Skip to content
Snippets Groups Projects
IGXMLScanner.cpp 102 KiB
Newer Older
Khaled Noaman's avatar
Khaled Noaman committed
    //  schema attributes.
    if (attCount)
      scanRawAttrListforNameSpaces(fRawAttrList, attCount);

    //  Also find any default or fixed xmlns attributes in DTD defined for
    //  this element.
    XMLElementDecl* elemDecl = 0;
    if (fGrammarType == Grammar::DTDGrammarType) {
        elemDecl = fGrammar->getElemDecl
        (
            fEmptyNamespaceId
            , 0
            , fQNameBuf.getRawBuffer()
            , Grammar::TOP_LEVEL_SCOPE
        );
        if (elemDecl) {
            if (elemDecl->hasAttDefs()) {
                XMLAttDefList& attDefList = elemDecl->getAttDefList();
                while (attDefList.hasMoreElements())
                {
                    // Get the current att def, for convenience and its def type
                    const XMLAttDef& curDef = attDefList.nextElement();
                    const XMLAttDef::DefAttTypes defType = curDef.getDefaultType();

                    // update the NSMap if there are any default/fixed xmlns attributes
                    if ((defType == XMLAttDef::Default)
                    ||  (defType == XMLAttDef::Fixed))
                    {
                        const XMLCh* rawPtr = curDef.getFullName();
                        if (!XMLString::compareNString(rawPtr, XMLUni::fgXMLNSColonString, 6)
                        ||  XMLString::equals(rawPtr, XMLUni::fgXMLNSString))
                            updateNSMap(rawPtr, curDef.getValue());
                    }
                }
            }
        }
    }

    //  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.
    int prefixColonPos = -1;
    const XMLCh* qnameRawBuf = fQNameBuf.getRawBuffer();
    unsigned int uriId = resolveQName
    (
        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);
        // 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.
    bool wasAdded = false;
    bool errorBeforeElementFound = false;
    bool laxBeforeElementFound = false;

Khaled Noaman's avatar
Khaled Noaman committed
    const XMLCh* nameRawBuf = &qnameRawBuf[prefixColonPos + 1];
    const XMLCh* original_uriStr = fGrammar->getTargetNamespace();
    unsigned orgGrammarUri = fURIStringPool->getId(original_uriStr);
Khaled Noaman's avatar
Khaled Noaman committed

    if (uriId != fEmptyNamespaceId) {

        // Check in current grammar before switching if necessary
        elemDecl = fGrammar->getElemDecl
        (
          uriId
          , nameRawBuf
          , qnameRawBuf
          , currentScope
        );

        if (!elemDecl && (orgGrammarUri != uriId)) {
Khaled Noaman's avatar
Khaled Noaman committed
            // not found, switch to the specified grammar
            const XMLCh* uriStr = getURIText(uriId);
            bool errorCondition = !switchGrammar(uriStr) && fValidate;
            if (errorCondition && !laxThisOne)
Khaled Noaman's avatar
Khaled Noaman committed
            {
                fValidator->emitError
                (
                    XMLValid::GrammarNotFound
                    ,uriStr
                );
Khaled Noaman's avatar
Khaled Noaman committed
            }
            else if(errorCondition)
                laxBeforeElementFound = true;
Khaled Noaman's avatar
Khaled Noaman committed

            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) {
                // 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) {
Khaled Noaman's avatar
Khaled Noaman committed
                    fValidator->emitError
                    (
                        XMLValid::ElementNotUnQualified
                        , elemDecl->getFullName()
                    );
Khaled Noaman's avatar
Khaled Noaman committed
                }
Khaled Noaman's avatar
Khaled Noaman committed
            }
        }

        if (!elemDecl) {
            // still not found, fault this in and issue error later
            // switch back to original grammar first
            switchGrammar(original_uriStr);
Khaled Noaman's avatar
Khaled Noaman committed
            elemDecl = fGrammar->putElemDecl(uriId
                        , nameRawBuf
                        , fPrefixBuf.getRawBuffer()
                        , qnameRawBuf
                        , currentScope
                        , true);
            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 && orgGrammarUri != fEmptyNamespaceId) {
            //not found, switch grammar and try globalNS
            bool errorCondition = !switchGrammar(XMLUni::fgZeroLenString) && fValidate;
            if (errorCondition && !laxThisOne)
Khaled Noaman's avatar
Khaled Noaman committed
            {
                fValidator->emitError
                (
                    XMLValid::GrammarNotFound
                  , XMLUni::fgZeroLenString
                );
Khaled Noaman's avatar
Khaled Noaman committed
            }
            else if(errorCondition)
                laxBeforeElementFound = true;
Khaled Noaman's avatar
Khaled Noaman committed

            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 && orgGrammarUri != fEmptyNamespaceId) {
                // still Not found in specified uri
                // go to original Grammar again to see if element needs to be fully qualified.
                const XMLCh* uriStr = getURIText(orgGrammarUri);
                bool errorCondition = !switchGrammar(original_uriStr) && fValidate;
                if (errorCondition && !laxThisOne)

                if (errorCondition && !laxThisOne)
Khaled Noaman's avatar
Khaled Noaman committed
                {
                    fValidator->emitError
                    (
                        XMLValid::GrammarNotFound
Khaled Noaman's avatar
Khaled Noaman committed
                    );
Khaled Noaman's avatar
Khaled Noaman committed
                }
                else if(errorCondition)
                    laxBeforeElementFound = true;
Khaled Noaman's avatar
Khaled Noaman committed

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

                if (elemDecl && elemDecl->getCreateReason() != XMLElementDecl::JustFaultIn && fValidate) {
                    fValidator->emitError
                    (
                        XMLValid::ElementNotQualified
                        , elemDecl->getFullName()
                    );
Khaled Noaman's avatar
Khaled Noaman committed
                }
            }
        }

        if (!elemDecl) {
            // still not found, fault this in and issue error later
            // switch back to original grammar first
            switchGrammar(original_uriStr);
Khaled Noaman's avatar
Khaled Noaman committed
            elemDecl = fGrammar->putElemDecl(uriId
                        , nameRawBuf
                        , fPrefixBuf.getRawBuffer()
                        , qnameRawBuf
                        , currentScope
                        , true);
            wasAdded = true;
        }
    }

    //  We do something different here according to whether we found the
    //  element or not.
    if (wasAdded)
    {
        if (laxThisOne) {
            fValidate = false;
            fElemStack.setValidationFlag(fValidate);
        }
        else if(fGrammarType == Grammar::SchemaGrammarType && fValidate) {
            ((SchemaElementDecl *)(elemDecl))->setValidationAttempted(PSVIDefs::FULL);
        }
Khaled Noaman's avatar
Khaled Noaman committed

        // 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()
            );
            if(fGrammarType == Grammar::SchemaGrammarType)
                ((SchemaElementDecl *)(elemDecl))->setValidity(PSVIDefs::INVALID);
        if(!laxBeforeElementFound && fGrammarType == Grammar::SchemaGrammarType) {
            if (fValidate) {
                ((SchemaElementDecl *)(elemDecl))->setValidationAttempted(PSVIDefs::FULL);
                ((SchemaElementDecl *)(elemDecl))->setValidity(PSVIDefs::VALID);
            }
        }

Khaled Noaman's avatar
Khaled Noaman committed
        // If its not marked declared and validating, then emit an error
        if (!elemDecl->isDeclared()) {
            if (laxThisOne) {
                fValidate = false;
                fElemStack.setValidationFlag(fValidate);
            }

Khaled Noaman's avatar
Khaled Noaman committed
            {
                fValidator->emitError
                (
                    XMLValid::ElementNotDefined
                    , elemDecl->getFullName()
                );

                if(fGrammarType == Grammar::SchemaGrammarType) {
                    ((SchemaElementDecl *)(elemDecl))->setValidity(PSVIDefs::INVALID);
                    ((SchemaElementDecl *)(elemDecl))->setValidationAttempted(PSVIDefs::FULL);
                }
        if (fGrammarType == Grammar::SchemaGrammarType) {
Khaled Noaman's avatar
Khaled Noaman committed
            ((SchemaElementDecl*)elemDecl)->setXsiComplexTypeInfo(0);
            ((SchemaElementDecl*)elemDecl)->setXsiSimpleTypeInfo(0);
        }
    }

    if(errorBeforeElementFound && fGrammarType == Grammar::SchemaGrammarType) {
        ((SchemaElementDecl *)(elemDecl))->setValidity(PSVIDefs::INVALID);
Khaled Noaman's avatar
Khaled Noaman committed
    //  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;

    //  Validate the element
    if (fValidate)
        fValidator->validateElement(elemDecl);

    if (fGrammarType == Grammar::SchemaGrammarType) {
        ComplexTypeInfo* typeinfo = ((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 XMLCh poundStr[] = {chPound, chNull};
            if (!XMLString::startsWith(typeName, poundStr)) {
                const int comma = XMLString::indexOf(typeName, chComma);
                if (comma > 0) {
                    XMLBuffer prefixBuf(comma+1);
                    prefixBuf.append(typeName, comma);
                    const XMLCh* uriStr = prefixBuf.getRawBuffer();

                    bool errorCondition = !switchGrammar(uriStr) && fValidate;
                    if (errorCondition && !laxThisOne)
Khaled Noaman's avatar
Khaled Noaman committed
                    {
                        fValidator->emitError
                        (
                            XMLValid::GrammarNotFound
                            , prefixBuf.getRawBuffer()
                        );
                        if(fGrammarType == Grammar::SchemaGrammarType)
                            ((SchemaElementDecl *)(elemDecl))->setValidity(PSVIDefs::INVALID);

                    }
                    else if(errorCondition) {
                        if(fGrammarType == Grammar::SchemaGrammarType) {
                            ((SchemaElementDecl *)(elemDecl))->setValidationAttempted(PSVIDefs::NONE);
                            ((SchemaElementDecl *)(elemDecl))->setValidity(PSVIDefs::UNKNOWN);
                        }
Khaled Noaman's avatar
Khaled Noaman committed
                    }
                }
            }
        }
        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)
    {
        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);
            if(fGrammarType == Grammar::SchemaGrammarType)
                ((SchemaElementDecl *)(elemDecl))->setValidity(PSVIDefs::INVALID);

Khaled Noaman's avatar
Khaled Noaman committed
        }
    }
    else if (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);
    }

    //  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);

    // activate identity constraints
    if (fValidate && fGrammar && fGrammarType == Grammar::SchemaGrammarType) {

        unsigned int count = ((SchemaElementDecl*) elemDecl)->getIdentityConstraintCount();

        if (count || fMatcherStack->getMatcherCount()) {

            fValueStoreCache->startElement();
            fMatcherStack->pushContext();
            fValueStoreCache->initValueStoresFor((SchemaElementDecl*) elemDecl, (int) elemDepth);

            for (unsigned int i = 0; i < count; i++) {
                activateSelectorFor(((SchemaElementDecl*) elemDecl)->getIdentityConstraintAt(i), (int) elemDepth);
            }

            // call all active identity constraints
            count = fMatcherStack->getMatcherCount();

            for (unsigned int j = 0; j < count; j++) {

                XPathMatcher* matcher = fMatcherStack->getMatcherAt(j);
                matcher->startElement(*elemDecl, 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
        );
    }

    //  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();

        // 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
                    , elemDecl->getFullName()
                    , elemDecl->getFormattedContentModel()
                );
                if(fGrammarType == Grammar::SchemaGrammarType)
                    ((SchemaElementDecl *)(elemDecl))->setValidity(PSVIDefs::INVALID);

Khaled Noaman's avatar
Khaled Noaman committed
            }

            if (fGrammarType == Grammar::SchemaGrammarType) {

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

                if (oldCount || ((SchemaElementDecl*) elemDecl)->getIdentityConstraintCount()) {

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

                        XPathMatcher* matcher = fMatcherStack->getMatcherAt(i);
                        matcher->endElement(*elemDecl, fContent.getRawBuffer());
Khaled Noaman's avatar
Khaled Noaman committed
                    }

                    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))
Khaled Noaman's avatar
Khaled Noaman committed
                            fValueStoreCache->transplant(ic, matcher->getInitialDepth());
                    }

                    // 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, matcher->getInitialDepth());

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

                    fValueStoreCache->endElement();
                }
            }
        }

        if(!isRoot && fGrammarType == Grammar::SchemaGrammarType)
           ((SchemaElementDecl *)fElemStack.topElement()->fThisElement)->updateValidityFromElement(elemDecl, fGrammarType);


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

        // reset xsi:type ComplexTypeInfo
        if (fGrammarType == Grammar::SchemaGrammarType) {
            ((SchemaElementDecl*)elemDecl)->reset();
            if (!isRoot)
                ((SchemaElementDecl*)(fElemStack.topElement()->fThisElement))->
                    setXsiComplexTypeInfo(((SchemaValidator*)fValidator)->getCurrentTypeInfo());
        }

Khaled Noaman's avatar
Khaled Noaman committed
        // 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();
            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();
        }
    }

    return true;
}


unsigned int
IGXMLScanner::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;
        }
    }
}

// ---------------------------------------------------------------------------
//  IGXMLScanner: Helper methos
// ---------------------------------------------------------------------------
void IGXMLScanner::resizeElemState() {

    unsigned int newSize = fElemStateSize * 2;
    unsigned int* newElemState = 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
    delete [] fElemState;
    fElemState = newElemState;
    fElemStateSize = newSize;
}

// ---------------------------------------------------------------------------
//  IGXMLScanner: IC activation methos
// ---------------------------------------------------------------------------
void IGXMLScanner::activateSelectorFor(IdentityConstraint* const ic, const int initialDepth) {

    IC_Selector* selector = ic->getSelector();

    if (!selector)
        return;

    XPathMatcher* matcher = selector->createMatcher(fFieldActivator, initialDepth);

    fMatcherStack->addMatcher(matcher);
    matcher->startDocumentFragment();
}

// ---------------------------------------------------------------------------
//  IGXMLScanner: Grammar preparsing
// ---------------------------------------------------------------------------
Grammar* IGXMLScanner::loadGrammar(const   InputSource& src
                                   , const short        grammarType
                                   , const bool         toCache)
{
Khaled Noaman's avatar
Khaled Noaman committed
    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;
        fSeeXsi = false;

        if (grammarType == Grammar::SchemaGrammarType) {
            loadedGrammar = loadXMLSchemaGrammar(src, toCache);
Khaled Noaman's avatar
Khaled Noaman committed
        }
        else if (grammarType == Grammar::DTDGrammarType) {
            loadedGrammar = loadDTDGrammar(src, toCache);
Khaled Noaman's avatar
Khaled Noaman committed
        }

        // 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(...)
        {
            // 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(...)
    {
        // Reset and rethrow
        fReaderMgr.reset();
        throw;
    }

Khaled Noaman's avatar
Khaled Noaman committed
}

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

    if (!fValidator->handlesDTD()) {
        if (fValidatorFromUser && fValidate)
            ThrowXML(RuntimeException, XMLExcepts::Gen_NoDTDValidator);
        else {
            fValidator = fDTDValidator;
        }
    }

    fDTDGrammar = new DTDGrammar();
    fGrammarResolver->putGrammar(XMLUni::fgDTDEntityString, fDTDGrammar);
    fGrammar = fDTDGrammar;
    fGrammarType = fGrammar->getGrammarType();
    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
    fIDRefList->removeAll();

    if (toCache) {

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

        fGrammarResolver->orphanGrammar(XMLUni::fgDTDEntityString);
        fGrammarResolver->putGrammar(sysIdStr, 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())
            ThrowXML1(RuntimeException, XMLExcepts::Scan_CouldNotOpenSource, src.getSystemId());
        else
            ThrowXML1(RuntimeException, XMLExcepts::Scan_CouldNotOpenSource_Warning, src.getSystemId());
    }

    //  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 DTDEntityDecl(gDTDStr);
    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 DTDElementDecl(gDTDStr, fEmptyNamespaceId);
        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);
    dtdScanner.setScannerInfo(this, &fReaderMgr, &fBufMgr);

    // Tell it its not in an include section
    dtdScanner.scanExtSubsetDecl(false, true);
Khaled Noaman's avatar
Khaled Noaman committed

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

    if (toCache)
        fGrammarResolver->cacheGrammars();

    return fDTDGrammar;
}


XERCES_CPP_NAMESPACE_END