lampepfl/dotty

Research platform for new language concepts and compiler technologies for Scala.

Miles Sabin

@milessabin

That would have happened whatever anyone did or said.

nafg

@nafg

yes

Guillaume Martres

@smarter

I guess no one is scared by-Y flags the same way no one is scared by import bla.experimental

nafg

@nafg

cf. macros

The bottom line is if you give people a feature they will enjoy it and then if you take it away they will complain

Miles Sabin

@milessabin

It's better in than out. macro-paradise relied on being able to monkey-patch the compiler it's loaded into as a plugin. It can break with the tiniest changes to the compiler.

nafg

@nafg

disclaimers don't change that much

That seems to be a general philosophy of dotty

Guillaume Martres

@smarter

The question is, how many other macro annotations like @Lenses can't be replaced by typeclass derivation / other features, and how popular are they

nafg

@nafg

Slinky?

Guillaume Martres

@smarter

what does slinky do ?

nafg

@nafg

it's a scalajs facade for React

Guillaume Martres

@smarter

but what does it use macro annotations for

nafg

@nafg

Something I would do if it was supported, in scalajs-react ( a different facade for it, which I use) you have this pattern of object Component { case class Props(...); val component = ... }; Component.component(Component.Props(...))) and it would be great to automatically generate an apply method matching the signature of Props that passes them to the component

https://slinky.dev/docs/writing-components/

This message was deleted

Guillaume Martres

@smarter

Aren't react components obsolete now ? :)

nafg

@nafg

Huh?

Oh you mean because of hooks?

Guillaume Martres

@smarter

https://reactjs.org/docs/hooks-faq.html

yeah

nafg

@nafg

heh

Kai

@neko-kai

@smarter

might be better to not have something that have something which people are unhappy with: scala/bug#10766

The subtype operator, if addressed on its own, is implementable without global state / unavoidable concurrency-problems. Unfortunately that might not be enough for hypothetical corporations with legacy scala-reflect code, but that would be enough for my company… And we’re actually implementing it right now – probably won’t copy the entire Scala typer and limit the checks to the common kinds of the types that actually appear in our codebase.

https://github.com/freechipsproject/chisel3 uses a few macro annotations, not sure if these would be hard to replace; seems like they’re just for renaming defs

Guillaume Martres

@smarter

A full implmentation of subtyping is extremely complex

Alexander Myltsev

@alexander-myltsev

what happenned to backticks? I’m trying to address the Akka's ContentTypes.`text/html(UTF-8)` , but the compiler states there is no such method

David Gregory

@DavidGregory084

Is there any plan to implement syntactic / untyped macros as part of the new macro system?

Kai

@neko-kai

@DavidGregory084 There’s inline whitebox - https://dotty.epfl.ch/docs/reference/metaprogramming/inline.html#specializing-inline-whitebox

Guillaume Martres

@smarter

@alexander-myltsev you're probably running into this bug: lampepfl/dotty#5936

@DavidGregory084 untyped macros is something we would like to not have

Guillaume Martres

@smarter

In fact, dotty doesn't even let you write a compiler plugin that runs before the typer phase

David Gregory

@DavidGregory084

Ah OK, why is that? It occurred to me that it would allow features like for comprehensions to be removed from the compiler

Guillaume Martres

@smarter

Because it effectively means you can make up your own language

Which is nice for research (and in fact we support "research plugins" which can run before typer but can only be used with nightly builds of the compiler), but not great for users or tools

Kai

@neko-kai

That just means better-monadic-for users will run the research compiler :^)

Guillaume Martres

@smarter

Better-monadic-for users should make a SIP so we can get the improvements in the compiler.

David Gregory

@DavidGregory084

Yeah I'm looking at that

Guillaume Martres

@smarter

Maybe that would have already happened if untyped macros were not a thing.

@DavidGregory084 looking forward to it!

David Gregory

@DavidGregory084

It's not trivial though, even the simplest changes (e.g. removing the superfluous map call) break lots of standard library code

(If it's implemented by changing the desugaring after parsing)

So I was wondering whether there might be other approaches in the long term

Guillaume Martres

@smarter

Scala 3 is trying to move Scala away from being a toolbox for making your own language

@DavidGregory084 you don't have to produce a document with all the problems worked out, even identifying the problems is already useful to start the process

David Gregory

@DavidGregory084

Yeah sure, I'm just working through what it's like to implement the simplest possible solution for each of the major features right now and identifying how much breakage it causes

Guillaume Martres

@smarter

Cool

Alexander Myltsev

@alexander-myltsev

@alexander-myltsev you're probably running into this bug: lampepfl/dotty#5936

@smarter ok, thanks

is that another compiler bug?

[error] (Compile / compileIncremental) java.lang.AssertionError: assertion failed: denotation module class annotation invalid in run 1. ValidFor: Period(1..5, run = 2)
[error] Total time: 0 s, completed Jul 24, 2019 10:33:32 PM

Guillaume Martres

@smarter

@CremboC Your confusion on the syntax of delegate is a useful data point on its current usability, I encourage you to comment on https://contributors.scala-lang.org/t/proposal-to-add-implied-instances-to-the-language/3070 with your experience

@alexander-myltsev if the compiler crashes, it's a bug.

nafg

@nafg

@benhutchison maybe lampepfl/dotty#6919 can help?

Where communities thrive

People

Repo info

Activity