Update column_spec to not put special characters inside of regular expression

[RockfordMankiniUCSD]

Used to fix the error that comes from running the following code in RMarkdown:

---
title: "Untitled"
output:
  pdf_document: default
date: "2025-05-15"
---

knitr::opts_chunk$set(echo = TRUE)

library(tidyverse)
library(kableExtra)

col1 <- c("Wave 1", "Wave 2", "Wave 3", "Wave 4", "Wave 5")
col2 <- c("ID", "ID", "ID", "ID", "ID")
col3 <- c("This is a test.", "This is a test.", "This is a test.", "This is a test", "This is a test.")
df <- data.frame(col1, col2, col3)



kbl(df[,2:3], booktabs=T, align="lll", format = "latex", col.names = NULL,
           row.names = FALSE, longtable = T,  escape=F) %>%
  kable_styling(position = "left", full_width=FALSE) %>%
  column_spec(1, width = "6cm") %>%  
  pack_rows(index=table(fct_inorder(df$col1)), bold=F) %>%
  add_indent(1:nrow(df),level_of_indent = 2) %>%
  sub("\\\\toprule", "", .) %>%  sub("\\\\bottomrule", "", .)

Error:

! Misplaced \noalign.
\addlinespace ->\noalign 
                         {\ifnum 0=`}\fi \@ifnextchar [{\@addspace }{\@addsp...
l.129 \addlinespace
                   [0.3em]
I expect to see \noalign only after the \cr of
an alignment. Proceed, and I'll ignore this case.

Error: LaTeX failed to compile testrmarkdown.tex. See https://yihui.org/tinytex/r/#debugging for debugging tips. See testrmarkdown.log for more info.
Execution halted

Analysis:
See this snippet from column_spec_latex in column_spec.R:

  for (i in rows) {
    target_row <- table_info$contents[i]
    new_row <- latex_cell_builder(
      target_row, column, table_info,
      bold[i], italic[i], monospace[i], underline[i],
      strikeout[i], color[i], background[i], link[i], image[i]
      # font_size, angle
      )
    temp_sub <- ifelse(i == 1 & (table_info$tabular == "longtable" |
                                   !is.null(table_info$repeat_header_latex)),
                       gsub, sub)
    out <- temp_sub(target_row, new_row, out, perl = T)
    table_info$contents[i] <- new_row
  }

target_row is used as a regular expression to match against data from the output of kbl(). If sub() is used with an unescaped ., it'll treat it as a wildcard and erroneously substitute out the 4th row, leading to the snippet generating broken LaTeX.

Here's the value of the piped kbl() output from above. You can see a period was erroneously inserted at ID & This is a test.\ along with other missing LaTeX syntax.

\begin{longtable}[l]{>{\raggedright\arraybackslash}p{6cm}l}

\addlinespace[0.3em]
\multicolumn{2}{l}{Wave 1}\\
\hspace{2em}\hspace{1em}ID & This is a \vphantom{3} test.\\
\addlinespace[0.3em]
\multicolumn{2}{l}{Wave 2}\\
\hspace{2em}\hspace{1em}ID & This is a \vphantom{2} test.\\
\addlinespace[0.3em]
\multicolumn{2}{l}{Wave 3}\\
\hspace{2em}\hspace{1em}ID & This is a \vphantom{1} test.\\
ID & This is a test.\
\addlinespace[0.3em]
\multicolumn{2}{l}{Wave 5}\\
\hspace{2em}\hspace{1em}ID & This is a test.\\

\end{longtable}

Solution: escape the regex before putting it into sub().

[dmurdoch]

I think your mods look as though they would work, but what worries me is I don't understand why this is necessary here and not elsewhere. The contents of table_info$contents[i] have already been processed by regex_escape(..., double_backslash = TRUE), which looks like this:

regex_escape <- function(x, double_backslash = FALSE) {
  if (double_backslash) {
    x <- gsub("\\\\", "\\\\\\\\", x)
  }
  x <- gsub("\\$", "\\\\\\$", x)
  x <- gsub("\\(", "\\\\(", x)
  x <- gsub("\\)", "\\\\)", x)
  x <- gsub("\\[", "\\\\[", x)
  x <- gsub("\\]", "\\\\]", x)
  x <- gsub("\\{", "\\\\{", x)
  x <- gsub("\\}", "\\\\}", x)
  x <- gsub("\\*", "\\\\*", x)
  x <- gsub("\\+", "\\\\+", x)
  x <- gsub("\\?", "\\\\?", x)
  x <- gsub("\\|", "\\\\|", x)
  x <- gsub("\\^", "\\\\^", x)
  return(x)
}

If dots need escaping, shouldn't that function be modified to do it?

[dmurdoch]

PR #911 (just merged) addresses this issue differently.