Add WriterT MonadError instance

[ceedubs]

This adds a MonadError instance for WriterT. However, it's not very useful in its current form. The following fails to compile, as the instance can't be found: MonadError[WriterT[String Xor ?, Long, ?], String]. Calling writerTMonadError[String Xor ?, Long, String] directly works though.

Does anyone know how to convince Scala to find this instance?

Even once it's found, there are some issues with the proper Arbitrary instances being found, but I guess we'll take it one step at a time.

[codecov-io]

Current coverage is 89.53% (diff: 100%)

Merging #689 into master will increase coverage by 0.32%

@@             master       #689   diff @@
==========================================
  Files           234        234          
  Lines          3141       3143     +2   
  Methods        3084       3089     +5   
  Messages          0          0          
  Branches         55         51     -4   
==========================================
+ Hits           2802       2814    +12   
+ Misses          339        329    -10   
  Partials          0          0          

Powered by Codecov. Last update 4f7b79a...172cf62

[lukiano]

I wonder if you should append the L from fa.run and the L that results from f(e).run ?
In other words, if the Log that came from fa will be lost (maybe it's ok for it to be forgotten)

[ceedubs]

@lukiano sorry I'm just seeing this. I don't think we can do what you have recommended. Let's consider Future as F (and thus Throwable as E). If the Future fails, then there is no log. We do have a log if the future succeeds, but in that case the function that we pass to handleErrorWith will never be called, because there is no error to handle.

Does that sound right?

[lukiano]

I think you are right. I was thinking of something like XorT -> WriterT -> Future, but in this case F is just Future so there's no monadError instance for it. XorT will provide the MonadError and will already do the right thing.

[non]

@ceedubs Is this something you want to keep open? Still working on it?

[ceedubs]

@non I'll check if this is found for an F[_] type like Future. If so it may be at least a little useful. I also am interested in whether the Xor example works when the SI-2712 fix plugin is used. I'll get back on this shortly.

[ceedubs]

@non @lukiano @kailuowang I just updated this PR. Thanks to @kailuowang's MonadError instance for Option, I can see that this can potentially be useful.

Sorry -- I force pushed a rebase because I had made a mess of things.

[ceedubs]

@non @lukiano @kailuowang updated again

[kailuowang]

LGTM. Did you test Xor with SI-2712 fix? I will be surprised if it doesn't work.

[ceedubs]

@non @kailuowang sorry about the delay on this one. I've resolved the merge conflict. I just did a check with a locally-published version, and MonadError[WriterT[String Xor ?, Long, ?], String] does indeed resolve just fine with the SI-2712 fix plugin. I think this should be good to merge if the Travis build succeeds.

[ceedubs]

I've added some (gnarly) tests for the ApplicativeError instance. I look forward to SI-2712 making this a lot cleaner.

[adelbertc]

LGTM 👍

[ceedubs]

@kailuowang does this still look good to you?

[kailuowang]

👍 LGTM