[ http://nagoya.apache.org/jira/browse/XALANJ-667?page=comments#action_56802 ] Joe Kesselman commented on XALANJ-667: --------------------------------------
See past discussion of this issue. There is *no* portable way to test Level 1 DOM nodes for identity. However, most DOM1 implementations do use a single object per node, and thus == works as an identity test for those. .equals(), semanticaly, ought to compare the *value* of two objects, not their identity. The concept of node value is not as clear as it should be -- note that an Element node officially has no node value -- but if it worked at all, node1.equals(node2) ought to mean the two subtrees were equivalent, not identitical -- so if anything it's inferior to == for our purposes. The *right* answer, as discussed in the past, is to test the DOM in question to see if it implements the DOM Level 3 isSameNode() method. If so, we should use this method to test node identity. If not, see above; there is no good portable solution and we either need to find some way of recognizing and handling specific implementations or fall back on == as a last resort. (Note that isSameNode() is *different* from isEqualNode(), also introduced in DOM3; that difference tracks the distinction I made above about identity versus equality.) Yes, it's ugly. DOM1, and DOM2, should never have gone out the door without the isSameNode() test. All I can offer in our defense is that we honestly didn't realize this was an issue until someone (myself, in fact) started creating a DOM where node identity was *not* isomorphic with object identity. That's the downside of developing in web-years; we all get to use the stuff sooner, but sometimes that's before it's really finished, and we have to deal with the fact that early versions may have gaps. Oh Well. -- Joe Kesselman, DOM Working Group Alumnus > using object identity as a substitute for isSameNode is bad idea > ---------------------------------------------------------------- > > Key: XALANJ-667 > URL: http://nagoya.apache.org/jira/browse/XALANJ-667 > Project: XalanJ2 > Type: Improvement > Components: DTM > Versions: 2.0.0 > Environment: Operating System: Other > Platform: Other > Reporter: Andrei Tchijov > Assignee: Joe Kesselman > > In org.apache.xml.dtm.ref.dom2dtm.DOM2DTM.getHandleOfNode "==" used to > test for two nodes been the same. In the comment just before this method > definition it is stated that this is "flaky" solution, and I can not agree > more. > The question is, what is wrong with using "equals" instead of "=="? I do > realize that "equals" is not part of the DOM, but same is true for "==". > Andrei Tchijov > PS These are snipets of code I am talking about > public int getHandleOfNode(Node node) > { > if (null != node) > { > // Is Node actually within the same document? If not, don't search! > // This would be easier if m_root was always the Document node, but > // we decided to allow wrapping a DTM around a subtree. > /* > * hyperNOC fix. > * used to be > if((m_root==node) || > (m_root.getNodeType()==DOCUMENT_NODE && > m_root==node.getOwnerDocument()) || > (m_root.getNodeType()!=DOCUMENT_NODE && > m_root.getOwnerDocument()==node.getOwnerDocument()) > ) > */ > if((m_root==node) || > (m_root.getNodeType()==DOCUMENT_NODE && > m_root.equals( node.getOwnerDocument())) || > (m_root.getNodeType()!=DOCUMENT_NODE && > m_root.getOwnerDocument().equals( node.getOwnerDocument())) > ) > { > // If node _is_ in m_root's tree, find its handle > // > // %OPT% This check may be improved significantly when DOM > // Level 3 nodeKey and relative-order tests become > // available! > for(Node cursor=node; > cursor!=null; > cursor= > (cursor.getNodeType()!=ATTRIBUTE_NODE) > ? cursor.getParentNode() > : ((org.w3c.dom.Attr)cursor).getOwnerElement()) > { > /* > * hyperNOC fix. > * used to be > if(cursor==m_root) > */ > if(cursor.equals( m_root )) > // We know this node; find its handle. > return getHandleFromNode(node); > } // for ancestors of node > } // if node and m_root in same Document > } // if node!=null > return DTM.NULL; > } > Same is true for getHandleFromNode > private int getHandleFromNode(Node node) > { > if (null != node) > { > int len = m_nodes.size(); > boolean isMore; > int i = 0; > do > { > for (; i < len; i++) > { > /* > * hyperNOC fix > * used to be > if (m_nodes.elementAt(i) == node) > */ > if (m_nodes.elementAt(i).equals( node )) > return i | m_dtmIdent; > } > isMore = nextNode(); > > len = m_nodes.size(); > > } > while(isMore || i < len); > } > > return DTM.NULL; > } -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://nagoya.apache.org/jira/secure/Administrators.jspa - If you want more information on JIRA, or have a bug to report see: http://www.atlassian.com/software/jira --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
