builder MarkdownView does not handle the self-closing part as expected.

message:

<tutorial-course-abandon-dismissal />
I see you're back on your user page! Perfect - now we're ready to continue with our course on creating a new project.

Let's go back to the homepage first, then we'll learn how to create a project. Please click the <highlight-link target-id="1ci8_wBp" tip="Click to go to homepage">Logo link</highlight-link> in the top navigation bar.

Once we're on the homepage, we'll use the Project menu to create our new project!

manifests as：

Nov 17 '25 11:11 Overu

原因

经过排查，在调用mdast-util-from-markdown的fromMarkdown后，上述的输入整体被识别为html_block。

这个行为其实符合CommonMark规范。这不是一个解析错误，而是源自于Markdown对HTML空白行的语义规则。

稍微改变下输入：

例子1：
<tutorial-course-abandon-dismissal />

I see you're back on your user page! Perfect - now we're ready to continue with our course on creating a new project.

或

例子2：
<tutorial-course-abandon-dismissal />I see you're back on your user page! Perfect - now we're ready to continue with our course on creating a new project.

那结果是符合预期的。

例子1中，html_block把空白行作为结束标志，所以<tutorial-course-abandon-dismissal />是html_block，第3行是paragraph

<html_block>&lt;tutorial-course-abandon-dismissal /&gt;</html_block>
<paragraph>
  <text>I see you</text>
  <text>'</text>
  <text>re back on your user page</text>
  <text>!</text>
  <text> Perfect - now we</text>
  <text>'</text>
  <text>re ready to continue with our course on creating a new project.</text>
</paragraph>

例子2中，无法识别的tag(或a、span等标签)被识别为html_inline，和剩余的部分放在一个paragraph中

<paragraph>
  <html_inline>&lt;tutorial-course-abandon-dismissal /&gt;</html_inline>
  <text>I see you</text>
  <text>'</text>
  <text>re back on your user page</text>
  <text>!</text>
  <text> Perfect - now we</text>
  <text>'</text>
  <text>re ready to continue with our course on creating a new project.</text>
</paragraph>

回过头看这个问题，输入中<tutorial-course-abandon-dismissal />和下面内容之间没有空白行，所以无法被解析为html_block或html_inline，而是将整个内容放到html_block中

<html_block>&lt;tutorial-course-abandon-dismissal /&gt;
I see you're back on your user page! Perfect - now we're ready to continue with our course on creating a new project.</html_block>

最终导致外层无法解析。

上述AST可以参考CommonMark playground

解决方案

扩展mdast，将custom element解析为html_inline
prompt规范输出格式
前置处理，确保输入fromMarkdown前，对于上述有问题的情况增加空白行或增加 反斜杠转义空行
后置处理，操作fromMarkdown的返回结果

Nov 17 '25 15:11 Overu

建议是走

前置处理，确保输入fromMarkdown前，对于上述有问题的情况增加空白行或增加反斜杠转义空行

这个处理方式

现在在这边应该已经有相关逻辑了，如 preprocessCustomRawComponents、preprocessIncompleteTags，应该往这边再加一个就好；不过我担心会影响到其他预期被 inline 的内容（比如 highlight-link），可能要也测一下；如果确实会影响，我们可能需要把 custom component 从现在的 custom / customRaw 再进一步细分为 customInline / customBlock / customRaw，然后只对 customBlock 做这个处理？

Nov 18 '25 03:11 nighca

目前先采用前置预处理，针对这种情况会增加\n\n，强制分为两行，考虑到只有self-closing tag加换行符后紧跟着字符会出现这个情况，暂时没有增加类似customInline。

Nov 18 '25 13:11 Overu

mdast-util-from-markdown行为分析

开/闭合标签

测试基于以下几种场景：

只有单标签

<tag-name></tag-name>

标签后面跟着文字

<tag-name></tag-name>hello

标签后面跟着换行

<tag-name></tag-name>
hello

标签前后跟着文字

hello<tag-name></tag-name>world

未知标签: inline
1. [paragraph -> [tag-name]]
2. [paragraph -> [tag-name, world]]
3. [paragraph -> [tag-name, \nworld]]
4. [paragraph -> [hello, tag-name, world]]
已知标签
- a、span：inline
  1. [paragraph -> [span]]
  2. [paragraph -> [span, world]]
  3. [paragraph -> [span, \nworld]]
  4. [paragraph -> [hello, span, world]]
- div、p(raw): block
  1. [html -> [div]]
  2. [html -> [divworld]]
  3. [html -> [div\nworld]]
  4. [paragraph -> [hello, div, world]]

自闭合

测试基于以下几种场景：

只有单标签

<tag-name />

标签后面跟着文字

<tag-name />world

标签后面跟着换行

<tag-name />
hello

标签前后跟着文字

hello<tag-name />world

未知标签: inline
1. [html -> [tag-name/]]
2. [paragraph -> [tag-name/, world]]
3. [html -> [tag-nam/\nworld]]
4. [paragraph -> [hello, tag-name/, world]]
已知标签
- a、span：inline
  1. [html -> [span/]]
  2. [paragraph -> [span/, world]]
  3. [html -> [span/\nworld]]
  4. [paragraph -> [hello, span/, world]]
- div、p: block
  1. [html -> [div/]]
  2. [html -> [div/world]]
  3. [html -> [div/\nworld]]
  4. [paragraph -> [hello, div/, world]]

raw

测试基于以下几种场景：

前面无内容

<tag-name>
component1

component2
component3
</tag-name>

前面有内容

hello<tag-name>
component1

component2
component3
</tag-name>world

后面有内容

hello<tag-name>
component1

component2
component3
</tag-name>world

pre
1. [html -> [<pre>\ncomponent1\n\ncomponent\n2component3\n]]
2. [paragraph -> [hello, pre, ...], paragraph -> [..., world, /pre]]
3. [html -> [<pre>\ncomponent1\n\ncomponent\n2component3\nworld]]
div
1. [html, paragraph, html]
2. [paragraph -> [hello, div, ...], paragraph -> [..., world, /div]]
3. [html, paragraph, html]

Dec 01 '25 11:12 Overu