Use stricter comment-detecting heuristics

java/org/chromium/distiller/webdocument/ElementAction.java

Index: java/org/chromium/distiller/webdocument/ElementAction.java
diff --git a/java/org/chromium/distiller/webdocument/ElementAction.java b/java/org/chromium/distiller/webdocument/ElementAction.java
index 0614179790a180d3b02424be9b90a00ab3499d75..589ed5c79bdc48e8613d5b07c70ce5e34a04f537 100644
--- a/java/org/chromium/distiller/webdocument/ElementAction.java
+++ b/java/org/chromium/distiller/webdocument/ElementAction.java
@@ -20,7 +20,7 @@ public class ElementAction {
     public JsArrayString labels = JavaScriptObject.createArray().<JsArrayString>cast();
 
     private static final RegExp REG_COMMENT = RegExp.compile("\\bcomments?\\b");
-    private static final int MAX_CLASS_COUNT = 5;
+    private static final int MAX_CLASS_COUNT = 2;

[mdjones, 2017/01/04 14:29:21]

nit: the way this is used seems to be as a minimum number, can we change this to
MIN_CLASS_COUNT?

[wychen, 2017/01/04 14:37:51]

What do you mean?
This constant is the largest allowed number of class names for this heuristic to
be enabled, so sounds like max more than min, right?

[mdjones, 2017/01/04 15:14:33]

Ah, I was looking at it from: "if class count doesn't meet this number, it's not
content".

 
     public static ElementAction getForElement(Element element) {
         Style style = DomUtil.getComputedStyle(element);
@@ -62,7 +62,7 @@ public class ElementAction {
                 break;
         }
 
-        if (!"HTML".equals(tagName) && !"BODY".equals(tagName)) {
+        if (!"HTML".equals(tagName) && !"BODY".equals(tagName) && !"ARTICLE".equals(tagName)) {
             String className = element.getAttribute("class");
             int classCount = DomUtil.getClassList(element).length();
             String id = element.getAttribute("id");