feat: external XML sitemaps as sources

nuxt-modules · Jan 17, 2025 · add57d3 · add57d3
1 parent c7b1ebd
commit add57d3
Show file tree

Hide file tree

Showing 6 changed files with 345 additions and 9 deletions.
diff --git a/docs/content/2.guides/0.data-sources.md b/docs/content/2.guides/0.data-sources.md
@@ -72,7 +72,8 @@ export default defineNuxtConfig({
 
 If you need your sitemap data to always be up-to-date at runtime, you will need to provide your own sources explicitly.
 
-A source is a URL that will be fetched and is expected to return an array of Sitemap URL entries.
+A source is a URL that will be fetched and is expected to return either JSON with an array of Sitemap URL entries or
+a XML sitemap.
 
 ::code-group
 

diff --git a/docs/content/2.guides/0.multi-sitemaps.md b/docs/content/2.guides/0.multi-sitemaps.md
@@ -166,7 +166,7 @@ export default defineSitemapEventHandler(() => {
 If you need to fetch the URLs from an endpoint for a sitemap, then you will need to use either the `urls` or `sources` option.
 
 - `urls` - Array of static URLs to include in the sitemap. You should avoid using this option if you have a lot of URLs
-- `sources` - Custom endpoint to fetch [dynamic URLs](/docs/sitemap/guides/dynamic-urls) from.
+- `sources` - Custom endpoint to fetch [dynamic URLs](/docs/sitemap/guides/dynamic-urls) from as JSON or XML.
 
 ```ts
 export default defineNuxtConfig({

diff --git a/docs/content/2.guides/2.dynamic-urls.md b/docs/content/2.guides/2.dynamic-urls.md
@@ -8,7 +8,22 @@ description: Use runtime API endpoints to generate dynamic URLs for your sitemap
 In some instances, like using a CMS, you may need to implement an endpoint to make
 all of your site URLs visible to the module.
 
-To do this, you can provide [user sources](/docs/sitemap/getting-started/data-sources) to the module.
+To do this, you can provide [user sources](/docs/sitemap/getting-started/data-sources) to the module. These can either be 
+a JSON response or an XML sitemap.
+
+## XML Sitemap
+
+If you're providing an XML sitemap, you can use the `sources` option to provide the URL to the sitemap.
+
+```ts [nuxt.config.ts]
+export default defineNuxtConfig({
+  sitemap: {
+    sources: [
+      'https://example.com/sitemap.xml',
+    ]
+  }
+})
+```
 
 ## Dynamic URLs from an external API
 

diff --git a/src/runtime/server/sitemap/urlset/sources.ts b/src/runtime/server/sitemap/urlset/sources.ts
@@ -2,12 +2,14 @@ import { getRequestHost } from 'h3'
 import type { H3Event } from 'h3'
 import type { FetchError } from 'ofetch'
 import { defu } from 'defu'
+import { parseURL } from 'ufo'
 import type {
   ModuleRuntimeConfig,
   SitemapSourceBase,
   SitemapSourceResolved,
   SitemapUrlInput,
 } from '../../../types'
+import { extractSitemapXML } from '#sitemap/server/sitemap/utils/extractSitemapXML'
 
 export async function fetchDataSource(input: SitemapSourceBase | SitemapSourceResolved, event?: H3Event): Promise<SitemapSourceResolved> {
   const context = typeof input.context === 'string' ? { name: input.context } : input.context || { name: 'fetch' }
@@ -21,24 +23,25 @@ export async function fetchDataSource(input: SitemapSourceBase | SitemapSourceRe
   const timeoutController = new AbortController()
   const abortRequestTimeout = setTimeout(() => timeoutController.abort(), timeout)
 
-  let isHtmlResponse = false
+  let isMaybeErrorResponse = false
+  const isXmlRequest = parseURL(url).pathname.endsWith('.xml')
+  const fetchContainer = (url.startsWith('/') && event) ? event : globalThis
   try {
-    const fetchContainer = (url.startsWith('/') && event) ? event : globalThis
-    const urls = await fetchContainer.$fetch(url, {
+    const res = await fetchContainer.$fetch(url, {
       ...options,
-      responseType: 'json',
+      responseType: isXmlRequest ? 'json' : 'text',
       signal: timeoutController.signal,
       headers: defu(options?.headers, {
         Accept: 'application/json',
       }, event ? { Host: getRequestHost(event, { xForwardedHost: true }) } : {}),
       // @ts-expect-error untyped
       onResponse({ response }) {
         if (typeof response._data === 'string' && response._data.startsWith('<!DOCTYPE html>'))
-          isHtmlResponse = true
+          isMaybeErrorResponse = true
       },
     })
     const timeTakenMs = Date.now() - start
-    if (isHtmlResponse) {
+    if (isMaybeErrorResponse) {
       context.tips.push('This is usually because the URL isn\'t correct or is throwing an error. Please check the URL')
       return {
         ...input,
@@ -48,6 +51,14 @@ export async function fetchDataSource(input: SitemapSourceBase | SitemapSourceRe
         error: 'Received HTML response instead of JSON',
       }
     }
+    let urls = []
+    if (typeof res === 'object') {
+      urls = res.urls || res
+    }
+    else if (typeof res === 'string' && parseURL(url).pathname.endsWith('.xml')) {
+      // fast pass XML extract all loc data, let's use
+      urls = extractSitemapXML(res)
+    }
     return {
       ...input,
       context,

diff --git a/src/runtime/server/sitemap/utils/extractSitemapXML.ts b/src/runtime/server/sitemap/utils/extractSitemapXML.ts
@@ -0,0 +1,101 @@
+import type { SitemapUrlInput } from '../../../types'
+
+export function extractSitemapXML(xml: string): SitemapUrlInput[] {
+  const urls = xml.match(/<url>[\s\S]*?<\/url>/g) || []
+  return urls.map((url) => {
+    const loc = url.match(/<loc>([^<]+)<\/loc>/)?.[1]
+    if (!loc) return null
+
+    const lastmod = url.match(/<lastmod>([^<]+)<\/lastmod>/)?.[1]
+    const changefreq = url.match(/<changefreq>([^<]+)<\/changefreq>/)?.[1]
+    const priority = url.match(/<priority>([^<]+)<\/priority>/) ? Number.parseFloat(url.match(/<priority>([^<]+)<\/priority>/)[1]) : undefined
+
+    const images = (url.match(/<image:image>[\s\S]*?<\/image:image>/g) || []).map((image) => {
+      const imageLoc = image.match(/<image:loc>([^<]+)<\/image:loc>/)?.[1]
+      return imageLoc ? { loc: imageLoc } : null
+    }).filter(Boolean)
+
+    const videos = (url.match(/<video:video>[\s\S]*?<\/video:video>/g) || []).map((video) => {
+      const videoObj: any = {}
+      const title = video.match(/<video:title>([^<]+)<\/video:title>/)?.[1]
+      const thumbnail_loc = video.match(/<video:thumbnail_loc>([^<]+)<\/video:thumbnail_loc>/)?.[1]
+      const description = video.match(/<video:description>([^<]+)<\/video:description>/)?.[1]
+      const content_loc = video.match(/<video:content_loc>([^<]+)<\/video:content_loc>/)?.[1]
+      if (!title || !thumbnail_loc || !description || !content_loc) return null
+
+      videoObj.title = title
+      videoObj.thumbnail_loc = thumbnail_loc
+      videoObj.description = description
+      videoObj.content_loc = content_loc
+
+      const player_loc = video.match(/<video:player_loc>([^<]+)<\/video:player_loc>/)?.[1]
+      if (player_loc) videoObj.player_loc = player_loc
+
+      const duration = video.match(/<video:duration>([^<]+)<\/video:duration>/) ? Number.parseInt(video.match(/<video:duration>([^<]+)<\/video:duration>/)[1], 10) : undefined
+      if (duration) videoObj.duration = duration
+
+      const expiration_date = video.match(/<video:expiration_date>([^<]+)<\/video:expiration_date>/)?.[1]
+      if (expiration_date) videoObj.expiration_date = expiration_date
+
+      const rating = video.match(/<video:rating>([^<]+)<\/video:rating>/) ? Number.parseFloat(video.match(/<video:rating>([^<]+)<\/video:rating>/)[1]) : undefined
+      if (rating) videoObj.rating = rating
+
+      const view_count = video.match(/<video:view_count>([^<]+)<\/video:view_count>/) ? Number.parseInt(video.match(/<video:view_count>([^<]+)<\/video:view_count>/)[1], 10) : undefined
+      if (view_count) videoObj.view_count = view_count
+
+      const publication_date = video.match(/<video:publication_date>([^<]+)<\/video:publication_date>/)?.[1]
+      if (publication_date) videoObj.publication_date = publication_date
+
+      const family_friendly = video.match(/<video:family_friendly>([^<]+)<\/video:family_friendly>/)?.[1]
+      if (family_friendly) videoObj.family_friendly = family_friendly
+
+      const restriction = video.match(/<video:restriction relationship="([^"]+)">([^<]+)<\/video:restriction>/)
+      if (restriction) videoObj.restriction = { relationship: restriction[1], restriction: restriction[2] }
+
+      const platform = video.match(/<video:platform relationship="([^"]+)">([^<]+)<\/video:platform>/)
+      if (platform) videoObj.platform = { relationship: platform[1], platform: platform[2] }
+
+      const price = (video.match(/<video:price [^>]+>([^<]+)<\/video:price>/g) || []).map((price) => {
+        const priceValue = price.match(/<video:price [^>]+>([^<]+)<\/video:price>/)?.[1]
+        const currency = price.match(/currency="([^"]+)"/)?.[1]
+        const type = price.match(/type="([^"]+)"/)?.[1]
+        return priceValue ? { price: priceValue, currency, type } : null
+      }).filter(Boolean)
+      if (price.length) videoObj.price = price
+
+      const requires_subscription = video.match(/<video:requires_subscription>([^<]+)<\/video:requires_subscription>/)?.[1]
+      if (requires_subscription) videoObj.requires_subscription = requires_subscription
+
+      const uploader = video.match(/<video:uploader info="([^"]+)">([^<]+)<\/video:uploader>/)
+      if (uploader) videoObj.uploader = { uploader: uploader[2], info: uploader[1] }
+
+      const live = video.match(/<video:live>([^<]+)<\/video:live>/)?.[1]
+      if (live) videoObj.live = live
+
+      const tag = (video.match(/<video:tag>([^<]+)<\/video:tag>/g) || []).map(tag => tag.match(/<video:tag>([^<]+)<\/video:tag>/)?.[1]).filter(Boolean)
+      if (tag.length) videoObj.tag = tag
+
+      return videoObj
+    }).filter(Boolean)
+
+    const alternatives = (url.match(/<xhtml:link[\s\S]*?\/>/g) || []).map((link) => {
+      const hreflang = link.match(/hreflang="([^"]+)"/)?.[1]
+      const href = link.match(/href="([^"]+)"/)?.[1]
+      return hreflang && href ? { hreflang, href } : null
+    }).filter(Boolean)
+
+    const news = url.match(/<news:news>[\s\S]*?<\/news:news>/)
+      ? {
+          title: url.match(/<news:title>([^<]+)<\/news:title>/)?.[1],
+          publication_date: url.match(/<news:publication_date>([^<]+)<\/news:publication_date>/)?.[1],
+          publication: {
+            name: url.match(/<news:name>([^<]+)<\/news:name>/)?.[1],
+            language: url.match(/<news:language>([^<]+)<\/news:language>/)?.[1],
+          },
+        }
+      : undefined
+
+    const urlObj: any = { loc, lastmod, changefreq, priority, images, videos, alternatives, news }
+    return Object.fromEntries(Object.entries(urlObj).filter(([_, v]) => v != null && v.length !== 0))
+  }).filter(Boolean) as any as SitemapUrlInput[]
+}