如何编组托管代码中的Subset byte []数组？

Question

Consider the following method (used in a factory method):

private Packet(byte[] rawBytes, int startIndex)
{
    m_packetId = BitConverter.ToUInt32(rawBytes, startIndex);
    m_dataLength = BitConverter.ToUInt16(rawBytes, startIndex + 4);
    if (this.Type != PacketType.Data)
        return;
    m_bytes = new byte[m_dataLength];
    rawBytes.CopyTo(m_bytes, startIndex + Packet.HeaderSize);
}

The last two lines of code strike me wasteful. Allocating more memory and populating it with values from memory seems silly.

With unmanaged code, something like this is possible:

    m_bytes = (rawBytes + (startIndex + Packet.HeaderSize));

(I didn't run it through a compiler so syntax is probably off, but you can see it's just a matter of pointer manipulation.)

I ran into a similar problem the other day when an API returned a byte[] array that was really a short[] array.

Are these types of array permutations just the cost of using managed code or is there a new school of thinking that I'm just missing?

Thanks in advance.

Answer 1

Have you considered restructuring so that maybe the copy is not necessary?

First option: store a reference to rawBytes in m_bytes and store the offset that needs to be added to all accesses into this byte array.

Second option: make m_bytes a MemoryStream instead, constructed from the buffer, an offset and a length; this constructor also does not copy the byte buffer and just allows access to the specified sub-segment.

Keep in mind though that the price of avoiding the copy operation is that the rawBytes and m_bytes (array or stream) will be aliases, so changes to one will change the other too.

Answer 2

是的，这在托管代码中成本更高，但是没有什么可以阻止您使此方法变得unsafe并进行所需的指针操作。