<code> not rendered correctly; incorrectly escapes when switching

[chwarr]

Describe the bug

The Stacks Editor does not properly render <code> elements and incorrectly escapes them when toggling from Markdown mode to rendered mode.

To Reproduce

1. Start a new post. Make sure you are in Markdown mode, not rendered/rich edit mode.
2. Enter the following Markdown: <code>this should be code formatted <i>with some italics inside</i></code>
   
3. Click the toggle to switch to rendered/rich edit mode.
4. Click the toggle to switch back to Markdown mode.

Expected behavior

• After step 3, the rendered post looks like: this should be code formatted with some italics inside
• After step 3, the Markdown should be exactly what was entered after step 2

Actual behavior

• After step 3, the rendered post has literal <code> and </code> shown. None of the sentence has typewriter formatting. The italics are shown correctly:
  

• After step 3, the Markdown has had <code> changed to `<code>` and similarly for </code>:

  `<code>`this should be code formatted <i>with some italics inside</i>`</code>`
  

  

Desktop (please complete the following information):

• OS: Windows 10
• Browser: Firefox 85.0, Edge 88.0.705.56

Additional context

If I submit the answer/edit without previewing, the <code> elements are rendered correctly.

I first encountered this on a StackOverflow Enterprise installation. I have the same experience on Meta Stack Exchange.

[b-kelly]

Thanks for the very clear report! I ran your input through the commonmark dingus and it is 100% in agreement with your expected behavior. Additionally, I am able to reproduce the behavior you are seeing.

I'll take a look into this issue along with my next round of bug fixes.

[LoganDark]

I'd like to point out here that as a side effect of this bug the editor is committing a grave sin: entering WYSIWYG mode in this case is changing the markup arbitrarily, which is not what one expects to see. Changes should only be done around the cursor/selection as a result of explicit user interaction :(

Many editors do this (eBay's editor minifies all HTML), but it is not expected behavior, and I would argue that it's not correct either.

[hamvocke]

@LoganDark Leaving markdown as is when switching between markdown and rich-text mode is something we've got on our radar but haven't solved yet. I totally agree that the editor should not change your markdown existing markdown syntax.

[b-kelly]

Did some investigation, reporting here so I remember where to pick up again later:

<code>this should be code formatted <i>with some italics inside</i></code> parses as [code_inline, text, em_open, text, em_close, code_inline].

`this should be code formatted <i>with some italics inside</i>`

is parsed as a single code_inline with the context set to the entire text (no html parsing).

What we're seeing is that code_inline does not have an _open/_close suffix like other marks, so the content between the bounding tokens is not wrapped in the code mark.

Easiest solution? Parse <code> as a different token (code_inline_split?) that opens/closes and is ultimately parsed to the same mark as code_inline.