SGXMLScanner.cpp

/*
 * Licensed to the Apache Software Foundation (ASF) under one or more
 * contributor license agreements.  See the NOTICE file distributed with
 * this work for additional information regarding copyright ownership.
 * The ASF licenses this file to You 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.
 */

/*
 * $Id$
 */


// ---------------------------------------------------------------------------
//  Includes
// ---------------------------------------------------------------------------
#include <xercesc/internal/SGXMLScanner.hpp>
#include <xercesc/util/RuntimeException.hpp>
#include <xercesc/util/UnexpectedEOFException.hpp>
#include <xercesc/util/XMLUri.hpp>
#include <xercesc/framework/LocalFileInputSource.hpp>
#include <xercesc/framework/URLInputSource.hpp>
#include <xercesc/framework/XMLDocumentHandler.hpp>
#include <xercesc/framework/XMLEntityHandler.hpp>
#include <xercesc/framework/XMLPScanToken.hpp>
#include <xercesc/framework/MemoryManager.hpp>
#include <xercesc/framework/XMLGrammarPool.hpp>
#include <xercesc/framework/psvi/PSVIElement.hpp>
#include <xercesc/framework/psvi/PSVIHandler.hpp>
#include <xercesc/framework/psvi/PSVIAttributeList.hpp>
#include <xercesc/framework/psvi/XSAnnotation.hpp>
#include <xercesc/internal/EndOfEntityException.hpp>
#include <xercesc/validators/common/ContentLeafNameTypeVector.hpp>
#include <xercesc/validators/schema/SchemaValidator.hpp>
#include <xercesc/validators/schema/TraverseSchema.hpp>
#include <xercesc/validators/schema/XSDDOMParser.hpp>
#include <xercesc/validators/schema/SubstitutionGroupComparator.hpp>
#include <xercesc/validators/schema/XMLSchemaDescriptionImpl.hpp>
#include <xercesc/validators/schema/identity/IdentityConstraintHandler.hpp>
#include <xercesc/validators/schema/identity/IC_Selector.hpp>
#include <xercesc/validators/schema/identity/ValueStore.hpp>
#include <xercesc/util/OutOfMemoryException.hpp>
#include <xercesc/util/HashPtr.hpp>
#include <xercesc/util/XMLStringTokenizer.hpp>

XERCES_CPP_NAMESPACE_BEGIN

inline XMLAttDefList& getAttDefList(ComplexTypeInfo* currType, XMLElementDecl* elemDecl);


typedef JanitorMemFunCall<SGXMLScanner> CleanupType;
typedef JanitorMemFunCall<ReaderMgr>    ReaderMgrResetType;


// ---------------------------------------------------------------------------
//  SGXMLScanner: Constructors and Destructor
// ---------------------------------------------------------------------------
SGXMLScanner::SGXMLScanner( XMLValidator* const valToAdopt
                          , GrammarResolver* const grammarResolver
                          , MemoryManager* const manager) :

    XMLScanner(valToAdopt, grammarResolver, manager)
    , fSeeXsi(false)
    , fGrammarType(Grammar::UnKnown)
    , fElemStateSize(16)
    , fElemState(0)
    , fContent(1023, manager)
    , fEntityTable(0)
    , fRawAttrList(0)
    , fRawAttrColonListSize(32)
    , fRawAttrColonList(0)
    , fSchemaGrammar(0)
    , fSchemaValidator(0)
    , fICHandler(0)
    , fElemNonDeclPool(0)
    , fElemCount(0)
    , fAttDefRegistry(0)
    , fUndeclaredAttrRegistryNS(0)
    , fPSVIAttrList(0)
    , fModel(0)
    , fPSVIElement(0)
    , fErrorStack(0)
{
    CleanupType cleanup(this, &SGXMLScanner::cleanUp);

    try
    {
         commonInit();
    }
    catch(const OutOfMemoryException&)
    {
        // Don't cleanup when out of memory, since executing the
        // code can cause problems.
        cleanup.release();

        throw;
    }

    cleanup.release();
}

SGXMLScanner::SGXMLScanner( XMLDocumentHandler* const docHandler
                          , DocTypeHandler* const     docTypeHandler
                          , XMLEntityHandler* const   entityHandler
                          , XMLErrorReporter* const   errHandler
                          , XMLValidator* const       valToAdopt
                          , GrammarResolver* const    grammarResolver
                          , MemoryManager* const      manager) :

    XMLScanner(docHandler, docTypeHandler, entityHandler, errHandler, valToAdopt, grammarResolver, manager)
    , fSeeXsi(false)
    , fGrammarType(Grammar::UnKnown)
    , fElemStateSize(16)
    , fElemState(0)
    , fContent(1023, manager)
    , fEntityTable(0)
    , fRawAttrList(0)
    , fRawAttrColonListSize(32)
    , fRawAttrColonList(0)
    , fSchemaGrammar(0)
    , fSchemaValidator(0)
    , fICHandler(0)
    , fElemNonDeclPool(0)
    , fElemCount(0)
    , fAttDefRegistry(0)
    , fUndeclaredAttrRegistryNS(0)
    , fPSVIAttrList(0)
    , fModel(0)
    , fPSVIElement(0)
    , fErrorStack(0)
{
    CleanupType cleanup(this, &SGXMLScanner::cleanUp);

    try
    {
        commonInit();
    }
    catch(const OutOfMemoryException&)
    {
        // Don't cleanup when out of memory, since executing the
        // code can cause problems.
        cleanup.release();

        throw;
    }

    cleanup.release();
}

SGXMLScanner::~SGXMLScanner()
{
    cleanUp();
}

// ---------------------------------------------------------------------------
//  XMLScanner: Getter methods
// ---------------------------------------------------------------------------
NameIdPool<DTDEntityDecl>* SGXMLScanner::getEntityDeclPool()
{
    return 0;
}

const NameIdPool<DTDEntityDecl>* SGXMLScanner::getEntityDeclPool() const
{
    return 0;
}

// ---------------------------------------------------------------------------
//  SGXMLScanner: Main entry point to scan a document
// ---------------------------------------------------------------------------
void SGXMLScanner::scanDocument(const InputSource& src)
{
    //  Bump up the sequence id for this parser instance. This will invalidate
    //  any previous progressive scan tokens.
    fSequenceId++;

    ReaderMgrResetType  resetReaderMgr(&fReaderMgr, &ReaderMgr::reset);

    try
    {
        //  Reset the scanner and its plugged in stuff for a new run. This
        //  resets all the data structures, creates the initial reader and
        //  pushes it on the stack, and sets up the base document path.
        scanReset(src);

        // If we have a document handler, then call the start document
        if (fDocHandler)
            fDocHandler->startDocument();

        //  Scan the prolog part, which is everything before the root element
        //  including the DTD subsets.
        scanProlog();

        //  If we got to the end of input, then its not a valid XML file.
        //  Else, go on to scan the content.
        if (fReaderMgr.atEOF())
        {
            emitError(XMLErrs::EmptyMainEntity);
        }
        else
        {
            // Scan content, and tell it its not an external entity
            if (scanContent())
            {
                // Do post-parse validation if required
                if (fValidate)
                {
                    //  We handle ID reference semantics at this level since
                    //  its required by XML 1.0.
                    checkIDRefs();

                    // Then allow the validator to do any extra stuff it wants
//                    fValidator->postParseValidation();
                }

                // That went ok, so scan for any miscellaneous stuff
                if (!fReaderMgr.atEOF())
                    scanMiscellaneous();
            }
        }

        // If we have a document handler, then call the end document
        if (fDocHandler)
            fDocHandler->endDocument();
    }
    //  NOTE:
    //
    //  In all of the error processing below, the emitError() call MUST come
    //  before the flush of the reader mgr, or it will fail because it tries
    //  to find out the position in the XML source of the error.
    catch(const XMLErrs::Codes)
    {
        // This is a 'first failure' exception, so fall through
    }
    catch(const XMLValid::Codes)
    {
        // This is a 'first fatal error' type exit, so fall through
    }
    catch(const XMLException& excToCatch)
    {
        //  Emit the error and catch any user exception thrown from here. Make
        //  sure in all cases we flush the reader manager.
        fInException = true;
        try
        {
            if (excToCatch.getErrorType() == XMLErrorReporter::ErrType_Warning)
                emitError
                (
                    XMLErrs::XMLException_Warning
                    , excToCatch.getCode()
                    , excToCatch.getMessage()
                );
            else if (excToCatch.getErrorType() >= XMLErrorReporter::ErrType_Fatal)
                emitError
                (
                    XMLErrs::XMLException_Fatal
                    , excToCatch.getCode()
                    , excToCatch.getMessage()
                );
            else
                emitError
                (
                    XMLErrs::XMLException_Error
                    , excToCatch.getCode()
                    , excToCatch.getMessage()
                );
        }
        catch(const OutOfMemoryException&)
        {
            // This is a special case for out-of-memory
            // conditions, because resetting the ReaderMgr
            // can be problematic.
            resetReaderMgr.release();

            throw;
        }
    }
    catch(const OutOfMemoryException&)
    {
        // This is a special case for out-of-memory
        // conditions, because resetting the ReaderMgr
        // can be problematic.
        resetReaderMgr.release();

        throw;
    }
}


bool SGXMLScanner::scanNext(XMLPScanToken& token)
{
    // Make sure this token is still legal
    if (!isLegalToken(token))
        ThrowXMLwithMemMgr(RuntimeException, XMLExcepts::Scan_BadPScanToken, fMemoryManager);

    // Find the next token and remember the reader id
    XMLSize_t orgReader;
    XMLTokens curToken;

    ReaderMgrResetType  resetReaderMgr(&fReaderMgr, &ReaderMgr::reset);

    bool retVal = true;

    try
    {
        while (true)
        {
            //  We have to handle any end of entity exceptions that happen here.
            //  We could be at the end of X nested entities, each of which will
            //  generate an end of entity exception as we try to move forward.
            try
            {
                curToken = senseNextToken(orgReader);
                break;
            }
            catch(const EndOfEntityException& toCatch)
            {
                // Send an end of entity reference event
                if (fDocHandler)
                    fDocHandler->endEntityReference(toCatch.getEntity());
            }
        }

        if (curToken == Token_CharData)
        {
            scanCharData(fCDataBuf);
        }
        else if (curToken == Token_EOF)
        {
            if (!fElemStack.isEmpty())
            {
                const ElemStack::StackElem* topElem = fElemStack.popTop();
                emitError
                (
                    XMLErrs::EndedWithTagsOnStack
                    , topElem->fThisElement->getFullName()
                );
            }

            retVal = false;
        }
        else
        {
            // Its some sort of markup
            bool gotData = true;
            switch(curToken)
            {
                case Token_CData :
                    // Make sure we are within content
                    if (fElemStack.isEmpty())
                        emitError(XMLErrs::CDATAOutsideOfContent);
                    scanCDSection();
                    break;

                case Token_Comment :
                    scanComment();
                    break;

                case Token_EndTag :
                    scanEndTag(gotData);
                    break;

                case Token_PI :
                    scanPI();
                    break;

                case Token_StartTag :
                    scanStartTag(gotData);
                    break;

                default :
                    fReaderMgr.skipToChar(chOpenAngle);
                    break;
            }

            if (orgReader != fReaderMgr.getCurrentReaderNum())
                emitError(XMLErrs::PartialMarkupInEntity);

            // If we hit the end, then do the miscellaneous part
            if (!gotData)
            {
                // Do post-parse validation if required
                if (fValidate)
                {
                    //  We handle ID reference semantics at this level since
                    //  its required by XML 1.0.
                    checkIDRefs();

                    // Then allow the validator to do any extra stuff it wants
//                    fValidator->postParseValidation();
                }

                // That went ok, so scan for any miscellaneous stuff
                scanMiscellaneous();

                if (toCheckIdentityConstraint())
                    fICHandler->endDocument();

                if (fDocHandler)
                    fDocHandler->endDocument();
            }
        }
    }
    //  NOTE:
    //
    //  In all of the error processing below, the emitError() call MUST come
    //  before the flush of the reader mgr, or it will fail because it tries
    //  to find out the position in the XML source of the error.
    catch(const XMLErrs::Codes)
    {
        // This is a 'first failure' exception, so return failure
        retVal = false;
    }
    catch(const XMLValid::Codes)
    {
        // This is a 'first fatal error' type exit, so return failure
        retVal = false;
    }
    catch(const XMLException& excToCatch)
    {
        //  Emit the error and catch any user exception thrown from here. Make
        //  sure in all cases we flush the reader manager.
        fInException = true;
        try
        {
            if (excToCatch.getErrorType() == XMLErrorReporter::ErrType_Warning)
                emitError
                (
                    XMLErrs::XMLException_Warning
                    , excToCatch.getCode()
                    , excToCatch.getMessage()
                );
            else if (excToCatch.getErrorType() >= XMLErrorReporter::ErrType_Fatal)
                emitError
                (
                    XMLErrs::XMLException_Fatal
                    , excToCatch.getCode()
                    , excToCatch.getMessage()
                );
            else
                emitError
                (
                    XMLErrs::XMLException_Error
                    , excToCatch.getCode()
                    , excToCatch.getMessage()
                );
        }
        catch(const OutOfMemoryException&)
        {
            // This is a special case for out-of-memory
            // conditions, because resetting the ReaderMgr
            // can be problematic.
            resetReaderMgr.release();

            throw;
        }

        retVal = false;
    }
    catch(const OutOfMemoryException&)
    {
        // This is a special case for out-of-memory
        // conditions, because resetting the ReaderMgr
        // can be problematic.
        resetReaderMgr.release();

        throw;
    }

    // If we are not at the end, release the object that will
    // reset the ReaderMgr.
    if (retVal)
        resetReaderMgr.release();

    return retVal;
}

// ---------------------------------------------------------------------------
//  SGXMLScanner: Private scanning methods
// ---------------------------------------------------------------------------

//  This method is called from scanStartTag() to handle the very raw initial
//  scan of the attributes. It just fills in the passed collection with
//  key/value pairs for each attribute. No processing is done on them at all.
XMLSize_t
SGXMLScanner::rawAttrScan(const   XMLCh* const                elemName
                          ,       RefVectorOf<KVStringPair>&  toFill
                          ,       bool&                       isEmpty)
{
    //  Keep up with how many attributes we've seen so far, and how many
    //  elements are available in the vector. This way we can reuse old
    //  elements until we run out and then expand it.
    XMLSize_t attCount = 0;
    XMLSize_t curVecSize = toFill.size();

    // Assume it is not empty
    isEmpty = false;

    //  We loop until we either see a /> or >, handling key/value pairs util
    //  we get there. We place them in the passed vector, which we will expand
    //  as required to hold them.
    while (true)
    {
        // Get the next character, which should be non-space
        XMLCh nextCh = fReaderMgr.peekNextChar();

        //  If the next character is not a slash or closed angle bracket,
        //  then it must be whitespace, since whitespace is required
        //  between the end of the last attribute and the name of the next
        //  one.
        //
        if (attCount)
        {
            if ((nextCh != chForwardSlash) && (nextCh != chCloseAngle))
            {
                if (fReaderMgr.getCurrentReader()->isWhitespace(nextCh))
                {
                    // Ok, skip by them and get another char
                    fReaderMgr.getNextChar();
                    fReaderMgr.skipPastSpaces();
                    nextCh = fReaderMgr.peekNextChar();
                }
                 else
                {
                    // Emit the error but keep on going
                    emitError(XMLErrs::ExpectedWhitespace);
                }
            }
        }

        //  Ok, here we first check for any of the special case characters.
        //  If its not one, then we do the normal case processing, which
        //  assumes that we've hit an attribute value, Otherwise, we do all
        //  the special case checks.
        if (!fReaderMgr.getCurrentReader()->isSpecialStartTagChar(nextCh))
        {
            //  Assume its going to be an attribute, so get a name from
            //  the input.
            int colonPosition;
            if (!fReaderMgr.getQName(fAttNameBuf, &colonPosition))
            {
                if (fAttNameBuf.isEmpty())
                    emitError(XMLErrs::ExpectedAttrName);
                else
                    emitError(XMLErrs::InvalidAttrName, fAttNameBuf.getRawBuffer());
                fReaderMgr.skipPastChar(chCloseAngle);
                return attCount;
            }

            const XMLCh* curAttNameBuf = fAttNameBuf.getRawBuffer();

            // And next must be an equal sign
            if (!scanEq())
            {
                static const XMLCh tmpList[] =
                {
                    chSingleQuote, chDoubleQuote, chCloseAngle
                    , chOpenAngle, chForwardSlash, chNull
                };

                emitError(XMLErrs::ExpectedEqSign);

                //  Try to sync back up by skipping forward until we either
                //  hit something meaningful.
                const XMLCh chFound = fReaderMgr.skipUntilInOrWS(tmpList);

                if ((chFound == chCloseAngle) || (chFound == chForwardSlash))
                {
                    // Jump back to top for normal processing of these
                    continue;
                }
                else if ((chFound == chSingleQuote)
                      ||  (chFound == chDoubleQuote)
                      ||  fReaderMgr.getCurrentReader()->isWhitespace(chFound))
                {
                    // Just fall through assuming that the value is to follow
                }
                else if (chFound == chOpenAngle)
                {
                    // Assume a malformed tag and that new one is starting
                    emitError(XMLErrs::UnterminatedStartTag, elemName);
                    return attCount;
                }
                else
                {
                    // Something went really wrong
                    return attCount;
                }
            }

            //  Next should be the quoted attribute value. We just do a simple
            //  and stupid scan of this value. The only thing we do here
            //  is to expand entity references.
            if (!basicAttrValueScan(curAttNameBuf, fAttValueBuf))
            {
                static const XMLCh tmpList[] =
                {
                    chCloseAngle, chOpenAngle, chForwardSlash, chNull
                };

                emitError(XMLErrs::ExpectedAttrValue);

                //  It failed, so lets try to get synced back up. We skip
                //  forward until we find some whitespace or one of the
                //  chars in our list.
                const XMLCh chFound = fReaderMgr.skipUntilInOrWS(tmpList);

                if ((chFound == chCloseAngle)
                ||  (chFound == chForwardSlash)
                ||  fReaderMgr.getCurrentReader()->isWhitespace(chFound))
                {
                    //  Just fall through and process this attribute, though
                    //  the value will be "".
                }
                else if (chFound == chOpenAngle)
                {
                    // Assume a malformed tag and that new one is starting
                    emitError(XMLErrs::UnterminatedStartTag, elemName);
                    return attCount;
                }
                else
                {
                    // Something went really wrong
                    return attCount;
                }
            }

            //  And now lets add it to the passed collection. If we have not
            //  filled it up yet, then we use the next element. Else we add
            //  a new one.
            KVStringPair* curPair = 0;
            if (attCount >= curVecSize)
            {
                curPair = new (fMemoryManager) KVStringPair
                (
                    curAttNameBuf
                    , fAttNameBuf.getLen()
                    , fAttValueBuf.getRawBuffer()
                    , fAttValueBuf.getLen()
                    , fMemoryManager
                );
                toFill.addElement(curPair);
            }
             else
            {
                curPair = toFill.elementAt(attCount);
                curPair->set
                (
                    curAttNameBuf
                    , fAttNameBuf.getLen()
                    , fAttValueBuf.getRawBuffer()
                    , fAttValueBuf.getLen()
                );
            }
            if (attCount >= fRawAttrColonListSize) {
                resizeRawAttrColonList();
            }
            fRawAttrColonList[attCount] = colonPosition;

            // And bump the count of attributes we've gotten
            attCount++;

            // And go to the top again for another attribute
            continue;
        }

        //  It was some special case character so do all of the checks and
        //  deal with it.
        if (!nextCh)
            ThrowXMLwithMemMgr(UnexpectedEOFException, XMLExcepts::Gen_UnexpectedEOF, fMemoryManager);

        if (nextCh == chForwardSlash)
        {
            fReaderMgr.getNextChar();
            isEmpty = true;
            if (!fReaderMgr.skippedChar(chCloseAngle))
                emitError(XMLErrs::UnterminatedStartTag, elemName);
            break;
        }
        else if (nextCh == chCloseAngle)
        {
            fReaderMgr.getNextChar();
            break;
        }
        else if (nextCh == chOpenAngle)
        {
            //  Check for this one specially, since its going to be common
            //  and it is kind of auto-recovering since we've already hit the
            //  next open bracket, which is what we would have seeked to (and
            //  skipped this whole tag.)
            emitError(XMLErrs::UnterminatedStartTag, elemName);
            break;
        }
        else if ((nextCh == chSingleQuote) || (nextCh == chDoubleQuote))
        {
            //  Check for this one specially, which is probably a missing
            //  attribute name, e.g. ="value". Just issue expected name
            //  error and eat the quoted string, then jump back to the
            //  top again.
            emitError(XMLErrs::ExpectedAttrName);
            fReaderMgr.getNextChar();
            fReaderMgr.skipQuotedString(nextCh);
            fReaderMgr.skipPastSpaces();
            continue;
        }
    }

    return attCount;
}


//  This method will kick off the scanning of the primary content of the
//  document, i.e. the elements.
bool SGXMLScanner::scanContent()
{
    //  Go into a loop until we hit the end of the root element, or we fall
    //  out because there is no root element.
    //
    //  We have to do kind of a deeply nested double loop here in order to
    //  avoid doing the setup/teardown of the exception handler on each
    //  round. Doing it this way we only do it when an exception actually
    //  occurs.
    bool gotData = true;
    bool inMarkup = false;
    while (gotData)
    {
        try
        {
            while (gotData)
            {
                //  Sense what the next top level token is. According to what
                //  this tells us, we will call something to handle that kind
                //  of thing.
                XMLSize_t orgReader;
                const XMLTokens curToken = senseNextToken(orgReader);

                //  Handle character data and end of file specially. Char data
                //  is not markup so we don't want to handle it in the loop
                //  below.
                if (curToken == Token_CharData)
                {
                    //  Scan the character data and call appropriate events. Let
                    //  him use our local character data buffer for efficiency.
                    scanCharData(fCDataBuf);
                    continue;
                }
                else if (curToken == Token_EOF)
                {
                    //  The element stack better be empty at this point or we
                    //  ended prematurely before all elements were closed.
                    if (!fElemStack.isEmpty())
                    {
                        const ElemStack::StackElem* topElem = fElemStack.popTop();
                        emitError
                        (
                            XMLErrs::EndedWithTagsOnStack
                            , topElem->fThisElement->getFullName()
                        );
                    }

                    // Its the end of file, so clear the got data flag
                    gotData = false;
                    continue;
                }

                // We are in some sort of markup now
                inMarkup = true;

                //  According to the token we got, call the appropriate
                //  scanning method.
                switch(curToken)
                {
                    case Token_CData :
                        // Make sure we are within content
                        if (fElemStack.isEmpty())
                            emitError(XMLErrs::CDATAOutsideOfContent);
                        scanCDSection();
                        break;

                    case Token_Comment :
                        scanComment();
                        break;

                    case Token_EndTag :
                        scanEndTag(gotData);
                        break;

                    case Token_PI :
                        scanPI();
                        break;

                    case Token_StartTag :
                        scanStartTag(gotData);
                        break;

                    default :
                        fReaderMgr.skipToChar(chOpenAngle);
                        break;
                }

                if (orgReader != fReaderMgr.getCurrentReaderNum())
                    emitError(XMLErrs::PartialMarkupInEntity);

                // And we are back out of markup again
                inMarkup = false;
            }
        }
        catch(const EndOfEntityException& toCatch)
        {
            //  If we were in some markup when this happened, then its a
            //  partial markup error.
            if (inMarkup)
                emitError(XMLErrs::PartialMarkupInEntity);

            // Send an end of entity reference event
            if (fDocHandler)
                fDocHandler->endEntityReference(toCatch.getEntity());

            inMarkup = false;
        }
    }

    // It went ok, so return success
    return true;
}


void SGXMLScanner::scanEndTag(bool& gotData)
{
    //  Assume we will still have data until proven otherwise. It will only
    //  ever be false if this is the end of the root element.
    gotData = true;

    //  Check if the element stack is empty. If so, then this is an unbalanced
    //  element (i.e. more ends than starts, perhaps because of bad text
    //  causing one to be skipped.)
    if (fElemStack.isEmpty())
    {
        emitError(XMLErrs::MoreEndThanStartTags);
        fReaderMgr.skipPastChar(chCloseAngle);
        ThrowXMLwithMemMgr(RuntimeException, XMLExcepts::Scan_UnbalancedStartEnd, fMemoryManager);
    }

    //  Pop the stack of the element we are supposed to be ending. Remember
    //  that we don't own this. The stack just keeps them and reuses them.
    unsigned int uriId = (fDoNamespaces)
        ? fElemStack.getCurrentURI() : fEmptyNamespaceId;

    // Make sure that its the end of the element that we expect
    const XMLCh *elemName = fElemStack.getCurrentSchemaElemName();
    const ElemStack::StackElem* topElem = fElemStack.topElement();
    if (!fReaderMgr.skippedString(elemName))
    {
        emitError
        (
            XMLErrs::ExpectedEndOfTagX
            , elemName
        );
        fReaderMgr.skipPastChar(chCloseAngle);
        fElemStack.popTop();
        return;
    }

    fPSVIElemContext.fErrorOccurred = fErrorStack->pop();

    // Make sure we are back on the same reader as where we started
    if (topElem->fReaderNum != fReaderMgr.getCurrentReaderNum())
        emitError(XMLErrs::PartialTagMarkupError);

    // Skip optional whitespace
    fReaderMgr.skipPastSpaces();

    // Make sure we find the closing bracket
    if (!fReaderMgr.skippedChar(chCloseAngle))
    {
        emitError
        (
            XMLErrs::UnterminatedEndTag
            , topElem->fThisElement->getFullName()
        );
    }

    if (fValidate && topElem->fThisElement->isDeclared())
    {
        fPSVIElemContext.fCurrentTypeInfo = ((SchemaValidator*) fValidator)->getCurrentTypeInfo();
        if(!fPSVIElemContext.fCurrentTypeInfo)
            fPSVIElemContext.fCurrentDV = ((SchemaValidator*) fValidator)->getCurrentDatatypeValidator();
        else
            fPSVIElemContext.fCurrentDV = 0;
        if (fPSVIHandler)
        {
            fPSVIElemContext.fNormalizedValue = ((SchemaValidator*) fValidator)->getNormalizedValue();

            if (XMLString::equals(fPSVIElemContext.fNormalizedValue, XMLUni::fgZeroLenString))
                fPSVIElemContext.fNormalizedValue = 0;

        }
    }
    else
    {
        fPSVIElemContext.fCurrentDV = 0;
        fPSVIElemContext.fCurrentTypeInfo = 0;
        fPSVIElemContext.fNormalizedValue = 0;
    }

    //  If validation is enabled, then lets pass him the list of children and
    //  this element and let him validate it.
    DatatypeValidator* psviMemberType = 0;
    if (fValidate)
    {
        XMLSize_t failure;
        bool res = fValidator->checkContent
        (
            topElem->fThisElement
            , topElem->fChildren
            , topElem->fChildCount
            , &failure
        );

        if (!res)
        {
            //  One of the elements is not valid for the content. NOTE that
            //  if no children were provided but the content model requires
            //  them, it comes back with a zero value. But we cannot use that
            //  to index the child array in this case, and have to put out a
            //  special message.
            if (!topElem->fChildCount)
            {
                fValidator->emitError
                (
                    XMLValid::EmptyNotValidForContent
                    , topElem->fThisElement->getFormattedContentModel()
                );
            }
            else if (failure >= topElem->fChildCount)
            {
                fValidator->emitError
                (
                    XMLValid::NotEnoughElemsForCM
                    , topElem->fThisElement->getFormattedContentModel()
                );
            }
            else
            {
                fValidator->emitError
                (
                    XMLValid::ElementNotValidForContent
                    , topElem->fChildren[failure]->getRawName()
                    , topElem->fThisElement->getFormattedContentModel()
                );
            }

        }

        // update PSVI info
        if (((SchemaValidator*) fValidator)->getErrorOccurred())
            fPSVIElemContext.fErrorOccurred = true;
        else if (fPSVIElemContext.fCurrentDV && fPSVIElemContext.fCurrentDV->getType() == DatatypeValidator::Union)
            psviMemberType = fValidationContext->getValidatingMemberType();
        if (fPSVIHandler)
        {
            fPSVIElemContext.fIsSpecified = ((SchemaValidator*) fValidator)->getIsElemSpecified();
            if(fPSVIElemContext.fIsSpecified)
                fPSVIElemContext.fNormalizedValue = ((SchemaElementDecl *)topElem->fThisElement)->getDefaultValue();
        }

        // call matchers and de-activate context
        if (toCheckIdentityConstraint())
        {
            fICHandler->deactivateContext
                        (
                         (SchemaElementDecl *) topElem->fThisElement
                       , fContent.getRawBuffer()
                        );
        }

    }

    // QName dv needed topElem to resolve URIs on the checkContent
    fElemStack.popTop();

    // See if it was the root element, to avoid multiple calls below
    const bool isRoot = fElemStack.isEmpty();

    if (fPSVIHandler)
    {
        endElementPSVI
        (
            (SchemaElementDecl*)topElem->fThisElement, psviMemberType
        );
    }
    // now we can reset the datatype buffer, since the
    // application has had a chance to copy the characters somewhere else
    ((SchemaValidator *)fValidator)->clearDatatypeBuffer();

    // If we have a doc handler, tell it about the end tag
    if (fDocHandler)
    {
        if (fGrammarType == Grammar::SchemaGrammarType) {
            if (topElem->fPrefixColonPos != -1)
                fPrefixBuf.set(elemName, topElem->fPrefixColonPos);
            else
                fPrefixBuf.reset();
        }
        else {
            fPrefixBuf.set(topElem->fThisElement->getElementName()->getPrefix());
        }
        fDocHandler->endElement
        (
            *topElem->fThisElement
            , uriId
            , isRoot
            , fPrefixBuf.getRawBuffer()
        );
    }

    if (!isRoot)
    {
        // update error information
        fErrorStack->push((fErrorStack->size() && fErrorStack->pop()) || fPSVIElemContext.fErrorOccurred);
    }

    // If this was the root, then done with content
    gotData = !isRoot;

    if (gotData) {

        // Restore the grammar
        fGrammar = fElemStack.getCurrentGrammar();
        fGrammarType = fGrammar->getGrammarType();
        fValidator->setGrammar(fGrammar);

        // Restore the validation flag
        fValidate = fElemStack.getValidationFlag();
    }
}


//  This method handles the high level logic of scanning the DOCType
//  declaration. This calls the DTDScanner and kicks off both the scanning of
//  the internal subset and the scanning of the external subset, if any.
//
//  When we get here the '<!DOCTYPE' part has already been scanned, which is
//  what told us that we had a doc type decl to parse.
void SGXMLScanner::scanDocTypeDecl()
{
    // Just skips over it
    // REVISIT: Should we issue a warning
    static const XMLCh doctypeIE[] =
    {
            chOpenSquare, chCloseAngle, chNull
    };
    XMLCh nextCh = fReaderMgr.skipUntilIn(doctypeIE);

    if (nextCh == chOpenSquare)
        fReaderMgr.skipPastChar(chCloseSquare);

    fReaderMgr.skipPastChar(chCloseAngle);
}

//  This method is called to scan a start tag when we are processing
//  namespaces. This method is called after we've scanned the < of a
//  start tag. So we have to get the element name, then scan the attributes,
//  after which we are either going to see >, />, or attributes followed
//  by one of those sequences.
bool SGXMLScanner::scanStartTag(bool& gotData)
{
    //  Assume we will still have data until proven otherwise. It will only
    //  ever be false if this is the root and its empty.
    gotData = true;

    // Reset element content
    fContent.reset();

    //  The current position is after the open bracket, so we need to read in
    //  in the element name.
    int prefixColonPos;
    if (!fReaderMgr.getQName(fQNameBuf, &prefixColonPos))
    {
        if (fQNameBuf.isEmpty())
            emitError(XMLErrs::ExpectedElementName);
        else
            emitError(XMLErrs::InvalidElementName, fQNameBuf.getRawBuffer());
        fReaderMgr.skipToChar(chOpenAngle);
        return false;
    }

    // See if its the root element
    const bool isRoot = fElemStack.isEmpty();

    // Skip any whitespace after the name
    fReaderMgr.skipPastSpaces();

    //  First we have to do the rawest attribute scan. We don't do any
    //  normalization of them at all, since we don't know yet what type they
    //  might be (since we need the element decl in order to do that.)
    const XMLCh* qnameRawBuf = fQNameBuf.getRawBuffer();
    bool isEmpty;
    XMLSize_t attCount = rawAttrScan
    (
        qnameRawBuf
        , *fRawAttrList
        , isEmpty
    );

    // save the contentleafname and currentscope before addlevel, for later use
    ContentLeafNameTypeVector* cv = 0;
    XMLContentModel* cm = 0;
    unsigned int currentScope = Grammar::TOP_LEVEL_SCOPE;
    bool laxThisOne = false;
    if (!isRoot)
    {
        // schema validator will have correct type if validating
        SchemaElementDecl* tempElement = (SchemaElementDecl*)
            fElemStack.topElement()->fThisElement;
        SchemaElementDecl::ModelTypes modelType = tempElement->getModelType();
        ComplexTypeInfo *currType = 0;

        if (fValidate)
        {
            currType = ((SchemaValidator*)fValidator)->getCurrentTypeInfo();
            if (currType)
                modelType = (SchemaElementDecl::ModelTypes)currType->getContentType();
            else // something must have gone wrong
                modelType = SchemaElementDecl::Any;
        }
        else
        {
            currType = tempElement->getComplexTypeInfo();
        }

        if ((modelType == SchemaElementDecl::Mixed_Simple)
          ||  (modelType == SchemaElementDecl::Mixed_Complex)
          ||  (modelType == SchemaElementDecl::Children))
        {
            cm = currType->getContentModel();
            cv = cm->getContentLeafNameTypeVector();
            currentScope = fElemStack.getCurrentScope();
        }
        else if (modelType == SchemaElementDecl::Any) {
            laxThisOne = true;
        }
    }

    //  Now, since we might have to update the namespace map for this element,
    //  but we don't have the element decl yet, we just tell the element stack
    //  to expand up to get ready.
    unsigned int elemDepth = fElemStack.addLevel();
    fElemStack.setValidationFlag(fValidate);
    fElemStack.setPrefixColonPos(prefixColonPos);

    //  Check if there is any external schema location specified, and if we are at root,
    //  go through them first before scanning those specified in the instance document
    if (isRoot
        && (fExternalSchemaLocation || fExternalNoNamespaceSchemaLocation)) {

        if (fExternalSchemaLocation)
            parseSchemaLocation(fExternalSchemaLocation);
        if (fExternalNoNamespaceSchemaLocation)
            resolveSchemaGrammar(fExternalNoNamespaceSchemaLocation, XMLUni::fgZeroLenString);
    }

    //  Make an initial pass through the list and find any xmlns attributes or
    //  schema attributes.
    if (attCount)
        scanRawAttrListforNameSpaces(attCount);

    //  Resolve the qualified name to a URI and name so that we can look up
    //  the element decl for this element. We have now update the prefix to
    //  namespace map so we should get the correct element now.
    unsigned int uriId = resolveQNameWithColon
    (
        qnameRawBuf
        , fPrefixBuf
        , ElemStack::Mode_Element
        , prefixColonPos
    );

    //if schema, check if we should lax or skip the validation of this element
    bool parentValidation = fValidate;
    if (cv) {
        QName element(fPrefixBuf.getRawBuffer(), &qnameRawBuf[prefixColonPos + 1], uriId, fMemoryManager);
        // elementDepth will be > 0, as cv is only constructed if element is not
        // root.
        laxThisOne = laxElementValidation(&element, cv, cm, elemDepth - 1);
    }

    //  Look up the element now in the grammar. This will get us back a
    //  generic element decl object. We tell him to fault one in if he does
    //  not find it.
    XMLElementDecl* elemDecl = 0;
    bool wasAdded = false;
    const XMLCh* nameRawBuf = &qnameRawBuf[prefixColonPos + 1];
    const XMLCh* original_uriStr = fGrammar->getTargetNamespace();

    if (uriId != fEmptyNamespaceId) {

        // Check in current grammar before switching if necessary
        elemDecl = fGrammar->getElemDecl
        (
          uriId
          , nameRawBuf
          , qnameRawBuf
          , currentScope
        );
        if(!elemDecl)
        {
            // look in the list of undeclared elements, as would have been done
            // before we made grammars stateless:
            elemDecl = fElemNonDeclPool->getByKey(nameRawBuf, uriId, currentScope);
        }
        // this is initialized correctly only if there is
        // no element decl.  The other uses in this scope will only
        // be encountered if there continues to be no element decl--which
        // implies that this will have been initialized correctly.
        unsigned int orgGrammarUri = uriId;
        if (!elemDecl && ( orgGrammarUri = fURIStringPool->getId(original_uriStr)) != uriId) {
            // not found, switch to the specified grammar
            const XMLCh* uriStr = getURIText(uriId);
            bool errorCondition = !switchGrammar(uriStr) && fValidate;
            if (errorCondition && !laxThisOne)
            {
                fValidator->emitError
                (
                    XMLValid::GrammarNotFound
                    ,uriStr
                );
            }

            elemDecl = fGrammar->getElemDecl
            (
              uriId
              , nameRawBuf
              , qnameRawBuf
              , currentScope
            );
        }

        if (!elemDecl && currentScope != Grammar::TOP_LEVEL_SCOPE) {
            // if not found, then it may be a reference, try TOP_LEVEL_SCOPE
            elemDecl = fGrammar->getElemDecl
                       (
                           uriId
                           , nameRawBuf
                           , qnameRawBuf
                           , Grammar::TOP_LEVEL_SCOPE
                       );
            if(!elemDecl)
            {
                // look in the list of undeclared elements, as would have been done
                // before we made grammars stateless:
                elemDecl = fElemNonDeclPool->getByKey(nameRawBuf, uriId, Grammar::TOP_LEVEL_SCOPE);
            }
            if(!elemDecl) {
                // still not found in specified uri
                // try emptyNamesapce see if element should be un-qualified.
                elemDecl = fGrammar->getElemDecl
                           (
                               fEmptyNamespaceId
                               , nameRawBuf
                               , qnameRawBuf
                               , currentScope
                           );
                bool errorCondition = elemDecl && elemDecl->getCreateReason() != XMLElementDecl::JustFaultIn;
                if (errorCondition && fValidate) {
                    fValidator->emitError
                    (
                        XMLValid::ElementNotUnQualified
                        , elemDecl->getFullName()
                    );
                }
            }
        }

        if (!elemDecl) {
            // still not found, fault this in and issue error later
            // switch back to original grammar first (if necessary)
            if(orgGrammarUri != uriId)
            {
                switchGrammar(original_uriStr);
            }
            elemDecl = new (fMemoryManager) SchemaElementDecl
            (
                fPrefixBuf.getRawBuffer()
                , nameRawBuf
                , uriId
                , SchemaElementDecl::Any
                , Grammar::TOP_LEVEL_SCOPE
                , fMemoryManager
            );
            elemDecl->setId(fElemNonDeclPool->put((void*)elemDecl->getBaseName(), uriId, currentScope, (SchemaElementDecl*)elemDecl));
            wasAdded = true;
        }
    }
    else if (!elemDecl)
    {
        //the element has no prefix,
        //thus it is either a non-qualified element defined in current targetNS
        //or an element that is defined in the globalNS

        //try unqualifed first
        elemDecl = fGrammar->getElemDecl
                   (
                      uriId
                    , nameRawBuf
                    , qnameRawBuf
                    , currentScope
                    );
        if(!elemDecl)
        {
            // look in the list of undeclared elements, as would have been done
            // before we made grammars stateless:
            elemDecl = fElemNonDeclPool->getByKey(nameRawBuf, uriId, currentScope);
        }
        // this is initialized correctly only if there is
        // no element decl.  The other uses in this scope will only
        // be encountered if there continues to be no element decl--which
        // implies that this will have been initialized correctly.
        unsigned int orgGrammarUri = fEmptyNamespaceId;
        if (!elemDecl && (orgGrammarUri = fURIStringPool->getId(original_uriStr)) != fEmptyNamespaceId) {
            //not found, switch grammar and try globalNS
            bool errorCondition = !switchGrammar(XMLUni::fgZeroLenString) && fValidate;
            if (errorCondition && !laxThisOne)
            {
                fValidator->emitError
                (
                    XMLValid::GrammarNotFound
                  , XMLUni::fgZeroLenString
                );
            }

            elemDecl = fGrammar->getElemDecl
            (
              uriId
              , nameRawBuf
              , qnameRawBuf
              , currentScope
            );
        }

        if (!elemDecl && currentScope != Grammar::TOP_LEVEL_SCOPE) {
            // if not found, then it may be a reference, try TOP_LEVEL_SCOPE
            elemDecl = fGrammar->getElemDecl
                       (
                           uriId
                           , nameRawBuf
                           , qnameRawBuf
                           , Grammar::TOP_LEVEL_SCOPE
                       );
            if(!elemDecl)
            {
                // look in the list of undeclared elements, as would have been done
                // before we made grammars stateless:
                elemDecl = fElemNonDeclPool->getByKey(nameRawBuf, uriId, Grammar::TOP_LEVEL_SCOPE);
            }
            if (!elemDecl && orgGrammarUri != fEmptyNamespaceId) {
                // still Not found in specified uri
                // go to original Grammar again to see if element needs to be fully qualified.
                bool errorCondition = !switchGrammar(original_uriStr) && fValidate;
                if (errorCondition && !laxThisOne)
                {
                    fValidator->emitError
                    (
                        XMLValid::GrammarNotFound
                        ,original_uriStr
                    );
                }

                elemDecl = fGrammar->getElemDecl
                           (
                               orgGrammarUri
                               , nameRawBuf
                               , qnameRawBuf
                               , currentScope
                           );
                if (elemDecl && elemDecl->getCreateReason() != XMLElementDecl::JustFaultIn && fValidate) {
                    fValidator->emitError
                    (
                        XMLValid::ElementNotQualified
                        , elemDecl->getFullName()
                    );
                }
            }
        }

        if (!elemDecl) {
            // still not found, fault this in and issue error later
            // switch back to original grammar first (if necessary)
            if(orgGrammarUri != fEmptyNamespaceId)
            {
                switchGrammar(original_uriStr);
            }
            elemDecl = new (fMemoryManager) SchemaElementDecl
            (
                fPrefixBuf.getRawBuffer()
                , nameRawBuf
                , uriId
                , SchemaElementDecl::Any
                , Grammar::TOP_LEVEL_SCOPE
                , fMemoryManager
            );
            elemDecl->setId(fElemNonDeclPool->put((void*)elemDecl->getBaseName(), uriId, currentScope, (SchemaElementDecl*)elemDecl));
            wasAdded = true;
        }
    }

    // this info needed for DOMTypeInfo
    fPSVIElemContext.fErrorOccurred = false;

    //  We do something different here according to whether we found the
    //  element or not.
    if (wasAdded)
    {
        if (laxThisOne) {
            fValidate = false;
            fElemStack.setValidationFlag(fValidate);
        }

        // If validating then emit an error
        if (fValidate)
        {
            // This is to tell the reuse Validator that this element was
            // faulted-in, was not an element in the grammar pool originally
            elemDecl->setCreateReason(XMLElementDecl::JustFaultIn);

            fValidator->emitError
            (
                XMLValid::ElementNotDefined
                , elemDecl->getFullName()
            );
            fPSVIElemContext.fErrorOccurred = true;
        }
    }
    else
    {
        // If its not marked declared and validating, then emit an error
        if (!elemDecl->isDeclared()) {
            if(elemDecl->getCreateReason() == XMLElementDecl::NoReason) {
                fPSVIElemContext.fErrorOccurred = true;
            }
            if (laxThisOne) {
                fValidate = false;
                fElemStack.setValidationFlag(fValidate);
            }

            if (fValidate)
                {
                fValidator->emitError
                (
                    XMLValid::ElementNotDefined
                    , elemDecl->getFullName()
                );
            }
        }
    }


    //  Now we can update the element stack to set the current element
    //  decl. We expanded the stack above, but couldn't store the element
    //  decl because we didn't know it yet.
    fElemStack.setElement(elemDecl, fReaderMgr.getCurrentReaderNum());
    fElemStack.setCurrentURI(uriId);

    if (isRoot)
    {
        fRootGrammar = fGrammar;
        fRootElemName = XMLString::replicate(qnameRawBuf, fMemoryManager);
    }

    if (fPSVIHandler)
    {
        fPSVIElemContext.fElemDepth++;

        if (elemDecl->isDeclared())
        {
            fPSVIElemContext.fNoneValidationDepth = fPSVIElemContext.fElemDepth;
        }
        else
        {
            fPSVIElemContext.fFullValidationDepth = fPSVIElemContext.fElemDepth;

            /******
             * While we report an error for historical reasons, this should
             * actually result in lax assessment - NG.
            if (isRoot && fValidate)
                fPSVIElemContext.fErrorOccurred = true;
            ******/
        }
    }

    //  Validate the element
    if (fValidate)
    {
        fValidator->validateElement(elemDecl);
        if (((SchemaValidator*) fValidator)->getErrorOccurred())
            fPSVIElemContext.fErrorOccurred = true;
    }

    // squirrel away the element's QName, so that we can do an efficient
    // end-tag match
    fElemStack.setCurrentSchemaElemName(fQNameBuf.getRawBuffer());

    ComplexTypeInfo* typeinfo = (fValidate)
        ? ((SchemaValidator*)fValidator)->getCurrentTypeInfo()
        : ((SchemaElementDecl*) elemDecl)->getComplexTypeInfo();

    if (typeinfo) {
        currentScope = typeinfo->getScopeDefined();

        // switch grammar if the typeinfo has a different grammar (happens when there is xsi:type)
        XMLCh* typeName = typeinfo->getTypeName();
        const int comma = XMLString::indexOf(typeName, chComma);
        if (comma > 0) {
            XMLBuffer prefixBuf(comma+1, fMemoryManager);
            prefixBuf.append(typeName, comma);
            const XMLCh* uriStr = prefixBuf.getRawBuffer();

            bool errorCondition = !switchGrammar(uriStr) && fValidate;
            if (errorCondition && !laxThisOne)
            {
                fValidator->emitError
                (
                    XMLValid::GrammarNotFound
                    , prefixBuf.getRawBuffer()
                );
            }
        }
        else if (comma == 0) {
            bool errorCondition = !switchGrammar(XMLUni::fgZeroLenString) && fValidate;
            if (errorCondition && !laxThisOne)
            {
                fValidator->emitError
                (
                    XMLValid::GrammarNotFound
                    , XMLUni::fgZeroLenString
                );
            }
        }
    }
    fElemStack.setCurrentScope(currentScope);

    // Set element next state
    if (elemDepth >= fElemStateSize) {
        resizeElemState();
    }

    fElemState[elemDepth] = 0;
    fElemStack.setCurrentGrammar(fGrammar);

    //  If this is the first element and we are validating, check the root
    //  element.
    if (!isRoot && parentValidation)
    {
        //  If the element stack is not empty, then add this element as a
        //  child of the previous top element. If its empty, this is the root
        //  elem and is not the child of anything.
        fElemStack.addChild(elemDecl->getElementName(), true);
    }

    // PSVI handling:  must reset this, even if no attributes...
    if(getPSVIHandler())
        fPSVIAttrList->reset();

    //  Now lets get the fAttrList filled in. This involves faulting in any
    //  defaulted and fixed attributes and normalizing the values of any that
    //  we got explicitly.
    //
    //  We update the attCount value with the total number of attributes, but
    //  it goes in with the number of values we got during the raw scan of
    //  explictly provided attrs above.
    attCount = buildAttList(*fRawAttrList, attCount, elemDecl, *fAttrList);

    if(attCount)
    {
        // clean up after ourselves:
        // clear the map used to detect duplicate attributes
        fUndeclaredAttrRegistryNS->removeAll();
    }

    // activate identity constraints
    if (toCheckIdentityConstraint())
    {
        fICHandler->activateIdentityConstraint
                        (
                          (SchemaElementDecl*) elemDecl
                        , (int) elemDepth
                        , uriId
                        , fPrefixBuf.getRawBuffer()
                        , *fAttrList
                        , attCount
                        );

    }

    // Since the element may have default values, call start tag now regardless if it is empty or not
    // If we have a document handler, then tell it about this start tag
    if (fDocHandler)
    {
        fDocHandler->startElement
        (
            *elemDecl
            , uriId
            , fPrefixBuf.getRawBuffer()
            , *fAttrList
            , attCount
            , false
            , isRoot
        );
    } // may be where we output something...

    // if we have a PSVIHandler, now's the time to call
    // its handleAttributesPSVI method:
    if(fPSVIHandler)
    {
        QName *eName = elemDecl->getElementName();
        fPSVIHandler->handleAttributesPSVI
        (
            eName->getLocalPart()
            , fURIStringPool->getValueForId(eName->getURI())
            , fPSVIAttrList
        );
    }

    //  If empty, validate content right now if we are validating and then
    //  pop the element stack top. Else, we have to update the current stack
    //  top's namespace mapping elements.
    if (isEmpty)
    {
        // Pop the element stack back off since it'll never be used now
        fElemStack.popTop();

        // reset current type info
        DatatypeValidator* psviMemberType = 0;
        if (fGrammarType == Grammar::SchemaGrammarType)
        {
            if (fValidate && elemDecl->isDeclared())
            {
                fPSVIElemContext.fCurrentTypeInfo = ((SchemaValidator*) fValidator)->getCurrentTypeInfo();
                if(!fPSVIElemContext.fCurrentTypeInfo)
                    fPSVIElemContext.fCurrentDV = ((SchemaValidator*) fValidator)->getCurrentDatatypeValidator();
                else
                    fPSVIElemContext.fCurrentDV = 0;
                if(fPSVIHandler)
                {
                    fPSVIElemContext.fNormalizedValue = ((SchemaValidator*) fValidator)->getNormalizedValue();

                    if (XMLString::equals(fPSVIElemContext.fNormalizedValue, XMLUni::fgZeroLenString))
                        fPSVIElemContext.fNormalizedValue = 0;
                }
            }
            else
            {
                fPSVIElemContext.fCurrentDV = 0;
                fPSVIElemContext.fCurrentTypeInfo = 0;
                fPSVIElemContext.fNormalizedValue = 0;
            }
        }

        // If validating, then insure that its legal to have no content
        if (fValidate)
        {
            XMLSize_t failure;
            bool res = fValidator->checkContent(elemDecl, 0, 0, &failure);
            if (!res)
            {
                // REVISIT:  in the case of xsi:type, this may
                // return the wrong string...
                fValidator->emitError
                (
                    XMLValid::ElementNotValidForContent
                    , elemDecl->getFullName()
                    , elemDecl->getFormattedContentModel()
                );
            }

            if (((SchemaValidator*) fValidator)->getErrorOccurred())
                fPSVIElemContext.fErrorOccurred = true;
            // note that if we're empty, won't be a current DV
            else
            {
                if (fPSVIHandler)
                {
                    fPSVIElemContext.fIsSpecified = ((SchemaValidator*) fValidator)->getIsElemSpecified();
                    if(fPSVIElemContext.fIsSpecified)
                        fPSVIElemContext.fNormalizedValue = ((SchemaElementDecl *)elemDecl)->getDefaultValue();
                }
                if (fPSVIElemContext.fCurrentDV && fPSVIElemContext.fCurrentDV->getType() == DatatypeValidator::Union)
                    psviMemberType = fValidationContext->getValidatingMemberType();
            }

            // call matchers and de-activate context
            if (toCheckIdentityConstraint())
            {
                fICHandler->deactivateContext
                       (
                        (SchemaElementDecl *) elemDecl
                      , fContent.getRawBuffer()
                       );
            }

        }
        else if (fGrammarType == Grammar::SchemaGrammarType) {
            ((SchemaValidator*)fValidator)->setNillable(false);
        }

        if (fPSVIHandler)
        {
            endElementPSVI
            (
                (SchemaElementDecl*)elemDecl, psviMemberType
            );
        }

        // If we have a doc handler, tell it about the end tag
        if (fDocHandler)
        {
            fDocHandler->endElement
            (
                *elemDecl
                , uriId
                , isRoot
                , fPrefixBuf.getRawBuffer()
            );
        }

        // If the elem stack is empty, then it was an empty root
        if (isRoot)
            gotData = false;
        else
        {
            // Restore the grammar
            fGrammar = fElemStack.getCurrentGrammar();
            fGrammarType = fGrammar->getGrammarType();
            fValidator->setGrammar(fGrammar);

            // Restore the validation flag
            fValidate = fElemStack.getValidationFlag();
        }
    }
    else    // not empty
    {

        // send a partial element psvi
        if (fPSVIHandler)
        {

            ComplexTypeInfo*   curTypeInfo = 0;
            DatatypeValidator* curDV = 0;
            XSTypeDefinition*  typeDef = 0;

            if (fValidate && elemDecl->isDeclared())
            {
                curTypeInfo = ((SchemaValidator*) fValidator)->getCurrentTypeInfo();

                if (curTypeInfo)
                {
                    typeDef = (XSTypeDefinition*) fModel->getXSObject(curTypeInfo);
                }
                else
                {
                    curDV = ((SchemaValidator*) fValidator)->getCurrentDatatypeValidator();

                    if (curDV)
                    {
                        typeDef = (XSTypeDefinition*) fModel->getXSObject(curDV);
                    }
                }
            }

            fPSVIElement->reset
                (
                  PSVIElement::VALIDITY_NOTKNOWN
                , PSVIElement::VALIDATION_NONE
                , fRootElemName
                , ((SchemaValidator*) fValidator)->getIsElemSpecified()
                , (elemDecl->isDeclared()) ? (XSElementDeclaration*) fModel->getXSObject(elemDecl) : 0
                , typeDef
                , 0 //memberType
                , fModel
                , ((SchemaElementDecl*)elemDecl)->getDefaultValue()
                , 0
                , 0
                , 0
                );


            fPSVIHandler->handlePartialElementPSVI
                (
                  elemDecl->getBaseName()
                , fURIStringPool->getValueForId(elemDecl->getURI())
                , fPSVIElement
                );

        }

        fErrorStack->push(fPSVIElemContext.fErrorOccurred);
    }

    return true;
}


unsigned int
SGXMLScanner::resolveQName(const   XMLCh* const qName
                           ,       XMLBuffer&   prefixBuf
                           , const short        mode
                           ,       int&         prefixColonPos)
{
    prefixColonPos = XMLString::indexOf(qName, chColon);
    return resolveQNameWithColon(qName, prefixBuf, mode, prefixColonPos);
}

unsigned int
SGXMLScanner::resolveQNameWithColon(const   XMLCh* const qName
                                    ,       XMLBuffer&   prefixBuf
                                    , const short        mode
                                    , const int          prefixColonPos)
{
    //  Lets split out the qName into a URI and name buffer first. The URI
    //  can be empty.
    if (prefixColonPos == -1)
    {
        //  Its all name with no prefix, so put the whole thing into the name
        //  buffer. Then map the empty string to a URI, since the empty string
        //  represents the default namespace. This will either return some
        //  explicit URI which the default namespace is mapped to, or the
        //  the default global namespace.
        bool unknown = false;

        prefixBuf.reset();
        return fElemStack.mapPrefixToURI(XMLUni::fgZeroLenString, (ElemStack::MapModes) mode, unknown);
    }
    else
    {
        //  Copy the chars up to but not including the colon into the prefix
        //  buffer.
        prefixBuf.set(qName, prefixColonPos);

        //  Watch for the special namespace prefixes. We always map these to
        //  special URIs. 'xml' gets mapped to the official URI that its defined
        //  to map to by the NS spec. xmlns gets mapped to a special place holder
        //  URI that we define (so that it maps to something checkable.)
        const XMLCh* prefixRawBuf = prefixBuf.getRawBuffer();
        if (XMLString::equals(prefixRawBuf, XMLUni::fgXMLNSString)) {

            // if this is an element, it is an error to have xmlns as prefix
            if (mode == ElemStack::Mode_Element)
                emitError(XMLErrs::NoXMLNSAsElementPrefix, qName);

            return fXMLNSNamespaceId;
        }
        else if (XMLString::equals(prefixRawBuf, XMLUni::fgXMLString)) {
            return  fXMLNamespaceId;
        }
        else
        {
            bool unknown = false;
            unsigned int uriId = fElemStack.mapPrefixToURI(prefixRawBuf, (ElemStack::MapModes) mode, unknown);

            if (unknown)
                emitError(XMLErrs::UnknownPrefix, prefixRawBuf);

            // check to see if uriId is empty
            if (fXMLVersion != XMLReader::XMLV1_0 &&
                uriId == fElemStack.getEmptyNamespaceId())
                emitError(XMLErrs::UnknownPrefix, prefixRawBuf);

            return uriId;
        }
    }
}

// ---------------------------------------------------------------------------
//  SGXMLScanner: Grammar preparsing
// ---------------------------------------------------------------------------
Grammar* SGXMLScanner::loadGrammar(const   InputSource& src
                                   , const short        grammarType
                                   , const bool         toCache)
{
    Grammar* loadedGrammar = 0;

    ReaderMgrResetType  resetReaderMgr(&fReaderMgr, &ReaderMgr::reset);

    try
    {
        fGrammarResolver->cacheGrammarFromParse(false);
		// if the new grammar has to be cached, better use the already cached
		// grammars, or the an exception will be thrown when caching an already
		// cached grammar
        fGrammarResolver->useCachedGrammarInParse(toCache);
        fRootGrammar = 0;

        if (fValScheme == Val_Auto) {
            fValidate = true;
        }

        // Reset some status flags
        fInException = false;
        fStandalone = false;
        fErrorCount = 0;
        fHasNoDTD = true;
        fSeeXsi = false;

        if (grammarType == Grammar::SchemaGrammarType) {
            loadedGrammar = loadXMLSchemaGrammar(src, toCache);
        }
    }
    //  NOTE:
    //
    //  In all of the error processing below, the emitError() call MUST come
    //  before the flush of the reader mgr, or it will fail because it tries
    //  to find out the position in the XML source of the error.
    catch(const XMLErrs::Codes)
    {
        // This is a 'first failure' exception, so fall through
    }
    catch(const XMLValid::Codes)
    {
        // This is a 'first fatal error' type exit, so fall through
    }
    catch(const XMLException& excToCatch)
    {
        //  Emit the error and catch any user exception thrown from here. Make
        //  sure in all cases we flush the reader manager.
        fInException = true;
        try
        {
            if (excToCatch.getErrorType() == XMLErrorReporter::ErrType_Warning)
                emitError
                (
                    XMLErrs::XMLException_Warning
                    , excToCatch.getCode()
                    , excToCatch.getMessage()
                );
            else if (excToCatch.getErrorType() >= XMLErrorReporter::ErrType_Fatal)
                emitError
                (
                    XMLErrs::XMLException_Fatal
                    , excToCatch.getCode()
                    , excToCatch.getMessage()
                );
            else
                emitError
                (
                    XMLErrs::XMLException_Error
                    , excToCatch.getCode()
                    , excToCatch.getMessage()
                );
        }
        catch(const OutOfMemoryException&)
        {
            // This is a special case for out-of-memory
            // conditions, because resetting the ReaderMgr
            // can be problematic.
            resetReaderMgr.release();

            throw;
        }
    }
    catch(const OutOfMemoryException&)
    {
        // This is a special case for out-of-memory
        // conditions, because resetting the ReaderMgr
        // can be problematic.
        resetReaderMgr.release();

        throw;
    }

    return loadedGrammar;
}

// ---------------------------------------------------------------------------
//  SGXMLScanner: Private helper methods
// ---------------------------------------------------------------------------
//  This method handles the common initialization, to avoid having to do
//  it redundantly in multiple constructors.
void SGXMLScanner::commonInit()
{
    //  Create the element state array
    fElemState = (unsigned int*) fMemoryManager->allocate
    (
        fElemStateSize * sizeof(unsigned int)
    ); //new unsigned int[fElemStateSize];

    //  And we need one for the raw attribute scan. This just stores key/
    //  value string pairs (prior to any processing.)
    fRawAttrList = new (fMemoryManager) RefVectorOf<KVStringPair>(32, true, fMemoryManager);
    fRawAttrColonList = (int*) fMemoryManager->allocate
    (
        fRawAttrColonListSize * sizeof(int)
    );

    //  Create the Validator and init them
    fSchemaValidator = new (fMemoryManager) SchemaValidator(0, fMemoryManager);
    initValidator(fSchemaValidator);

    // Create IdentityConstraint info
    fICHandler = new (fMemoryManager) IdentityConstraintHandler(this, fMemoryManager);

    //  Add the default entity entries for the character refs that must always
    //  be present.
    fEntityTable = new (fMemoryManager) ValueHashTableOf<XMLCh>(11, fMemoryManager);
    fEntityTable->put((void*) XMLUni::fgAmp, chAmpersand);
    fEntityTable->put((void*) XMLUni::fgLT, chOpenAngle);
    fEntityTable->put((void*) XMLUni::fgGT, chCloseAngle);
    fEntityTable->put((void*) XMLUni::fgQuot, chDoubleQuote);
    fEntityTable->put((void*) XMLUni::fgApos, chSingleQuote);
    fElemNonDeclPool = new (fMemoryManager) RefHash3KeysIdPool<SchemaElementDecl>(29, true, 128, fMemoryManager);
    fAttDefRegistry = new (fMemoryManager) RefHashTableOf<unsigned int>
    (
        131, false, new (fMemoryManager)HashPtr(), fMemoryManager
    );
    fUndeclaredAttrRegistryNS = new (fMemoryManager) RefHash2KeysTableOf<unsigned int>
    (
        7, false, fMemoryManager
    );
    fPSVIAttrList = new (fMemoryManager) PSVIAttributeList(fMemoryManager);

    if (fValidator)
    {
        if (!fValidator->handlesSchema())
            ThrowXMLwithMemMgr(RuntimeException, XMLExcepts::Gen_NoSchemaValidator, fMemoryManager);
    }
    else
    {
        fValidator = fSchemaValidator;
    }
}

void SGXMLScanner::cleanUp()
{
    fMemoryManager->deallocate(fElemState); //delete [] fElemState;
    delete fSchemaGrammar;
    delete fEntityTable;
    delete fRawAttrList;
    fMemoryManager->deallocate(fRawAttrColonList);
    delete fSchemaValidator;
    delete fICHandler;
    delete fElemNonDeclPool;
    delete fAttDefRegistry;
    delete fUndeclaredAttrRegistryNS;
    delete fPSVIAttrList;
    if (fPSVIElement)
        delete fPSVIElement;

    if (fErrorStack)
        delete fErrorStack;
}

void SGXMLScanner::resizeElemState() {

    unsigned int newSize = fElemStateSize * 2;
    unsigned int* newElemState = (unsigned int*) fMemoryManager->allocate
    (
        newSize * sizeof(unsigned int)
    ); //new unsigned int[newSize];

    // Copy the existing values
    unsigned int index = 0;
    for (; index < fElemStateSize; index++)
        newElemState[index] = fElemState[index];

    for (; index < newSize; index++)
        newElemState[index] = 0;

    // Delete the old array and udpate our members
    fMemoryManager->deallocate(fElemState); //delete [] fElemState;
    fElemState = newElemState;
    fElemStateSize = newSize;
}

void SGXMLScanner::resizeRawAttrColonList() {

    unsigned int newSize = fRawAttrColonListSize * 2;
    int* newRawAttrColonList = (int*) fMemoryManager->allocate
    (
        newSize * sizeof(int)
    ); //new int[newSize];

    // Copy the existing values
    unsigned int index = 0;
    for (; index < fRawAttrColonListSize; index++)
        newRawAttrColonList[index] = fRawAttrColonList[index];

    // Delete the old array and udpate our members
    fMemoryManager->deallocate(fRawAttrColonList); //delete [] fRawAttrColonList;
    fRawAttrColonList = newRawAttrColonList;
    fRawAttrColonListSize = newSize;
}

//  This method is called from scanStartTag() to build up the list of
//  XMLAttr objects that will be passed out in the start tag callout. We
//  get the key/value pairs from the raw scan of explicitly provided attrs,
//  which have not been normalized. And we get the element declaration from
//  which we will get any defaulted or fixed attribute defs and add those
//  in as well.
XMLSize_t
SGXMLScanner::buildAttList(const  RefVectorOf<KVStringPair>&  providedAttrs
                          , const XMLSize_t                   attCount
                          ,       XMLElementDecl*             elemDecl
                          ,       RefVectorOf<XMLAttr>&       toFill)
{
    //  Ask the element to clear the 'provided' flag on all of the att defs
    //  that it owns, and to return us a boolean indicating whether it has
    //  any defs.
    DatatypeValidator *currDV = 0;
    ComplexTypeInfo *currType = 0;

    if (fValidate)
    {
        currType = ((SchemaValidator*)fValidator)->getCurrentTypeInfo();
        if (!currType) {
            currDV = ((SchemaValidator*)fValidator)->getCurrentDatatypeValidator();
        }
    }

    const bool hasDefs = (currType && fValidate)
            ? currType->hasAttDefs()
            : elemDecl->hasAttDefs();

    fElemCount++;

    //  If there are no expliclitily provided attributes and there are no
    //  defined attributes for the element, the we don't have anything to do.
    //  So just return zero in this case.
    if (!hasDefs && !attCount)
        return 0;

    // Keep up with how many attrs we end up with total
    XMLSize_t retCount = 0;

    //  And get the current size of the output vector. This lets us use
    //  existing elements until we fill it, then start adding new ones.
    const XMLSize_t curAttListSize = toFill.size();

    //  We need a buffer into which raw scanned attribute values will be
    //  normalized.
    XMLBufBid bbNormal(&fBufMgr);
    XMLBuffer& normBuf = bbNormal.getBuffer();

    //  Loop through our explicitly provided attributes, which are in the raw
    //  scanned form, and build up XMLAttr objects.
    XMLSize_t index;
    for (index = 0; index < attCount; index++)
    {
        PSVIItem::VALIDITY_STATE attrValid = PSVIItem::VALIDITY_VALID;
        PSVIItem::ASSESSMENT_TYPE attrAssessed = PSVIItem::VALIDATION_FULL;
        const KVStringPair* curPair = providedAttrs.elementAt(index);

        //  We have to split the name into its prefix and name parts. Then
        //  we map the prefix to its URI.
        const XMLCh* const namePtr = curPair->getKey();
        ArrayJanitor <XMLCh> janName(0);

        // use a stack-based buffer when possible.
        XMLCh tempBuffer[100];

        const int colonInd = fRawAttrColonList[index];
        const XMLCh* prefPtr = XMLUni::fgZeroLenString;
        const XMLCh* suffPtr = XMLUni::fgZeroLenString;
        if (colonInd != -1)
        {
            // We have to split the string, so make a copy.
            if (XMLString::stringLen(namePtr) < sizeof(tempBuffer) / sizeof(tempBuffer[0]))
            {
                XMLString::copyString(tempBuffer, namePtr);
                tempBuffer[colonInd] = chNull;
                prefPtr = tempBuffer;
            }
            else
            {
                janName.reset(XMLString::replicate(namePtr, fMemoryManager), fMemoryManager);
                janName[colonInd] = chNull;
                prefPtr = janName.get();
            }

            suffPtr = namePtr + colonInd + 1;
        }
        else
        {
            // No colon, so we just have a name with no prefix
            suffPtr = namePtr;
        }

        //  Map the prefix to a URI id. We tell him that we are mapping an
        //  attr prefix, so any xmlns attrs at this level will not affect it.
        const unsigned int uriId = resolvePrefix(prefPtr, ElemStack::Mode_Attribute);

        //  If the uri comes back as the xmlns or xml URI or its just a name
        //  and that name is 'xmlns', then we handle it specially. So set a