`.spl_context` has 0 rows and missing columns when no row split is defined.

[wwojciech]

Hi,

I’m writing a custom cfun whose behavior depends on the current column, but .spl_context is empty when there is no row-split defined. See below:

c_test <- function(x, labelstr = "", .spl_context) {
  cat("nrow(.spl_context) is:", nrow(.spl_context), "\n")
  cat("colnames(.spl_context) are:", colnames(.spl_context), "\n\n")

  rcell(1, label = labelstr)
}

lyt <- basic_table() |>
  split_cols_by("ARM") |>
  summarize_row_groups(format = NULL, cfun = c_test)

tbl <- build_table(lyt, DM)

nrow(.spl_context) is: 0 
colnames(.spl_context) are: split value full_parent_df all_cols_n A: Drug X B: Placebo C: Combination 

nrow(.spl_context) is: 0 
colnames(.spl_context) are: split value full_parent_df all_cols_n A: Drug X B: Placebo C: Combination 

nrow(.spl_context) is: 0 
colnames(.spl_context) are: split value full_parent_df all_cols_n A: Drug X B: Drug X B: Placebo C: Combination 

I’m wondering what the reasoning is behind the fact that, when there is no row split in the layout, .spl_context has 0 rows and some of its usual columns are missing. Is this the expected behavior? I would have expected .spl_context to contain all of its usual columns, and also with a single row representing the root row-split context.

I understand that using summarize_row_groups() without any row split in the layout is a legitimate use case. For example:

lyt2 <- basic_table() |>
  split_cols_by("ARM") |>
  summarize_row_groups(label_fstr = "Overall")
tbl2 <- build_table(lyt2, DM)
tbl2
           A: Drug X      B: Placebo    C: Combination
——————————————————————————————————————————————————————
Overall   121 (100.0%)   106 (100.0%)    129 (100.0%)

In this case, summarize_row_groups() is effectively operating at the root level, so I would expect the root context to be represented in .spl_context, even though there are no row splits.

PS. I know I can work around this by using analyze(some_var, afun = c_test).