这个问题参照了 chatgpt 的答案并加以修改，直接看源码。
探究
根据其回答思路，主要是编写了一个插件，通过遍历其中的 h2 和 h3 元素，如果有，则作为指示器，跳过其子节点，然后按顺序将相邻节点筛选出来。
源码
import { unified } from 'unified'
import parse from 'rehype-parse'
import stringify from 'rehype-stringify'
import { SKIP, visit } from 'unist-util-visit'
const html = `
<h2>Section 1</h2>
<p>This is content for section 1.</p>
<p>More content for section 1.</p>
<h3>Subsection 1.1</h3>
<p>This is content for subsection 1.1.</p>
<h3>Subsection 1.2</h3>
<p>This is content for subsection 1.2.</p>
<h2>Section 2</h2>
<p>This is content for section 2.</p>
<h3>Subsection 2.1</h3>
<p>This is content for subsection 2.1.</p>
<p>More content for subsection 2.1.</p>
<h2>Section 3</h2>
<p>This is content for section 3.</p>
`
const processor = unified()
.use(parse)
.use(function splitSections() {
const sections = []
let currentSection = null
return (tree) => {
visit(tree, (node) => {
if (
node.type === 'element' &&
(node.tagName === 'h2' || node.tagName === 'h3')
) {
if (currentSection) {
sections.push(currentSection)
}
currentSection = {
type: 'element',
tagName: 'div',
properties: {},
children: [node],
}
return SKIP
} else if (currentSection) {
currentSection.children.push(node)
return SKIP
}
})
if (currentSection) {
sections.push(currentSection)
}
tree.children = sections
}
})
.use(stringify)
const processedHtml = processor.processSync(html).toString()
console.log(processedHtml)
输出：
<div>
<h2>Section 1</h2>
<p>This is content for section 1.</p>
<p>More content for section 1.</p>
</div>
<div>
<h3>Subsection 1.1</h3>
<p>This is content for subsection 1.1.</p>
</div>
<div>
<h3>Subsection 1.2</h3>
<p>This is content for subsection 1.2.</p>
</div>
<div>
<h2>Section 2</h2>
<p>This is content for section 2.</p>
</div>
<div>
<h3>Subsection 2.1</h3>
<p>This is content for subsection 2.1.</p>
<p>More content for subsection 2.1.</p>
</div>
<div>
<h2>Section 3</h2>
<p>This is content for section 3.</p>
</div>
问题
当然，以上代码也有问题，具体有两个
-
只支持顶层的h2和h3，如果是被引用框包裹住的这类的是不管的
-
会从首个h2或者h3元素开始，其以上的元素会丢失，这个比较严重
当然解决办法也是有的，第一个应该不算是个问题，我也就不太管了。主要是第二个问题，在本主题中已经解决了，其源码在此，可以点击查看。