Markdown: Leave Gutenberg content intact

[tyxla]

This PR updates the Markdown module to leave any Gutenberg content untouched, in order to not mess up the block markup validation and parsing.

Fixes #10634

Changes proposed in this Pull Request:

• Update the Markdown module to leave untouched any content that contains Gutenberg blocks.

Testing instructions:

• Make sure you have the latest WordPress 5.0 beta.
• Checkout this branch.
• Make sure your site is connected.
• Make sure Markdown is enabled.
• Write a Gutenberg post with various blocks.
• After saving and refreshing, verify there are no validation errors.
• Try the Markdown block.
• Verify it works well with various content.
• Install and activate the Classic Editor plugin.
• Write a post in the classic editor and verify Markdown still works well there.

Proposed changelog entry for your changes:

• Make Markdown module compatible with the Gutenberg editor in WordPress 5.0.

[matticbot]

Caution: This PR has changes that must be merged to WordPress.com
Please review this diff before merging: D20897-code. (newly created revision)

[jetpackbot]

Thank you for the great PR description!

When this PR is ready for review, please apply the [Status] Needs Review label. If you are an a11n, please have someone from your team review the code if possible. The Jetpack team will also review this PR and merge it to be included in the next Jetpack release.

Scheduled Jetpack release: December 4, 2018.
Scheduled code freeze: November 27, 2018

Generated by 🚫 dangerJS

[kraftbj]

I tested this on my personal site (which was impacted by this bug). With this PR, the text is not transformed (resulting in the paragraph breaks between blocks being left intact).

[jeherve]

It works well on my end!

[mdawaffe]

@dmsnell suggested updating the PHP Markdown code to run on Classic Blocks when the post was a Gutenberg post.

Dennis, is there a simple API for that these days?

Even if there is, I'm happy with the solution in this PR for now. I just flagged it as "Request Changes" so we could discuss it before merging.

Let's discuss the fancier alternative in #10642.

[mdawaffe]

An alternative solution has been proposed in D20852-code. Let's consolidate on one.

[brandonpayton]

The alternate solution in D20852 is to allow markdown processing but stop the transform from mangling paragraphs. I think I may prefer the simplicity and clarity of this PR. Markdown processing could break blocks in others ways, and it would be good to avoid that.

[mdawaffe]

What tests do we need to do for existing Markdown content? (That is, Markdown content that was published prior to Gutenberg/WordPress 5.0.)

Edit: I mean - what should we add to the Testing instructions for this PR or whatever eventual fix?

[brandonpayton]

One other thing to note is that D20852 fixes an issue with the markdown plugin not being properly disabled when Gutenberg is enabled by a user. I'm planning to separate that fix into a separate patch and land that soon.

The issue we were troubleshooting was for mixed content that results when a user:

1. Enables Gutenberg.
2. Adds block content to an existing post with classic content.
3. Disables Gutenberg.
4. Edits the same post with the classic editor and finds their block content to be missing.

If we fix WP.com to properly disable the markdown transform when inserting posts from Gutenberg, the above issue is resolved.

But then we have another issue when editing mixed block/non-block in the classic editor:
Markdown processing removes p tags, but wpautop isn't called to re-add them because the post contains block content.

[brandonpayton]

What tests do we need to do for existing Markdown content? (That is, Markdown content that was published prior to Gutenberg/WordPress 5.0.)

I'm not sure I understand but will add some comments in case they help.

When an existing markdown post is edited in Gutenberg, the transformed content is loaded by Gutenberg from post_content. This is because the easy-markdown plugin puts the original markdown content in post_content_filtered and the output of the markdown transform in post_content. When loading a post in the classic editor, easy-markdown uses the edit_post_content filter to provide the original markdown content, but this swap does not occur via the way Gutenberg loads posts. This means that editing a markdown post via Gutenberg uses the output of the markdown transform and not the original markdown.

[mdawaffe]

When an existing markdown post is edited in Gutenberg, the transformed content is loaded by Gutenberg from post_content. This is because the easy-markdown plugin puts the original markdown content in post_content_filtered and the output of the markdown transform in post_content. When loading a post in the classic editor, easy-markdown uses the edit_post_content filter to provide the original markdown content, but this swap does not occur via the way Gutenberg loads posts. This means that editing a markdown post via Gutenberg uses the output of the markdown transform and not the original markdown.

Interesting, thanks for the insight. I'm not sure if that behavior makes backward compatibility easier or harder :)

I feel the most important thing is to preserve the content as viewed on the site. The behavior above seems to ensure that.

I feel it's less important to preserve the content as viewed in the editor. That is, if we Gutenbergify the for-edit-content, as long as it's "semantically" the same even though it's been syntactically changed is maybe OK?

But basically, I want us to be sure to test these scenarios (like the complicated one you outlined above) :)

[tyxla]

Thanks for the alternate solution! I agree we'll need thorough testing in order to decide which solution works best and has as little negative impact as possible, on both old and new content. Perhaps @Automattic/flowpatrol could help with this?

[jeherve]

I'll merge this for now, while leaving #10642 open so we can discuss and implement a fancier alternative in the future.

Discussing alternatives in a new issue

[jeherve]

Cherry-picked to branch-6.8 in c395a4f