deserialiseMd created an invalid TElement array

[naorzr]

Description

when trying to deserialise markdown, the function produces invalid TElement array structure that results in
Cannot read properties of undefined (reading 'map') error.

it seems like it breaks with images, it converts them to nodes that contain caption and link instead of url.
the caption part seems to be the breaking part.

example:

{
    "type": "img",
    "children": [
        {
            "text": ""
        }
    ],
    "link": "https://sampleurl.com/8b0a-2c3b0f0f3e3e.png",
    "caption": "image"
}

Steps

1. go to https://platejs.org
2. paste the markdown from the example https://markdown-it.github.io/

Funding

• You can sponsor this specific effort via a Polar.sh pledge below
• We receive the pledge once the issue is completed & verified

[12joan]

Thanks for the bug report!

As you say, pasting an image is enough to cause the error. There's a slightly different error message if the alt text is empty.

![Minion](https://octodex.github.com/images/minion.png)
![](https://octodex.github.com/images/minion.png)

[12joan]

There are a few issues on remark-slate, the package used by Plate to convert markdown to Slate nodes, relating to this problem. See #44, for example.

Ideally, the configuration in Plate should work out of the box. We might need to open a PR on remark-slate to add the configuration options necessary for that.

I've tested PR #45, which fixes the error but still produces an invalid image node.

[12joan]

PR #45 combined with imageSourceKey: 'url' is enough to produce a valid image. The only weird thing is it's added as a child of a paragraph node. @zbeyens Is this a problem?

Node produced by HTML deserializer

  {
    "type": "img",
    "url": "https://octodex.github.com/images/minion.png",
    "children": [
      {
        "text": ""
      }
    ]
  },

Node produced by markdown deserializer

  {
    "type": "p",
    "children": [
      {
        "type": "img",
        "children": [
          {
            "text": ""
          }
        ],
        "url": "https://octodex.github.com/images/minion.png",
        "caption": [
          {
            "text": "Minion"
          }
        ]
      }
    ]
  }

[zbeyens]

@12joan It looks like remark-slate author is inactive – one PR has been open for a year now. Let's fork its src into the deserializer-md package.

img should not be inside a p: this is valid only for inline images. Let's make it a block by default, and support inline with an option.