C#: Is it possible to check that a simple expression will always return true?

Question

And I mean this in the easiest way. Say you have a function with the following signature:

public static Expression<Func<T, bool>> CreateExpression<T>(string value)
{
    // ...
}

Usually it will create a more complex expression of some sort, but if the value is null the method should return a constant, always true expression. In other words:

public static Expression<Func<T, bool>> CreateExpression<T>(string value)
{
    if(value == null)
        return x => true;

    // ...
}

Is this something I can create a unit test for? That when I send in null as the value, I do get back a constant true expression?

Answer 1

There's no way to do so that I know of. However, if you're willing to refactor a bit:

class Sample<T>
{
    public static Func<T, bool> Identity = x => true;

    public static Func<T, bool> CreateExpression(string value)
    {
        if(value == "foo")
            return Identity;

        return x => value.Length % 2 == 0;
    }
}

class Test
{
    public void DoTest()
    {
        Debug.Assert(Sample<string>.CreateExpression("foo") == Sample<string>.Identity);
    }
}

Answer 2

It would be easy enough to test for exactly that expression (the body will be a ConstantExpression with value true). But in the general case? No - too complex.

static bool IsConstantTrue(LambdaExpression lambda)
{
    return lambda.Body.NodeType == ExpressionType.Constant
        && true.Equals(((ConstantExpression)lambda.Body).Value);
}

with

Expression<Func<SomeType, bool>> exp = x => true; // or some method that 
                                                  // returns a lambda expression
Assert.IsTrue(IsConstantTrue(exp));

Answer 3

What do you mean be "simple" here? It's kind of a woolly term...

In general, the only thing we can seay is that this is yet another manifestation of the halting problem. Consider, how can you you determine the result of a function until all possible parameters, unless you actually run it under all possible parameters? Apart from being practically infeasible, you can't even guarantee a result because of the nature of the halting problem (you don't know the method will even terminate, or what path it may take in the indefinite future).