feat: support for traits as type arguments with variance on datatypes in Java #3072
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
fabiomadge
previously approved these changes
Nov 29, 2022
| if (!isCoCall) { | ||
| // For an ordinary constructor (that is, one that does not guard any co-recursive calls), generate: | ||
| // Dt.<T>create_Cons( args ) | ||
| wr.Write($"{dtName}.{typeParams}{DtCreateName(ctor)}({arguments})"); | ||
| } else { | ||
| wr.Write($"new {dt.CompileName}__Lazy("); | ||
| wr.Write("() -> { return "); | ||
| wr.Write($"new {DtCtorName(ctor)}({arguments})"); | ||
| wr.Write($"new {DtCtorName(ctor)}{typeParams}({arguments})"); |
There was a problem hiding this comment.
This looked fishy to me but is fine; Nevertheless, I found a bug along the way.
module M {
codatatype Stream<T> = Next(shead: T, stail: Stream)
}
function method CoUp(n: int, b: bool): M.Stream<int>
{
if b then
CoUp(n, false) // recursive, not co-recursive, call
else
M.Next(n, CoUp(n+1, true)) // CoUp is co-recursive call
}
method Main(){
print CoUp(0, false), "\n";
}It's a one-liner. L1968 needs a public.
There was a problem hiding this comment.
Good find! It was handled incorrectly in 3 compilers. I fixed it.
| var t: set<nat> := s; | ||
| var u: set<Even> := s; | ||
| t := u; | ||
| u := t; |
There was a problem hiding this comment.
Suggested change
| u := t; | |
| s := t; |
Both seem like a decent idea.
There was a problem hiding this comment.
For further testing, I prefer u, since the types Even and nat are not subtypes of each other.
There was a problem hiding this comment.
I added your test, too (here and in a few more places).
docs/dev/news/3072.feat
Outdated
| @@ -0,0 +1 @@ | |||
| feat: support for traits as type arguments with variance on datatypes in Java | |||
There was a problem hiding this comment.
Suggested change
| feat: support for traits as type arguments with variance on datatypes in Java | |
| feat: support for traits as type arguments by allowing variance on datatypes in Java |
There was a problem hiding this comment.
I also add "fully allowing"
Co-authored-by: Fabio Madge <[email protected]>
Co-authored-by: Fabio Madge <[email protected]>
Co-authored-by: Fabio Madge <[email protected]>
Co-authored-by: Fabio Madge <[email protected]>
fabiomadge
approved these changes
Nov 29, 2022
robin-aws
added a commit
that referenced
this pull request
Nov 30, 2022
…t now (#3141) Ideally the feature would have been removed in the PR that implemented it in Java (#3072), but unfortunately the `Feature` mechanism only allows a compiler to cleanly opt-out of supporting a feature, and we don't yet have a complementary mechanism to test that a compiler that claims to not support a feature does not in fact support that feature! <small>By submitting this pull request, I confirm that my contribution is made under the terms of the [MIT license](https://github.com/dafny-lang/dafny/blob/master/LICENSE.txt).</small>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR removes that pesky "not supported" error for Java where it wouldn't allow a
traitto be passed in as a variant type parameter of a (co)datatype.Fixes #2013
The solution is pretty simple. Basically, it just emits casts when needed. To be accepted by Java's static type checker, these casts first upcast to
Objectand then downcast to the desired type. Since the JVM doesn't keep track of type parameters, it'll never know what hit it. (In particular, this means that we don't need the expensiveDowncastClonemechanism that we're using for C# to satisfy .NET's more informed runtime.)To make it work, this PR also makes more precise the
IsTargetSupertypemethod for Java. In particular, to check ifA<X, Y>is a supertype ofB<U, V, W>, first keep converting the latter up thetraitparents ofBuntil it gets toA(if ever). This is already done for all the compilers. Call the resultA<X', Y'>. For the other compilers, one would now start to see if the parametersX, Yare target supertypes ofX', Y'with variance in mind (for example, if the first type parameter is contravariant, we would actually check ifX'is a supertype ofX). For Java, however, we continue with variance in mind only ifAis a collection type (because those already have run-time support for variance, using Java's bounded existential types). For non-collection types, we instead check ifX, Yare target-type equal toX', Y'. If they are not, the compiler callsEmitDowncast. (To read these in the code, look atEmitDowncastIfNecessaryinSinglePassCompiler.cs.)We could have used simple casts like this for collection types, too. But since I already have those, I left those alone.
By submitting this pull request, I confirm that my contribution is made under the terms of the MIT license.