typelevel / typelevel/cats

`Applicative[OneAnd[ZipLazyList, *]]` works incorrectly

Open
#4,840 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Scala
Stars
5.5k
Forks
1.2k
Avg merge
2d 10h
Merged PRs (30d)
5

Description

A follow-up to #4830.

Consider the snippet:

//> using scala 3.8.2
//> using options -Xkind-projector
//> using dep org.typelevel::cats-core:2.13.0

import cats.Applicative
import cats.data.OneAnd
import cats.data.ZipLazyList

val s1 = LazyList(1, 2, 3)
val s2 = LazyList(4, 5)
val sR = Applicative[LazyList].product(s1, s2)
println("-      LasyList: " + sR.mkString(","))

val ns1 = OneAnd(s1.head, s1.tail)
val ns2 = OneAnd(s2.head, s2.tail)
val nsR = Applicative[OneAnd[LazyList, *]].product(ns1, ns2)
println("-    NeLasyList: " + (nsR.head #:: nsR.tail).mkString(","))

val z1 = ZipLazyList(s1)
val z2 = ZipLazyList(s2)
val zR = Applicative[ZipLazyList].product(z1, z2)
println("-   ZipLasyList: " + zR.value.mkString(","))

val nz1 = OneAnd(s1.head, ZipLazyList(s1.tail))
val nz2 = OneAnd(s2.head, ZipLazyList(s2.tail))
val nzR = Applicative[OneAnd[ZipLazyList, *]].product(nz1, nz2)
println("- NeZipLasyList: " + (nzR.head #:: nzR.tail.value).mkString(","))

It prints the following:

-      LasyList: (1,4),(1,5),(2,4),(2,5),(3,4),(3,5)
-    NeLasyList: (1,4),(1,5),(2,4),(2,5),(3,4),(3,5)
-   ZipLasyList: (1,4),(2,5)
- NeZipLasyList: (1,4),(1,5),(2,4),(3,4)

Note: it works the same way on Scala 2.13 and for Stream type on Scala 2.12

As you can see, while Applicative[F].product produces identical results for both LazyList and OneAnd[LazyList, *],
for ZipLazyList and OneAnd[ZipLazyList, *] it doesn't. Moreover, the latter is neither the Cartesian product nor element-wise composition – it's something different.

I would argue that OneAnd for a collection type should behave as the non-empty variant of that collection. That is not the case for ZipLazyList, apparently.

The reason appears to be the Alternative[ZipLazyList] instance discussed in #4830 – it seems to be unlawful. The problem for OneAnd is that its Applicative instance requires an Alternative for the underlying type F:
https://github.com/typelevel/cats/blob/8baa9bd2ff2185ea8faf88849ecee1b2da337464/core/src/main/scala/cats/data/OneAnd.scala#L248

There's something off with OneAnd:

  • if we assume that Alternative[ZipLazyList] is incorrect indeed, then the current Applicative implementation for OneAnd implies that it cannot work correctly for Zip* collections.
  • if we still believe that Alternative[ZipLazyList] is good enough for OneAnd, then Applicative for OneAnd is not correctly implemented.
  • or, perhaps, we cannot treat OneAnd as just a non-empty "drop-off" for the underlying F.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Reproduce the supplied Scala snippet with cats-core 2.13.0 and compare the LazyList, OneAnd[LazyList, *], ZipLazyList, and OneAnd[ZipLazyList, *] results. Read the OneAnd Applicative implementation in core/src/main/scala/cats/data/OneAnd.scala around line 248 and the Alternative discussion in #4830. Done requires a decided and documented correct relationship between these instances.

Written by the indexing model from the issue text.

Assessment

Tech stack
scala
Domain
backend-api-design
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.