Support coercing on a function type #6821
Comments
|
One observation: the coercion in the example seems to work if the argument type is not polymorphic, e.g. |
|
Ah, that makes sense. Unfortunately the way I'm trying to use it with Array sort requires polymorphism. For now I've just used an obj magic. |
See #6821 Free variables are not allowed in coercion, and if that happens, simple unification is performed without ever attempting coercion. This seems to go back a very long time. There's probably a good reason why free vars were not allowed. At the same time, there are no objects/classes supported in the language anymore, so it's not clear how those reasons would adapt. This just marks the place where one could investigate.
See #6821 Free variables are not allowed in coercion, and if that happens, simple unification is performed without ever attempting coercion. This seems to go back a very long time. There's probably a good reason why free vars were not allowed. At the same time, there are no objects/classes supported in the language anymore, so it's not clear how those reasons would adapt. This just marks the place where one could investigate.
See #6821 Free variables are not allowed in coercion, and if that happens, simple unification is performed without ever attempting coercion. This seems to go back a very long time. There's probably a good reason why free vars were not allowed. At the same time, there are no objects/classes supported in the language anymore, so it's not clear how those reasons would adapt. This just marks the place where one could investigate.
|
@TheSpyder any more tests for #6828 ? |
|
@cristianoc the second example is the one I pulled from my codebase, I haven't come across any others (although I haven't really been looking). I appreciate you looking at this! |
Great, that example is covered already. |
* Test: allow free vars in types for coercion. See #6821 Free variables are not allowed in coercion, and if that happens, simple unification is performed without ever attempting coercion. This seems to go back a very long time. There's probably a good reason why free vars were not allowed. At the same time, there are no objects/classes supported in the language anymore, so it's not clear how those reasons would adapt. This just marks the place where one could investigate. * Add tests of type coercion with free vars. * Clean up code for coercion and free variables. * Update CHANGELOG.md * Remove unused code for self coercion. * Remove unused error that suggests double coercion.
|
Done in #6828 |
I recently discovered that coercion works for variants where everything is represented by the same primitive type:
https://rescript-lang.org/docs/manual/latest/variant#coercing-variants-to-primitives
This is great, but I have a function that returns one of these variants and I can't coerce that. It seems to work for no-argument functions, but not functions with arguments.
https://rescript-lang.org/try?version=v11.1.1&code=C4TwDgpgBAhgzgSQHbCgXigHygAXgCgEYBKKAeSWmzznwCZSAVAdwHstcCBmJgCwCcIEAFDCANhFStKyYAC5YiFOnKVxkqMFayV+aRB1yAfFACWKYqImpBwAK78ksgKpJTqDPlJoTFEdahbBycUAGVzAHMJXQB9b181dVQtWQAxQl0gxxc3VGMofDtc9BNzYGJLAJSUVLpMyWDZcKQo6Hz8AHIYErMLYiA
The specific example I'm trying to model is array sort, so I'd need it to work on 2-argument functions as well.
https://rescript-lang.org/try?version=v11.1.1&code=C4TwDgpgBAxg9gWzAQwE4EsDOcB2UC8AUFFAD5QACymAFALQCMAlFADISaYAqAFsjsTKVqNAAwsAogEcArsgA2g8lVrMoAcVQRkwCKl79ChOJDwApTADoAgqlTIQAJkLyIwKNlTv8UGmlQANFAAfgBmAFy+AOTIQTEs+AB8sIgoGNg4CYmCnsAAkjgACvLIMBAA6ujAPH52dImY8uhlAGKoiGJMQTShUOHJ6DjATExAA
The text was updated successfully, but these errors were encountered: