Rmarkdown is printing the latex code in the knited PDF-CodePudding

I am running in an odd result. I have inserted a number of tables in my document in Rmarkdown. This is the YAML:

---
output:
  pdf_document:
    toc: true
    toc_depth: 2
    number_sections: true
header-includes:
 \usepackage{float}
 \floatplacement{figure}{H}
 \floatplacement{table}{H}
---

And the first chunk:

{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE,
                      fig.pos = "H",
                      fig.width=8,
                      fig.height=7,
                      tab.pos = "H",
                      collapse = TRUE,
                      message = FALSE,
                      warning = FALSE,
                      comment = "#>"
)

knitr::opts_knit$set(root.dir = rprojroot::find_rstudio_root_file())

options(knitr.kable.NA = "-")

The tables are inserted in this manner:

kable(tableX, caption = "Caption", booktabs = TRUE) %>% 
  row_spec(0, bold=TRUE) %>% 
  kable_styling(bootstrap_options = "condensed") %>% 
  column_spec(1, italic = TRUE)

Now, most tables are inserted correctly, but there is just one which is printed with the latex code, \begin{table}, then \caption{}, then the table, and at the end \end{table}. How is it possible that one table is being printed in this manner with the code being the same?

Thank you

CodePudding user response：

As the author of rmarkdown Yihui Xie noted here, you can disable the character-escaping argument in kable() to include LaTeX's spacial characters in the table caption, or column/row names, or both. Then you need to manually escape such characters by adding \\ before each special character.

```{r tab1, echo=F}
tableX=data.frame(col1 = "Empty data frame")
kable(
  tableX,
  caption  = "Caption 20\\%",
  booktabs = TRUE,
  escape   = FALSE
  ) |> 
  row_spec(0, bold=TRUE) |>
  kable_styling(bootstrap_options = "condensed") %>% 
  column_spec(1, italic = TRUE)
```

CodePudding user response：

R markdown will produce latex code when the output option is set to pdf_document (the latex code can be viewed by selecting the option keep_tex and inspecting the produced .tex file) or when the output is set to latex_fragment.

knitr::kable output the full latex output for a table (when in tex mode), but consumes any caption provided literally. The knitr interpreter then consumes the produced latex block. When somewhere in between the \begin and \end environment statement something illegal is produced (such as a caption with an unescaped % in there), the interpreter instead produces escaped latex code.

The valid code

```{r tab1, echo=F}
tableX=data.frame(col1 = "Empty data frame")
kable(tableX, caption = "Caption 20pct", booktabs = TRUE) |> 
  row_spec(0, bold=TRUE) |>
  kable_styling(bootstrap_options = "condensed") %>% 
  column_spec(1, italic = TRUE)
```

produces latex code as expected:

\begin{table}

\caption{\label{tab:tab1}Caption 20pct}
\centering
\begin{tabular}[t]{>{}l}
\toprule
\textbf{col1}\\
\midrule
\em{Empty data frame}\\
\bottomrule
\end{tabular}
\end{table}

But when a % is added to the caption, the latex code generation breaks. The code block

```{r tab1, echo=F}
tableX=data.frame(col1 = "Empty data frame")
kable(tableX, caption = "Caption 20%", booktabs = TRUE) |> 
  row_spec(0, bold=TRUE) |>
  kable_styling(bootstrap_options = "condensed") %>% 
  column_spec(1, italic = TRUE)
```

produces

\textbackslash begin\{table\}

\textbackslash caption\{\label{tab:tab1}Caption 20\%\} \centering

\begin{tabular}[t]{>{}l}
\toprule
\textbf{col1}\\
\midrule
\em{Empty data frame}\\
\bottomrule
\end{tabular}

\textbackslash end\{table\}

which just renders in the pdf as literal latex code.