我在一个大型 Java 项目（移动应用程序）上工作，并且由于性能不佳（高复杂性）而获得了审查和优化/重构代码的“感谢”工作。注意：我对 Java 完全陌生（我的背景是 C/C++），因此我申请任何愚蠢的问题。我做的第一件事是使用 Findbugs 并修复所有报告的问题。之后我使用了度量工具Understand来大致了解一下哪些方法的圈复杂度很高。不幸的是，有很多方法的圈复杂度高于 2^20 范围 :-( 其中之一是我需要一些帮助或好主意的地方......
简短描述：对于与服务器的通信，数据必须被序列化。此移动框架上没有可用的 Serializable 接口。因此，编写所有代码（单独）的同事实现了一个包含一个名为的方法的 Serializable 接口toByteArray()。例如class Customer：
class Customer
{
Address address;
AttributeCollection attributes;
LocationCollection locations;
int recId;
int recStatus;
DateTime recCreated;
String recCreatedBy;
String recCreatedByProg;
DateTime recChanged;
String recChangedBy;
String recChangedByProg;
int refAddressesId;
int refMandatorsId;
CustomerPropertyUsage usage;
/**
* Serialize the properties of a class into a byte array.
* @param destData Byte array, where the serialized data should be stored. Minimum 2 bytes.
* @param serializationIndex Offset within the passed byte array, from which the serialized data of the class
* should be entered. The offset is increased by the registered number of bytes so that after this method the
* next call points to the serialized data subsequent byte.
*/
void toByteArray(byte[] destData, IntClass serializationIndex)
{
if (this.address == null)
this.usage.value &= ~CustomerPropertyUsage.ADDRESS;
if (this.attributes == null)
this.usage.value &= ~CustomerPropertyUsage.ATTRIBUTES;
if (this.locations == null)
this.usage.value &= ~CustomerPropertyUsage.LOCATIONS;
this.usage.toByteArray(destData, serializationIndex);
CatrString catrString = null;
if ((this.usage.value & CustomerPropertyUsage.RECORD_HEADER) != CustomerPropertyUsage.NONE)
{
// Call static method getBytes from SerializationHelper class
SerializationHelper.getBytes(this.recId, 4, destData, serializationIndex.value);
serializationIndex.value += 4;
SerializationHelper.getBytes(this.recStatus, 4, destData, serializationIndex.value);
serializationIndex.value += 4;
// recChanged is a DateTime object. For the serialization we need minimum a 7 bytes array,
// Call method toByteArray() from DateTime class.
this.recChanged.toByteArray(destData, serializationIndex);
// call toByteArray of CatrString class
catrString = new CatrString(this.recChangedBy);
catrString.toByteArray(destData, serializationIndex);
catrString.setValue(this.recChangedByProg);
catrString.toByteArray(destData, serializationIndex);
// Same as recChanged
this.recCreated.toByteArray(destData, serializationIndex);
catrString = new CatrString(this.recCreatedBy);
catrString.toByteArray(destData, serializationIndex);
catrString.setValue(this.recCreatedByProg);
catrString.toByteArray(destData, serializationIndex);
SerializationHelper.getBytes(this.refAddressesId, 4, destData, serializationIndex.value);
serializationIndex.value += 4;
SerializationHelper.getBytes(this.refMandatorsId, 4, destData, serializationIndex.value);
serializationIndex.value += 4;
}
if (next property...)
{
... Serialization ...
}
if (next property...)
{
... Serialization ...
}
}
}
为了保持较低的 GPRS 费用，服务器在 this.usage.value 中设置了一个值，因此只有一个特定的属性将被序列化并传输回服务器 --> 传输的消息很小。这种方法会创建很多 if-case，具体取决于类中有多少属性，因此路径计数会越来越高。我认为这不是一个完美的解决方案，但没关系。我想改变的是 if-cases 中的序列化调用。目前它们看起来像这样：
---- class SerializationHelper ----
static void getBytes(long valueToConvert, int numOfBytesToConvert, byte[] destinationBytes, int destinationBytesOffset)
{
destinationBytes[destinationBytesOffset] = (byte)(valueToConvert & 0x000000FF);
destinationBytes[destinationBytesOffset + 1] = (byte)((valueToConvert & 0x0000FF00) >> 8);
if (numOfBytesToConvert > 2)
{
destinationBytes[destinationBytesOffset + 2] = (byte)((valueToConvert & 0x00FF0000) >> 16);
destinationBytes[destinationBytesOffset + 3] = (byte)((valueToConvert & 0xFF000000) >> 24);
if (numOfBytesToConvert > 4)
{
destinationBytes[destinationBytesOffset + 4] = (byte)((valueToConvert & 0x000000FF00000000L) >> 32);
destinationBytes[destinationBytesOffset + 5] = (byte)((valueToConvert & 0x0000FF0000000000L) >> 40);
destinationBytes[destinationBytesOffset + 6] = (byte)((valueToConvert & 0x00FF000000000000L) >> 48);
destinationBytes[destinationBytesOffset + 7] = (byte)((valueToConvert & 0xFF00000000000000L) >> 56);
}
}
}
---- class CatrString ----
void toByteArray(byte[] destData, IntClass serializationIndex)
{
// Number of unicode characters
SerializationHelper.getBytes(this.textLength, 2, destData, serializationIndex.value);
serializationIndex.value += 2;
// Text UTF-16 unicode characters
for (int charIndex = 0; charIndex < this.textLength; charIndex++)
{
destData[serializationIndex.value] = (byte)(this.charCodes[charIndex] & 0x00FF);
serializationIndex.value++;
destData[serializationIndex.value] = (byte)((this.charCodes[charIndex] & 0xFF00) >> 8);
serializationIndex.value++;
}
// Code End of string as UTF-16 unicode character
destData[serializationIndex.value] = 0x00;
serializationIndex.value++;
destData[serializationIndex.value] = 0x00;
serializationIndex.value++;
}
---- class DateTime ----
void toByteArray(byte[] destData, IntClass serializationIndex)
{
destData[serializationIndex.value + 0] = (byte) (m_year % 0x0100); // year low-Byte.
destData[serializationIndex.value + 1] = (byte) (m_year / 0x0100); // year high-Byte.
destData[serializationIndex.value + 2] = (byte) (m_month);
destData[serializationIndex.value + 3] = (byte) m_day;
destData[serializationIndex.value + 4] = (byte) m_hour;
destData[serializationIndex.value + 5] = (byte) m_minute;
destData[serializationIndex.value + 6] = (byte) m_second;
serializationIndex.value += 7;
}
应该可以编写一个更“通用”的类来完成所有序列化工作，我说序列化 xy 字节就是这样。但我不明白的是“特殊”toByteArray()方法，如字符串（UTF-16 编码）或日期和时间？如果我把它们放在一个班级里，这是一个好的解决方案吗？我有什么收获吗？可维护代码？高性能代码？？你的方法是什么？
谢谢