Allow single-line lambdas after :

[odersky]

Previously, we need to indent after the arrow, e.g.

xs.map: x =>
  x + 1

We now also allow to write the lambda on a single line:

xs.map: x => x + 1

The lambda extends to the end of the line.

This is a trial balloon to see whether anything breaks. If we want to pursue this it needs a SIP.

[soronpo]

Will it also work for self types?

trait Foo: self =>
end Foo

[lihaoyi]

I wonder if xs.map: case Some(x) => x + 1 could be made to work as well?

[som-snytt]

The case for case is at #22016. I think the parens SIP opens the floodgates of consistency.

[lihaoyi]

If we need a SIP for this, maybe we could fold it into https://contributors.scala-lang.org/t/sip-xx-allow-partial-function-literals-to-be-defined-with-parentheses/7207/16

[odersky]

if we need a SIP for this, maybe we could fold it into https://contributors.scala-lang.org/t/sip-xx-allow-partial-function-literals-to-be-defined-with-parentheses/7207/16

Yes, I think the two go well together. This PR now also implements the single-case version of sip-xx. I have some reservations against admitting multiple cases between parentheses. I'll explain more on the SIP.

[odersky]

Will it also work for self types?

Right now, no. I'd like to re-think self types in general. So I'm not keen on fiddling with them until that's done.

[GLI-RK0]

Awesome. Will this extend to lambdas with by-name parameters and/or context parameters?

You can currently write

// after delay, execute fn
def delay(ms: Double)(fn: => Unit) = ???

delay(300):
  println("hello")

Will we also be able to write

def delay(ms: Double)(fn: => Unit) = ???

delay(300): println("hello")

and

def delay(ms: Double)(fn: (timestamp : Double) ?=> Unit) = ???

delay(300): println(s"time ${timestamp}")

?

[odersky]

Awesome. Will this extend to lambdas with by-name parameters and/or context parameters?

Currently, no. We should discuss what we want. Some choices:

1. no extension to by-name or context-parameters
2. allow : syntax for these parameters only
3. allow : syntax for all arguments.

Possible reason for (2): The situation is similar to single-line-lambdas with parameters.
Possible reason for (3): We don't distincguish syntactically between by-name and by-value elsewhere.

[lihaoyi]

If we allow delay(300): println("hello"), wouldn't we then have to allow println: "hello", and wouldn't println: "hello" be indistinguishable from a type ascription?

[GLI-RK0]

If we allow delay(300): println("hello"), wouldn't we then have to allow println: "hello", and wouldn't println: "hello" be indistinguishable from a type ascription?

Sorry I might be missing something, why would we have to allow println: "hello"? I was under the impression that : would only trigger an attempt to interpret the following rhs as a lambda when the parameter is typed as such? I assume that's what Martin meant by option 2?

[lihaoyi]

Sure, println is not a higher order function, and delay(300): is. But do you know that information during parsing? Because this distinction between "this is a type ascription" and "this is a function taking a code block on the right" needs to happen super early even before typechecking happens.

If you don't like println, how about delay(300): "hello": is this a function call with a type ascription, or a function call taking two parameter lists?

[odersky]

@lihaoyi Yes, indeed. Allowing arbitrary expressions after: would conflict with type ascriptions. So I think we need to bury the idea.

[odersky]

I believe we have a problem with the new build. It does not pick up the scala-library changes in the bootstrapped build.
I see:

-- [E008] Not Found Error: tests/pos/closure-args.scala:1:29 -----------------------------------------------------------
1 |import language.experimental.relaxedLambdaSyntax
  |                             ^^^^^^^^^^^^^^^^^^^
  |                             value relaxedLambdaSyntax is not a member of object language.experimental

Yet, relaxedLambdaSyntax was added to language.experimental in the commit.

[som-snytt]

The reproduction was a parser crash instead of (some) error (only under -rewrite).

I haven't looked yet, but I wonder if it is worth trying to preserve the migration help (in 2025 almost '26)

parentheses are required around the parameter of a lambda

They can use copilot.

[odersky]

This was accepted as ready or implementation in the SIP meeting of Oct 24th, 2025.

[WojciechMazur]

No new regressions found in the OpenCB when comparing against current main