I generate markdown tables with pander R package that have pipes in some column titles (like : P > |t| ). It seems (unless I missed something) that they are not correctly processed neither by pandoc nor by Rmarkdown due to the confusion between column separators and "true" pipes.
Consider the following Rmd example:
```{r message = FALSE}
library(pander)
panderOptions("table.style" , "rmarkdown")
panderOptions("table.split.table" , Inf) # avoid to split the tables
data(iris)
mod <- lm(Sepal.Length ~ Species, data = iris)
```
```{r results='asis'}
pandoc.table(summary(mod)$coefficients[,-4])
```
```{r results='asis'}
pandoc.table(summary(mod)$coefficients)
```
The last table generated by pander looks like this (note the pipes in the last column name):
| | Estimate | Std. Error | t value | Pr(>|t|) |
|------------------------:|:----------:|:------------:|:---------:|:----------:|
| **(Intercept)** | 5.006 | 0.0728 | 68.76 | 1.134e-113 |
| **Speciesversicolor** | 0.93 | 0.103 | 9.033 | 8.77e-16 |
| **Speciesvirginica** | 1.582 | 0.103 | 15.37 | 2.215e-32 |
If I knit this to html (via Rstudio button that uses Rmarkdown to generate the HTML if I'm not wrong), the last table is not displayed as a table but as plain text in the HTML output.
If I use the md generated by knitr and transform it into html with pandoc the output is a table but the last column name becomes "Pr(>".
The first table without the last column is displayed correctly.
You can escape the pipes with a backslash (\|).