feat(route): add sitemap transform (#13297)

* feat(route): add sitemap transform

* chore: resolve lint errors

* chore: change router order

* chore: apply the suggestion (xml2js -> cheerio)

* chore: fix errors

* chore: add maintainer

* fix: sticks to sitemap protocol

---------

lib/v2/rsshub/maintainer.js

@@ -3,4 +3,5 @@ module.exports = {
     '/rsshub/sponsors': ['DIYgod'],
     '/transform/html/:url/:routeParams': ['ttttmr'],
     '/transform/json/:url/:routeParams': ['ttttmr'],
+    '/transform/sitemap/:url/:routeParams?': ['flrngel'],
 };

lib/v2/rsshub/router.js

@@ -3,4 +3,5 @@ module.exports = (router) => {
     router.get('/sponsors', require('./sponsors'));
     router.get('/transform/html/:url/:routeParams', require('./transform/html'));
     router.get('/transform/json/:url/:routeParams', require('./transform/json'));
+    router.get('/transform/sitemap/:url/:routeParams?', require('./transform/sitemap'));
 };

lib/v2/rsshub/transform/sitemap.js

@@ -0,0 +1,52 @@
+const got = require('@/utils/got');
+const cheerio = require('cheerio');
+const config = require('@/config').value;
+
+module.exports = async (ctx) => {
+    if (!config.feature.allow_user_supply_unsafe_domain) {
+        ctx.throw(403, `This RSS is disabled unless 'ALLOW_USER_SUPPLY_UNSAFE_DOMAIN' is set to 'true'.`);
+    }
+    const { url } = ctx.params;
+    const response = await got({
+        method: 'get',
+        url,
+    });
+
+    const routeParams = new URLSearchParams(ctx.params.routeParams);
+    const $ = cheerio.load(response.data, { xmlMode: true });
+
+    const rssTitle = routeParams.get('title') ? routeParams.get('title') : ($('urlset url').length && $('urlset url').first().find('loc').text() ? $('urlset url').first().find('loc').text() : 'Sitemap');
+
+    let items;
+    const urls = $('urlset url').toArray();
+    if (urls && urls.length) {
+        items = urls.map((item) => {
+            try {
+                const title = $(item).find('loc').text() || '';
+                const link = $(item).find('loc').text() || '';
+                const description = $(item).find('loc').text() || '';
+                const pubDate = $(item).find('lastmod').text() || undefined;
+
+                return {
+                    title,
+                    link,
+                    description,
+                    pubDate,
+                };
+            } catch (e) {
+                return null;
+            }
+        })
+        .filter(Boolean);
+    }
+    else {
+        items = [];
+    }
+
+    ctx.state.data = {
+        title: rssTitle,
+        link: url,
+        description: `Proxy ${url}`,
+        item: items,
+    };
+};

website/docs/routes/other.md

@@ -646,6 +646,25 @@ Parsing of `routeParams` parameter:
 
 </Route>
 
+### Sitemap {#transformation-sitemap}
+
+Specify options (in the format of query string) in parameter `routeParams` parameter to extract data from Sitemap. (Follows Sitemap Protocol 0.9)
+
+| Key             | Meaning                                                        | Accepted Values | Default                |
+|-----------------|----------------------------------------------------------------|-----------------|------------------------|
+| `title`         | The title of the RSS                                           | `string`        | Extract from `<title>` |
+
+<Route author="flrngel" example="/rsshub/transform/xml/https%3A%2F%2Fwww.sitemaps.org%2Fsitemap.xml/" path="/rsshub/transform/html/:url/:routeParams?" paramsDesc={['`encodeURIComponent`ed URL address', 'Transformation rules, requires URL encode']} selfhost="1">
+
+Parameters parsing in the above example:
+
+| Parameter     | Value                                     |
+|---------------|-------------------------------------------|
+| `url`         | `https://www.sitemaps.org/sitemap.xml` |
+| `routeParams` | `title=Example`      |
+
+</Route>
+
 ## TSSstatus (iOS downgrade channel) {#tssstatus-ios-downgrade-channel}
 
 ### Status {#tssstatus-ios-downgrade-channel-status}