Add type constructors

[ekmett]

Pull Request Description

This is a work-in-progress draft PR of what we need to get type constructors to exist for real in the compiler.

Important Notes

This removes ResolvedTypeName, and merges it with ResolvedName. It then expands AtomConstructor with parentType and variants information. variants being the name used internally for member types of the parent type.

Open issues:

Status: This now pulls the information from the right place, but doesn't yet populate all of that information.

We need to populate the parent side of the code I ported from Marcin's branch and the child portion of my code to make the two engines work together.

I've yet to fully fix the test suite, though I started with a couple of errors due to resource exhaustion that don't appear to be caused by me.

Checklist

Please include the following checklist in your PR:

• The documentation has been updated if necessary.
• All code conforms to the Scala, Java, and Rust style guides.
• All code has been tested:
  • Unit tests have been written where possible.
  • If GUI codebase was changed: Enso GUI was tested when built using BOTH ./run dist and ./run watch.

[JaroslavTulach]

Please avoid adding Scala data structures into the engine code that is supposed to be Partially Evaluated. Converting to plain @CompilationFinal(dimension=1) AtomConstructor[] shall be enough.

Please make the parentType and variants technically immutable after parsing or make sure changes to the fields properly invalidate partially evaluated code depending on them.

[JaroslavTulach]

I am trying to reproduce the current (with e47d05a) failures by doing:

enso$ sbt --java-home /graalvm-ce-java11-21.3.0/
sbt:enso> buildEngineDistribution
sbt:enso> buildEngineDistribution
sbt:enso> project runtime
sbt:runtime> testOnly *TextTest*
[info]   - should support converting values to display texts *** FAILED *** (62 milliseconds)
[info]     org.enso.interpreter.test.InterpreterException: Compile error: The name Panic is ambiguous. Possible candidates are:
[info]     Type Panic defined in module Standard.Base.Error.Common;
[info]     Type Panic defined in module Standard.Base.Error.Common;
[info]     at <enso>.argument<0>(Test:15)
[info]     at <enso>.argument<1>(Test:15)
[info]     at <enso>.Test.main(Test:15)
[info]     at org.graalvm.sdk/org.graalvm.polyglot.Value.execute(Value.java:841)
[info]     at org.enso.polyglot.Function.execute(Function.scala:16)
[info]     at org.enso.interpreter.test.InterpreterRunner.$anonfun$eval$1(InterpreterTest.scala:179)
[info]     at org.enso.interpreter.test.InterpreterException$.rethrowPolyglot(InterpreterException.scala:18)
[info]     at org.enso.interpreter.test.InterpreterRunner.eval(InterpreterTest.scala:177)
[info]     at org.enso.interpreter.test.InterpreterRunner.eval$(InterpreterTest.scala:174)
[info]     at org.enso.interpreter.test.semantic.TextTest.eval(TextTest.scala:5)

There are other tests failing with similar errors. Finding the cause could fix them all.

I can debug this by telling my IDE to Socket Listen on port 5005 and then executing:

sbt:runtime> withDebug --debugger testOnly -- *TextTest*

when the breakpoint stops at engine/runtime/src/main/scala/org/enso/compiler/pass/resolve/MethodDefinitions.scala:110 I can see err.candidates list contains the same Module as well as BindingsMap$ModuleReference$Concrete twice.

Eliminating the duplicity by:

enso.master$ git diff
diff --git a/engine/runtime/src/main/scala/org/enso/compiler/data/BindingsMap.scala b/engine/runtime/src/main/scala/org/enso/compiler/data/BindingsMap.scala
index 4e90a9fc7..d82cc6580 100644
--- a/engine/runtime/src/main/scala/org/enso/compiler/data/BindingsMap.scala
+++ b/engine/runtime/src/main/scala/org/enso/compiler/data/BindingsMap.scala
@@ -195,7 +195,7 @@ case class BindingsMap(
   private def handleAmbiguity(
     candidates: List[ResolvedName]
   ): Either[ResolutionError, ResolvedName] = {
-    candidates.distinct match {
+    candidates.distinctBy(_.module) match {
       case List()   => Left(ResolutionNotFound)
       case List(it) => Right(it)
       case items    => Left(ResolutionAmbiguous(items))

seems to fix the TextTest. Now:

sbt:runtime> testOnly *TextTest*

passes OK.

[JaroslavTulach]

What does the _Tp suffix stand for? ToP level type?

[radeusgd]

I'd like to understand this rename.

Is it temporary? If not, I'm a bit skeptical about the _Tp suffix. Also we'd probably need to change all type signatures taking Any to be Any_Tp instead, right? (And analogously for all the others)

[JaroslavTulach]

Maybe it is _Tp as _TemPorary`?