An argument (lol) for the use of explicit argument names in C# function calls even when they seem redundant.
- You can assert correctness when reviewing the call site in a dumb text display (e.g. a patch in a github pull request)
- Parameter-order refactoring won’t introduce subtle bugs and breaking changes.
This is particularly important when you have multiple arguments of the same type.
It’s also a good idea when the call doesn’t immediately make it obvious what the meaning of the parameter is. E.g. Foo(3) vs Foo(sizeInMeters: 3).
No argument names (bad)
Starting point:
void Main()
{
Foo("Home", "Blah"); // don't do this
}
void Foo(string importantThing, string otherThing)
{
switch (importantThing)
{
case "Home":
DoSomething();
break;
default:
DoSomethingElse();
break;
}
Log(otherThing);
}
If Foo() is defined in a different file it’s hard to eyeball Main() and tell if the arguments line up properly.
If I now re-order the arguments of Foo() can you easily spot that the call site in Main() is now incorrect?
void Main()
{
Foo("Home", "Blah"); // Now broken!! But can you tell?
}
void Foo(string otherThing, string importantThing) // <-- parameters flipped
{
switch (importantThing)
{
case "Home":
DoSomething();
break;
default:
DoSomethingElse();
break;
}
Log(otherThing);
}
With argument names (good)
Starting point:
void Main()
{
Foo(importantThing: "Home", otherThing: "Blah");
}
void Foo(string importantThing, string otherThing)
{
switch (importantThing)
{
case "Home":
DoSomething();
break;
default:
DoSomethingElse();
break;
}
Log(otherThing);
}
You can now easily see that "Home" is going to "importantThing", even in a plain-text view of the file.
If I now re-order the arguments of Foo() again, it doesn’t break because the compiler knows which string to connect to which parameter. Win.
void Main()
{
Foo(importantThing: "Home", otherThing: "Blah"); // still valid!
}
void Foo(string otherThing, string importantThing) // <-- refactored
{
switch (importantThing)
{
case "Home":
DoSomething();
break;
default:
DoSomethingElse();
break;
}
Log(otherThing);
}
Making Wrong Code Look Wrong
This piece of code-style opinion falls under “Making Wrong Code Look Wrong” as Joel Spolsky teaches.
Using types
Having many parameters of the same types can be a code smell that it’s time to move to stronger typing.
There’s two ways you can do this:
- Create a wrapper type for all the data you need and only have one parameter of that type.
- Create meaningful types that wrap primitive types and pass those instead so that the compiler will check the right type of thing is going in as the right argument.
E.g. this:
void Foo(string importantThing, string otherThing)
becomes
void Foo(SomeThings things)
or
void Foo(ImportantThing importantThing, OtherThing otherThing)
Thus making the arguments impossible to mix up through being out of order.