Escaping pipe (`|`) characters when `tablefmt="pipe"`

[timvink]

Thanks for the epic libary!

I have an edge-case that produces invalid markdown tables. Let's say I have a CSV file like this.

col1,col2|withpipe,col3
y,y|n|unknown,y

when I would read it (with pd.read_csv()) and then convert it to markdown with .to_markdown(tablefmt='pipe'), it would return:

|    | col1   | col2|withpipe   | col3   |
|---:|:-------|:----------------|:-------|
|  0 | y      | y|n|unknown     | y      |

Which is of course not a valid table because the rows don't have equal numbers of columns, and there are too many columns.

Of course, this could be fixed at the source. But I maintain mkdocs-table-reader-plugin which depends on tabulate, and I would prefer for users that it 'just works', or at least gives an informative error.

A work around is to escape all pipe characters before conversion (| -> \|):

df = pd.read_csv(...)
df.columns = [x.replace('|','\\|') for x in df.columns]
df = df.applymap(lambda s: s.replace('|','\\|') if isinstance(s, str) else s)
df.to_markdown(tablefmt="pipe")

There is however a slight performance hit for larger tables of course.

Is this something you would like to address in tabulate?

[PyroGenesis]

I have the same issue when using df.to_markdown() for a pandas DataFrame. Any pipe characters in the DataFrame values are not escaped, resulting in an invalid markdown table. Is there any workaround?