Bug: Trailing whitespace being formatted produces invalid Common Mark Markdown

[ldrick]

Lexical version: 0.31.2

Steps To Reproduce

1. Go to https://playground.lexical.dev/
2. Mark in the sentence Make "sure to check "out the various[...] (only mark the Text in double quotes and the trailing whitespace is important) and format it bold by pressing "B" Symbol
3. The trailing whitespace get's converted to HTML representation   and behind the result will become Make **sure to check **out the various[...]
4. That is invalid common mark (and also gfm) and would other editors to print the surrounding "**" before and after.
5. Please compare to this playground https://spec.commonmark.org/dingus/?text=test%20**test**%20test%5C%0Atest%20**test%20**%20test%5C%0Atest%20**test%20**test%5C%0Atest%20**test%26%2332%3B**test%5C%0Atest%20**test%26%2332%3B**%20test

Link to code example:

The current behavior

HTML space   is added, but space to following word is removed - which breaks Common Mark Syntax

The expected behavior

Any valid Common Mark dealing the trailing whitespace, see Step 5 in reproduction, to get a idea, what it could be (maybe examples are incomplete and there is much more)

Impact of fix

Happens always if within a text some words are marked to be formatted and makes it impossible to process the resulting Common Mark properly with other tools.

[Gnative]

Caused by the change in this merged.. https://github.com/facebook/lexical/pull/7400/files.

[Gnative]

We had to revert to prior 0.30.0 because of this issue and to 0.28.0 to avoid and issue with escaped underscores in markdown.

[obraud]

Hello! Is there any news on this issue 🙏 ?
Double clicking on a word in the editor will include any trailing whitespace, so it's easy to make a mistake that will output invalid markdown.

[aleksandrsjevdokimovswv]

We had to end up using find and replace regex after converting to MD to avoid downgrading. We have the same issue with  .

This is still present in v0.36.2

[etrepum]

Feel free to submit a PR with tests that fixes this issue

[mjolivet-lucca]

the same issue happens if a lonely space is set to bold. qsd** **qsd becomes qsd** **qsd

[reinseth]

@etrepum: I see from the discussion and related PRs in #6526 that the solution of escaping leading and trailing whitespace with html entities comes from your suggestions. I would like to open a PR to fix this issue, but that would mean reverting #7400 and instead change the behaviour to strip formatting of whitespace alone (e.g. * * or ** **). Would you be open for that?

[etrepum]

There’s already an open PR #7972

[reinseth]

@etrepum: I'm not seeing anything fixed in that PR. Is there something I'm missing?

Anyways, I took a stab it and ended up with a fix that is really small:
21d87d6

I'll open a PR if you're ok with that.

[etrepum]

Go ahead and open a PR