Daily Archives: 19 May 2010

2 posts

LINQ to XML and reading large XML files

LINQ to XML makes it relatively easy to read and query XML files. For example consider the following XML file:

<xml version="1.0" encoding="utf-8" ?>
    <users>
        <user name="User1" groupid="4" />
        <user name="User2" groupid="1" />
        <user name="User3" groupid="3" />
        <user name="User4" groupid="1" />
        <user name="User5" groupid="1" />
        <user name="User6" groupid="2" />
        <user name="User7" groupid="1" />
    </users>

Suppose you would like to find all records with groupid > 2. You could be tempted to issue the following query:

XElement doc = XElement.Load("users.xml");
    var users = from u in doc.Elements("user")
                where u.Attribute("groupid") != null &&
                int.Parse(u.Attribute("groupid").Value) > 2
                select u;
    Console.WriteLine("{0} users match query", users.Count());

There’s a flaw in this method. XElement.Load method will load the whole XML file in memory and if this file is quite big, Continue reading

Dynamic methods in .NET

Using reflection to invoke methods which are not known at compile time might be problematic in performance critical applications. It is roughly 2.5-3.0 times slower than direct method calls. Here’s a sample test I’ve conducted:

class Program
    {
        [DllImport("kernel32.dll")]
        static extern void QueryPerformanceCounter(ref long ticks);

        static PropertyInfo _intProp = typeof(Foo).GetProperty("IntProp", BindingFlags.Public | BindingFlags.Instance);

        static void Main(string[] args)
        {
            Foo foo = new Foo { IntProp = 10 };
            const int COUNT = 1;
            Console.WriteLine(Measure(() => ReadPropertyWithReflection(foo), COUNT));
            Console.WriteLine(Measure(() => ReadPropertyDirectly(foo), COUNT));
        }

        static void ReadPropertyWithReflection(Foo foo)
        {
            int intProp = (int)_intProp.GetValue(foo, null);
        }

        static void ReadPropertyDirectly(Foo foo)
        {
            int intProp = foo.IntProp;
        }

        static long Measure(Action action, int count)
        {
            long startTicks = 0;
            QueryPerformanceCounter(ref startTicks);
            for (int i = 0; i < count; i++)
            {
                action();
            }
            long endTicks = 0;
            QueryPerformanceCounter(ref endTicks);
            return endTicks - startTicks;
        }

        class Foo
        {
            public int IntProp { get; set; }
        }
    }

Here are the results:

Method CPU units
Direct method invocation 796
Reflection method invocation 1986

So using reflection to perform a single property read is 2.5 slower than direct property access.

Dynamic methods can be used to generate and execute a method at run time, without having to declare a dynamic assembly and a dynamic type to contain the method. They are the most efficient way to generate and execute small amounts of code. Continue reading