Add support for figure element

java/org/chromium/distiller/extractors/embeds/ImageExtractor.java

 // Copyright 2015 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 package org.chromium.distiller.extractors.embeds;
 
 import com.google.gwt.dom.client.Element;
 import com.google.gwt.dom.client.ImageElement;
+import com.google.gwt.dom.client.NodeList;
+import org.chromium.distiller.webdocument.WebFigure;
+import org.chromium.distiller.webdocument.WebImage;
 import org.chromium.distiller.LogUtil;

[wychen, 2016/06/07 17:39:19]

The sorting is case sensitive. Capital ones come first.

-import org.chromium.distiller.webdocument.WebImage;
 
 import java.util.HashSet;
 import java.util.Set;
 
 /**
  * This class treats images as another type of embed and provides heuristics for lead image
  * candidacy.
  */
 public class ImageExtractor implements EmbedExtractor {
     private static final Set<String> relevantTags = new HashSet<>();
+    private String imgSrc;
+    private int width;
+    private int height;
+
     static {
         // TODO(mdjones): Add "DIV" to this list for css images and possibly captions.
         relevantTags.add("IMG");
+        relevantTags.add("FIGURE");
     }
+
     private static final String[] LAZY_IMAGE_ATTRIBUTES =
-        {"data-src", "data-original", "datasrc", "data-url"};
+            {"data-src", "data-original", "datasrc", "data-url"};
 
     @Override
     public Set<String> getRelevantTagNames() {
         return relevantTags;
     }
 
     @Override
     public WebImage extract(Element e) {
         if (!relevantTags.contains(e.getTagName())) {
             return null;
         }
-        String imgSrc = "";
-        // Getting OffSetWidth/Height as default values, even they are
-        // affected by padding, border, etc.
-        int width = e.getOffsetWidth();
-        int height = e.getOffsetHeight();
+        imgSrc = "";
+
         if ("IMG".equals(e.getTagName())) {
-            // This will get the absolute URL of the image and
-            // the displayed image dimension.
-            ImageElement imageElement = ImageElement.as(e);
-            // Try to get lazily-loaded images before falling back to get the src attribute.
-            for(String attr: LAZY_IMAGE_ATTRIBUTES) {
-                imgSrc = imageElement.getAttribute(attr);
-                if (!imgSrc.isEmpty())
-                    break;
-            }
-            if (!imgSrc.isEmpty()) {
-                // We cannot trust the dimension if the image is not loaded yet.
-                // In some cases there are 1x1 placeholder images.
-                width = 0;
-                height = 0;
-            } else {
-                imgSrc = imageElement.getSrc();
-                // As an ImageElement is manipulated here, it is possible
-                // to get the real dimensions.
-                width = imageElement.getWidth();
-                height = imageElement.getHeight();
+            extractImageAttributes(ImageElement.as(e));
+            return new WebImage(e, width, height, imgSrc);
+        } else if ("FIGURE".equals(e.getTagName())) {
+            Element img = getFirstElementByTagName(e, "IMG");
+            if (img != null) {
+                String caption;
+                extractImageAttributes(ImageElement.as(img));
+                Element cap = getFirstElementByTagName(e, "FIGCAPTION");
+                if (cap != null) {
+                    caption = getFirstElementByTagName(cap, "A") != null ?
+                            cap.getInnerHTML() : cap.getInnerText();

[wychen, 2016/06/07 17:39:19]

This innerText should be escaped. Or a vulnerability like XSS could be
introduced here. Be very careful about this.

Getting innerHTML doesn't look right. For example, links don't get converted to
absolute urls. You can check how WebTable handles this.

[marcelorcorrea, 2016/06/08 14:34:15]

Done.

[wychen, 2016/06/20 18:54:20]

WebTable stores an element, and use DomUtil.generateOutputFromTree() when
generating its output. I think that looks more robust than doing
innerHTML/innerText with manual escaping.

+                } else {
+                    caption = e.getInnerText();
+                }
+                return new WebFigure(img, width, height, imgSrc, caption);
             }
         }
+        return null;
+    }
 
+    private void extractImageAttributes(ImageElement imageElement) {
+        // This will get the absolute URL of the image and
+        // the displayed image dimension.
+        // Try to get lazily-loaded images before falling back to get the src attribute.
+        for (String attr : LAZY_IMAGE_ATTRIBUTES) {
+            imgSrc = imageElement.getAttribute(attr);
+            if (!imgSrc.isEmpty())
+                break;
+        }
+        if (!imgSrc.isEmpty()) {
+            // We cannot trust the dimension if the image is not loaded yet.
+            // In some cases there are 1x1 placeholder images.
+            width = 0;
+            height = 0;
+        } else {
+            imgSrc = imageElement.getSrc();
+            // As an ImageElement is manipulated here, it is possible
+            // to get the real dimensions.
+            width = imageElement.getWidth();
+            height = imageElement.getHeight();
+        }
         if (LogUtil.isLoggable(LogUtil.DEBUG_LEVEL_VISIBILITY_INFO)) {
             LogUtil.logToConsole("Extracted WebImage: " + imgSrc);
         }
-        return new WebImage(e, width, height, imgSrc);
+    }
+
+    private Element getFirstElementByTagName(Element e, String tagName) {
+        NodeList<Element> elements = e.getElementsByTagName(tagName);
+        if (elements.getLength() > 0) {
+            return elements.getItem(0);
+        }
+        return null;
     }
 }