Returning instance created in 'using' statement

Question

If I have this method:

public StreamReader GetResourceStream(string filename)
{
    using (Stream stream = this.GetType().Assembly.GetManifestResourceStream(filename))
    using (StreamReader sr = new StreamReader(stream))
    {
        return sr;
    }
}

When I call it, should I call it this way

StreamReader resource = GetResourceStream("blah.xml");

or this way:

using (StreamReader resource = GetResourceStream("blah.xml"))
{
    // Do stuff
}

If second way, does this means the using (StreamReader sr = new StreamReader(stream)) line is not making any difference by using a using statement?

Answer 1

You should change the method itself not to include a using statement, but then make sure that the calling code uses a using statement:

public StreamReader GetResourceStream(string filename)
{
    return new StreamReader(GetType().Assembly
                                     .GetManifestResourceStream(filename));
}

using (StreamReader resource = GetResourceStream("blah.xml"))
{
    // Do stuff
}

That way it will only dispose of the resource when you've actually used it. Basically the method is giving ownership of the returned resource to the caller. That much isn't too bad - it's reasonably common, even (think of File.OpenText etc).

Where it gets more tricky is when you pass a resource into a method or constructor - should the "called" code take ownership or not? Usually not, but occasionally it's appropriate - for example, the Bitmap(Stream) documentation states that you must keep the stream alive for as long as the bitmap is required. In fact, disposing of the bitmap also disposes of the stream, so you still only have one resource to keep track of... but you do need to know that you mustn't close the stream.