do you use extension methods intensively in your projects?

BillWoodruff

Hi Sandor, I'd really appreciate seeing an example of Extension Methods for an Interface. cheers, Bill

«There is a spectrum, from "clearly desirable behaviour," to "possibly dodgy behavior that still makes some sense," to "clearly undesirable behavior." We try to make the latter into warnings or, better, errors. But stuff that is in the middle category you don’t want to restrict unless there is a clear way to work around it.» Eric Lippert, May 14, 2008

Sander Rossel

The entire LINQ library depends upon it. For example the Count() extension method on collections may look as follows:

public static class Extensions
{
public static int Count(this IEnumerable collection)
{
int count = 0;
using (var enumerator = collection.GetEnumerator())
{
while (enumerator.MoveNext())
{
count += 1;
}
}
return count;
}
}

The actual extension method checks for null and tries to cast to ICollection<T> and ICollection for the Count property first, but it's an extension on an interface. There are LOTS of them...

Read my (free) ebook Object-Oriented Programming in C# Succinctly. Visit my blog at Sander's bits - Writing the code you need. Or read my articles here on CodeProject.

Simplicity is prerequisite for reliability. — Edsger W. Dijkstra

Regards, Sander

Southmountain

very inspirational

diligent hands rule....

Southmountain

this is a good one..

diligent hands rule....

Southmountain

thanks for this example!

diligent hands rule....

Mark_Wallace

Marc Clifton wrote:

public static bool Yes<T>(this T foo) {return true;}

That looks suspiciously like an advert for Ty-phoo tea!

I wanna be a eunuchs developer! Pass me a bread knife!

Forogar

Extensively, yes. Intensively, no.

- I would love to change the world, but they won’t give me the source code.

Chris Maunder

We use them in order to have POCO's and then attach extension methods to them to give them some mojo.

cheers Chris Maunder

Fabio Franco

Sander Rossel wrote:

On a few occasions I've created an interface, ISomething, and before writing any implementation I wrote a few extension methods I knew I needed.

Still doesn't make sense to me. If the interface is your own, you should write the method to the interface as it makes sense to be in its scope. If you need behavior without having to implement everywhere, you're looking for an abstract class. Extension methods provide a clean way to have helper methods to class you cannot modify. Otherwise you're just creating spaghetti code and killing a few principles. Although M$ does with it's own classes, it makes sense, specially to not break backwards compatibility when implementing new frameworks (like LINQ), so yes, they own the code, but they should not modify it, as LINQ is implemented in a whole different scope. If you are developing frameworks, with attachable components, that extend other standalone components, yes, it makes sense to have extension methods on your own code, otherwise I agree that's "silly".

To alcohol! The cause of, and solution to, all of life's problems - Homer Simpson ---- Our heads are round so our thoughts can change direction - Francis Picabia

Fabio Franco

Extensively sounds funny to me. Unless you are creating a framework of extension methods for built-in or third party types, using it intensively may indicate you're doing something wrong with your code. They have a purpose, be sure you understand it before using it everywhere. It may create unorganized, hard to maintain code.

To alcohol! The cause of, and solution to, all of life's problems - Homer Simpson ---- Our heads are round so our thoughts can change direction - Francis Picabia

Fabio Franco

Why not implement in the object themselves and keep them contained and organized? Is there anything preventing the code to be where it belongs?

To alcohol! The cause of, and solution to, all of life's problems - Homer Simpson ---- Our heads are round so our thoughts can change direction - Francis Picabia

Sander Rossel

I wholeheartedly disagree :D Extension methods are nothing more than static helper methods that are treated as instance methods by the editor (and the compiler?). I don't think there is a principle against creating static helper methods and it's certainly not creating spaghetti, especially since the methods read as instance methods (editor trick or not). An abstract class is heavy weaponry, every implementor must now inherit the abstract class and since you can only inherit once this puts a serious limitation on implementors. If you can't inherit you don't get the functionality. You could create a wrapper, but that's just a lot of boilerplate code. Besides, inheritance is added complexity/spaghetti too!

public interface ISomething
{
IEnumerable Stuff { get; }
}

public static class SomethingExtensions
{
public static bool HasStuff(this ISomething something)
{
return something.Stuff.Any();
}
}How is this not awesome? Every implementor now gets HasStuff for free without the need for inheritance!

You agree that extension methods are clean for LINQ and when dealing with third party code, but for your own code it's suddenly spaghetti? Sounds like double standards to me ;)

Read my (free) ebook Object-Oriented Programming in C# Succinctly.
Visit my blog at Sander's bits - Writing the code you need.
Or read my articles here on CodeProject.

Simplicity is prerequisite for reliability.
— Edsger W. Dijkstra

Regards,
Sander

PSU Steve

Yes, quite a bit. I love the extension method concept. Now if they could just implement "extension properties". Yes, I know you can use extension methods to pseudo-implement properties, but a full-blown extension property implementation would be sweet...

Cameron Oltmann

For some reason I never thought of writing them for enums, though I use them for interfaces all the time. Can you give an example where an extension method for an enum would be useful? Can't think of any myself off hand.

Chris Maunder

Separation of concerns, mostly.

cheers Chris Maunder

Harley L Pebley

The core products for the company I work for are mainly COM based libraries. I've written .Net libraries of extension methods that help hide the COM ugliness and provide a more .Net friendly interface.

Matthew Dennis

I have a .NET Core based Redis Client that I've been working on for a while that heavily uses extension methods. The 'client' just knows how to send a command and receive a response. All the Redis commands are implemented as extension methods on the 'client'. So as static class for the Key Commands, another for the Hash Commands, ... in Redis Client

/// <summary>
/// Sends a command and returns the response string.
/// </summary>
/// <param name="command">The Command.</param>
/// <param name="parameters">The parameters.</param>
/// <returns>The response.</returns>
public Task SendAsync(string command, params object[] parameters)
{
return SendAsync(command, (IEnumerable<object>)parameters);
}

/// <summary>
/// Sends a command and returns the response string.
/// </summary>
/// <param name="command">The Command.</param>
/// <param name="parameters">The parameters.</param>
/// <returns>The awaitable Task.</returns>
public async Task SendAsync(string command, IEnumerable<object> parameters = null)
{
await EnsureConnected().ConfigureAwait(false);
await _commandWriter.WriteRedisCommandAsync(command, parameters).ConfigureAwait(false);
}

in HashCommands

public static Task SendHGetAsync(this RedisClient client, string key, string field)
{
if (string.IsNullOrWhiteSpace(key))
throw new ArgumentNullException(nameof(key));

if (string.IsNullOrWhiteSpace(field))
    throw new ArgumentNullException(nameof(field));

return client.SendAsync("HGet", key, field);

}

Each commands are small and trivial. The core client has no tricky dependencies on the commands. Classic Open for Extension but Closed for Modification. What has been delaying me is the number of tests to test all the commands, plus an update to the latest .NET Core bits. Another great thing about Extension Methods is the ability to extend 3rd party libraries.

"Time flies like an arrow. Fruit flies like a banana."

Lost User

Well it depends on what that enum is about, it's not always useful. I've used it for example to give an `enum ConditionCode { etc` the fake-methods `Invert` and `SwapOperands`.

Jim Balter

I used them extensively.

Cameron Oltmann

Ah gotcha. Thanks!