Include literal curly brackets in tbl_summary()

[ddsjoberg]

Reported here: https://stackoverflow.com/questions/79362375

library(magrittr)

gtsummary::tbl_summary(mtcars, statistic = all_continuous() ~ "\\makecell{{{mean}}}")
#> Error in `gtsummary::tbl_summary()`:
#> ! Error processing `statistic` argument.
#> ! Caused by error in `all_continuous()`: ! could not find function
#>   "all_continuous"
#> ℹ Select among columns "mpg", "cyl", "disp", "hp", "drat", "wt", "qsec", "vs",
#>   "am", "gear", and "carb"

.extract_glue_elements <- function(x) {
  regmatches(x, gregexpr("\\{([^\\}]*)\\}", x)) |>
    unlist() %>%
    {substr(., 2, nchar(.) - 1)} # styler: off
}

.extract_glue_elements("\\makecell{mean}")     # should return `c('mean')`
#> [1] "mean"
.extract_glue_elements("\\makecell{{mean}}")   # should return `character(0L)`
#> [1] "{mean"
.extract_glue_elements("\\makecell{{{mean}}}") # should return `c('mean')`
#> [1] "{{mean"

# these will eventually be processed with glue()
glue::glue("{letters[1]}")
#> a
glue::glue("{{letters[1]}}")
#> {letters[1]}
glue::glue("{{{letters[1]}}}")
#> {a}

^{Created on 2025-01-16 with reprex v2.1.1}

[raffaem]

Can't we just remove {{ and }} before extract_glue_elements?

[ddsjoberg]

That's a good point @raffaem, thanks !

Perhaps something like this would work, where we iteratively remove pairs of double curlies...

library(magrittr)

# attaching two internal functions form gtsummary
str_detect <- gtsummary:::str_detect
str_remove <- gtsummary:::str_remove

.extract_glue_elements <- function(x) {
  # remove PAIRS fo double curly brackets 
  while (str_detect(x, pattern = "{{", fixed = TRUE) && str_detect(x, pattern = "}}", fixed = TRUE)) {
    x <- str_remove(x, "{{", fixed = TRUE) |> str_remove(x, "}}", fixed = TRUE)
  }
  
  regmatches(x, gregexpr("\\{([^\\}]*)\\}", x)) |>
    unlist() %>%
    {substr(., 2, nchar(.) - 1)} # styler: off
}

.extract_glue_elements("\\makecell{mean}") # should return `c('mean')`
#> [1] "mean"
.extract_glue_elements("\\makecell{{mean}}") # should return `character(0L)`
#> character(0)
.extract_glue_elements("\\makecell{{{mean}}}") # should return `c('mean')`
#> [1] "mean"
.extract_glue_elements("{{{mean}}}}") # should return `c('mean')` (unbalanced curlies)
#> [1] "mean"

^{Created on 2025-01-16 with reprex v2.1.1}

[ddsjoberg]

Welp, it almost works with the code below....if we remove pairs of double curly brackets, it'll work in a well-defined space, but if curly brackets appear in another space we still see the parsing error.... 🤔🤔🤔

.extract_glue_elements <- function(x) {
  # remove PAIRS of double curly brackets
  while (isTRUE(any(str_detect(x, pattern = "{{", fixed = TRUE) & str_detect(x, pattern = "}}", fixed = TRUE)))) {
    idx <- str_detect(x, pattern = "{{", fixed = TRUE) & str_detect(x, pattern = "}}", fixed = TRUE)
    x[idx] <- str_remove(x[idx], "{{", fixed = TRUE) |> str_remove("}}", fixed = TRUE)
  }

  regmatches(x, gregexpr("\\{([^\\}]*)\\}", x)) |>
    unlist() %>%
    {substr(., 2, nchar(.) - 1)} # styler: off
}

devtools::load_all(path = "/Users/sjobergd/Documents/GitHub/gtsummary")
#> ℹ Loading gtsummary

tbl_summary(mtcars["mpg"], statistic = all_continuous() ~ "\\makecell{{{mean}}}") |> as.data.frame()
#>   **Characteristic**       **N = 32**
#> 1                mpg \\makecell{20.1}
tbl_summary(mtcars["mpg"], statistic = all_continuous() ~ "\\mak{{ecell{{{mean}}}") |> as.data.frame()
#> Error in `tbl_summary()`:
#> ! Problem with the `statistic` argument.
#> Error converting string "{{mean" to a function.
#> ℹ Is the name spelled correctly and available?

^{Created on 2025-01-16 with reprex v2.1.1}

[ddsjoberg]

Probably should investigate the glue code and steal what they have there....

[raffaem]

I don't understand what "well defined space" and "another space" mean.
I don't think the case of unbalanced curlies is legal.
Why we are using those functions instead of stringr?

[raffaem]

.extract_glue_elements <- function(x) {
  x <- stringr::str_replace_all(x, "(\\{\\{)|(\\}\\})", "")
  regmatches(x, gregexpr("\\{([^\\}]*)\\}", x)) |>
    unlist() %>%
    {
      substr(., 2, nchar(.) - 1)
    }
}

> tbl_summary(mtcars["mpg"], statistic = all_continuous() ~ "\\makecell{{{mean}}}") |> as.data.frame()
  **Characteristic**       **N = 32**
1                mpg \\makecell{20.1}
> tbl_summary(mtcars["mpg"], statistic = all_continuous() ~ "\\mak{{ecell{{{mean}}}") |> as.data.frame()
  **Characteristic**        **N = 32**
1                mpg \\mak{ecell{20.1}

[ddsjoberg]

Unfortunately, that doesn't quite solve the problem. it's tricky to mimic glue::glue()

library(magrittr)

.extract_glue_elements <- function(x) {
  x <- stringr::str_replace_all(x, "(\\{\\{)|(\\}\\})", "")
  regmatches(x, gregexpr("\\{([^\\}]*)\\}", x)) |>
    unlist() %>%
    substr(start = 2, stop = nchar(.) - 1)
}
.extract_glue_elements("{{{{{mean}}}}}}")
#> character(0)
glue::glue("{{{{{letters[1]}}}}}}")
#> {{a}}}

^{Created on 2025-01-29 with reprex v2.1.1}

[ddsjoberg]

@raffaem I think I have something working here: #2133

If you have a moment, can you please try to break it?