Breaking Changes

[magnus-madsen]

We are currently in a period of breaking changes. I just wanted to collect these in one place.

Recent, current, and on-going breaking changes:

• (Records) We recently changed the record type syntax from f :: Int32 to f = Int32.
• (Concurrency) We recently changed the syntax for using channels and spawning threads. This is now mostly in place.
  • Please avoid the use of the Static region (esp. for channels / processes).
• (Casts) We recently updated the cast syntax (and some more changes are expected).
• (Namespaces) We are replacing the whole namespace mechanism. Three changes will become visible:
  • We will switch to a uniform use syntax, e.g. use X.Y.Z.W.
  • Enums cases will have to be explicitly used.
  • The namespace A/B/C syntax will become mod A.B.C.
• (Effects) We will (eventually) be merging the two effect systems into one.
• (Result) We are discussing some changes to Result (and how we can make it into a Monad etc.)
  • This may affect let* etc.
• (Rel and Lat pseudo type aliases) We plan to deprecated rel and lat in favor of using an ordinary type alias (but this is not yet possible).

EDIT: The good news is that the "core" language (other than channels/processes) remain stable.

[magnus-madsen]

CC @stephentetley I hope this is not too painful.

[JonathanStarup]

Maybe we could also use the minor version more for non-breaking versions instead of 0.X.0 for everything.

[stephentetley]

Thanks Magnus, the changes are fine.

One thing that is a problem is "unexpected non-enum" in Resolver.scala gets throw crashing the compiler if use statements try to access enum constructors that don't exist. I don't think Flix checked use imports before so I have quite a few cases in my libraries that were never caught as code got refactored. The stacktrace from the compiler crash doesn't help locate the problem so I've had to append the source location to the "unexpected non-enum" message in a hacked version of the compiler.

[magnus-madsen]

Thanks Magnus, the changes are fine.

One thing that is a problem is "unexpected non-enum" in Resolver.scala gets throw crashing the compiler if use statements try to access enum constructors that don't exist. I don't think Flix checked use imports before so I have quite a few cases in my libraries that were never caught as code got refactored. The stacktrace from the compiler crash doesn't help locate the problem so I've had to append the source location to the "unexpected non-enum" message in a hacked version of the compiler.

I suspect this will be fixed. We are kind of "in between" the old and new system.