java - 从字节缓冲区构建 xml 消息

Question

我有一个传递字节数组的函数。这些字节可能包含 xml 消息的一部分、一个整体或一个整体和一部分。我已经成功地使用状态机尝试解析消息，直到今天我们意识到如果第二个来源正在向我们发送数据，我可以得到消息的前半部分，并且在第二个来源的另一条消息开始之后立即. 我希望能够通过丢弃出现故障的消息来纠正这一点。我认为我不应该存储它并保留它直到完成。

这是我的状态机：

for (int i = 0; i < bytes.length; i++)
    {
        byte data = bytes[i];

        //based on the read state, perform an action.  This is attempting to 
        //determine if we are reading the beginning of a message, or the end of the message.
        if (bufferIndex == BmsConstants.MAX_DRAWING_LENGTH)
        {
            resetState();               
        }

        switch (readState)
        {
            case SYNC:
                //we've read:  <
                if (data == '<')
                {
                    readState = DrawingDeviceParserState.START_LT;
                    buff[bufferIndex++] = (char) data;
                } else
                {
                    resetState();
                }
                break;
            case START_LT:
                //we've read:  <?
                if (data == '?')
                {
                    readState = DrawingDeviceParserState.START_QM;
                    buff[bufferIndex++] = (char) data;
                } else
                {
                    resetState();
                }
                break;
            case START_QM:
                //we've read:  <?x
                if (data == 'x')
                {
                    readState = DrawingDeviceParserState.START_X;
                    buff[bufferIndex++] = (char) data;
                } else
                {
                    resetState();
                }
                break;
            case START_X:
                //we've read:  <, now just read until we get to the </event>
                if (data == '<')
                {
                    readState = DrawingDeviceParserState.END_LT;
                    buff[bufferIndex++] = (char) data;
                }
                else
                {
                    buff[bufferIndex++] = (char) data;
                }
                break;
            case END_LT:
                //we've read:  </
                if (data == '/')
                {
                    readState = DrawingDeviceParserState.END_SLASH;
                    buff[bufferIndex++] = (char) data;
                }
                else
                {
                    readState = DrawingDeviceParserState.START_X;
                    buff[bufferIndex++] = (char) data;
                }
                break;
            case END_SLASH:
                //we've read:  </e
                if (data == 'e')
                {
                    readState = DrawingDeviceParserState.END_E;
                    buff[bufferIndex++] = (char) data;
                }
                else
                {
                    readState = DrawingDeviceParserState.START_X;
                    buff[bufferIndex++] = (char) data;
                }
                break;
            case END_E:
                //we've read:  </ev
                if (data == 'v')
                {
                    readState = DrawingDeviceParserState.END_V;
                    buff[bufferIndex++] = (char) data;
                }
                else
                {
                    readState = DrawingDeviceParserState.START_X;
                    buff[bufferIndex++] = (char) data;
                }
                break;
            case END_V:
                //we've read:  </eve
                if (data == 'e')
                {
                    readState = DrawingDeviceParserState.END_E1;
                    buff[bufferIndex++] = (char) data;
                }
                else
                {
                    readState = DrawingDeviceParserState.START_X;
                    buff[bufferIndex++] = (char) data;
                }
                break;
            case END_E1:
                //we've read:  </even
                if (data == 'n')
                {
                    readState = DrawingDeviceParserState.END_N;
                    buff[bufferIndex++] = (char) data;
                }
                else
                {
                    readState = DrawingDeviceParserState.START_X;
                    buff[bufferIndex++] = (char) data;
                }
                break;
            case END_N:
                //we've read:  </event
                if (data == 't')
                {
                    readState = DrawingDeviceParserState.END_T;
                    buff[bufferIndex++] = (char) data;
                }
                else
                {
                    readState = DrawingDeviceParserState.START_X;
                    buff[bufferIndex++] = (char) data;
                }
                break;
            case END_T:
                //we've read:  </event>  
                //End of the message, now we can parse it.
                if (data == '>')
                {
                    //readState = DrawingDeviceParserState.END_GT;
                    buff[bufferIndex++] = (char) data;

                    parseXml(new String(buff, 0, bufferIndex));
                    resetState();
                }
                else
                {
                    readState = DrawingDeviceParserState.START_X;
                    buff[bufferIndex++] = (char) data;
                }
                break;
        }
    }

弱点是此状态机位于案例 START_X 内。我将继续阅读，直到找到标签，在 2 条消息乱七八糟地塞在一起的情况下，该标签将位于第二条消息的末尾。我希望有人能做的是要么告诉我状态机不是我应该做的，要么修改我的代码/逻辑以更好地处理这种边缘情况。