(2.10.4) whitebox macros are now first typechecked against outerPt #3256
Conversation
Even though whitebox macros are supposed to be used to produce expansions that refine advertised return types of their macro definitions, sometimes those more precise types aren’t picked up by the typechecker. It all started with Travis generating structural types with macros and noticing that typer needs an extra nudge in order to make generated members accessible to the outside world. I didn’t understand the mechanism of the phenomenon back then, and after some time I just gave up. Afterwards, when this issue had been brought up again in a different StackOverflow question, we discussed it at reflection meeting, figured out that typedBlock provides some special treatment to anonymous classes, and it became clear that the first macro typecheck (the one that types the expansion against the return type of the corresponding macro def) is at fault here. The thing is that if we have a block that stands for a desugard anonymous class instantiation, and we typecheck it with expected type different from WildcardType, then typer isn’t going to include decls of the anonymous class in the resulting structural type: https://github.com/scala/scala/blob/master/src/compiler/scala/tools/nsc/typechecker/Typers.scala#L2350. I tried to figure it out at https://groups.google.com/forum/#!topic/scala-internals/eXQt-BPm4i8, but couldn’t dispel the mystery, so again I just gave up. But today I had a profound WAT experience that finally tipped the scales. It turns out that if we typecheck an if, providing a suitable pt, then the resulting type of an if is going to be that pt, even though the lub of the branch types might be more precise. I’m sure that reasons for this behavior are also beyond my understanding, so I decided to sidestep this problem. upd. Here’s Jason’s clarification: Doing thing differently would require us to believe that "'Tis better to have lubbed and lost than never to have lubbed at all." But the desire for efficiency trumps such sentimentality. Now expansions of whitebox macros are first typechecked against pt, the expected type that comes from the enclosing context, before being typechecked against expectedTpe, the expected type that comes from the return type of the macro def. This means that now pt provides the correct expected type for the initial, most important typecheck, which makes types more precise.
ghost
assigned 
|
Small process suggestion: I think it is better not to submit the backport PR until the original change is reviewed and fine tuned. I find that having both in queue makes me feel under some pressure to avoid finding faults during review, as I know you'll have to rework the backport. |
|
@retronym Fair enough, I will keep that in mind in the future. Do you want me to close the backports that are currently in the queue? |
|
Yes, that would be great. I'm happy that you have them ready, I just want you to keep them to yourself for a little while :) I suppose it would also be handy to declare your intention to backport in the main PR; maybe the reviewer will develop thoughts on the pros and cons of that while reviewing. |
|
Closes this PR for now until its master counterpart is merged. |
|
Sorry, the queue was empty from where I was sitting because these were closed. |
Even though whitebox macros are supposed to be used to produce expansions
that refine advertised return types of their macro definitions, sometimes
those more precise types aren’t picked up by the typechecker.
It all started with Travis generating structural types with macros
and noticing that typer needs an extra nudge in order to make generated
members accessible to the outside world. I didn’t understand the mechanism
of the phenomenon back then, and after some time I just gave up.
Afterwards, when this issue had been brought up again in a different
StackOverflow question, we discussed it at reflection meeting, figured out
that typedBlock provides some special treatment to anonymous classes,
and it became clear that the first macro typecheck (the one that types
the expansion against the return type of the corresponding macro def)
is at fault here.
The thing is that if we have a block that stands for a desugard anonymous
class instantiation, and we typecheck it with expected type different from
WildcardType, then typer isn’t going to include decls of the anonymous class
in the resulting structural type: https://github.com/scala/scala/blob/master/src/compiler/scala/tools/nsc/typechecker/Typers.scala#L2350.
I tried to figure it out at https://groups.google.com/forum/#!topic/scala-internals/eXQt-BPm4i8,
but couldn’t dispel the mystery, so again I just gave up.
But today I had a profound WAT experience that finally tipped the scales.
It turns out that if we typecheck an if, providing a suitable pt, then
the resulting type of an if is going to be that pt, even though the lub
of the branch types might be more precise. I’m sure that reasons for this
behavior are also beyond my understanding, so I decided to sidestep this problem.
upd. Here’s Jason’s clarification: Doing thing differently would require
us to believe that "'Tis better to have lubbed and lost than never to have
lubbed at all." But the desire for efficiency trumps such sentimentality.
Now expansions of whitebox macros are first typechecked against pt,
the expected type that comes from the enclosing context, before being
typechecked against expectedTpe, the expected type that comes from the return
type of the macro def. This means that now pt provides the correct
expected type for the initial, most important typecheck, which makes
types more precise.