为什么XmlReader读取不同的块大小

Question

我注意到XmlReader并不总是读取标准4096块中的XML输入。 特别是，如果XmlReader发现第一个缓冲区以部分XML元素名称结尾，则XML读取器使用来自第一个缓冲区的部分数据填充第二个缓冲区，然后才将更多数据读入该缓冲区。

为什么XmlReader会这样？

为什么XmlReader不会总是请求用4096个字符填充标准缓冲区？

重现奇怪的XmlReader缓冲行为的代码：

[TestClass]
public class XmlReaderTests
{
    public class DebugStreamReader : StreamReader
    {
        public DebugStreamReader(Stream stream) : base(stream) { }

        public override int Read(
            char[] buffer,
            int index,
            int count)
        {
            int readLength = base.Read(buffer, index, count);
            Trace.WriteLine("readLength: " + readLength + "; buffer.Length: " + buffer.Length + "; index: " + index + "; count: " + count);
            //Debug.Write("readLength: " + readLength);
            return readLength;
        }
    }

    [TestMethod]
    public void SuperLongXmlElementNameXmlReaderTest()
    {
        string elementName = new string('a', 20000);
        string xmlWithSuperLongElementName = "<" + elementName + ">test</" + elementName + ">";
        MemoryStream inputStream = new MemoryStream(Encoding.UTF8.GetBytes(xmlWithSuperLongElementName));
        StreamReader sr = new DebugStreamReader(inputStream);
        using (XmlReader reader = XmlReader.Create(sr))
        {
            while (reader.Read())
            {
                //string value = 
                reader.ReadString();
            }
        }
    }
}

输出：

readLength: 4096; buffer.Length: 4097; index: 0; count: 4096
readLength: 1; buffer.Length: 4097; index: 4095; count: 1
readLength: 4097; buffer.Length: 8194; index: 4096; count: 4097
readLength: 8194; buffer.Length: 16388; index: 8193; count: 8194
readLength: 16388; buffer.Length: 32776; index: 16387; count: 16388
readLength: 7233; buffer.Length: 32776; index: 12768; count: 20007
readLength: 0; buffer.Length: 32776; index: 0; count: 32775