Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • 06:29
    SethTisue synchronize #1488
  • 06:29
    scala-jenkins milestoned #9765
  • 06:29
    SethTisue synchronize #1488
  • 06:26
    jxnu-liguobin opened #9765
  • 06:03

    SethTisue on 2.13.x

    add fetch (#1491) (compare)

  • 06:03
    SethTisue closed #1491
  • 00:18

    eed3si9n on main

    显示 -> 显式 Hello, I am a chinese… Merge pull request #2185 from f… (compare)

  • 00:18
    eed3si9n closed #2185
  • Sep 21 21:45
    SethTisue synchronize #1491
  • Sep 21 21:43
    SethTisue synchronize #1488
  • Sep 21 21:43

    SethTisue on 2.13.x

    2.13: add catbird (#1490) (compare)

  • Sep 21 21:43
    SethTisue closed #1490
  • Sep 21 21:22
    SethTisue edited #1490
  • Sep 21 21:21
    SethTisue synchronize #1490
  • Sep 21 21:09
    SethTisue synchronize #1491
  • Sep 21 21:07
    SethTisue synchronize #1491
  • Sep 21 21:05
    SethTisue converted_to_draft #1489
  • Sep 21 21:05
    SethTisue commented #1489
  • Sep 21 21:04
    SethTisue edited #1491
  • Sep 21 21:04
    SethTisue labeled #1491
Rob Norris
@tpolecat
There is a case for B.
I understand the irony of not understanding the error message in a blog post about improved error messages.
Tom Grigg
@griggt
I took it to mean that B cannot be shown to be disjoint from A, but I have no idea what I'm talking about.
Rob Norris
@tpolecat
That’s what it says but why not? X is exactly B which is unrelated to A other than both subtyping Any, which is true of any pair of Scala types.
Martijn
@martijnhoekstra:matrix.org
[m]
I think the point is that there may be some C extends A with B, but I don't understand what the point is anymore
Rob Norris
@tpolecat
But X is B. We know it’s not C.
Tom Grigg
@griggt
yes, it reduces if at least one of A or B is sealed
Martijn
@martijnhoekstra:matrix.org
[m]
there is more documentation on the documentation of match types: https://dotty.epfl.ch/docs/reference/new-types/match-types.html, which mentions
The compiler implements the following reduction algorithm:

If the scrutinee type `S` is an empty set of values (such as Nothing or String & Int), do not reduce.
Sequentially consider each pattern Pi
If S <: Pi reduce to Ti.
Otherwise, try constructing a proof that S and Pi are disjoint, or, in other words, that no value s of type S is also of type Pi.
If such proof is found, proceed to the next case (Pi+1), otherwise, do not reduce.
Rob Norris
@tpolecat
We know B is not a subtype of A.
Why do we care if there may some day be a value that’s both an A and a B?
Match types are absolutely baffling.
Martijn
@martijnhoekstra:matrix.org
[m]
They are covariant in some way it seems
trait Base
class Derived extends Base

type M[X] = X match {
  case Base => Int
  case Derived => String
}

def testit: Int = 7: M[Derived]
this compiles
Rob Norris
@tpolecat
...
ooookay
Martijn
@martijnhoekstra:matrix.org
[m]
😶
Bjorn Regnell
@bjornregnell
Would it be possible to also allow the new control syntax in 2.13.x? if then, while do, for do
Guillaume Martres
@smarter
@tpolecat, it's not very different from term match, if you write case x: Base => ...; case x: Derived => ... you're always going to hit the first case
in fact the whole point is that you can write your term match to mirror your match type
(and give your term match an expected type which is that match type, and have that infer correctly)
and the compiler prevents you from doing the reduction in situations where it cannot know at compile-time which case would be taken at runtime
Rob Norris
@tpolecat
Thanks, that’s helpful.
Guillaume Martres
@smarter
I'm glad :)
Rob Norris
@tpolecat
It seems awkward to me to have type and term-level representations of the same program, plus all the machinery to ensure that they are consistent, as opposed to the shapeless style where they are woven together. I will have to play around with it some more.
Seth Tisue
@SethTisue

Would it be possible to also allow the new control syntax in 2.13.x? if then, while do, for do

if a pull request came in adding that under -Xsource:3 I can't see why it wouldn't be accepted

Bjorn Regnell
@bjornregnell
@SethTisue That would be really cool in these times of migration.

if a pull request came in adding that under -Xsource:3 I can't see why it wouldn't be accepted

Would it be sensible if I create an issue on scala/bug for this enhancement?

Oron Port
@soronpo
There is something very annoying yet very satisfying in minimizing compiler crash issues. It can take up to several hours of mind-numbing work, but still it calls to me like a moth to a flame.
Jasper Moeys
@Jasper-M

It seems awkward to me to have type and term-level representations of the same program, plus all the machinery to ensure that they are consistent, as opposed to the shapeless style where they are woven together.

That's my biggest issue with the type-level stuff in scala 3. The main issue against shapeless style type-level programming seemed to be that you have to switch to a Prologish mindset. But at least in that style all the type-level and value-level code beautifully works together. With scala 3 match types and mirrors and all that stuff, types and implementations are separated and often you have to use casts to make them agree.

cmhteixeira
@cmhteixeira

Dear type-level masters, how is the below possible?

class Door[State <: DoorState] {
    def doorKey: Int = 1
    def open[Phatom >: State <: Closed.type]: Door[Open.type] = new Door[Open.type]
    def close[Phatom >: State <: Open.type]: Door[Closed.type] = new Door[Closed.type]
}

sealed trait DoorState
case object Closed extends DoorState
case object Open extends DoorState

println(aClosedDoor.open.close.close.doorKey == 1)  // How come this is compiling?

I am able to close a closed door. The compiler is not enforcing type bounds
Is this a bug? Am I missing something?
(also published on scala/scala gitter channel)

Rob Norris
@tpolecat
You can make Open and Closed extend some supertrait and parameterized it as [A <: DoorState] if you like but it probably doesn't matter.
Upper bounds are rarely necessary.
Guillaume Martres
@smarter
@tpolecat conversation on this issue was moved to scala/scala
Rob Norris
@tpolecat
ah
Guillaume Martres
@smarter
(spoiler: it's fixed in dotty)
cmhteixeira
@cmhteixeira
@tpolecat As smarter says, its on scala/scala. Thanks. The question was also more with the objective of understanding why that was happening. I was already aware of the implicit alternative. I appreciate it though!
For reference, we ended up opening this ticket: scala/bug#12413
Rob Norris
@tpolecat
ok
Swoorup Joshi
@Swoorup
this should ideally fail type check isn’t it?
import scala.language.implicitConversions

type Timeframe = "1m" | "2m" | "1H"
type TimeframeN = 1 | 2 | 60

def manualConvertToN(tf: Timeframe): TimeframeN = tf match
  case "1m" => 1
  case "2m" => 2
  case "1H" => 60
  case "4H" => 60 // incorrect
Rob Norris
@tpolecat
Seems like a bug to me. Looks like it's checking that the LHS is a string (case true => ... fails) but it needs to check that it's a Timeframe.
Swoorup Joshi
@Swoorup
yep logged a bug: lampepfl/dotty#12805
seems to prematurly widen
sinanspd
@sinanspd
hmm
aren't Enums a better use for this ?
like i am actually not sure what the expected behavior here should be
You are creating a union type with concrete values, no?
sinanspd
@sinanspd
nvm it seems like this is in fact expected to work
Swoorup Joshi
@Swoorup
they are good near the boundaries interacting with other languages, or doing things like
def acceptFontSize(fontSize: 9 | 10 | “11rem”) = ???
widening/or unwidening are purely a compile time construct