kotlinlang #multiplatform

It turns out even spaces are not allowed in backticked identifiers in the JS target. But IntelliJ offers an autocorrect to rename test methods to backtick identifiers with spaces between words!

spierce7

07/11/2017, 4:26 AM

kapt only works for JVM right? It doesn't work for javascript does it?

sandwwraith

08/09/2017, 10:23 AM

Hello! How can I configure stdlib dependency in common module in IDEA? Currently it complains about "Kotlin not configured", but suggests only Java or JavaScript configurators

karelpeeters

08/14/2017, 8:16 PM

I want to have a project that runs on the JVM and one that's embedded in a webpage as JS, but they share some common code. How would I organise this with gradle?

yoavst

08/18/2017, 1:59 PM

https://youtrack.jetbrains.com/issue/KT-17005#tab=Changes I'm on 1.2-M2 and still doesn't work

08/20/2017, 7:37 PM

Not yet

ilya.gorbunov

08/21/2017, 5:35 AM

What do you like least in the java.io.* API?

sandwwraith

08/22/2017, 1:23 PM

I'm trying to make common abstraction over

java.io.Writer

class. Given the

header abstract class Writer protected constructor()

in common module and

impl typealias Writer = java.io.Writer

in jvm module, it fails to compile with:

The following declaration is incompatible because some members are not implemented:
    public typealias Writer = Writer
No implementations are found for members listed below:

    protected constructor Writer()

    The following declaration is incompatible because visibility is different:
        protected/*protected and package*/ constructor Writer()
    The following declaration is incompatible because number of value parameters is different:
        protected/*protected and package*/ constructor Writer(p0: Any!)

Is it a bug or I'm missing something? Don't know how to make "protected/*protected and package*/ constructor"...

Deactivated User

09/19/2017, 11:54 PM

while doing this I have noticed that a project called kotlinx.typedarrays or so using Js Typed Arrays+ArrayBuffer in JS, NIO Buffers in Java and pointers in Konan would be nice. Since I’m going to implement it (JS and JVM), if you like the idea I can extract it to a separate project/repo and then move/donate to Kotlin org so it can evolve there

Deactivated User

09/20/2017, 8:25 AM

I noticed the

impl -> actual

rename yesterday in the github’s history. Is it going to be available in 1.1.50 final? Also I can’t find

KClass

inside common. Right now I have this:

// @TODO: Hack since we don't have access to KClass from common
typealias KClass<T> = Any

Deactivated User

09/20/2017, 9:07 AM

Other nice thing I would suggest: Allow to actual implement a header with a default implementation. A possible use case is this one:

header fun ByteArray.copyRangeTo(srcPos: Int, dst: ByteArray, dstPos: Int, count: Int) {
	for (n in 0 until count) dst[dstPos + n] = this[srcPos + n]
}

Here I provide a generic implementation, that can be improved using System.arraycopy in JVM or the former for JS typed arrays, but as an optimization, not mandatory. So for example I wouldn’t have to implement for Konan at the very beginning and implement it later when optimizing things

orangy

09/22/2017, 12:10 PM

There is no “static” in some runtimes, e.g. in JavaScript.

Deactivated User

09/24/2017, 10:42 PM

Do you know why I’m getting a

Duplicate source root

? https://travis-ci.org/korlibs/korio/builds/279319112#L724 ---- https://github.com/korlibs/korio/blob/d1fdb1e153a9fe3644f3b1e2883d0f6fd467e3eb/build.gradle

Deactivated User

09/28/2017, 5:57 AM

Wishlist for Kotlin 1.2 (conversion between int-float and long-double in all targets including JS). Here a JS implementation I just did using typed arrays: https://github.com/korlibs/korma/blob/992800b59a333fec7c8cb942fed17b9bfab85128/korma-js/src/main/kotlin/com/soywiz/korma/math/NativeMath.kt#L5-L29

Deactivated User

09/28/2017, 6:01 AM

Also System.arraycopy and Arrays.fill would be great. I have seen Kotlin code for it. If nobody is doing this. I can do a PR adding them. I would suggest something like

Array.copyTo(srcPos, dst, dstPos, count)

. As far as I understand the right place to put it would be: https://github.com/JetBrains/kotlin/blob/4a0ad54829aa7b2e4c22eca08dbd42131b732c9c/libraries/tools/kotlin-stdlib-gen/src/templates/SpecialJVM.kt

gaetan

09/30/2017, 12:34 AM

@orangy maybe it’s time to change this channel description 😉

gaetan

09/30/2017, 2:19 PM

@raniejade I have the same behaviour on https://github.com/data2viz/data2viz No problem with gradle compilation but not working in intellij. However, it’s better now with the beta than previously. Before lastest release I had problems like “unresolved reference” during parsing of code in editor. Now it’s ok during edition and I can live without intellij idea compilation.

ilya.gorbunov

09/30/2017, 8:02 PM

@raniejade you need to have a project that actually implements your

common

project. Even if you have no platform specific code, you have to create a platform-specific project that implements common and produces the artifacts for that platform. Then jvm project needs also to have a compile dependency on that platform project with the compiled

Foo

class.

eddie

10/01/2017, 9:03 PM

@ilya.gorbunov posting passwords in public channels? 🙂

ilya.gorbunov

10/01/2017, 9:26 PM

Though

actual

modifier is technically can be omitted, its presence helps to improve the IDE experience.

ilya.gorbunov

10/01/2017, 11:48 PM

@uli That's a tempting idea. We considered it, but decided to start from something simpler. The problem with these semi-implemented library modules is that it might be impossible to compile them to bytecode for JVM because some actual types are missing and unknown until provided by the "terminal" modules.

orangy

10/03/2017, 7:45 AM

Given that we still need two modifiers, if you were designing it, would you come up with different names? Bikeshed in a thread :thread-please:

eddie

10/03/2017, 3:30 PM

@orangy so I wanted to see a KEEP because 1) it seemed that would be the way changes (especially ones as significant as this) would be introduced to Kotlin going forward and 2) so far I hear that you need two modifiers for some reason, but no one is able to explain why much beyond "well it helps in the IDE", and that's just not a valid reason to bloat a language IMO.

Title

eddie

10/03/2017, 3:30 PM

Ruckus

10/03/2017, 3:36 PM

I'd just like to point out (again) that "it helps the IDE" is incredibly important in a language like Kotlin where the tooling is as much a part of the language experience as the code itself. I'm not saying it's the only reason, but don't dismiss it as insignificant.

abreslav

10/03/2017, 4:21 PM

+1 The IDE is really important So far we’ve been keeping it safe and required a handshake between expecting/actual code wo make sure no changes can mess things up without the user noticing. This is much analogous to why we require the

override

modifier

➕ 5

eddie

10/03/2017, 5:22 PM

@Ruckus @abreslav I don't mean to dismiss it; I really enjoy the integration with IntelliJ. However, I'd argue that you can remove the keyword on the implementation side while retaining the help the IDE would provide (like a special syntax highlighting variation or something in the gutter, for instance). This way the IDE could provide help in the same way that it provides Inspections and Intentions: not everyone needs or wants all of those turned on, so they're configurable. If you require a keyword in the language you lose that flexibility.

IntelliJ's newer parameter hints feature is another example of this done well, IMO.

abreslav

10/06/2017, 10:49 AM

It’s more about what can and c=not be done fast in the IDE. Some things can’t be computed quickly if they are not in the code

eddie

10/08/2017, 12:24 AM

But you already have to mark the module in your gradle build script as one that provides implementations, no? At the end of the day, I just need to know if an expected implementation is available or not.

View count: 1