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). |
| @@ -66,16 +66,19 @@ import dotty.tools.dotc.core.quoted._ | |||
| * ``` | |||
| * 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
| @@ -625,6 +629,9 @@ class ReifyQuotes extends MacroTransformWithImplicits with InfoTransformer { | |||
| } | |||
| } | |||
|
|
|||
| 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.
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.