DGXMLScanner.cpp

                if (rootDecl)
                    ((DTDGrammar*)fGrammar)->setRootElemId(rootDecl->getId());
                else {
                    rootDecl = new (fGrammarPoolMemoryManager) DTDElementDecl
                    (
                        bbRootName.getRawBuffer()
                        , fEmptyNamespaceId
                        , DTDElementDecl::Any
                        , fGrammarPoolMemoryManager
                    );
                    rootDecl->setCreateReason(DTDElementDecl::AsRootElem);
                    rootDecl->setExternalElemDeclaration(true);
                    ((DTDGrammar*)fGrammar)->setRootElemId(fGrammar->putElemDecl(rootDecl));
                }
                *********/

                return;
            }
        }

        if (fLoadExternalDTD || fValidate)
        {
            // And now create a reader to read this entity
            InputSource* srcUsed;
            XMLReader* reader = fReaderMgr.createReader
            (
                sysId
                , pubId
                , false
                , XMLReader::RefFrom_NonLiteral
                , XMLReader::Type_General
                , XMLReader::Source_External
                , srcUsed
                , fCalculateSrcOfs
            );

            // Put a janitor on the input source
            Janitor<InputSource> janSrc(srcUsed);

            //  If it failed then throw an exception
            if (!reader)
                ThrowXMLwithMemMgr1(RuntimeException, XMLExcepts::Gen_CouldNotOpenDTD, srcUsed->getSystemId(), fMemoryManager);

            if (fToCacheGrammar) {

                unsigned int stringId = fGrammarResolver->getStringPool()->addOrFind(srcUsed->getSystemId());
                const XMLCh* sysIdStr = fGrammarResolver->getStringPool()->getValueForId(stringId);

                fGrammarResolver->orphanGrammar(XMLUni::fgDTDEntityString);
                ((XMLDTDDescription*) (fGrammar->getGrammarDescription()))->setRootName(sysIdStr);
                fGrammarResolver->putGrammar(fGrammar);
            }

            //  In order to make the processing work consistently, we have to
            //  make this look like an external entity. So create an entity
            //  decl and fill it in and push it with the reader, as happens
            //  with an external entity. Put a janitor on it to insure it gets
            //  cleaned up. The reader manager does not adopt them.
            const XMLCh gDTDStr[] = { chLatin_D, chLatin_T, chLatin_D , chNull };
            DTDEntityDecl* declDTD = new (fGrammarPoolMemoryManager) DTDEntityDecl(gDTDStr, false, fGrammarPoolMemoryManager);
            declDTD->setSystemId(sysId);
            Janitor<DTDEntityDecl> janDecl(declDTD);

            // Mark this one as a throw at end
            reader->setThrowAtEnd(true);

            // And push it onto the stack, with its pseudo name
            fReaderMgr.pushReader(reader, declDTD);

            // Tell it its not in an include section
            dtdScanner.scanExtSubsetDecl(false, true);
        }
    }
}

bool DGXMLScanner::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;

    //  Get the QName. In this case, we are not doing namespaces, so we just
    //  use it as is and don't have to break it into parts.
    if (!fReaderMgr.getName(fQNameBuf))
    {
        emitError(XMLErrs::ExpectedElementName);
        fReaderMgr.skipToChar(chOpenAngle);
        return false;
    }

    // Assume it won't be an empty tag
    bool isEmpty = false;

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

    //  Lets try to look up the element in the validator's element decl pool
    //  We can pass bogus values for the URI id and the base name. We know that
    //  this can only be called if we are doing a DTD style validator and that
    //  he will only look at the QName.
    //
    //  We *do not* tell him to fault in a decl if he does not find one - NG.
    bool wasAdded = false;
    const XMLCh* qnameRawBuf = fQNameBuf.getRawBuffer(); 

    XMLElementDecl* elemDecl = fGrammar->getElemDecl
    (
        fEmptyNamespaceId
        , 0
        , qnameRawBuf
        , Grammar::TOP_LEVEL_SCOPE
    );
    // look in the undeclared pool:
    if(!elemDecl) 
    {
        elemDecl = fDTDElemNonDeclPool->getByKey(qnameRawBuf);
    }
    if(!elemDecl) 
    {
        wasAdded = true;
        elemDecl = new (fMemoryManager) DTDElementDecl 
        (
            qnameRawBuf
            , fEmptyNamespaceId
            , DTDElementDecl::Any
            , fMemoryManager
        );
        elemDecl->setId(fDTDElemNonDeclPool->put((DTDElementDecl*)elemDecl));
    }

    if (fValidate) {

        if (wasAdded)
        {
            // This is to tell the reuse Validator that this element was
            // faulted-in, was not an element in the validator pool originally
            elemDecl->setCreateReason(XMLElementDecl::JustFaultIn);

            fValidator->emitError
            (
                XMLValid::ElementNotDefined
                , qnameRawBuf
            );
        }
        // If its not marked declared, then emit an error
        else if (!elemDecl->isDeclared())
        {
            fValidator->emitError
            (
                XMLValid::ElementNotDefined
                , qnameRawBuf
            );
        }


        fValidator->validateElement(elemDecl);
    }

    // Expand the element stack and add the new element
    fElemStack.addLevel(elemDecl, fReaderMgr.getCurrentReaderNum());

    //  If this is the first element and we are validating, check the root
    //  element.
    if (isRoot)
    {
        fRootGrammar = fGrammar;

        if (fValidate)
        {
            //  If a DocType exists, then check if it matches the root name there.
            if (fRootElemName && !XMLString::equals(qnameRawBuf, fRootElemName))
                fValidator->emitError(XMLValid::RootElemNotLikeDocType);

            //  Some validators may also want to check the root, call the
            //  XMLValidator::checkRootElement
            if (fValidatorFromUser && !fValidator->checkRootElement(elemDecl->getId()))
                fValidator->emitError(XMLValid::RootElemNotLikeDocType);
        }
    }
    else if (fValidate)
    {
        //  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);
    }

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

    //  We loop until we either see a /> or >, handling attribute/value
    //  pairs until we get there.
    unsigned int    attCount = 0;
    unsigned int    curAttListSize = fAttrList->size();
    wasAdded = false;
    // clear the map used to detect duplicate attributes
    fUndeclaredAttrRegistry->removeAll();

    fElemCount++;

    while (true)
    {
        // And get the next non-space character
        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 peek another char
                    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.
            if (!fReaderMgr.getName(fAttNameBuf))
            {
                emitError(XMLErrs::ExpectedAttrName);
                fReaderMgr.skipPastChar(chCloseAngle);
                return false;
            }

            // 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, elemDecl->getFullName());
                    return false;
                }
                else
                {
                    // Something went really wrong
                    return false;
                }
            }

            //  See if this attribute is declared for this element. If we are
            //  not validating of course it will not be at first, but we will
            //  fault it into the pool (to avoid lots of redundant errors.)
            XMLAttDef* attDef = ((DTDElementDecl *)elemDecl)->getAttDef ( fAttNameBuf.getRawBuffer());
            XMLCh * namePtr = fAttNameBuf.getRawBuffer();

            //  Add this attribute to the attribute list that we use to
            //  pass them to the handler. We reuse its existing elements
            //  but expand it as required.
            // Note that we want to this first since this will
            // make a copy of the namePtr; we can then make use of
            // that copy in the hashtable lookup that checks
            // for duplicates.  This will mean we may have to update
            // the type of the XMLAttr later.
            XMLAttr* curAtt;
            if (attCount >= curAttListSize)
            {
                if (fDoNamespaces) {
                    curAtt = new (fMemoryManager) XMLAttr
                    (
                        fEmptyNamespaceId
                        , fAttNameBuf.getRawBuffer()
                        , XMLUni::fgZeroLenString
                        , (attDef)?attDef->getType():XMLAttDef::CData
                        , true
                        , fMemoryManager
                    );
                }
                else
                {
                    curAtt = new (fMemoryManager) XMLAttr
                    (
                        -1
                        , fAttNameBuf.getRawBuffer()
                        , XMLUni::fgZeroLenString
                        , XMLUni::fgZeroLenString
                        , (attDef)?attDef->getType():XMLAttDef::CData
                        , true
                        , fMemoryManager
                    );
                }
                fAttrList->addElement(curAtt);
            }
            else
            {
                curAtt = fAttrList->elementAt(attCount);

                if (fDoNamespaces)
                {
                    curAtt->set
                    (
                        fEmptyNamespaceId
                        , fAttNameBuf.getRawBuffer()
                        , XMLUni::fgZeroLenString
                        , (attDef)?attDef->getType():XMLAttDef::CData
                    );
                }
                else
                {
                    curAtt->set
                    (
                        -1
                        , fAttNameBuf.getRawBuffer()
                        , XMLUni::fgZeroLenString
                        , XMLUni::fgZeroLenString
                        , (attDef)?attDef->getType():XMLAttDef::CData
                    );
                }
                curAtt->setSpecified(true);
            }
            // reset namePtr so it refers to newly-allocated memory
            namePtr = (XMLCh *)curAtt->getName();

            // now need to prepare for duplicate detection
            if(attDef)
            {
                unsigned int *curCountPtr = fAttDefRegistry->get(attDef);
                if(!curCountPtr)
                {
                    curCountPtr = getNewUIntPtr();
                    *curCountPtr = fElemCount;
                    fAttDefRegistry->put(attDef, curCountPtr);
                }
                else if(*curCountPtr < fElemCount)
                    *curCountPtr = fElemCount;
                else
                {
                    emitError
                    ( 
                        XMLErrs::AttrAlreadyUsedInSTag
                        , attDef->getFullName()
                        , elemDecl->getFullName()
                    );
                }
            }
            else
            {
                unsigned int *curCountPtr = fUndeclaredAttrRegistry->get(namePtr);
                if(!curCountPtr)
                {
                    curCountPtr = getNewUIntPtr();
                     *curCountPtr = fElemCount;
                    fUndeclaredAttrRegistry->put((void *)namePtr, curCountPtr);
                }
                else if(*curCountPtr < fElemCount)
                    *curCountPtr = fElemCount;
                else
                {
                    emitError
                    ( 
                        XMLErrs::AttrAlreadyUsedInSTag
                        , namePtr
                        , elemDecl->getFullName()
                    );
                }
            }
            if (fValidate)
            {
                if (!attDef)
                {

                    fValidator->emitError
                    (
                        XMLValid::AttNotDefinedForElement
                        , fAttNameBuf.getRawBuffer()
                        , qnameRawBuf
                    );
                }
            }

            //  Skip any whitespace before the value and then scan the att
            //  value. This will come back normalized with entity refs and
            //  char refs expanded.
            fReaderMgr.skipPastSpaces();
            if (!scanAttValue(attDef, 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)
                ||  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, elemDecl->getFullName());
                    return false;
                }
                else
                {
                    // Something went really wrong
                    return false;
                }
            }
            // must set the newly-minted value on the XMLAttr:
            curAtt->setValue(fAttValueBuf.getRawBuffer());

            //  Now that its all stretched out, lets look at its type and
            //  determine if it has a valid value. It will output any needed
            //  errors, but we just keep going. We only need to do this if
            //  we are validating.
            if (attDef)
            {
                // Let the validator pass judgement on the attribute value
                if (fValidate)
                {
                    fValidator->validateAttrValue
                    (
                        attDef
                        , fAttValueBuf.getRawBuffer()
                        , false
                        , elemDecl
                    );
                }
            }

            if (fDoNamespaces)
            {
                //  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;
                    }
                }
            }

            attCount++;

            // And jump back to the top of the loop
            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, elemDecl->getFullName());
            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, elemDecl->getFullName());
            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;
        }
    }

    //  Make an initial pass through the list and find any xmlns attributes.
    if (fDoNamespaces && attCount)
      scanAttrListforNameSpaces(fAttrList, attCount, elemDecl);

    //  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(attCount, elemDecl, *fAttrList);

    //  If we have a document handler, then tell it about this start tag. We
    //  don't have any URI id to send along, so send fEmptyNamespaceId. We also do not send
    //  any prefix since its just one big name if we are not doing namespaces.
    unsigned int uriId = fEmptyNamespaceId;
    if (fDocHandler)
    {
        if (fDoNamespaces)
        {
            uriId = resolvePrefix
            (
                elemDecl->getElementName()->getPrefix()
                , ElemStack::Mode_Element
            );
        }

        fDocHandler->startElement
        (
            *elemDecl
            , uriId
            , (fDoNamespaces) ? elemDecl->getElementName()->getPrefix() : 0
            , *fAttrList
            , attCount
            , false
            , isRoot
        );
    }

    //  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)
    {
        // If validating, then insure that its legal to have no content
        if (fValidate)
        {
            const int res = fValidator->checkContent(elemDecl, 0, 0);
            if (res >= 0)
            {
                fValidator->emitError
                (
                    XMLValid::ElementNotValidForContent
                    , qnameRawBuf
                    , elemDecl->getFormattedContentModel()
                );
            }
        }

        // If we have a doc handler, tell it about the end tag
        if (fDocHandler)
        {
            fDocHandler->endElement
            (
                *elemDecl
                , uriId
                , isRoot
                , (fDoNamespaces) ? elemDecl->getElementName()->getPrefix()
                                  : XMLUni::fgZeroLenString
            );
            // pass back type name information
            fDocHandler->elementTypeInfo(0, 0); 
        }

        // Pop the element stack back off since it'll never be used now
        fElemStack.popTop();

        // If the elem stack is empty, then it was an empty root
        if (isRoot)
            gotData = false;
    }

    return true;
}

unsigned int
DGXMLScanner::resolveQName(const   XMLCh* const qName
                           ,       XMLBuffer&   prefixBuf
                           , const short        mode
                           ,       int&         prefixColonPos)
{
    //  Lets split out the qName into a URI and name buffer first. The URI
    //  can be empty.
    prefixColonPos = XMLString::indexOf(qName, chColon);
    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);

            return uriId;
        }
    }
}

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

    try
    {
        fGrammarResolver->cacheGrammarFromParse(false);
        fGrammarResolver->useCachedGrammarInParse(false);
        fRootGrammar = 0;

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

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

        if (grammarType == Grammar::DTDGrammarType) {
            loadedGrammar = loadDTDGrammar(src, toCache);
        }

        // Reset the reader manager to close all files, sockets, etc...
        fReaderMgr.reset();
    }
    //  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 fatal error' type exit, so reset and fall through
        fReaderMgr.reset();
    }
    catch(const XMLValid::Codes)
    {
        // This is a 'first fatal error' type exit, so reset and fall through
        fReaderMgr.reset();

    }
    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::DisplayErrorMessage
                    , excToCatch.getMessage()
                );
            else if (excToCatch.getErrorType() >= XMLErrorReporter::ErrType_Fatal)
                emitError
                (
                    XMLErrs::XMLException_Fatal
                    , excToCatch.getType()
                    , excToCatch.getMessage()
                );
            else
                emitError
                (
                    XMLErrs::XMLException_Error
                    , excToCatch.getType()
                    , excToCatch.getMessage()
                );
        }
        catch(const OutOfMemoryException&)
        {
            throw;
        }
        catch(...)
        {
            // Flush the reader manager and rethrow user's error
            fReaderMgr.reset();
            throw;
        }

        // If it returned, then reset the reader manager and fall through
        fReaderMgr.reset();
    }
    catch(const OutOfMemoryException&)
    {
        throw;
    }
    catch(...)
    {
        // Reset and rethrow
        fReaderMgr.reset();
        throw;
    }

    return loadedGrammar;
}

Grammar* DGXMLScanner::loadDTDGrammar(const InputSource& src,
                                      const bool toCache)
{
    // Reset the validators
    fDTDValidator->reset();
    if (fValidatorFromUser)
        fValidator->reset();

    fDTDGrammar = new (fGrammarPoolMemoryManager) DTDGrammar(fGrammarPoolMemoryManager);
    fGrammarResolver->putGrammar(fDTDGrammar);
    fGrammar = fDTDGrammar;
    fValidator->setGrammar(fGrammar);

    //  And for all installed handlers, send reset events. This gives them
    //  a chance to flush any cached data.
    if (fDocHandler)
        fDocHandler->resetDocument();
    if (fEntityHandler)
        fEntityHandler->resetEntities();
    if (fErrorReporter)
        fErrorReporter->resetErrors();

    // Clear out the id reference list
    resetValidationContext();

    if (toCache) {

        unsigned int sysId = fGrammarResolver->getStringPool()->addOrFind(src.getSystemId());
        const XMLCh* sysIdStr = fGrammarResolver->getStringPool()->getValueForId(sysId);

        fGrammarResolver->orphanGrammar(XMLUni::fgDTDEntityString);
        ((XMLDTDDescription*) (fGrammar->getGrammarDescription()))->setRootName(sysIdStr);
        fGrammarResolver->putGrammar(fGrammar);
    }

    //  Handle the creation of the XML reader object for this input source.
    //  This will provide us with transcoding and basic lexing services.
    XMLReader* newReader = fReaderMgr.createReader
    (
        src
        , false
        , XMLReader::RefFrom_NonLiteral
        , XMLReader::Type_General
        , XMLReader::Source_External
        , fCalculateSrcOfs
    );
    if (!newReader) {
        if (src.getIssueFatalErrorIfNotFound())
            ThrowXMLwithMemMgr1(RuntimeException, XMLExcepts::Scan_CouldNotOpenSource, src.getSystemId(), fMemoryManager);
        else
            ThrowXMLwithMemMgr1(RuntimeException, XMLExcepts::Scan_CouldNotOpenSource_Warning, src.getSystemId(), fMemoryManager);
    }

    //  In order to make the processing work consistently, we have to
    //  make this look like an external entity. So create an entity
    //  decl and fill it in and push it with the reader, as happens
    //  with an external entity. Put a janitor on it to insure it gets
    //  cleaned up. The reader manager does not adopt them.
    const XMLCh gDTDStr[] = { chLatin_D, chLatin_T, chLatin_D , chNull };
    DTDEntityDecl* declDTD = new (fGrammarPoolMemoryManager) DTDEntityDecl(gDTDStr, false, fGrammarPoolMemoryManager);
    declDTD->setSystemId(src.getSystemId());
    Janitor<DTDEntityDecl> janDecl(declDTD);

    // Mark this one as a throw at end
    newReader->setThrowAtEnd(true);

    // And push it onto the stack, with its pseudo name
    fReaderMgr.pushReader(newReader, declDTD);

    //  If we have a doc type handler and advanced callbacks are enabled,
    //  call the doctype event.
    if (fDocTypeHandler) {

        // Create a dummy root
        DTDElementDecl* rootDecl = new (fGrammarPoolMemoryManager) DTDElementDecl
        (
            gDTDStr
            , fEmptyNamespaceId
            , DTDElementDecl::Any
            , fGrammarPoolMemoryManager
        );
        rootDecl->setCreateReason(DTDElementDecl::AsRootElem);
        rootDecl->setExternalElemDeclaration(true);
        Janitor<DTDElementDecl> janSrc(rootDecl);

        fDocTypeHandler->doctypeDecl(*rootDecl, src.getPublicId(), src.getSystemId(), false, true);
    }

    // Create DTDScanner
    DTDScanner dtdScanner
    (
        (DTDGrammar*)fGrammar
        , fDocTypeHandler
        , fGrammarPoolMemoryManager
        , fMemoryManager
    );
    dtdScanner.setScannerInfo(this, &fReaderMgr, &fBufMgr);

    // Tell it its not in an include section
    dtdScanner.scanExtSubsetDecl(false, true);

    if (fValidate) {
        //  validate the DTD scan so far
        fValidator->preContentValidation(false, true);
    }

    if (toCache)
        fGrammarResolver->cacheGrammars();

    return fDTDGrammar;
}


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

    //  And we need one for the raw attribute scan. This just stores key/
    //  value string pairs (prior to any processing.)
    fAttrNSList = new (fMemoryManager) ValueVectorOf<XMLAttr*>(8, fMemoryManager);

    //  Create the Validator and init them
    fDTDValidator = new (fMemoryManager) DTDValidator();
    initValidator(fDTDValidator);
    fDTDElemNonDeclPool = new (fMemoryManager) NameIdPool<DTDElementDecl>(29, 128, fMemoryManager);
    fAttDefRegistry = new (fMemoryManager) RefHashTableOf<unsigned int>
    (
        131, false, new (fMemoryManager)HashPtr(), fMemoryManager
    );
    fUndeclaredAttrRegistry = new (fMemoryManager) RefHashTableOf<unsigned int>
    (
        7, false, new (fMemoryManager)HashXMLCh(), fMemoryManager
    );
}

void DGXMLScanner::cleanUp()
{
    delete fAttrNSList;
    delete fDTDValidator;
    delete fDTDElemNonDeclPool;
    delete fAttDefRegistry;
    delete fUndeclaredAttrRegistry;
}


//  This method is called from scanStartTagNS() 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.
unsigned int
DGXMLScanner::buildAttList(const unsigned int           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.
    const bool hasDefs = elemDecl->hasAttDefs();

    //  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
    unsigned int retCount = attCount;

    //  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 unsigned int curAttListSize = toFill.size();

    //  Ok, so lets get an enumerator for the attributes of this element
    //  and run through them for well formedness and validity checks. But
    //  make sure that we had any attributes before we do it, since the list
    //  would have have gotten faulted in anyway.
    if (hasDefs)
    {
        XMLAttDefList& attDefList = elemDecl->getAttDefList();
        for(unsigned int i=0; i<attDefList.getAttDefCount(); i++)
        {
            // Get the current att def, for convenience and its def type
            XMLAttDef& curDef = attDefList.getAttDef(i);

            unsigned int *attCountPtr = fAttDefRegistry->get(&curDef);
            if (!attCountPtr || *attCountPtr < fElemCount)
            { // did not occur
                const XMLAttDef::DefAttTypes defType = curDef.getDefaultType();

                if (fValidate)
                {
                    // If we are validating and its required, then an error
                    if (defType == XMLAttDef::Required)
                    {
                        fValidator->emitError
                        (
                            XMLValid::RequiredAttrNotProvided
                            , curDef.getFullName()
                        );
                    }
                    else if ((defType == XMLAttDef::Default) ||
		                       (defType == XMLAttDef::Fixed)  )
                    {
                        if (fStandalone && curDef.isExternal())
                        {
                            // XML 1.0 Section 2.9