add parTraverseN and parSequenceN

[sideeffffect]

/cc @jendakol @augi

[sideeffffect]

Is there a way to work around the issue

[error]  * abstract method catsEffectSyntaxAsyncObj(cats.effect.Async)cats.effect.Async in trait cats.effect.syntax.AsyncSyntax is inherited by class AllCatsEffectSyntax in current version.
[error]    filter with: ProblemFilters.exclude[InheritedNewAbstractMethodProblem]("cats.effect.syntax.AllCatsEffectSyntax.catsEffectSyntaxAsyncObj")
[error]  * abstract method toAsyncOps(java.lang.Object,cats.effect.Async)cats.effect.Async#Ops in trait cats.effect.Async#ToAsyncOps is inherited by class AllCatsEffectSyntax in current version.
[error]    filter with: ProblemFilters.exclude[InheritedNewAbstractMethodProblem]("cats.effect.syntax.AllCatsEffectSyntax.toAsyncOps")
[error] java.lang.RuntimeException: cats-effect: Binary compatibility check failed!

that happens only with Scala 2.11?

[sideeffffect]

@SystemFw any tips for solving the problem with Scala 2.11?

[djspiewak]

@sideeffffect It can't really be resolved on Scala 2.11. What you're doing here is you're adding functions to a trait (AllCatsEffectSyntax) which previously existed. Anyone inheriting from that trait is going to get that function. The problem with this is that, in Scala 2.11 (or more specifically, prior to Java 8), traits were encoded through explicit function delegation to a nested class (which contained the definition) in the implementing subtypes. This is problematic because the implementing subtypes cannot delegate to the nested class on these new functions (such as parTraverseN) since it didn't exist when they were compiled! (since they were compiled against previous versions of cats-effect) So any code which calls parTraverseN would then get an AbstractMethodError. This is what MiMa is detecting.

You basically cannot add syntax like this in the 1.x timeline. Or rather, you could add AsyncSyntax, but you can't add it to the main trait. The good news is that 2.x is right around the corner. The bad news is we really don't want to change too much, so I'm not sure if there's time for this to make it.

[rossabaker]

Is it any worse for compatibility than the changes proposed in #262?

[djspiewak]

@rossabaker Do you mean #519? Broadly-speaking no, it should be about the same if I fully comprehend both PRs. I would definitely be comfortable with both making it.

[djspiewak]

@sideeffffect I would recommend waiting for #546 to land (which flushes out mima), then merge with master (please don't rebase open PRs) and see if that resolves the issues.

[djspiewak]

@sideeffffect Any update on this? We're going to be in compatibility-freeze very shortly.

[sideeffffect]

@djspiewak rebased on the newest master
OK to merge?

[codecov-io]

Codecov Report

Merging #498 into master will decrease coverage by 0.04%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #498      +/-   ##
==========================================
- Coverage   89.09%   89.05%   -0.05%     
==========================================
  Files          73       75       +2     
  Lines        2119     2137      +18     
  Branches      129      125       -4     
==========================================
+ Hits         1888     1903      +15     
- Misses        231      234       +3

[djspiewak]

@sideeffffect Reminder: as I said, do not rebase PRs. It's in the contributing guidelines and everything. :-)

I'll take a final look at things and merge (hopefully later today). Thanks for all your work!

[sideeffffect]

@djspiewak facepalm sorry, I totally missed that 😄 I'll pay better attention next time
thanks for looking at it!