type aliases

[nrc]

Again, I'd like to discuss generics separately (#29), so that should make this one pretty trivial.

[strega-nil]

// generally
type X /* <- one line */ = Foo;
// need to break
type LongName = /* <- don't break here */
  FooBar; /* <- pull this down */

[nrc]

Pretty simple one, I think, ready for FCP.

[nrc]

Questions:

If there is a where clause, should we leave on one line, e.g.,

type Foo<T> where T: Foo = Bar;

I propose not, since that is what we agreed for where clauses in general.

In which case how should we break this?

type Foo<T>
where
    T: Foo,
= Bar;

Would be my suggestion, the only downside is that we had previously decided to break after the = for the non-where case if we need to break the line. Perhaps we should revisit that?

And to check - trailing comma here is OK?

[joshtriplett]

@nrc Type aliases with where clauses seem like a sufficiently different case to be worth evaluating separately from functions with where clauses. And they have kind of an awful order of terms, which we can't do anything about.

I think I'd argue that in the majority of cases type aliases should inline the constraints inside the <>, to improve readability. In the case of a where clause, though, I'd agree with your proposed formatting. The standard response to someone going "ewww" in reaction can be to suggest not using a where clause in a type alias. :)

I don't think we should change the decision to break after the = in the non-where case.