How to preserve spaces in GFM tables?

[peterdell]

Hello,
in my project there are many table of the form

| A |B|
|---|-|
| 1 |2|

When I use the GFM table extension they are correctly parsed, but all spaces are removed and that make editing the table impossible afterwards. Is there a way to prevent this condensation of the existing spaces?

[robinst]

You mean after rendering the table back to markdown using MarkdownRenderer?

We're probably not preserving spaces there. The fix is going to be preserving that information on parsing and then taking it into account on rendering as well.

[peterdell]

You mean after rendering the table back to markdown using MarkdownRenderer?
Yes.

I think not putting the spaces is valid behavior for programmaticly generated MDs and if nobdy complained so for, it should probably a builder option.
I also wanted to have automatic spacing and formatting of the "|---|---|" separator lines. Not sure if that is worth to be in the standard.
I've implemented that in my modified copy here: https://github.com/atariwiki-org/wiki-tools/tree/main/src/org/commonmark/ext/gfm/tables
It for now only looks into the width of the header text. Mid-term I want to check also the really used column-width in all rows (with a maximum of N chars).

[robinst]

Right, so what we want is not to preserve the input spaces, but instead an option to render any table with visual alignment. Yeah, I think that makes sense!

Mid-term I want to check also the really used column-width in all rows (with a maximum of N chars).

Yes, so it will need an additional pass over all cells, but that's ok. We'll need to actually render all the cells to strings first, then visually align the resulting strings (bonus points for handling Unicode combining marks, zero-width joiners and full-width characters correctly). Do you want to raise a PR for that?

[robinst]

bonus points for handling Unicode combining marks, zero-width joiners and full-width characters correctly

Btw I looked into that, it's not easy to do without extra Unicode data (e.g. ICU4J). I recommend we:

• By default count grapheme clusters (assuming one cluster = 1 column), using BreakIterator
• Allow consumers to pass in a Function<String, Integer> for overriding width calculation (e.g. using ICU4J's EAST_ASIAN_WIDTH data).