Adds a MonadTrans typeclass

[stew]

This is similar to MonadTrans in scalaz. It allows you to lift a Monad into a transformer wrapping that monad

I skipped Hoist, since I've never seen that actually used in the wild.

I was shamed into doing this by @djspiewak :)

[codecov-io]

Current coverage is 89.12%

Merging #844 into master will decrease coverage by -0.14% as of ece069f

@@            master    #844   diff @@
======================================
  Files          168     170     +2
  Stmts         2319    2336    +17
  Branches        75      75       
  Methods          0       0       
======================================
+ Hit           2070    2082    +12
  Partial          0       0       
- Missed         249     254     +5

Review entire Coverage Diff as of ece069f

Powered by Codecov. Updated on successful CI builds.

[ceedubs]

Missing return type

[ceedubs]

Looks good. I just left a couple very minor comments.

I think we could also have instances for Kleisli and StateT, though that could be a separate PR.

[ceedubs]

I think this could just delegate to liftF

[ceedubs]

I think that MonadTrans is really helpful. However, there is one thing that bothers me a little bit about it. It seems overly-specific to Monad. Most of these transformers actually have constraints that are weaker than Monad in order to lift into them. For example, we can lift any F[A] into a Kleisli without constraints on F. We can lift any F[A] where F has a Functor into an OptionT.

I'm pondering whether it might make sense to have something like Transformer[F[_], T[_[_], _]] that also abstracts over what the type class needed by the constructor is. I might be trying to get a little too creative, and it might run into all sorts of inference issues. But I'd kind of like to give it a try. Any thoughts?

[ceedubs]

Hmm I'm confused by this. Your branch seems to not have the latest from master, so Cartesian is still called Monoidal. Yet you've added CartesianSyntax and somehow the build passed?

[travisbrown]

@ceedubs Travis CI will use GitHub's merge commit, which in this case is ece069f, which has 7c9c209 for a parent and includes the Cartesian change. (I thought there was some way to have the merge commit automatically update for new commits on the target branch, but I might have been making that up.)

[ceedubs]

Thanks @travisbrown. It works fine if I merge in the latest from master.

I tried out my idea and I think it looks like it could work. I'll probably submit a PR (maybe to Stew's branch) tomorrow for discussion.

[aryairani]

thanks @stew, this helped me finally understand MonadTrans.

[stew]

OMG @ceedubs YES, I have it. I'm closing this PR in favor of #853 which generalizes this to not always need monads to lift