HTML过滤和补齐（二）

private static StringBuffer getHTML(Node node, boolean check) {
   short type = node.getNodeType();
   String name = StringUtil.defaultIfBlank(node.getNodeName());
   String value2 = node.getNodeValue();
   NamedNodeMap attrs = node.getAttributes();

   if (check) {
    try { // 如果是表格并且没有在表格里,忽略

     if ((type == Node.ELEMENT_NODE)
       && "tr".equalsIgnoreCase(name)
       && !"table".equalsIgnoreCase(node.getParentNode()
         .getNodeName())) {
      type = Node.TEXT_NODE;
     } else if ((type == Node.ELEMENT_NODE)
       && "td".equalsIgnoreCase(name)
       && (!"tr".equalsIgnoreCase(node.getParentNode()
         .getNodeName()) || !"table"
         .equalsIgnoreCase(node.getParentNode()
           .getParentNode().getNodeName()))) {
      type = Node.TEXT_NODE;
     } else if ((type == Node.ELEMENT_NODE)
       && "a".equalsIgnoreCase(name)) { // 如果是链接，则是能是taobao内部链接.

      if (attrs != null) {
       Node hrefNode = attrs.getNamedItem("href");

       if (hrefNode != null) {
        String href = hrefNode.getNodeValue();

        if ((escapeSpecialHTMLPattern == null)
          || !(new Perl5Matcher()).matches(href,
            escapeSpecialHTMLPattern)) {
         type = Node.TEXT_NODE;
        }
       }
      }
     }
    } catch (Exception e) {
     type = Node.TEXT_NODE;
    }
   }

   StringBuffer sb = new StringBuffer();
   boolean hasVal = false;

   if (type == Node.ELEMENT_NODE) {
    sb.append("<");
    sb.append(StringUtil.toLowerCase(name));

    if (attrs != null) {
     for (int i = 0; i < attrs.getLength(); i++) {
      Node n = attrs.item(i);

      if ((n.getNodeName() != null) && (n.getNodeValue() != null)) {
       sb.append(" ");
       sb.append(StringUtil.toLowerCase(n.getNodeName()));
       sb.append("=\"");
       sb.append(StringEscapeUtil.escapeHtml(n.getNodeValue()
         .trim()));
       sb.append("\"");
      }
     }
    }

    if (StringUtil.isNotBlank(value2)) {
     hasVal = true;
     sb.append(">");
     sb.append(escapeHTML(value2.trim()));
     sb.append("");
    }
   } else if ((type == Node.TEXT_NODE) && StringUtil.isNotBlank(value2)) {
    sb.append(escapeHTML(value2.trim()));

    // sb.append("\n");
    return sb;
   }

   Node child = node.getFirstChild();

   if ((type == Node.ELEMENT_NODE) && (child == null)) {
    if (!hasVal
      && INLINE_CLOSED_TAG.contains(StringUtil.toLowerCase(name))) {
     sb.append(" />");
    } else if (!hasVal) {
     sb.append("></");
     sb.append(StringUtil.toLowerCase(name));
     sb.append(">");
    }
   } else {
    if (!hasVal && (type == Node.ELEMENT_NODE)) {
     sb.append(">");
    }

    while (child != null) {
     sb.append(getHTML(child, check));
     child = child.getNextSibling();
    }

    if (type == Node.ELEMENT_NODE) {
     sb.append("</");
     sb.append(StringUtil.toLowerCase(name));
     sb.append(">");
    }
   }

   return sb;
}

/**
* for display convenience we must cut the length of title if the lenght of
* title greater than ITEM_TITLE_MAX_SIZE, it will return the
* ITEM_TITLE_MAX_SIZE characters of the origianl title, otherwise, the
* whole title.
*
* @param t -
*            title
* @param maxSize -
*            title max size
*
* @return - title returned
*/
public static String getFormattedTitle(String t, int maxSize) {
   StringBuffer title = new StringBuffer();

   if (StringUtil.isBlank(t)) {
    return "";
   }

   if (t.length() > maxSize) {
    // title = t.trim().substring(0,ITEM_TITLE_MAX_SIZE)+"...";
    double weight = 0;
    int i = 0;
    char c = ' ';

    do {
     try {
      c = t.charAt(i);
     } catch (IndexOutOfBoundsException e) {
      break;
     }

     title.append(c);

     if (Character.isLetterOrDigit(c)) {
      if (Character.getType(c) == Character.OTHER_LETTER) {
       weight++; // chinese char weight 1
      } else {
       weight += 0.5; // english char only weight 0.5
      }
     } else if (Character.isWhitespace(c)) {
      weight += 0.5; // white space char weight 0.5
     } else {
      weight++; // chinese char weight 1
     }

     i++;
    } while (weight < (maxSize));

    if (weight >= maxSize) {
     try {
      i++;
      c = t.charAt(i);
      title.append("..."); // still has character, append ...
            // to ignore the rest
     } catch (IndexOutOfBoundsException e) {
      // do nothing;
     }
    }
   } else {
    title.append(t.trim());
   }

   return title.toString();
}