sgxmlscanner.cpp

来自「IBM的解析xml的工具Xerces的源代码」· C++ 代码 · 共 1,881 行 · 第 1/5 页

                case Token_PI :
                    scanPI();
                    break;

                case Token_StartTag :
                    scanStartTag(gotData);
                    break;

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

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

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

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

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

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

                if (fDocHandler)
                    fDocHandler->endDocument();
            }
        }
    }
    //  NOTE:
    //
    //  In all of the error processing below, the emitError() call MUST come
    //  before the flush of the reader mgr, or it will fail because it tries
    //  to find out the position in the XML source of the error.
    catch(const XMLErrs::Codes)
    {
        // This is a 'first failure' exception, so reset and return failure
        fReaderMgr.reset();
        return false;
    }
    catch(const XMLValid::Codes)
    {
        // This is a 'first fatal error' type exit, so reset and reuturn failure
        fReaderMgr.reset();
        return false;
    }
    catch(const XMLException& excToCatch)
    {
        //  Emit the error and catch any user exception thrown from here. Make
        //  sure in all cases we flush the reader manager.
        fInException = true;
        try
        {
            if (excToCatch.getErrorType() == XMLErrorReporter::ErrType_Warning)
                emitError
                (
                    XMLErrs::XMLException_Warning
                    , excToCatch.getType()
                    , 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(...)
        {
            // Reset and rethrow user error
            fReaderMgr.reset();
            throw;
        }

        // Reset and return failure
        fReaderMgr.reset();
        return false;
    }
    catch(const OutOfMemoryException&)
    {
        throw;
    }
    catch(...)
    {
        // Reset and rethrow original error
        fReaderMgr.reset();
        throw;
    }

    // If we hit the end, then flush the reader manager
    if (!retVal)
        fReaderMgr.reset();

    return retVal;
}

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

//  This method is called from scanStartTag() to handle the very raw initial
//  scan of the attributes. It just fills in the passed collection with
//  key/value pairs for each attribute. No processing is done on them at all.
unsigned int
SGXMLScanner::rawAttrScan(const   XMLCh* const                elemName
                          ,       RefVectorOf<KVStringPair>&  toFill
                          ,       bool&                       isEmpty)
{
    //  Keep up with how many attributes we've seen so far, and how many
    //  elements are available in the vector. This way we can reuse old
    //  elements until we run out and then expand it.
    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 (fReaderMgr.getCurrentReader()->isWhitespace(nextCh))
                {
                    // Ok, skip by them and get another char
                    fReaderMgr.getNextChar();
                    fReaderMgr.skipPastSpaces();
                    nextCh = fReaderMgr.peekNextChar();
                }
                 else
                {
                    // Emit the error but keep on going
                    emitError(XMLErrs::ExpectedWhitespace);
                }
            }
        }

        //  Ok, here we first check for any of the special case characters.
        //  If its not one, then we do the normal case processing, which
        //  assumes that we've hit an attribute value, Otherwise, we do all
        //  the special case checks.
        if (!fReaderMgr.getCurrentReader()->isSpecialStartTagChar(nextCh))
        {
            //  Assume its going to be an attribute, so get a name from
            //  the input.
            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)
                      ||  fReaderMgr.getCurrentReader()->isWhitespace(chFound))
                {
                    // Just fall through assuming that the value is to follow
                }
                else if (chFound == chOpenAngle)
                {
                    // Assume a malformed tag and that new one is starting
                    emitError(XMLErrs::UnterminatedStartTag, elemName);
                    return attCount;
                }
                else
                {
                    // Something went really wrong
                    return attCount;
                }
            }

            //  Next should be the quoted attribute value. We just do a simple
            //  and stupid scan of this value. The only thing we do here
            //  is to expand entity references.
            if (!basicAttrValueScan(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, 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 (fMemoryManager) KVStringPair
                (
                    fAttNameBuf.getRawBuffer()
                    , fAttValueBuf.getRawBuffer()
                    , fMemoryManager
                );
                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)
            ThrowXMLwithMemMgr(UnexpectedEOFException, XMLExcepts::Gen_UnexpectedEOF, fMemoryManager);

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

    return attCount;
}


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