Question

In .Net, you can chain methods returning a value or by using a void. Is one of them the “right way”?

So you could say

1)

Foo myFoo = new Foo();
myfoo.Bars = 
  myBars.DoSomethingCool(x)
  .DoSomethingElse(y)
  .AndSomethingElse(z);

public static IList<IBar> DoSomethingCool(this IList<IBar> source, object x)
{
  IList<IBar> result = //some fn(source)
  return result;
}

In this case, all 3 extension methods need to return IList (the type for myFoo.Bars)

or it could also be written as

2)

myBars.DoSomethingCool(x)
.DoSomethingElse(y)
.AndSomethingElse(z);

public static void DoSomethingCool(this IList<IBar> source, object x)
{
  //Modify source
  source = //some fn(source)
  //Don't return anything
}

in which case, the extension methods return a void, but do the work on the source object that’s coming in?

UPDATE Simon was correct in his answer that 2) won’t compile. Here is how that could be re-written:

DoSomethingCool(myBars)
.DoSomethingElse(myBars)
.AndSomethingElse(myBars);

myBars would then be changing inside the call of each method and the methods would be returning void.

Answer 1

1) wins. Simon correctly answered that 2) wouldn’t compile, but he did it as a comment, so I can’t gie him credit for answering. The updated solution for 2) does everything as a side effect an I can’t think of a reason why you would ever want to do that in a statically typed language.

The points made about chaining’s debugging issue are something to consider, though I find chaining to be especially useful when filtering.

mybars.FilterByHasHappyHour() is much nicer than

BigGiantUtilityClass.GetBarsWithHappyHours(myBars)