dom2dtm.java

来自「Mobile 应用程序使用 Java Micro Edition (Java M」· Java 代码 · 共 1,764 行 · 第 1/4 页

/*
 * Copyright 1999-2004 The Apache Software Foundation.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
/*
 * $Id: DOM2DTM.java,v 1.2.4.1 2005/09/15 08:15:10 suresh_emailid Exp $
 */
package com.sun.org.apache.xml.internal.dtm.ref.dom2dtm;

import java.util.Vector;

import javax.xml.transform.SourceLocator;
import javax.xml.transform.dom.DOMSource;

import com.sun.org.apache.xml.internal.dtm.DTM;
import com.sun.org.apache.xml.internal.dtm.DTMManager;
import com.sun.org.apache.xml.internal.dtm.DTMWSFilter;
import com.sun.org.apache.xml.internal.dtm.ref.DTMDefaultBaseIterators;
import com.sun.org.apache.xml.internal.dtm.ref.DTMManagerDefault;
import com.sun.org.apache.xml.internal.dtm.ref.ExpandedNameTable;
import com.sun.org.apache.xml.internal.dtm.ref.IncrementalSAXSource;
import com.sun.org.apache.xml.internal.res.XMLErrorResources;
import com.sun.org.apache.xml.internal.res.XMLMessages;
import com.sun.org.apache.xml.internal.utils.FastStringBuffer;
import com.sun.org.apache.xml.internal.utils.QName;
import com.sun.org.apache.xml.internal.utils.StringBufferPool;
import com.sun.org.apache.xml.internal.utils.TreeWalker;
import com.sun.org.apache.xml.internal.utils.XMLCharacterRecognizer;
import com.sun.org.apache.xml.internal.utils.XMLString;
import com.sun.org.apache.xml.internal.utils.XMLStringFactory;
import org.w3c.dom.Attr;
import org.w3c.dom.Document;
import org.w3c.dom.DocumentType;
import org.w3c.dom.Element;
import org.w3c.dom.Entity;
import org.w3c.dom.NamedNodeMap;
import org.w3c.dom.Node;
import org.xml.sax.ContentHandler;

/** The <code>DOM2DTM</code> class serves up a DOM's contents via the
 * DTM API.
 *
 * Note that it doesn't necessarily represent a full Document
 * tree. You can wrap a DOM2DTM around a specific node and its subtree
 * and the right things should happen. (I don't _think_ we currently
 * support DocumentFrgment nodes as roots, though that might be worth
 * considering.)
 *
 * Note too that we do not currently attempt to track document
 * mutation. If you alter the DOM after wrapping DOM2DTM around it,
 * all bets are off.
 * */
public class DOM2DTM extends DTMDefaultBaseIterators
{
  static final boolean JJK_DEBUG=false;
  static final boolean JJK_NEWCODE=true;
  
  /** Manefest constant
   */
  static final String NAMESPACE_DECL_NS="http://www.w3.org/XML/1998/namespace";
  
  /** The current position in the DOM tree. Last node examined for
   * possible copying to DTM. */
  transient private Node m_pos;
  /** The current position in the DTM tree. Who children get appended to. */
  private int m_last_parent=0;
  /** The current position in the DTM tree. Who children reference as their 
   * previous sib. */
  private int m_last_kid=NULL;

  /** The top of the subtree.
   * %REVIEW%: 'may not be the same as m_context if "//foo" pattern.'
   * */
  transient private Node m_root;

  /** True iff the first element has been processed. This is used to control
      synthesis of the implied xml: namespace declaration node. */
  boolean m_processedFirstElement=false;
        
  /** true if ALL the nodes in the m_root subtree have been processed;
   * false if our incremental build has not yet finished scanning the
   * DOM tree.  */
  transient private boolean m_nodesAreProcessed;

  /** The node objects.  The instance part of the handle indexes
   * directly into this vector.  Each DTM node may actually be
   * composed of several DOM nodes (for example, if logically-adjacent
   * Text/CDATASection nodes in the DOM have been coalesced into a
   * single DTM Text node); this table points only to the first in
   * that sequence. */
  protected Vector m_nodes = new Vector();

  /**
   * Construct a DOM2DTM object from a DOM node.
   *
   * @param mgr The DTMManager who owns this DTM.
   * @param domSource the DOM source that this DTM will wrap.
   * @param dtmIdentity The DTM identity ID for this DTM.
   * @param whiteSpaceFilter The white space filter for this DTM, which may 
   *                         be null.
   * @param xstringfactory XMLString factory for creating character content.
   * @param doIndexing true if the caller considers it worth it to use 
   *                   indexing schemes.
   */
  public DOM2DTM(DTMManager mgr, DOMSource domSource, 
                 int dtmIdentity, DTMWSFilter whiteSpaceFilter,
                 XMLStringFactory xstringfactory,
                 boolean doIndexing)
  {
    super(mgr, domSource, dtmIdentity, whiteSpaceFilter, 
          xstringfactory, doIndexing);

    // Initialize DOM navigation
    m_pos=m_root = domSource.getNode();
    // Initialize DTM navigation
    m_last_parent=m_last_kid=NULL;
    m_last_kid=addNode(m_root, m_last_parent,m_last_kid, NULL);

    // Apparently the domSource root may not actually be the
    // Document node. If it's an Element node, we need to immediately
    // add its attributes. Adapted from nextNode().
    // %REVIEW% Move this logic into addNode and recurse? Cleaner!
    //
    // (If it's an EntityReference node, we're probably scrod. For now
    // I'm just hoping nobody is ever quite that foolish... %REVIEW%)
		//
		// %ISSUE% What about inherited namespaces in this case?
		// Do we need to special-case initialize them into the DTM model?
    if(ELEMENT_NODE == m_root.getNodeType())
    {
      NamedNodeMap attrs=m_root.getAttributes();
      int attrsize=(attrs==null) ? 0 : attrs.getLength();
      if(attrsize>0)
      {
        int attrIndex=NULL; // start with no previous sib
        for(int i=0;i<attrsize;++i)
        {
          // No need to force nodetype in this case;
          // addNode() will take care of switching it from
          // Attr to Namespace if necessary.
          attrIndex=addNode(attrs.item(i),0,attrIndex,NULL);
          m_firstch.setElementAt(DTM.NULL,attrIndex);
        }
        // Terminate list of attrs, and make sure they aren't
        // considered children of the element
        m_nextsib.setElementAt(DTM.NULL,attrIndex);

        // IMPORTANT: This does NOT change m_last_parent or m_last_kid!
      } // if attrs exist
    } //if(ELEMENT_NODE)

    // Initialize DTM-completed status 
    m_nodesAreProcessed = false;
  }

  /**
   * Construct the node map from the node.
   *
   * @param node The node that is to be added to the DTM.
   * @param parentIndex The current parent index.
   * @param previousSibling The previous sibling index.
   * @param forceNodeType If not DTM.NULL, overrides the DOM node type.
   *	Used to force nodes to Text rather than CDATASection when their
   *	coalesced value includes ordinary Text nodes (current DTM behavior).
   *
   * @return The index identity of the node that was added.
   */
  protected int addNode(Node node, int parentIndex,
                        int previousSibling, int forceNodeType)
  {
    int nodeIndex = m_nodes.size();

    // Have we overflowed a DTM Identity's addressing range?
    if(m_dtmIdent.size() == (nodeIndex>>>DTMManager.IDENT_DTM_NODE_BITS))
    {
      try
      {
        if(m_mgr==null)
          throw new ClassCastException();
                                
                                // Handle as Extended Addressing
        DTMManagerDefault mgrD=(DTMManagerDefault)m_mgr;
        int id=mgrD.getFirstFreeDTMID();
        mgrD.addDTM(this,id,nodeIndex);
        m_dtmIdent.addElement(id<<DTMManager.IDENT_DTM_NODE_BITS);
      }
      catch(ClassCastException e)
      {
        // %REVIEW% Wrong error message, but I've been told we're trying
        // not to add messages right not for I18N reasons.
        // %REVIEW% Should this be a Fatal Error?
        error(XMLMessages.createXMLMessage(XMLErrorResources.ER_NO_DTMIDS_AVAIL, null));//"No more DTM IDs are available";
      }
    }

    m_size++;
    // ensureSize(nodeIndex);
    
    int type;
    if(NULL==forceNodeType)
        type = node.getNodeType();
    else
        type=forceNodeType;
        
    // %REVIEW% The Namespace Spec currently says that Namespaces are
    // processed in a non-namespace-aware manner, by matching the
    // QName, even though there is in fact a namespace assigned to
    // these nodes in the DOM. If and when that changes, we will have
    // to consider whether we check the namespace-for-namespaces
    // rather than the node name.
    //
    // %TBD% Note that the DOM does not necessarily explicitly declare
    // all the namespaces it uses. DOM Level 3 will introduce a
    // namespace-normalization operation which reconciles that, and we
    // can request that users invoke it or otherwise ensure that the
    // tree is namespace-well-formed before passing the DOM to Xalan.
    // But if they don't, what should we do about it? We probably
    // don't want to alter the source DOM (and may not be able to do
    // so if it's read-only). The best available answer might be to
    // synthesize additional DTM Namespace Nodes that don't correspond
    // to DOM Attr Nodes.
    if (Node.ATTRIBUTE_NODE == type)
    {
      String name = node.getNodeName();

      if (name.startsWith("xmlns:") || name.equals("xmlns"))
      {
        type = DTM.NAMESPACE_NODE;
      }
    }
    
    m_nodes.addElement(node);
    
    m_firstch.setElementAt(NOTPROCESSED,nodeIndex);
    m_nextsib.setElementAt(NOTPROCESSED,nodeIndex);
    m_prevsib.setElementAt(previousSibling,nodeIndex);
    m_parent.setElementAt(parentIndex,nodeIndex);
    
    if(DTM.NULL != parentIndex && 
       type != DTM.ATTRIBUTE_NODE && 
       type != DTM.NAMESPACE_NODE)
    {
      // If the DTM parent had no children, this becomes its first child.
      if(NOTPROCESSED == m_firstch.elementAt(parentIndex))
        m_firstch.setElementAt(nodeIndex,parentIndex);
    }
    
    String nsURI = node.getNamespaceURI();

    // Deal with the difference between Namespace spec and XSLT
    // definitions of local name. (The former says PIs don't have
    // localnames; the latter says they do.)
    String localName =  (type == Node.PROCESSING_INSTRUCTION_NODE) ? 
                         node.getNodeName() :
                         node.getLocalName();
                         
    // Hack to make DOM1 sort of work...
    if(((type == Node.ELEMENT_NODE) || (type == Node.ATTRIBUTE_NODE)) 
        && null == localName)
      localName = node.getNodeName(); // -sb
      
    ExpandedNameTable exnt = m_expandedNameTable;

    // %TBD% Nodes created with the old non-namespace-aware DOM
    // calls createElement() and createAttribute() will never have a
    // localname. That will cause their expandedNameID to be just the
    // nodeType... which will keep them from being matched
    // successfully by name. Since the DOM makes no promise that
    // those will participate in namespace processing, this is
    // officially accepted as Not Our Fault. But it might be nice to
    // issue a diagnostic message!
    if(node.getLocalName()==null &&
       (type==Node.ELEMENT_NODE || type==Node.ATTRIBUTE_NODE))
      {
        // warning("DOM 'level 1' node "+node.getNodeName()+" won't be mapped properly in DOM2DTM.");
      }
    
    int expandedNameID = (null != localName) 
       ? exnt.getExpandedTypeID(nsURI, localName, type) :
         exnt.getExpandedTypeID(type);

    m_exptype.setElementAt(expandedNameID,nodeIndex);
    
    indexNode(expandedNameID, nodeIndex);

    if (DTM.NULL != previousSibling)
      m_nextsib.setElementAt(nodeIndex,previousSibling);

    // This should be done after m_exptype has been set, and probably should
    // always be the last thing we do
    if (type == DTM.NAMESPACE_NODE)
        declareNamespaceInContext(parentIndex,nodeIndex);

    return nodeIndex;
  }
  
  /**
   * Get the number of nodes that have been added.
   */
  public int getNumberOfNodes()
  {
    return m_nodes.size();
  }
  
 /**
   * This method iterates to the next node that will be added to the table.
   * Each call to this method adds a new node to the table, unless the end
   * is reached, in which case it returns null.
   *
   * @return The true if a next node is found or false if 
   *         there are no more nodes.
   */
  protected boolean nextNode()
  {
    // Non-recursive one-fetch-at-a-time depth-first traversal with 
    // attribute/namespace nodes and white-space stripping.
    // Navigating the DOM is simple, navigating the DTM is simple;
    // keeping track of both at once is a trifle baroque but at least
    // we've avoided most of the special cases.
    if (m_nodesAreProcessed)
      return false;
        
    // %REVIEW% Is this local copy Really Useful from a performance
    // point of view?  Or is this a false microoptimization?
    Node pos=m_pos; 
    Node next=null;
    int nexttype=NULL;

    // Navigate DOM tree
    do
      {
        // Look down to first child.
        if (pos.hasChildNodes()) 
          {
            next = pos.getFirstChild();

            // %REVIEW% There's probably a more elegant way to skip
            // the doctype. (Just let it go and Suppress it?
            if(next!=null && DOCUMENT_TYPE_NODE==next.getNodeType())
              next=next.getNextSibling();

            // Push DTM context -- except for children of Entity References, 
            // which have no DTM equivalent and cause no DTM navigation.
            if(ENTITY_REFERENCE_NODE!=pos.getNodeType())
              {
                m_last_parent=m_last_kid;
                m_last_kid=NULL;
                // Whitespace-handler context stacking
                if(null != m_wsfilter)
                {
                  short wsv =
                    m_wsfilter.getShouldStripSpace(makeNodeHandle(m_last_parent),this);
                  boolean shouldStrip = (DTMWSFilter.INHERIT == wsv) 
                    ? getShouldStripWhitespace() 
                    : (DTMWSFilter.STRIP == wsv);
                  pushShouldStripWhitespace(shouldStrip);
                } // if(m_wsfilter)
              }
          }

        // If that fails, look up and right (but not past root!)
        else 
          {
            if(m_last_kid!=NULL)
              {
                // Last node posted at this level had no more children
                // If it has _no_ children, we need to record that.
                if(m_firstch.elementAt(m_last_kid)==NOTPROCESSED)
                  m_firstch.setElementAt(NULL,m_last_kid);
              }
                        
            while(m_last_parent != NULL)
              {
                // %REVIEW% There's probably a more elegant way to
                // skip the doctype. (Just let it go and Suppress it?
                next = pos.getNextSibling();
                if(next!=null && DOCUMENT_TYPE_NODE==next.getNodeType())
                  next=next.getNextSibling();

                if(next!=null)
                  break; // Found it!
                
                // No next-sibling found. Pop the DOM.
                pos=pos.getParentNode();
                if(pos==null)
                  {
                    // %TBD% Should never arise, but I want to be sure of that...
                    if(JJK_DEBUG)
                      {
                        System.out.println("***** DOM2DTM Pop Control Flow problem");
                        for(;;); // Freeze right here!
                      }
                  }
                
                // The only parents in the DTM are Elements.  However,
                // the DOM could contain EntityReferences.  If we
                // encounter one, pop it _without_ popping DTM.
                if(pos!=null && ENTITY_REFERENCE_NODE == pos.getNodeType())
                  {
                    // Nothing needs doing
                    if(JJK_DEBUG)
                      System.out.println("***** DOM2DTM popping EntRef");
                  }
                else
                  {
                    popShouldStripWhitespace();
                    // Fix and pop DTM
                    if(m_last_kid==NULL)
                      m_firstch.setElementAt(NULL,m_last_parent); // Popping from an element
                    else
                      m_nextsib.setElementAt(NULL,m_last_kid); // Popping from anything else
                    m_last_parent=m_parent.elementAt(m_last_kid=m_last_parent);
                  }
              }
            if(m_last_parent==NULL)
              next=null;
          }
                
        if(next!=null)
          nexttype=next.getNodeType();
                
        // If it's an entity ref, advance past it.
        //
        // %REVIEW% Should we let this out the door and just suppress it?
        // More work, but simpler code, more likely to be correct, and
        // it doesn't happen very often. We'd get rid of the loop too.
        if (ENTITY_REFERENCE_NODE == nexttype)
          pos=next;
      }
    while (ENTITY_REFERENCE_NODE == nexttype);