start = choices_selected(
        choices = variable_choices(ADCM, "ASTDY"),
        selected = c("ASTDY")
      ),
      dsrelday_end = choices_selected(
        choices = variable_choices(ADCM, "AENDY"),
        selected = c("AENDY")
      )
    )
  )
)

shinyApp(app$ui, app$server)

Notice that ASTDY and AENDY are columns that are passed in from both ADAE and ADCM

Inside of srv_g_patient_timeline:

  p_timeline_merged_data <- data_merge_module(
    datasets = datasets,
    data_extract = list(
      dsrelday_start, dsrelday_end,
      aerelday_start, aerelday_end,
      aeterm, aetime_start,
      aetime_end, dstime_start, dstime_end, cmtrt),
    input_id = c(
      "dsrelday_start", "dsrelday_end",
      "aerelday_start", "aerelday_end",
      "aeterm", "aetime_start",
      "aetime_end", "dstime_start", "dstime_end", "cmtrt")
  )

If you press the Show R code button on the app:

data_merge_module tries to resolve this conflict by renaming. However, the renaming is unpredictable.

If instead, the developer rearranges the order:

  p_timeline_merged_data <- data_merge_module(
    datasets = datasets,
    data_extract = list(
      aeterm, aetime_start,
      aetime_end, dstime_start, dstime_end, cmtrt,
      dsrelday_start, dsrelday_end,
      aerelday_start, aerelday_end),
    input_id = c(
      "aeterm", "aetime_start",
      "aetime_end", "dstime_start", "dstime_end", "cmtrt",
      "dsrelday_start", "dsrelday_end",
      "aerelday_start", "aerelday_end")
  )

Pressing SRC:

Solution:

ASTDY -> ADAE.ASTDY
AENDY -> ADAE.AENDY

ASTDY -> ADCM.ASTDY
AENDY -> ADCM.AENDY

This actually makes more sense, too.

Provenance:

Creator: junlue

[cicdguy]

It adds input_id as a prefix. This should be unique inside the call (i.e. repetitions are not allowed). This however isn't true for a dataname. What if you need to do merge of multiple selectors from the same dataset?

SELECTOR1:
id: id1
ADAE
select: X, Y, Z
filter: COL_1 = VAL_1
key: X, Y

SELECTOR2:
id: id2
ADAE
select:  X, Y, Z
filter: COL_1 = VAL_2
key: X, Y

Provenance:

Creator: pawelru

[cicdguy]

@pawelru

I agree that my solution is too simplistic and would not work always work.

But then we should consider other solutions, because the issue still remains: the current renaming scheme is unpredictable.

Provenance:

Creator: junlue

[kpagacz]

It adds input_id as a prefix. This should be unique inside the call (i.e. repetitions are not allowed). This however isn't true for a dataname. What if you need to do merge of multiple selectors from the same dataset?

SELECTOR1:
id: id1
ADAE
select: X, Y, Z
filter: COL_1 = VAL_1
key: X, Y

SELECTOR2:
id: id2
ADAE
select:  X, Y, Z
filter: COL_1 = VAL_2
key: X, Y

Provenance:

Creator: pawelru

It does, but it doesn't add the same input_id to the same data_extract_spec if the order of input is scrambled. And it probably should, right?