From a476a773f10264734ce1735d9690c3668f5a6de4 Mon Sep 17 00:00:00 2001
From: Noel De Martin <noel@moodle.com>
Date: Thu, 23 Jan 2020 10:33:43 +0100
Subject: [PATCH] MOBILE-3286 login: Improve guessing site domains

---
 src/classes/utils/url.ts          | 61 ++++++++++++++++++++++++-------
 src/core/login/pages/site/site.ts | 16 +++-----
 2 files changed, 54 insertions(+), 23 deletions(-)

diff --git a/src/classes/utils/url.ts b/src/classes/utils/url.ts
index 579d8cbb0..2539d19c6 100644
--- a/src/classes/utils/url.ts
+++ b/src/classes/utils/url.ts
@@ -27,6 +27,11 @@ interface UrlParts {
      */
     domain?: string;
 
+    /**
+     * Url port.
+     */
+    port?: string;
+
     /**
      * Url path.
      */
@@ -49,39 +54,69 @@ interface UrlParts {
  */
 export class CoreUrl {
 
-    // Avoid creating singleton instances
+    // Avoid creating singleton instances.
     private constructor() {}
 
     /**
      * Parse parts of a url, using an implicit protocol if it is missing from the url.
      *
      * @param url Url.
-     * @param implicitProtocol Protocol to be used if the url doesn't have any.
      * @return Url parts.
      */
-    static parse(url: string, implicitProtocol?: string): UrlParts | null {
-        // Prepare url before parsing
-        url = url.trim();
-
-        if (implicitProtocol && !url.match(/^[a-zA-Z]+:\/\//)) {
-            url = `${implicitProtocol}://${url}`;
-        }
-
-        // Regular expression taken from RFC 3986: https://tools.ietf.org/html/rfc3986#appendix-B
+    static parse(url: string): UrlParts | null {
+        // Parse url with regular expression taken from RFC 3986: https://tools.ietf.org/html/rfc3986#appendix-B.
         const match = url.trim().match(/^(([^:/?#]+):)?(\/\/([^/?#]*))?([^?#]*)(\?([^#]*))?(#(.*))?/);
 
         if (!match) {
             return null;
         }
 
-        // Prepare parts replacing empty strings with undefined
+        // Split host into domain and port.
+        const host = match[4] || '';
+        const [domain, port]: string[] = host.indexOf(':') === -1 ? [host] : host.split(':');
+
+        // Prepare parts replacing empty strings with undefined.
         return {
             protocol: match[2] || undefined,
-            domain: match[4] || undefined,
+            domain: domain || undefined,
+            port: port || undefined,
             path: match[5] || undefined,
             query: match[7] || undefined,
             fragment: match[9] || undefined,
         };
     }
 
+    /**
+     * Guess the Moodle domain from a site url.
+     *
+     * @param url Site url.
+     * @return Guessed Moodle domain.
+     */
+    static guessMoodleDomain(url: string): string | null {
+        // Add protocol if it was missing. Moodle can only be served through http or https, so this is a fair assumption to make.
+        if (!url.match(/^https?:\/\//)) {
+            url = `https://${url}`;
+        }
+
+        // Match using common suffixes.
+        const knownSuffixes = [
+            '\/my\/?',
+            '\/\\\?redirect=0',
+            '\/index\\\.php',
+            '\/course\/view\\\.php',
+            '\/login\/index\\\.php',
+            '\/mod\/page\/view\\\.php',
+        ];
+        const match = url.match(new RegExp(`^https?:\/\/(.*?)(${knownSuffixes.join('|')})`));
+
+        if (match) {
+            return match[1];
+        }
+
+        // If nothing else worked, parse the domain.
+        const urlParts = CoreUrl.parse(url);
+
+        return urlParts && urlParts.domain ? urlParts.domain : null;
+    }
+
 }
diff --git a/src/core/login/pages/site/site.ts b/src/core/login/pages/site/site.ts
index 27db57687..5a231d27c 100644
--- a/src/core/login/pages/site/site.ts
+++ b/src/core/login/pages/site/site.ts
@@ -87,6 +87,8 @@ export class CoreLoginSitePage {
             return;
         }
 
+        url = url.trim();
+
         const modal = this.domUtils.showModalLoading(),
             siteData = this.sitesProvider.getDemoSiteData(url);
 
@@ -113,18 +115,12 @@ export class CoreLoginSitePage {
         } else {
             // Not a demo site.
             this.sitesProvider.checkSite(url)
-
-                // Attempt parsing the domain after initial check failed
                 .catch((error) => {
-                    const urlParts = CoreUrl.parse(url, 'http');
+                    // Attempt guessing the domain if the initial check failed
+                    const domain = CoreUrl.guessMoodleDomain(url);
 
-                    if (!urlParts || !urlParts.domain) {
-                        throw error;
-                    }
-
-                    return this.sitesProvider.checkSite(urlParts.domain);
+                    return domain ? this.sitesProvider.checkSite(domain) : Promise.reject(error);
                 })
-
                 .then((result) => this.login(result))
                 .catch((error) => this.showLoginIssue(url, error))
                 .finally(() => modal.dismiss());
@@ -177,7 +173,7 @@ export class CoreLoginSitePage {
      *
      * @return Promise resolved after logging in.
      */
-    private async login(response: CoreSiteCheckResponse): Promise<void> {
+    protected async login(response: CoreSiteCheckResponse): Promise<void> {
         return this.sitesProvider.checkRequiredMinimumVersion(response.config).then(() => {
             if (response.warning) {
                 this.domUtils.showErrorModal(response.warning, true, 4000);