WIP Fix SIP-23 (Literal Types)

[valencik]

This PR cleans up some incorrect code examples in the SIP-23 document.

a PR bringing the SIP in line with the implementation would be welcome; so would a PR merging the whole SIP into the spec - @SethTisue (via gitter)

• update comments showing return types to match repl session
• confirm all examples in proposal details section
• figure out if we need to update the history section to reflect these changes

Additional commentary

This PR better addresses the issues discussed in #1570
I do not think this addresses all the confusion one could run into when playing with literal types.
I think we need to improve the spec and write some documentation elsewhere, perhaps an overview or section in the tour.
However, hopefully this PR serves as an incremental improvement over the current state.

[milessabin]

I think the example use of .type with subtypes of AnyVal should be tweaked, but should stay. We can prevent the widening with a singleton expected type, or by assigning to a final val, eg.,

scala> def foo[T](t: T): t.type = t
foo: [T](t: T)t.type

scala> final val bar = foo(23)
bar: 23 = 23

Otherwise LGTM.

[valencik]

Hi @milessabin, thanks for the comment.
Unfortunately I am not seeing that behavior in scala 2.13.1:

Welcome to Scala 2.13.1 (OpenJDK 64-Bit Server VM, Java 1.8.0_222).
Type in expressions for evaluation. Or try :help.

scala> def foo[T](t: T): t.type = t
foo: [T](t: T)t.type

scala> foo(23)
res0: Int = 23

scala> final var bar = foo(23)
bar: Int = 23

In each case I am seeing the widened type.

[valencik]

Oh how embarrassing, look how that dreaded var snuck in and ruined the fun.
Using a val not a var we see the expected behavior return:

Welcome to Scala 2.13.0 (OpenJDK 64-Bit Server VM, Java 1.8.0_222).
Type in expressions for evaluation. Or try :help.

scala> def foo[T](t: T): t.type = t
foo: [T](t: T)t.type

scala> foo(23)
res0: Int = 23

scala> final val bar = foo(23)
bar: 23 = 23

Although it is perhaps interesting to note that the type gets widened for a var...

I'll restore the example with .type

[valencik]

A remaining question is whether we should update the history section of the SIP or not.

[SethTisue]

I don't think the history section needs updating; that section's purpose is to alert the SIP committee to design changes in the proposal. we have the git history for those interested in full details

thanks @valencik for tackling this!