XMLScanner.cpp

//
//  This method handles the common initialization, to avoid having to do
//  it redundantly in multiple constructors.
//
void XMLScanner::commonInit()
{

    //
    //  We have to do a little init that involves statics, so we have to
    //  use the mutex to protect it.
    //
    {
        XMLMutexLock lockInit(&gScannerMutex());

        // If we haven't loaded our message yet, then do that
        if (!gMsgLoader)
        {
            gMsgLoader = XMLPlatformUtils::loadMsgSet(XMLUni::fgXMLErrDomain);
            if (!gMsgLoader)
                XMLPlatformUtils::panic(XMLPlatformUtils::Panic_CantLoadMsgDomain);

            // Register this object to be cleaned up at termination
			cleanupMsgLoader.registerCleanup(reinitMsgLoader);
        }

        // And assign ourselves the next available scanner id
        fScannerId = ++gScannerId;
    }

    //
    //  Create the element state array
    //
    fElemState = new unsigned int[fElemStateSize];

    //
    //  Create the attribute list, which is used to store attribute values
    //  during start tag processing. Give it a reasonable initial size that
    //  will serve for most folks, though it will grow as required.
    //
    fAttrList = new RefVectorOf<XMLAttr>(32);

    //
    //  And we need one for the raw attribute scan. This just stores key/
    //  value string pairs (prior to any processing.)
    //
    fRawAttrList = new RefVectorOf<KVStringPair>(32);

    //
    //  Create the id ref list. This is used to enforce XML 1.0 ID ref
    //  semantics, i.e. all id refs must refer to elements that exist
    //
    fIDRefList = new RefHashTableOf<XMLRefInfo>(109);

    //  Create the GrammarResolver
    fGrammarResolver = new GrammarResolver();

    //  Create the Validator and init them
    fDTDValidator = new DTDValidator();
    initValidator(fDTDValidator);
    fSchemaValidator = new SchemaValidator();
    initValidator(fSchemaValidator);

    // Create IdentityConstraint info
    fMatcherStack = new XPathMatcherStack();
    fValueStoreCache = new ValueStoreCache();
    fFieldActivator = new FieldActivator(fValueStoreCache, fMatcherStack);
    fValueStoreCache->setScanner(this);
}



void XMLScanner::initValidator(XMLValidator* theValidator) {
    //
    //  Tell the validator about the stuff it needs to know in order to
    //  do its work.
    //
    theValidator->setScannerInfo(this, &fReaderMgr, &fBufMgr);
    theValidator->setErrorReporter(fErrorReporter);

    if (theValidator->handlesSchema()) {

        ((SchemaValidator*) theValidator)->setGrammarResolver(fGrammarResolver);
        ((SchemaValidator*) theValidator)->setExitOnFirstFatal(fExitOnFirstFatal);
    }
}

void XMLScanner::resetURIStringPool() {
    fURIStringPool->flushAll();

    fEmptyNamespaceId   = fURIStringPool->addOrFind(XMLUni::fgZeroLenString);
    fUnknownNamespaceId = fURIStringPool->addOrFind(XMLUni::fgUnknownURIName);
    fXMLNamespaceId     = fURIStringPool->addOrFind(XMLUni::fgXMLURIName);
    fXMLNSNamespaceId   = fURIStringPool->addOrFind(XMLUni::fgXMLNSURIName);
    fSchemaNamespaceId  = fURIStringPool->addOrFind(SchemaSymbols::fgURI_XSI);
}

// ---------------------------------------------------------------------------
//  XMLScanner: Error emitting methods
// ---------------------------------------------------------------------------

//
//  These methods are called whenever the scanner wants to emit an error.
//  It handles getting the message loaded, doing token replacement, etc...
//  and then calling the error handler, if its installed.
//
void XMLScanner::emitError(const XMLErrs::Codes toEmit)
{
    // Bump the error count if it is not a warning
    if (XMLErrs::errorType(toEmit) != XMLErrorReporter::ErrType_Warning)
        incrementErrorCount();

    if (fErrorReporter)
    {
        // Load the message into a local for display
        const unsigned int msgSize = 1023;
        XMLCh errText[msgSize + 1];

        // Lock the mutex and load the text
        {
            XMLMutexLock lockInit(&gScannerMutex());
            if (!gMsgLoader->loadMsg(toEmit, errText, msgSize))
            {
                // <TBD> Probably should load a default msg here
            }
        }

        //
        //  Create a LastExtEntityInfo structure and get the reader manager
        //  to fill it in for us. This will give us the information about
        //  the last reader on the stack that was an external entity of some
        //  sort (i.e. it will ignore internal entities.
        //
        ReaderMgr::LastExtEntityInfo lastInfo;
        fReaderMgr.getLastExtEntityInfo(lastInfo);

        fErrorReporter->error
        (
            toEmit
            , XMLUni::fgXMLErrDomain
            , XMLErrs::errorType(toEmit)
            , errText
            , lastInfo.systemId
            , lastInfo.publicId
            , lastInfo.lineNumber
            , lastInfo.colNumber
        );
    }

    // Bail out if its fatal an we are to give up on the first fatal error
    if (XMLErrs::isFatal(toEmit) && fExitOnFirstFatal && !fInException)
        throw toEmit;
}

void XMLScanner::emitError( const   XMLErrs::Codes    toEmit
                            , const XMLCh* const        text1
                            , const XMLCh* const        text2
                            , const XMLCh* const        text3
                            , const XMLCh* const        text4)
{
    // Bump the error count if it is not a warning
    if (XMLErrs::errorType(toEmit) != XMLErrorReporter::ErrType_Warning)
        incrementErrorCount();

    if (fErrorReporter)
    {
        //
        //  Load the message into alocal and replace any tokens found in
        //  the text.
        //
        const unsigned int maxChars = 2047;
        XMLCh errText[maxChars + 1];

        // Lock the mutex and load the text
        {
            XMLMutexLock lockInit(&gScannerMutex());
            if (!gMsgLoader->loadMsg(toEmit, errText, maxChars, text1, text2, text3, text4))
            {
                // <TBD> Should probably load a default message here
            }
        }

        //
        //  Create a LastExtEntityInfo structure and get the reader manager
        //  to fill it in for us. This will give us the information about
        //  the last reader on the stack that was an external entity of some
        //  sort (i.e. it will ignore internal entities.
        //
        ReaderMgr::LastExtEntityInfo lastInfo;
        fReaderMgr.getLastExtEntityInfo(lastInfo);

        fErrorReporter->error
        (
            toEmit
            , XMLUni::fgXMLErrDomain
            , XMLErrs::errorType(toEmit)
            , errText
            , lastInfo.systemId
            , lastInfo.publicId
            , lastInfo.lineNumber
            , lastInfo.colNumber
        );
    }

    // Bail out if its fatal an we are to give up on the first fatal error
    if (XMLErrs::isFatal(toEmit) && fExitOnFirstFatal && !fInException)
        throw toEmit;
}

void XMLScanner::emitError( const   XMLErrs::Codes    toEmit
                            , const char* const         text1
                            , const char* const         text2
                            , const char* const         text3
                            , const char* const         text4)
{
    // Bump the error count if it is not a warning
    if (XMLErrs::errorType(toEmit) != XMLErrorReporter::ErrType_Warning)
        incrementErrorCount();

    if (fErrorReporter)
    {
        //
        //  Load the message into alocal and replace any tokens found in
        //  the text.
        //
        const unsigned int maxChars = 2047;
        XMLCh errText[maxChars + 1];

        // Lock the mutex and load the text
        {
            XMLMutexLock lockInit(&gScannerMutex());
            if (!gMsgLoader->loadMsg(toEmit, errText, maxChars, text1, text2, text3, text4))
            {
                // <TBD> Should probably load a default message here
            }
        }

        //
        //  Create a LastExtEntityInfo structure and get the reader manager
        //  to fill it in for us. This will give us the information about
        //  the last reader on the stack that was an external entity of some
        //  sort (i.e. it will ignore internal entities.
        //
        ReaderMgr::LastExtEntityInfo lastInfo;
        fReaderMgr.getLastExtEntityInfo(lastInfo);

        fErrorReporter->error
        (
            toEmit
            , XMLUni::fgXMLErrDomain
            , XMLErrs::errorType(toEmit)
            , errText
            , lastInfo.systemId
            , lastInfo.publicId
            , lastInfo.lineNumber
            , lastInfo.colNumber
        );
    }

    // Bail out if its fatal an we are to give up on the first fatal error
    if (XMLErrs::isFatal(toEmit) && fExitOnFirstFatal && !fInException)
        throw toEmit;
}


// ---------------------------------------------------------------------------
//  XMLScanner: Getter methods
// ---------------------------------------------------------------------------

//
//  This method allows the caller to query the current location of the scanner.
//  It will return the sys/public ids of the current entity, and the line/col
//  position within it.
//
//  NOTE: This API returns the location with the last external file. So if its
//  currently scanning an entity, the position returned will be the end of
//  the entity reference in the file that had the reference.
//
bool
XMLScanner::getLastExtLocation(         XMLCh* const    sysIdToFill
                                , const unsigned int    maxSysIdChars
                                ,       XMLCh* const    pubIdToFill
                                , const unsigned int    maxPubIdChars
                                ,       XMLSSize_t&     lineToFill
                                ,       XMLSSize_t&     colToFill)
{
    // Create a local info object and get it filled in by the reader manager
    ReaderMgr::LastExtEntityInfo lastInfo;
    fReaderMgr.getLastExtEntityInfo(lastInfo);

    // Fill in the line and column number
    lineToFill = lastInfo.lineNumber;
    colToFill = lastInfo.colNumber;

    // And copy over as much of the ids as will fit
    sysIdToFill[0] = 0;
    if (lastInfo.systemId)
    {
        if (XMLString::stringLen(lastInfo.systemId) > maxSysIdChars)
            return false;
        XMLString::copyString(sysIdToFill, lastInfo.systemId);
    }

    pubIdToFill[0] = 0;
    if (lastInfo.publicId)
    {
        if (XMLString::stringLen(lastInfo.publicId) > maxPubIdChars)
            return false;
        XMLString::copyString(pubIdToFill, lastInfo.publicId);
    }
    return true;
}


// ---------------------------------------------------------------------------
//  XMLScanner: 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.
//
unsigned int
XMLScanner::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.
    //
    unsigned int attCount = 0;
    unsigned int 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 (XMLReader::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 (!XMLReader::isSpecialStartTagChar(nextCh))
        {
            //
            //  Assume its going to be an attribute, so get a name from
            //  the input.
            //
            if (!fReaderMgr.getName(fAttNameBuf))
            {
                emitError(XMLErrs::ExpectedAttrName);
                fReaderMgr.skipPastChar(chCloseAngle);
                return attCount;
            }

            // 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)
                      ||  XMLReader::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(fAttNameBuf.getRawBuffer(), 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)
                ||  XMLReader::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;
                }
            }

            //
            //  Make sure that the name is basically well formed for namespace
            //  enabled rules. It either has no colons, or it has one which
            //  is neither the first or last char.
            //
            const int colonFirst = XMLString::indexOf(fAttNameBuf.getRawBuffer(), chColon);
            if (colonFirst != -1)
            {
                const int colonLast = XMLString::lastIndexOf(fAttNameBuf.getRawBuffer(), chColon);

                if (colonFirst != colonLast)
                {
                    emitError(XMLErrs::TooManyColonsInName);
                    continue;
                }
                 else if ((colonFirst == 0)
                      ||  (colonLast == (int)fAttNameBuf.getLen() - 1))
                {
                    emitError(XMLErrs::InvalidColonPos);
                    continue;
                }
            }


            //
            //  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 KVStringPair
                (
                    fAttNameBuf.getRawBuffer()
                    , fAttValueBuf.getRawBuffer()
                );
                toFill.addElement(curPair);
            }
             else
            {
                curPair = toFill.elementAt(attCount);
                curPair->set(fAttNameBuf.getRawBuffer(), fAttValueBuf.getRawBuffer());
            }

            // 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)
            ThrowXML(UnexpectedEOFException, XMLExcepts::Gen_UnexpectedEOF);

        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 XMLScanner::scanContent(const bool extEntity)
{
    //
    //  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.
                //
                unsigned int 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 :
                        if (fDoNamespaces)
                            scanStartTagNS(gotData);
                        else
                            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 XMLScanner::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);
        ThrowXML(RuntimeException, XMLExcepts::Scan_UnbalancedStartEnd);
    }

    // After the </ is the element QName, so get a name from the input
    XMLBufBid bbQName(&fBufMgr);
    XMLBuffer& qnameBuf = bbQName.getBuffer();
    if (!fReaderMgr.getName(qnameBuf))
    {
        // It failed so we can't really do anything with it
        emitError(XMLErrs::ExpectedElementName);
        fReaderMgr.skipPastChar(chCloseAngle);
        return;
    }

    unsigned int uriId = fEmptyNamespaceId;
    XMLBufBid bbName(&fBufMgr);
    XMLBufBid bbPrefix(&fBufMgr);
    if (fDoNamespaces)
    {
        uriId = resolveQName
        (
            qnameBuf.getRawBuffer()
            , bbName.getBuffer()
            , bbPrefix.getBuffer()
            , ElemStack::Mode_Element
        );
    }

    //
    //  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.
    //
    //  NOTE: We CANNOT do this until we've resolved the element name because
    //  the element stack top contains the prefix to URI mappings for this
    //  element.
    //
    unsigned int topUri = fElemStack.getCurrentURI();
    const ElemStack::StackElem* topElem = fElemStack.popTop();

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

    // Make sure that its the end of the element that we expect
    XMLElementDecl* tempElement = topElem->fThisElement;
    if (fDoNamespaces && fGrammarType == Grammar::SchemaGrammarType) {
        if ((topUri != uriId) || (XMLString::compareString(tempElement->getBaseName(), bbName.getRawBuffer())))
        {
            emitError
            (
                XMLErrs::ExpectedEndOfTagX
                , topElem->fThisElement->getFullName()
            );
        }
    }
    else {
        if (XMLString::compareString(tempElement->getFullName(), qnameBuf.getRawBuffer()))
        {
            emitError
            (
                XMLErrs::ExpectedEndOfTagX
                , topElem->fThisElement->getFullName()
            );
        }
    }


    // 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 validation is enabled, then lets pass him the list of children and
    //  this element and let him validate it.
    //
    if (fValidate)
    {
        int res = fValidator->checkContent
        (
            topElem->fThisElement
            , topElem->fChildren
            , topElem->fChildCount
        );

        if (res >= 0)
        {
            //
            //  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 ((unsigned int)res >= topElem->fChildCount)
            {
                fValidator->emitError
                (
                    XMLValid::NotEnoughElemsForCM
                    , topElem->fThisElement->getFormattedContentModel()
                );
            }
             else
            {
                fValidator->emitError
                (
                    XMLValid::ElementNotValidForContent
                    , topElem->fChildren[res]->getRawName()
                    , topElem->fThisElement->getFormattedContentModel()
                );
            }
        }

        // reset xsi:type ComplexTypeInfo
        if (fGrammarType == Grammar::SchemaGrammarType) {
            ((SchemaElementDecl*)topElem->fThisElement)->setXsiComplexTypeInfo(0);

            // call matchers and de-activate context
            int oldCount = fMatcherStack->getMatcherCount();

            if (oldCount ||
                ((SchemaElementDecl*)topElem->fThisElement)->getIdentityConstraintCount()) {

                for (int i = oldCount - 1; i >= 0; i--) {

                    XPathMatcher* matcher = fMatcherStack->getMatcherAt(i);
                    matcher->endElement(*(topElem->fThisElement));
                }

                if (fMatcherStack->size() > 0) {
                    fMatcherStack->popContext();
                }

                // handle everything *but* keyref's.
                int newCount = fMatcherStack->getMatcherCount();

                for (int j = oldCount - 1; j >= newCount; j--) {

                    XPathMatcher* matcher = fMatcherStack->getMatcherAt(j);
                    IdentityConstraint* ic = matcher->getIdentityConstraint();

                    if (ic  && (ic->getType() != IdentityConstraint::KEYREF)) {

                        matcher->endDocumentFragment();
                        fValueStoreCache->transplant(ic);
                    }
                    else if (!ic) {
                        matcher->endDocumentFragment();
                    }
                }

                // now handle keyref's...
                for (int k = oldCount - 1; k >= newCount; k--) {

                    XPathMatcher* matcher = fMatcherStack->getMatcherAt(k);
                    IdentityConstraint* ic = matcher->getIdentityConstraint();

                    if (ic && (ic->getType() == IdentityConstraint::KEYREF)) {

                        ValueStore* values = fValueStoreCache->getValueStoreFor(ic);

                        if (values) { // nothing to do if nothing matched!
                            values->endDcocumentFragment(fValueStoreCache);
                        }

                        matcher->endDocumentFragment();
                    }
                }

                fValueStoreCache->endElement();
            }
        }
    }

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

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

    if (gotData) {
        if (fDoNamespaces) {
            // Restore the grammar
            fGrammar = fElemStack.getCurrentGrammar();
            fGrammarType = fGrammar->getGrammarType();
            if (fGrammarType == Grammar::SchemaGrammarType && !fValidator->handlesSchema()) {
                if (fValidatorFromUser)
                    ThrowXML(RuntimeException, XMLExcepts::Gen_NoSchemaValidator);
                else {
                    fValidator = fSchemaValidator;
                }
            }
            else if (fGrammarType == Grammar::DTDGrammarType && !fValidator->handlesDTD()) {
                if (fValidatorFromUser)
                    ThrowXML(RuntimeException, XMLExcepts::Gen_NoDTDValidator);
                else {
                    fValidator = fDTDValidator;
                }
            }

            fValidator->setGrammar(fGrammar);
        }

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


//
//  This method is called after the end of the root element, to handle
//  any miscellaneous stuff hanging around.
//
void XMLScanner::scanMiscellaneous()
{
    // Get a buffer for this work
    XMLBufBid bbCData(&fBufMgr);

    while (true)
    {
        try
        {
            const XMLCh nextCh = fReaderMgr.peekNextChar();

            // Watch for end of file and break out
            if (!nextCh)
                break;

            if (nextCh == chOpenAngle)
            {
                if (checkXMLDecl(true))
                {
                    // Can't have an XML decl here
                    emitError(XMLErrs::NotValidAfterContent);
                    fReaderMgr.skipPastChar(chCloseAngle);
                }
                 else if (fReaderMgr.skippedString(XMLUni::fgPIString))
                {
                    scanPI();
                }