Static file download parsing error

[qylwhetc]

Describe the bug

Why is the image link correct but the download link incorrect ？

markdown--> html

Reproduction

Folder situation：

\public\dn\
    b.zip
\subdir\
    demo.md
    a.zip

config.mjs:

export default defineConfig({
  title: "xxx系统",
  description: "A VitePress Site",
  base:'/vp/',
  themeConfig: {}
})

The content of the file：demo.md

<a href="./a.zip" download> dn</a>
[a.zip](./a.zip){download}
[b.zip](/dn/b.zip){download}

build result:

.vitepress\dist\subdir\
    demo.html
     a.zip：does not exist 
.vitepress\dist\dn\
    b.zip，ok，exist 

The content of the file：demo.html

<p><a href="./a.zip" download>dn</a></p>
<p><a href="./a.zip" download>a.zip</a></p>
<p><a href="/dn/b.zip" download>b.zip</a></p>

Expected behavior

What I hope for is：

1. The file "a.zip" should be copied to ".vitepress\dist\subdir\a.zip"

2. Hyperlinks are built based on the {base}：

<p><a href="/vp/dn/b.zip" down>b.zip</a></p>

but The MD below parsing and file copying are both correct：

![pic](./c.png) 
![pic](/img/d.png) //public/img/d.png

to

<p><img src="/vp/assets/c.BbAI9XcS.png" alt="pic"></p>
<p><img src="/vp/img/d.png" alt="pic"></p>

My configuration is incomplete or there are bug?

System Info

windows 11，iis

Additional context

No response

Validations

• Check if you're on the latest VitePress version.
• Follow our Code of Conduct
• Read the docs.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.

[brc-dd]

Hmm, image and videos, etc. have more advanced asset handling (https://github.com/vitejs/vite-plugin-vue/tree/main/packages/plugin-vue#asset-url-handling, https://vite.dev/guide/assets). Normal links aren't transformed like that.

I'd recommend keeping such files in public directory and use withBase helper. Something like this should work:

<a :href="withBase('/dn/b.zip')" download>b.zip</a>

<script setup>
import { withBase } from 'vitepress'
</script>

---

Alternatively, you can manually copy such files later, that will match the dev behavior more closely:

import { copyFile, glob, mkdir } from 'node:fs/promises'
import path from 'node:path'
import { defineConfig } from 'vitepress'

export default defineConfig({
  // ...

  async buildEnd(siteConfig) {
    const files = glob('**/*.zip', { cwd: siteConfig.srcDir })

    for await (const file of files) {
      const srcPath = path.join(siteConfig.srcDir, file)
      const destPath = path.join(siteConfig.outDir, file)

      console.log(`Copying ${srcPath} to ${destPath}`)

      await mkdir(path.dirname(destPath), { recursive: true })
      await copyFile(srcPath, destPath)
    }
  }
})

^ This will work with relative paths:

<a href="./a.zip" download> dn</a>
[a.zip](./a.zip){download}

---

Another alternative:

import { defineConfig } from 'vitepress'

const base = '/vp/'

export default defineConfig({
  base,

  markdown: {
    config(md) {
      const orig = md.renderer.rules.link_open
      md.renderer.rules.link_open = (tokens, idx, options, env, self) => {
        const hrefIndex = tokens[idx].attrIndex('href')
        const downloadIndex = tokens[idx].attrIndex('download')
        if (hrefIndex >= 0 && downloadIndex >= 0) {
          const href = tokens[idx].attrs![hrefIndex][1]
          if (href.startsWith('/')) {
            tokens[idx].attrs![hrefIndex][1] = base.slice(0, -1) + href
          }
        }
        return orig ? orig(tokens, idx, options, env, self) : self.renderToken(tokens, idx, options)
      }
    }
  }
})

^ This will work with public paths in markdown (html in markdown won't be transformed even if it has href starting with /):

[b.zip](/dn/b.zip){download}

[qylwhetc]

first method is too difficult for people who only use markdown。
I still hope to be able to analyze it correctly：

[b.zip](/dn/b.zip){download}

I need to test the following method, thank you！

[qylwhetc]

在你的代码基础是稍作修改：

  async buildEnd(siteConfig) {
    //--  const files = glob('**/*.zip', { cwd: siteConfig.srcDir })
    const pattern = '**/*.{pdf,zip,docx,xlsx,txt}'  //++
    const files = glob(pattern, { cwd: siteConfig.srcDir, ignore: ['**/node_modules/**'] })  //++

    for await (const file of files) {
      if (file.startsWith('node_modules')) { continue; }  //++
      //...
    }
  }

修改：

  markdown: {
    config(md) {
      const orig = md.renderer.rules.link_open
      md.renderer.rules.link_open = (tokens, idx, options, env, self) => {
        const hrefIndex = tokens[idx].attrIndex('href')
        const downloadIndex = tokens[idx].attrIndex('download')

        const targetIndex = tokens[idx].attrIndex('target')  //++

        if (hrefIndex >= 0 && (downloadIndex >= 0 || targetIndex >= 0)) {
           //...
        }
        return orig ? orig(tokens, idx, options, env, self) : self.renderToken(tokens, idx, options)
      }
    }
  }

测试：

<a href="./q.txt" download>下载txt</a>  
[下载zip](/dn/q.zip){download}  
[链接pdf](/dn/q.pdf){target=_blank}

生成

<p><a href="./q.txt" download>下载txt</a></p>
<p><a href="/vp/dn/q.zip" download="">下载zip</a></p>
<p><a href="/vp/dn/q.pdf" target="_blank">链接pdf</a></p>

基本可行（download="" 不爽）；同时，查看拷贝的文件，也是正确的。

谢谢了！

[brc-dd]

download="" 不爽

Both download and download="" behave the same in HTML. There are just few more characters in the latter. But I don't think that should cause any issue.

const files = glob(pattern, { cwd: siteConfig.srcDir, ignore: ['**/node_modules/**'] })

It should be exclude instead of ignore - https://nodejs.org/api/fs.html#fspromisesglobpattern-options. I think you can remove that file.startsWith('node_modules') thing after that. (Needs Node.js v22.14.0 or higher. People on older Node.js versions can use some third party library like fast-glob / tinyglobby.)

[qylwhetc]

啊，知道了。
谢谢！