Merge pull request #135 from gijsk/links

Bug 1147584 - Don't strip unlikely <a>s, and replace useless <a>s with textContent

JSDOMParser.js

@@ -48,6 +48,12 @@
     "'": "'",
   };
 
+  function encodeTextContentHTML(s) {
+    return s.replace(/[&<>]/g, function(x) {
+      return reverseEntityTable[x];
+    });
+  }
+
   function encodeHTML(s) {
     return s.replace(/[&<>'"]/g, function(x) {
       return reverseEntityTable[x];
@@ -535,7 +541,7 @@
     },
     get innerHTML() {
       if (typeof this._innerHTML === "undefined") {
-        this._innerHTML = encodeHTML(this._textContent || "");
+        this._innerHTML = encodeTextContentHTML(this._textContent || "");
       }
       return this._innerHTML;
     },
@@ -583,7 +589,13 @@
     createElement: function (tag) {
       var node = new Element(tag);
       return node;
-    }
+    },
+
+    createTextNode: function (text) {
+      var node = new Text();
+      node.textContent = text;
+      return node;
+    },
   };
 
   var Element = function (tag) {

Readability.js

@@ -197,20 +197,28 @@ Readability.prototype = {
       return pathBase + uri;
     }
 
-    function convertRelativeURIs(tagName, propName) {
+    var links = articleContent.getElementsByTagName("a");
-      var elems = articleContent.getElementsByTagName(tagName);
+    this._forEachNode(links, function(link) {
-      this._forEachNode(elems, function(elem) {
+      var href = link.getAttribute("href");
-        var relativeURI = elem.getAttribute(propName);
+      if (href) {
-        if (relativeURI != null)
+        // Replace links with javascript: URIs with text content, since
-          elem.setAttribute(propName, toAbsoluteURI(relativeURI));
+        // they won't work after scripts have been removed from the page.
-      });
+        if (href.indexOf("javascript:") === 0) {
+          var text = this._doc.createTextNode(link.textContent);
+          link.parentNode.replaceChild(text, link);
+        } else {
+          link.setAttribute("href", toAbsoluteURI(href));
         }
+      }
+    });
 
-     // Fix links.
+    var imgs = articleContent.getElementsByTagName("img");
-    convertRelativeURIs.call(this, "a", "href");
+    this._forEachNode(imgs, function(img) {
-
+      var src = img.getAttribute("src");
-     // Fix images.
+      if (src) {
-    convertRelativeURIs.call(this, "img", "src");
+        img.setAttribute("src", toAbsoluteURI(src));
+      }
+    });
   },
 
   /**
@@ -587,7 +595,8 @@ Readability.prototype = {
         if (stripUnlikelyCandidates) {
           if (this.REGEXPS.unlikelyCandidates.test(matchString) &&
               !this.REGEXPS.okMaybeItsACandidate.test(matchString) &&
-              node.tagName !== "BODY") {
+              node.tagName !== "BODY" &&
+              node.tagName !== "A") {
             this.log("Removing unlikely candidate - " + matchString);
             node = this._removeAndGetNext(node);
             continue;

test/test-jsdomparser.js

@@ -218,8 +218,10 @@ describe("Test HTML escaping", function() {
     // let's manipulate via textContent in order to test that it alters
     // the innerHTML correctly.
     txtNode.textContent = txtNode.textContent + " ";
-    expect("<p>" + txtNode.innerHTML + "</p>").eql(baseStr.replace("</p>", " </p>"));
+    txtNode.textContent = txtNode.textContent.trim();
-    expect("<p>" + p.innerHTML + "</p>").eql(baseStr.replace("</p>", " </p>"));
+    var expectedHTML = baseStr.replace("&quot;", '"').replace("&apos;", "'");
+    expect("<p>" + txtNode.innerHTML + "</p>").eql(expectedHTML);
+    expect("<p>" + p.innerHTML + "</p>").eql(expectedHTML);
 
   });
 

test/test-pages/clean-links/expected-metadata.json

@@ -0,0 +1,6 @@
+{
+  "title": "Bartleby the Scrivener Web Study Text",
+  "byline": null,
+  "excerpt": "Ere introducing the scrivener, as he first appeared to me, it is fit \n          I make some mention of myself, my employees, my business, my chambers, \n          and general surroundings; because some such description is indispensable \n          to an adequate understanding of the chief character about to be presented.",
+  "readerable": true
+}