Fix #4431: Change semantics of inline params in macro #4460
Conversation
|
For context: the motivation for this semantics in #4431 is that this is a limited form of cross-stage persistence (maybe in reverse) or inspection of trees, which is easy for literals/primitive types but hard in general. (Tho, I was thinking of using a typeclass for extensibility). 👍 for the semantic change (and for updating docs). |
| * to | ||
| * ``` | ||
| * inline def foo[T1, ...](inline x1: X, ..., y1: Y, ....): Seq[Any] => Object = { (args: Seq[Any]) => { | ||
| * inline def foo[T1, ...](inline x1: X, ..., y1: Y, ..., inline f1: X => Y): Seq[Any] => Object = { (args: Seq[Any]) => { |
There was a problem hiding this comment.
Maybe add f1 at the original form as well in the comment.
There was a problem hiding this comment.
I removed it and made it part of the comment below
| } | ||
|
|
||
| private def isStage0Value(sym: Symbol)(implicit ctx: Context): Boolean = | ||
| sym.is(Inline) && sym.owner.is(Macro) && !defn.isFunctionType(sym.info) |
There was a problem hiding this comment.
The addition of !defn.isFunctionType(sym.info) is the only real change in this PR, the rest is refactoring.
There was a problem hiding this comment.
After discussion with Aggelos, also LGTM for me.
@Blaisorblade "cross stage persistence in reverse" is a good way to characterize this.
4 participants
inlinemacro parameters in stage 0 if they have type Boolean, Byte, Short, Int, Long, Float, Double, Char or StringThis change in semantics allows us to support
inlineparameters of function type.