crawler.js: crawler http header

crawler.js

@@ -3,6 +3,7 @@
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 var fs = require("fs");
+const cheerio = require('cheerio');
 var args = require("yargs")
   .usage("Usage : crawler.js -link website_link -f file_name")
   .demandOption(['link'])
@@ -39,15 +40,14 @@ crawler.maxConcurrency = 5;
 crawler.maxDepth = depth;
 
 crawler.on("fetchcomplete", function(queueItem, responseBuffer, response) {
-  var mime = response.headers["content-type"];
-  if (mime.startsWith("text/html") || mime.startsWith("text/plain")) {
-    readline.cursorTo(process.stdout, 0);
-    process.stdout.write(count + "");
-    count+=1;
-    url_list.push({
-      "loc": queueItem.url
-    });
-  }
+  readline.cursorTo(process.stdout, 0);
+  process.stdout.write(count + "");
+  count+=1;
+  url_list.push({
+    "loc": queueItem.url,
+    "stateData": queueItem.stateData,
+    "referrer": queueItem.referrer
+  });
 });
 
 // Fire callback
@@ -56,7 +56,6 @@ crawler.on("complete", function() {
   console.log("crawled " + url_list.length + " urls");
   var xml_dict = {
     "urlset": {
-      "@xmlns": "http://www.sitemaps.org/schemas/sitemap/0.9",
       "url": url_list
     }
   };
@@ -66,10 +65,32 @@ crawler.on("complete", function() {
 });
 
 crawler.on("fetcherror", function(queueItem, response) {
-  console.log("");
   console.log("Error " + response.statusCode + " while fetching " + queueItem.url);
+  url_list.push({
+    "loc": queueItem.url,
+    "stateData": queueItem.stateData,
+    "referrer": queueItem.referrer
+  });
 });
 
+
+crawler.discoverResources = function(buffer, queueItem) {
+  var $ = cheerio.load(buffer.toString("utf8"));
+
+  var tag_a =  $("a[href]").map(function () {
+      return $(this).attr("href");
+  }).get();
+
+  var tag_link = $("link[href]").map(function () {
+    return $(this).attr("href");
+  }).get();
+
+  var tag_script = $("script[src]").map(function () {
+    return $(this).attr("src");
+  }).get();
+  return tag_a.concat(tag_link).concat(tag_script);
+};
+
 // Start Crawl
 crawler.start();
 

package.json

@@ -12,6 +12,7 @@
   "dependencies": {
     "simplecrawler": "^1.1.9",
     "xmlbuilder": "^15.1.1",
-    "yargs": "^16.2.0"
+    "yargs": "^16.2.0",
+    "cheerio": "^0.22.0"
   }
 }