How Developers Are Using var in C#
Date Published: 14 June 2011
The var keyword was introduced in C# 3.0, and has since gained quite a bit of popularity. There is also a fair bit of contention over how it should be used, with posts like this one (with which I happen to agree) being not uncommon. Over the last week I posted a couple of polls on twitter that asked about specific scenarios in which one might use var, trying to address the two scenarios outlined by the author of that post.
This was meant to depict the case where the type would appear twice on one line, and so using var makes the code more succinct while no less clear.
75% of respondents agree that this is a scenario in which you should use the var keyword. I’ve heard from more than one person (including the article linked in the first paragraph) that even better would be if you could instantiate a variable more easily, as in String s = new(); or alternately to not require the type at all on the left:
Of course, neither of those options are going to fly with C# (without some significant changes to the compiler), so getting back to “things we can actually do today with C#” we come to the second scenario.
In this scenario I’m going for the case where there is some method that is returning a type, and the type isn’t obvious.
So the question comes down to, should you use var to declare variables that are initialized from a method call such that the type being returned is not obvious? 64% of respondents still favor using var in this case. Personally, I prefer the somewhat more clear red option, since it’s apparent immediately what the type of result is, but obviously I’m in the minority (for this poll) in this case.
The third question was in response to some comments and questions I had about the first scenario. Some folks contended that in the case of var s = new String(); it wasn’t really worth it to use var, but that they would use var if it was a very long, complex, or convoluted type. So that’s what this question was meant to reflect.
In this case 78% (3% more than scenario one) agree that one should use var here. A number of people suggested that a custom type would be better than the ugly Dictionary of Dictionaries that I used in my example, resulting in some of the 3% Other votes (and I agree, but I was just going for a long and complex type, so work with me here). Less than 20% of respondents felt that it was worthwhile to type out the complex type twice on this particular line of code.
A scenario that I’m obviously leaving out is where anonymous types are required. These can only be represented by var, and are the main reason for its introduction into the C# language in the first place. Thus, I didn’t see much value in polling people about its usage in this case.
Steve is an experienced software architect and trainer, focusing on code quality and Domain-Driven Design with .NET.