kotlinlang #arrow-contributors

I believe this file is very specific to JVM, and we can't move it to multiplatform. Do you guys have any idea on this? Is this file actually needed on all subprojects? I'm only touching

arrowAnnotations

LeoColman

04/30/2019, 10:21 PM

So I don't implement

GenOf

at Multiplatform/Common, but I do it in JVM

Bob Glamm

05/24/2019, 12:28 PM

If you'd prefer I make them consistently authored by 'glamm' just revert and I'll put everything together from an entirely separate machine. Messing around switching git credentials on OS X is aggravating, at best

Bob Glamm

06/04/2019, 3:06 AM

@raulraja @pakoito @simon.vergauwen ok, based on a conversation with @raulraja: https://github.com/arrow-kt/arrow/pull/1472/files to change effect to match cats; rationale in pull request. Please let me know if I need to change anything

Bob Glamm

06/04/2019, 3:09 AM

and ++ on

gradlew

. I guess I could have tried downloading gradle-5.0 and running that instead. Gradle 5.2.1 and 5.4.1 definitely fail the same way, though

pakoito

06/13/2019, 9:49 AM

https://github.com/arrow-kt/arrow/issues/1484

pakoito

06/13/2019, 9:49 AM

one super easy ticket to add a function and its docs!

Imran/Malic

06/14/2019, 10:25 AM

Hey guys I am working on this PR https://github.com/arrow-kt/arrow/pull/1481 And I detected the Problem why I always get this error:

e: error: Arrow's annotations can only be used on Kotlin classes. Not valid for error.NonExistentClass

In the Debug stacktrace it occurs that the failure happens in

arrow-core-extensions

more specifically, the kapt Task

kaptKotlin

I have 2 assumptions: - Either there is a problem with codegen and I need to make this PR bigger than expected. Moving everything out of arrow-extras-data. - Or I am missing an import in gradle

-.kt

carbaj0

06/14/2019, 3:07 PM

the 0.9.1-SNAPSHOT doesn´t work, do i wait to the stable version?

Imran/Malic

06/19/2019, 10:10 AM

I suggest reverting the last commit in masters and start from there, because if we merge right away others will have the same problem. And I am not sure if the Artifact is built if the CI does not pass

pakoito

07/08/2019, 3:13 PM

@Imran/Malic https://github.com/arrow-kt/arrow/pull/1509#issuecomment-509267111 Good job with FunctorFilter! Let’s do the others now 😄

👍 1

pakoito

07/10/2019, 11:00 AM

you should add the error here too, it may shed some light

pakoito

07/10/2019, 11:00 AM

maybe you need to hint the type at someInt

pakoito

07/10/2019, 11:00 AM

I'd suggest testing the snippet on a test first

addamsson

07/14/2019, 11:07 AM

@raulraja I'm working on a Kotlin MPP library for which I need some FP utilities which is provided by Arrow Core. My problem is that it is not MPP yet and now I'm trying to determine which is a more time-efficient solution: roll my own FP lib including the essentials or help Arrow become MPP. I've heard from @pakoito that you're working on it. Can you give me a rundown on what is needed to get there?

addamsson

07/14/2019, 4:01 PM

I'm looking at the source code of Arrow and I've noted that there are interfaces like

Functor

which are very well documented but there are things like

Profunctor

which have literally no docs. The functions are also a bit problematic because their kdocs are mostly hard to understand or non-existent. Do you plan on improving this? It would be useful for contributors who didn't come from a Scala background.

addamsson

07/14/2019, 4:07 PM

I'm looking at

Semigroup

and I'm wondering why

combine

is not just an alias for

operator fun plus

addamsson

07/14/2019, 4:25 PM

@simon.vergauwen I've checked the typeclasses and I think I can do

PersistentCollection

PersistentMap

only with

Semigroup

Monoid

Foldable

Functor

and

FunctorFilter

. There is one question though: Implementing removal with

FunctorFilter

isn't going to be ineffective? I mean it is

O(n)

O(log n)

or even

O(1)

in some of the cases. Is there a typeclass for explicit removal instead of filtering?

addamsson

07/15/2019, 12:13 PM

When do you think KEEP-87 will be out?

Imran/Malic

07/15/2019, 3:08 PM

Hi @raulraja, @Jorge Castillo came across some issues in the Docs. Some parts have the old artifact structure. Should I change them immediately. Hi Jorge, please share in this thread some of your discoveries and we can sort them out 🙂

raulraja

07/15/2019, 3:13 PM

Next release is 0.10.0 as soon as the par traverse changes are in, then we can release

simon.vergauwen

07/15/2019, 4:05 PM

We should also verify that this is fixed. https://github.com/arrow-kt/arrow/issues/1510

Imran/Malic

07/16/2019, 11:11 PM

Even though, this compiles it’s clear that there are cornerCases where this doesn’t work.

fun <B> mapFilter(f: (A) -> Option<B>): NonEmptyList<B> =
    map(f).foldLeft(NonEmptyList(listOf())) { acc, o -> o.fold({ acc }, { acc + it }) }

Well I suggest to change the returnType to something like

Option<NonEmptyList<B>>

, because if there doesn’t exist any value which fulfills the f function and returns an Some<B> the return List is empty thus it’s not an NonEmptyList<B>. I also consider something like an

Either<List<Nothing>, NonEmptyList<B>>

, which is in my opinion more granular. I am happy for Feedback 🙂

pakoito

07/16/2019, 11:42 PM

NEL doesn't have a FunctorFilter instance, you have to transform it to List first

addamsson

07/17/2019, 8:22 AM

Which modules do you want to transform to a MPP project first? I've identified some modules which use some features which won't work in Javascript targets because they inherit from functions (which is not supported by the JS target)

Jorge Castillo

07/19/2019, 12:33 PM

0.9.1 is not gonna be a thing is it?

👍 1

addamsson

07/21/2019, 5:29 PM

@simon.vergauwen do you think that persistent data structures synergize really well with lenses / optics?

Title

addamsson

07/21/2019, 5:29 PM

@simon.vergauwen do you think that persistent data structures synergize really well with lenses / optics?

im' just reading the docs and it occurred to me

simon.vergauwen

07/21/2019, 7:10 PM

Optics are meant to be used with immutable data so it's a perfect fit. However it's solves a different problem than the typeclass hierarchy.

addamsson

07/22/2019, 7:25 AM

it solves the problem of mutating deeply nested immutable structures, no?

assoc-in

in clojure

simon.vergauwen

07/22/2019, 7:27 AM

Not only that, it also solves the problem that those operations aren't reusable or composable.

Not sure if that's similar in Clojure or not tho.

addamsson

07/22/2019, 7:57 AM

does this also solve thread safety?

like atoms/refs in clojure?

simon.vergauwen

07/22/2019, 7:58 AM

We have

Ref

in Arrow Fx 🙂

addamsson

07/22/2019, 7:59 AM

what does it do?

simon.vergauwen

07/22/2019, 8:00 AM

It’s a functional atomic ref (or a higher kind version of AtomicReference). It’s built on top of effects so we can take into account exceptions and async jumps and we can guarantee safety across that.

addamsson

07/22/2019, 8:19 AM

how about performance?

simon.vergauwen

07/22/2019, 8:20 AM

It’s uses a similar principle as typeclasses. All Optics are interfaces so you can delegate to whatever preforms best but in general we can derive all this boilerplate at compile time. I.e. data classes just use

copy

to update immutable fields and

copy

gets generated by the compiler.

Sealed classes use

when

which is equivalent to

TABLE_SWITCH

so super fast

Complexer structures with Optics like

Fold

Traversal

can be derived from

Foldable

and

Traverse

so they always should be the highest performing.

The idea is also that you use code gen with Optics so you get composable (re-useable) and boilerplate free code to work with immutable data.

A very nice example when working with a complex structure like Json here. https://github.com/47deg/helios/blob/master/helios-sample/src/main/kotlin/helios/sample/sample.kt

I need to work on docs there 2…. but the code for Optics is vey small (less than 200 loc) so definitely interesting to check out if you’re curios about the internals. I hand wrote everything there to produce an artefact with small bytecode 🙂

how about performance?

Or did you mean something else?

addamsson

07/22/2019, 9:12 AM

yep I understand how the code is generated / can be overridden

I was asking about the atomic ref

i'm interested in the performance characteristics of ref vs actor or simple thread confinement

simon.vergauwen

07/22/2019, 9:12 AM

Oh, it just wraps

AtomicReference

. And for MPP we’ll wrap kotlinx.atomicfu which is also a simply a typealiased

AtomicReference

for Java.

i’m interested in the performance characteristics of ref vs actor or simple thread confinement

Oh, that’s more related to architecture than it’s related to individual building blocks of Arrow. In Arrow Fx you’ll typically find all low level building blocks but in the context of

Ref

AtomicReference

Promise

has many use-cases but a good example is that you can use it as a latch to do synchronisation and

Semaphore

. So it’s all low level building blocks concurrency people are typically familiar with.

Wish I could comment more on your actual question but I am not knowledge enough on those subjects.

Feel free to share any lecture you find 🙂

addamsson

07/22/2019, 11:55 AM

i'll write some prototypes after my vacation

i've tried Kotlin's

actor

but its performance is abysmal

i need shared mutable state

but the state needs to remain consistent

simon.vergauwen

07/22/2019, 11:55 AM

Ye, I imagine. I have an

actor

implementation on top of

IO

and concurrency tools in Arrow

addamsson

07/22/2019, 11:56 AM

that's why i need the persistent data structures

so I don't have to confine reads only writes

but with atomic-fu I might not even need to confine writes

simon.vergauwen

07/22/2019, 11:56 AM

Then my solution might work for you, although it’s rather an example as an actual fast impl.

It’s a combination of

Queue

Ref

and

Promise

. Where you basically have a

Queue

Promise

and you can

take

in the actor to act on a message and you can

put

in the

Queue

to send a message.

addamsson

07/22/2019, 11:57 AM

i see

i figured something similar

that writes are wrapped in an

Eval

and put in a queue

and a single threaded coroutine executes them in sequence

i don't even need a queue/eval for this since this functionality is provided by coroutines out of the box

simon.vergauwen

07/22/2019, 11:59 AM

I based it on this. https://github.com/vlovgr/actors-cats-effect-fs2 The 20min live coding session is very enlightening on how to build advance concurrency stuff with Arrow Fx (Cats-effects).

addamsson

07/22/2019, 11:59 AM

thanks I'll take a look

👍

did you have a chance to review the changes to my PR?

simon.vergauwen

07/22/2019, 12:07 PM

I am reviewing it now 🙂

addamsson

07/22/2019, 12:08 PM

👍

View count: 1