Polish to_text on intersection types
#12192
Conversation
JaroslavTulach
added
the
CI: No changelog needed
JaroslavTulach
requested review from
jdunkerley,
radeusgd,
GregoryTravis,
AdRiley,
marthasharkey,
4e6,
hubertp and
Akirathan
as code owners
JaroslavTulach
changed the title
to_text on _intersection types_to_text on intersection types
…s via StructsLibrary
JaroslavTulach
force-pushed
the
wip/jtulach/MultiValueToText11827
branch
from
91597b5 to
3196767
Compare
| @@ -302,7 +302,7 @@ Object doMultiValue( | |||
| if (fnAndType != null) { | |||
| var ctx = EnsoContext.get(this); | |||
| if (ctx.getBuiltins().any() != fnAndType.getRight()) { | |||
| var unwrapSelf = castTo.findTypeOrNull(fnAndType.getRight(), self, false, false); | |||
| var unwrapSelf = castTo.findTypeOrNull(fnAndType.getRight(), self, true, false); | |||
There was a problem hiding this comment.
Reorders the intersection types to make sure the value of the type we dispatch the method to is at EnsoMultiValue.firstDispatch.
| ab.a_id_unchecked . is_a A . should_be_true | ||
| ab.a_id_unchecked . is_a B . should_be_true | ||
| ab.a_id_unchecked.a_method . should_equal "A method" | ||
| ab.a_id_unchecked.b_method . should_equal "B method" | ||
| Test.expect_panic No_Such_Method (ab.a_id_unchecked.b_method) |
There was a problem hiding this comment.
As a result of 3196767 we have to make sure that dispatching a method on self moves the type of the method to the firstDispatch. This is done by https://github.com/enso-org/enso/pull/12192/files#r1935731256, but:
- right now the reordering hides all the other types
- semantically it is not necessary to hide the other types
- we could keep them visible
It is just simpler to hide the other types as there is an implementation doing that available right now. Opinions? Wanna better behavior at no slowdown? CCing @radeusgd.
There was a problem hiding this comment.
I'm not 100% sure yet as all of this has many implications that only come out in practice.
But at first, I think that hiding actually seems consistent with type checking requirements - inside of method A.foo self I actually do want to assume that self:A - so it makes sense for other 'parts' to be hidden.
|
Now, when dispatch of an instance method allows To do so we have to change the "builtin prelude code" - e.g. the way we prepare the values before invoking the |
|
Jaroslav Tulach reports a new STANDUP for yesterday (2025-02-04): Progress: .
|
|
Jaroslav Tulach reports a new STANDUP for yesterday (2025-02-05): Progress: .
|
|
Jaroslav Tulach reports a new STANDUP for yesterday (2025-02-06): Progress: .
|
|
Jaroslav Tulach reports a new STANDUP for today (2025-02-07): Progress: .
|
#12201) - as #12192 (comment) states: - there is a need to extract values from `EnsoMultiValue` in the _"builtin method prelude code"_ - to enable _specializations based on type of arguments_, we need to wrap such _prelude_ by a `Node` - hence introducing `ArgNode` & co. - it kinda replaces the _"static code emitted"_ by annotation processor - with Truffle `@Specialization`s done inside of `ArgNode` - making this all more Truffle-like
Pull Request Description
Fixes #11827 by better coordination of dispatch to
EnsoMultiValue.Checklist
Please ensure that the following checklist has been satisfied before submitting the PR:
Scala,
Java,