Percent symbol in qwraps2 summary table not rendering properly in bookdown

Question

I am struggling to get a qwraps2 summary table to render properly in a markdown file knitting to a pdf document using bookdown. The issue arises when using the percent sign in the row names.

Sample markdown document:

---
title: "Untitled"
author: "TEST"
output:
  pdf_document: default
  bookdown::pdf_document2: null
---

```{r results = 'asis'}
library(qwraps2)
library(knitr)
library(kableExtra)

# Create som sample data
set.seed(2)
df <- data.frame(
  group = as.factor(sample(c("A","B"),8, replace = T)),
  value = sample(2:10,8, replace=F),
  pos = as.factor(sample(c("pos.","neg."),8, replace = T))
)

# Define the table for summary_table
summary <- list(
  "Groups" = list(
    "Value" = ~ median_iqr(value),
    "Pos. (\\\\%)" = ~ qwraps2::n_perc(pos == "pos."))
)

new_table <- summary_table(df, summaries = summary, by = "group")

new_table

I am aware that the percent sign needs to be escaped with backslash, but with two backslashes the rendering fails. With four the table is not rendered correctly. See below.

How do I add a percent sign to the qwraps2 summary table row?

Answer 1

This looks like a bug in R triggered by bad code in qwraps2.

The qwraps2 package converts the summaries argument to a character string and then parses it, purely for the purpose of giving a warning about using data in summaries. This doesn't make sense: summaries is already a list, so it shouldn't need parsing. That code should be fixed.

However, the bug in R is that the conversion to a character string is done incorrectly when you have "Pos. (\\%)" = ~ qwraps2::n_perc(pos == "pos."), and that's why you get the error.

The easiest fix I can think of is to remove that warning from the qwraps2:::summary_table.data.frame function. If you insert this code into your document, you'll do that:

summary_table.data.frame <- qwraps2:::summary_table.data.frame
body(summary_table.data.frame)[[2]] <- NULL

After that change, you can use

summary <- list(
  "Groups" = list(
    "Value" = ~ median_iqr(value),
    "Pos. (\\%)" = ~ qwraps2::n_perc(pos == "pos."))
)

and things will be fine:

However, this fix is not very robust: if the author of qwraps2 changes that function, this could fail badly. So your workaround of editing the result of the summary_table call is probably safer.