Bold/Strong text not parsed correctly when ending with punctuation followed by CJK characters

[Red-Asuka]

Description

Bold syntax (**text**) fails to render when:

1. The bold text ends with a punctuation mark
2. Immediately followed by a non-space character

Reproduction

https://stackblitz.com/edit/github-ezruxrr6?file=app%2Fapp.vue

Test Cases

**文字：**后续    <!-- ❌ Fails -->
**文字：** 后续   <!-- ✅ Works (space after **) -->
**文字**：后续    <!-- ✅ Works (punctuation outside **) -->
**English:**next  <!-- ❌ Fails -->
**English:** next <!-- ✅ Works -->

Affected Punctuation

Chinese: ： ！ ？ 。 ， ） "
English: : ! ? . , ) "

Workaround

Add a space after the closing **:

**标题：** 内容

[Red-Asuka]

I've implemented a workaround in my project using mdc.config.ts. Since remark plugins run after micromark has already parsed the markdown into AST, modifying file.value at that stage has no effect.

My solution is to post-process the AST and find unprocessed **...** patterns that remain as text nodes (instead of being converted to strong nodes), then manually convert them.

Here's the workaround:

// mdc.config.ts
import type { Parent, Root, Text } from 'mdast'
import { defineConfig } from '@nuxtjs/mdc/config'
import { SKIP, visit } from 'unist-util-visit'

const boldPattern = /\*\*([^*]+)\*\*/g

function remarkFixCjkEmphasis() {
  return (tree: Root) => {
    visit(tree, 'text', (node: Text, index, parent: Parent | undefined) => {
      if (!parent || index === undefined)
        return

      const text = node.value

      // Quick check: skip if no ** at all
      if (!text.includes('**'))
        return

      const matches = [...text.matchAll(boldPattern)]
      if (matches.length === 0)
        return

      // Build new nodes array
      const newNodes: (Text | { type: 'strong', children: Text[] })[] = []
      let lastIndex = 0

      for (const match of matches) {
        const matchStart = match.index!
        const matchEnd = matchStart + match[0].length
        const boldContent = match[1]

        // Text before the match
        if (matchStart > lastIndex) {
          newNodes.push({ type: 'text', value: text.slice(lastIndex, matchStart) } as Text)
        }

        // The strong node
        newNodes.push({
          type: 'strong',
          children: [{ type: 'text', value: boldContent } as Text],
        })

        lastIndex = matchEnd
      }

      // Text after the last match
      if (lastIndex < text.length) {
        newNodes.push({ type: 'text', value: text.slice(lastIndex) } as Text)
      }

      // Replace the original node with new nodes
      parent.children.splice(index, 1, ...newNodes as never[])

      return SKIP
    })
  }
}

export default defineConfig({
  unified: {
    remark(processor) {
      return processor.use(remarkFixCjkEmphasis)
    },
  },
})

This works because:

• Only text nodes are processed (code blocks become code nodes, so they're safe)
• Failed emphasis parsing leaves **...** as raw text in text nodes
• Successfully parsed emphasis becomes strong nodes and won't be affected

This is a temporary workaround. A proper fix would need to be in micromark's attention parsing logic.