我正在编写一个宏,可以在编译期间记录一条短消息,使用 scala 2.13 的模式匹配和常量类型特性:
class EmitMsg[T, SS <: EmitMsg.EmitLevel] {}
object EmitMsg {
trait EmitLevel
trait Info extends EmitLevel
trait Warning extends EmitLevel
trait Error extends EmitLevel
trait Abort extends EmitLevel
def create[A, SS <: EmitMsg.EmitLevel]: EmitMsg[A, SS] = new EmitMsg[A, SS]
implicit def emit[A, SS <: EmitMsg.EmitLevel]: EmitMsg[A, SS] = macro Macros.emit[A, SS]
final class Macros(val c: whitebox.Context) {
val u = c.universe
import u._
def outer: EmitMsg.type = EmitMsg.this
def emit[A: c.WeakTypeTag, LL: c.WeakTypeTag]: c.Tree = {
val aa: Type = weakTypeOf[A]
val v = aa match {
case v: u.ConstantType => v.value.value
case _ =>
throw new UnsupportedOperationException(
s"type $aa is not a constant"
)
}
val ss = "" + v
val ll: Type = weakTypeOf[LL]
// if inherited from multiple traits, take the most serious one
if (ll <:< weakTypeOf[Abort]) {
c.abort(c.enclosingPosition, ss)
} else if (ll <:< typeOf[Error]) {
c.error(c.enclosingPosition, ss)
} else if (ll <:< typeOf[Warning]) {
c.warning(c.enclosingPosition, ss)
} else if (ll <:< typeOf[Info]) {
c.info(c.enclosingPosition, ss, force = true)
} else {
throw new UnsupportedOperationException(
s"type $ll is not an EmitLevel"
)
}
q"$liftOuter.create[$aa, $ll]"
}
}
}
在测试的时候,我发现直接调用宏大部分时间都是有效的:
EmitMsg.emit["ABC", EmitMsg.Error]
EmitMsg.emit["ABC", EmitMsg.Warning]
...
(这会生成正确的编译消息):
[Error] /home/peng/git/shapesafe/macro/src/test/scala/org/shapesafe/m/EmitMsgSpec.scala:19: ABC
[Warn] /home/peng/git/shapesafe/macro/src/test/scala/org/shapesafe/m/EmitMsgSpec.scala:20: ABC
one warning found
...但如果是隐式宏模式的一部分(https://docs.scala-lang.org/overviews/macros/implicits.html),则很少起作用:
it("can emit error")
type TT = EmitMsg["ABC", EmitMsg.Error]
implicitly[TT] //(EmitMsg.emit)
}
/*
Generates the following Message:
[Error] /home/peng/git/shapesafe/macro/src/test/scala/org/shapesafe/m/EmitMsgSpec.scala:15: could not find implicit value for parameter e: TT
*/
it("can emit warning") {
type TT = EmitMsg["ABC", EmitMsg.Warning]
implicitly[TT] //(EmitMsg.emit)
}
/*
Doesn't do a thing
*/
it("can emit info") {
type TT = EmitMsg["ABC", EmitMsg.Info]
implicitly[TT] //(EmitMsg.emit)
}
/*
Doesn't do a thing
*/
所以我的问题是:
调用 c.info/warning/error 是登录编译时的正确方法吗?
如果是这样,为什么它们永远不会作为隐式宏模式的一部分?
非常感谢您的建议!
更新 1刚刚在https://github.com/fthomas/singleton-ops/blob/204195838ada34de7e453401fb06810ace2c99b0/src/main/scala/singleton/ops/impl/GeneralMacros.scala#L102找到了一个案例(错误)的解决方案
val tree0 =
c.typecheck(
q"""
new _root_.scala.annotation.implicitNotFound("dummy")
""",
silent = false
)
class SubstMessage extends Transformer {
val global = c.universe.asInstanceOf[scala.tools.nsc.Global]
override def transform(tree: Tree): Tree = {
super.transform {
tree match {
case Literal(Constant("dummy")) => Literal(Constant(msg))
case t => t
}
}
}
}
val tree = new SubstMessage().transform(tree0)
annotatedSym.setAnnotations(Annotation(tree))
()
这是一个复杂的 hack,它通过在隐式函数的定义上添加 @implicitNotFound 注释来工作。到目前为止,我不知道警告和信息案例有任何类似的解决方案。然而,更简单的解决方案始终是首选