Support parsing remote sitemap indexes

helpers/api/sitemap/parse.js

@@ -7,12 +7,25 @@ import {
     sitemapLocation,
     sitemapIndexFileName,
 } from '~/helpers/constants.js'
+import { isValidHttpUrl } from '~/helpers/check-types.js'
 
 const sitemapFilesToTry = [
     sitemapIndexFileName,
     'sitemap.xml'
 ]
 
+function getDomainFromString ( urlString ) {
+    const url = new URL( urlString )
+
+    return url.hostname
+}
+
+function getPathFromUrl ( urlString ) {
+    const url = new URL( urlString )
+
+    return url.pathname
+}
+
 export function parseSitemapXml ( sitemapXml ) {
     // Get URLs from index
     const sitemapRoot = parse( sitemapXml )
@@ -39,8 +52,23 @@ export function parseSitemapXml ( sitemapXml ) {
 }
 
 export async function getAllUrlsFromLocalSitemap ( sitemapPath ) {
+    return await getAllUrlsFromSitemap( sitemapPath, {
+        getMethod: async ( sitemapPath ) => {
+            const sitemapXml = await fs.readFile( sitemapPath, 'utf8' )
+
+            return sitemapXml
+        }
+    })
+}
+
+export async function getAllUrlsFromSitemap ( sitemapSource, { getMethod } = {} ) {
+
+    const sitemapPath = isValidHttpUrl( sitemapSource ) ? getPathFromUrl( sitemapSource ) : sitemapSource
+
+    // console.log('sitemapPath', sitemapPath)
+
     // Get intial sitemap
-    const sitemapXml = await fs.readFile( sitemapPath, 'utf8' )
+    const sitemapXml = await getMethod( sitemapSource )// fs.readFile( sitemapPath, 'utf8' )
     const sitemapDirectory = path.dirname( sitemapPath )
 
     // Get URLs from index
@@ -59,15 +87,58 @@ export async function getAllUrlsFromLocalSitemap ( sitemapPath ) {
 
         const childSitemapPath = path.join( sitemapDirectory, sitemapUrl.pathname )
 
-        return await getAllUrlsFromLocalSitemap( childSitemapPath )
+        return await getAllUrlsFromSitemap( childSitemapPath, { getMethod } )
     }))
 
     // Flatten array
     return sitemaps.flat()
 }
 
-export async function fetchParsedSitemapXmlForDomain ( domain ) {
+export async function fetchAllUrlsFromSitemaps ( urlString ) {
+    // const domain = getDomainFromString( urlString )
+
+    const allUrls = new Map()
+
     for ( const sitemapFile of sitemapFilesToTry ) {
 
+        const sitemapUrl = new URL( sitemapFile, urlString )
+
+        // console.log( 'sitemapUrl', sitemapUrl.href )
+
+        // Just do a quich HEAD request to see if the file exists with getting the whole body
+        const exists = await axios.head( sitemapUrl.href )
+            .catch( () => false )
+            .then( response => {
+                // console.log( 'response', response.status )
+                return response.status < 300
+            } )
+
+        // console.log( 'exists', exists )
+
+        // Skip if sitemap doesn't exist
+        if ( !exists ) continue
+
+        // Fetch the whole sitemap
+        const urls = await getAllUrlsFromSitemap( sitemapUrl.href , {
+            getMethod: async sitemapPath => {
+                const sitemapUrl = new URL( sitemapPath, urlString )
+
+                const sitemapXml = await axios.get( sitemapUrl.href )
+                    .then( response => response.data )
+
+                return sitemapXml
+            }
+        })
+
+        // console.log( 'Sitemap urls', urls )
+
+        // Set our urls to the map
+        urls.forEach( tag => allUrls.set( tag.loc, new URL( tag.loc ) ) )
+
+        // Store sitemap urls to context
+        // const urlsMap = new Map( urls.map( tag => [ tag.loc, new URL( tag.loc )] ) )
+
     }
+
+    return allUrls
 }

test/main.js

@@ -14,7 +14,8 @@ import {
 import { logArraysDifference } from '~/helpers/array.js'
 import {
     parseSitemapXml,
-    getAllUrlsFromLocalSitemap
+    getAllUrlsFromLocalSitemap,
+    fetchAllUrlsFromSitemaps
 } from '~/helpers/api/sitemap/parse.js'
 
 require('dotenv').config()
@@ -122,11 +123,12 @@ test('Sitemap mostly matches production', async (t) => {
     const urlsNotOnLive = new Set()
     // const newLocalUrls = new Set()
 
-    const liveSitemapXml = await axios( 'https://doesitarm.com/sitemap.xml' ).then( response => response.data )
+    const liveSitemapUrls = await fetchAllUrlsFromSitemaps( 'https://doesitarm.com' )
-    const liveSitemap = parser.parse( liveSitemapXml )
 
-    // Store sitemap urls to context
+    // Assert that any sitemap urls exist on live
-    const liveSitemapUrls = new Map( liveSitemap.urlset.url.map( tag => [ tag.loc, new URL( tag.loc )] ) )
+    t.assert( liveSitemapUrls.size > 0, 'No sitemap urls found on live.' )
+
+    // console.log( 'liveSitemapUrls', liveSitemapUrls )
 
 
     for ( const localUrl of t.context.sitemapUrls ) {