Question

If I have a class like this:

public class Foo
{
    public IEnumerable<Bar> Bars { get; set; }

    public Foo()
    {
        Bars = new List<Bar>();
    }
}

At some stage I re-factor the class and add a secondary constructor which implements the first one like this:

public class Foo
{
    public IEnumerable<Bar> Bars { get; set; }

    // some more properties were added

    public Foo()
    {
        Bars = new List<Bar>();
    }

    public Foo(string parameter): this()
    {
        .... some code here
    }
}

I could have also written it similar to this:

public class Foo
{
    public IEnumerable<Bar> Bars { get; set; }

    // some more properties were added too

    public Foo()
    {
        InitilizeFoo();
    }

    public Foo(string parameter)
    {
        InitilizeFoo();
        .... some code here
    }

    private void InitializeFoo()
    {
        Bars = new List<Bar>();
    }
}

Seeing both approaches work in this scenario, is there a benefit or drawback in using one over the other?

Is inheriting constrcutors more efficient and making that code execute faster or is there a drawback which I don’t know about making the second implementation more efficient instead?

Answer 1

One of the key benefits in having one constructor call another constructor is that you can set read-only fields that way, you can’t do that by calling a non-constructor method.

For example:

public class Foo
{
    private readonly int myNumber;

    public Foo() : this(42)
    {
    }

    public Foo(int num)
    {
        myNumber = num;
    }
}

Performance wise, it’s probably no more or less efficient to call another constructor than to call another method, but it is more readable, in my opinion, for a constructor to call another constructor than to call a separate, private method whose only point is to be called by a constructor.

There could, of course, be situations when having a separate method makes sense, and it’s certainly not “wrong” per se. Chaining constructors just reads better to many for most uses, and there is no negative performance impact.

UPDATE: I performed 10,000,000 iterations of each way (chained vs private initialization method) and the results were so close they were nearly indistinguishable:

Initializer Method took: 84 ms for 10,000,000 iterations, 8.4E-06 ms/each.
Chained Constructors took: 81 ms for 10,000,000 iterations, 8.1E-06 ms/each.

So really, performance-wise there is nearly no benefit either way. The main benefit is with chained constructors you can set readonly fields, and in most cases it is more readable.