Replaced hard-coded rules by a DSL

configs/ghostery.js

@@ -17,7 +17,8 @@ module.exports = {
   publish: publish.toEdge('browser-core', 'ghostery'),
   sourceMaps: false,
   format: 'common',
-  settings: { ...urls,
+  settings: {
+    ...urls,
     channel: 'CH80',
     MSGCHANNEL: 'web-extension',
     URL_CHANGED_EVENT_DEBOUNCE: 500,
@@ -33,7 +34,10 @@ module.exports = {
       },
     },
     HUMAN_WEB_LITE_COLLECTOR_VIA_PROXY: 'https://collector-hpn.ghostery.net',
-    HUMAN_WEB_LITE_COLLECTOR_DIRECT: 'https://collector-hpn.ghostery.net' },
+    HUMAN_WEB_LITE_COLLECTOR_DIRECT: 'https://collector-hpn.ghostery.net',
+    HUMAN_WEB_LITE_PATTERNS: 'https://cdn2.ghostery.com/human-web-android/patterns.json',
+    HUMAN_WEB_LITE_AUTO_TRIGGER: true
+  },
   default_prefs: {
     'modules.human-web.enabled': true,
     'modules.antitracking.enabled': true,

modules/human-web-lite/sources/html-parser.es

@@ -6,23 +6,10 @@
  * file, You can obtain one at https://mozilla.org/MPL/2.0/.
  */
 
-// TODO: here we need something that works in react-native.
-// The following code is copied from human-web and probably
-// will not work outside of the browser.
-//
-// If there are problems, maybe one of the libraries mentioned here
-// can be used on Mobile:
-// * https://stackoverflow.com/q/38343951/783510
-//
-// Note: In jsdom, this implementation should work:
-//
-//   return new JSDOM(html).window.document;
-//
-// However, it is hard to use jsdom outside of NodeJs.
-// I failed to get it working in the Browser, for example.
-//
 import window from '../core/globals-window';
 
+// TODO: consider using linkedom here as well
+// (Note: we need something that works in react-native)
 export default function parseHtml(html) {
   if (!parseHtml.domParser) {
     parseHtml.domParser = new window.DOMParser();

modules/human-web-lite/sources/http.es

@@ -9,23 +9,90 @@
 
 const SECOND = 1000;
 
-// TODO: Can we use the "AbortController" API on Mobile? Otherwise,
-// running jobs in background might be difficult because we risk that
-// the app will be killed.
-export async function anonymousHttpGet(url, { timeout = 15 * SECOND } = {}) {
+/**
+ * Performs a HTTP Get.
+ *
+ * Optional:
+ * - headers: allows to overwrite headers
+ * - redirect: 'follow' or 'manual' (default; i.e. redirects are errors)
+ */
+export async function anonymousHttpGet(url, {
+  headers = null,
+  redirect = null,
+  timeout = 15 * SECOND,
+} = {}) {
   const options = {
     credentials: 'omit',
     mode: 'no-cors',
-    redirect: 'manual',
+    redirect: redirect || 'manual',
 
     // TODO: Or maybe this does work? It is not part of the fetch standard,
     // but I have seen it in some react-native examples.
     // If it works, it could be used if AbortController is not available.
     timeout,
   };
-  const response = await fetch(url, options);
-  if (!response.ok) {
-    throw new Error(`Failed to fetch url ${url}: ${response.statusText}`);
+
+  // The following code overwrite the headers of the request.
+  // Note that "fetch" allows to overwrite headers in a simple declarative way,
+  // but unfortunately it is limited. For example, it is not possible to
+  // overwrite the cookie headers. The following code will work for all
+  // type of headers.
+  //
+  // The matching logic is not perfect but should be fairly accurate.
+  // Ideally, we would want to run the handler only for the request that we
+  // are about to trigger, but not for any other requests to avoid unintended
+  // side-effects. To mitigate the risk, uninstall the handler at the first
+  // opportunity: either if it is called or if the request finished
+  // (and we know the handle will never be called).
+  let webRequestHandler;
+  const uninstallHandler = () => {
+    if (webRequestHandler) {
+      chrome.webRequest.onBeforeSendHeaders.removeListener(webRequestHandler);
+      webRequestHandler = null;
+    }
+  };
+  const headerNames = Object.keys(headers || {});
+  if (headerNames.length > 0) {
+    webRequestHandler = (details) => {
+      if (details.url !== url || details.type !== 'xmlhttprequest' || details.method !== 'GET') {
+        // does that match the request that we intended to trigger
+        return {};
+      }
+
+      // match: now we can already deregister the listener
+      // (it should not be executed multiple times)
+      uninstallHandler();
+      const normalizedHeaders = headerNames.map(x => x.toLowerCase());
+
+      /* eslint-disable no-param-reassign */
+      details.requestHeaders = details.requestHeaders.filter(
+        header => !normalizedHeaders.includes(header.name.toLowerCase())
+      );
+
+      headerNames.forEach((name) => {
+        details.requestHeaders.push({
+          name,
+          value: headers[name],
+        });
+      });
+
+      return {
+        requestHeaders: details.requestHeaders
+      };
+    };
+    chrome.webRequest.onBeforeSendHeaders.addListener(webRequestHandler, {
+      urls: [url]
+    }, ['blocking', 'requestHeaders']);
+  }
+
+  try {
+    const response = await fetch(url, options);
+    if (!response.ok) {
+      throw new Error(`Failed to fetch url ${url}: ${response.statusText}`);
+    }
+    uninstallHandler();
+    return response.text();
+  } finally {
+    uninstallHandler();
   }
-  return response.text();
 }

modules/human-web-lite/sources/human-web-lite.es

@@ -6,6 +6,8 @@
  * file, You can obtain one at https://mozilla.org/MPL/2.0/.
  */
 
+import Patterns from './patterns';
+import PatternsUpdater from './patterns-updater';
 import Sanitizer from './sanitizer';
 import UrlAnalyzer from './url-analyzer';
 import MessageSender from './message-sender';
@@ -21,8 +23,16 @@ export default class HumanWebLite {
     // to collect data.
     this.isActive = false;
 
+    this.patterns = new Patterns();
+    this.patternsUpdater = new PatternsUpdater({
+      config,
+      patterns: this.patterns,
+      storage,
+      storageKey: 'patterns',
+    });
+
     this.sanitizer = new Sanitizer(config);
-    this.urlAnalyzer = new UrlAnalyzer();
+    this.urlAnalyzer = new UrlAnalyzer(this.patterns);
     this.persistedHashes = new PersistedHashes({
       storage,
       storageKey: 'deduplication_hashes',
@@ -32,16 +42,15 @@ export default class HumanWebLite {
     this.messageSender = new MessageSender(this.duplicateDetector, hpn);
     this.searchExtractor = new SearchExtractor({
       config,
+      patterns: this.patterns,
       sanitizer: this.sanitizer,
       persistedHashes: this.persistedHashes,
     });
     this.jobScheduler = new JobScheduler(this.messageSender, this.searchExtractor);
   }
 
   async init() {
-    // TODO: In a feature-complete implementation, you would need
-    // to have a mechanism to keep the extraction patterns up-to-date.
-    // As we have hard-coded patterns, there is nothing to do here.
+    await this.patternsUpdater.init();
     this.isActive = true;
   }
 
@@ -78,7 +87,19 @@ export default class HumanWebLite {
     return true;
   }
 
-  processPendingJobs() {
+  async processPendingJobs() {
+    await this._ensurePatternsAreUpToDate();
     return this.jobScheduler.processPendingJobs();
   }
+
+  async _ensurePatternsAreUpToDate() {
+    // Currently, the PatternsUpdater needs to be externally triggered.
+    // This implementation detail could be avoided, if the PatternsUpdater
+    // could use a browser API like timers in persistent background pages
+    // or the Alert API (Manifest V3).
+    // The "update" function is a quick operation unless for the rare
+    // situation that the patterns are outdated and need to be fetched.
+    // Thus, there should be no harm in calling it here.
+    await this.patternsUpdater.update();
+  }
 }