Functor.fmap

[fosskers]

This PR adds a .fmap method to Functor as an alias of .map. This is in the spirit of Foldable.combineAll which is an alias for .fold, as many types have a built-in .fold with different behaviour, preventing .fold from being injected.

PROs

• .fmap is injectable, exposing lawful Functorness to implementing classes (i.e. Map has a built-in .map that breaks Functor laws, but its injected .fmap would be lawful). This also benefits authors of new Functor instances.
• Haskellers get a warm, fuzzy feeling of familiarity.

CONs

• Increased API surface area for a method that is not strictly necessary (but then you'd have to say the same for .combineAll)
• Potential confusion with .flatMap for non-Haskellers.

[johnynek]

If we add this it should be final. But could we just add it as syntax and not to the typeclass?

[fosskers]

final so that it can't be overridden, yeah?

[fosskers]

Foldable.combineAll is included directly in the typeclass too.

[SystemFw]

I don't have any strong opinions on this, and avoid having to do Functor[Blah].map seems useful enough, however wrt "haskellers get a warm-fuzzy feeling of familiarity", this is a case where actually (as a Haskeller), things got fixed here.
Haskell used to have fmap be called just map. Then a map specialised to Lists was added in the early days to avoid complicating things for beginners that don't know about typeclasses. However, after the FTP proposal has brought polymorphism back into the Prelude, Functor fmap and a special map for List stick out like a sore thumb.

[LukaJCB]

Hi @fosskers, can you make a mima exception, please? :)

[fosskers]

@SystemFw The fmap / map situation in Haskell is weird, yes. Hopefully we can settle on something nice there too. For Cats, it's a problem of "Well .map can't be injected. Okay let's make an alias. What to call it? Okay, .fmap because of prescedent and intuition: fmap = Functor map."

[codecov-io]

Codecov Report

Merging #2056 into master will decrease coverage by 0.3%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #2056      +/-   ##
==========================================
- Coverage   94.95%   94.64%   -0.31%     
==========================================
  Files         311      318       +7     
  Lines        5270     5364      +94     
  Branches      133      119      -14     
==========================================
+ Hits         5004     5077      +73     
- Misses        266      287      +21

Impacted Files	Coverage Δ
core/src/main/scala/cats/Functor.scala	100% <100%> (ø)	⬆️
core/src/main/scala/cats/instances/eq.scala	80% <0%> (-20%)	⬇️
core/src/main/scala/cats/instances/order.scala	83.33% <0%> (-16.67%)	⬇️
core/src/main/scala/cats/data/IndexedStateT.scala	89.36% <0%> (-9.45%)	⬇️
core/src/main/scala/cats/instances/function.scala	92.3% <0%> (-7.7%)	⬇️
core/src/main/scala/cats/data/OptionT.scala	97.46% <0%> (-2.54%)	⬇️
core/src/main/scala/cats/data/WriterT.scala	93.13% <0%> (-0.62%)	⬇️
core/src/main/scala/cats/syntax/either.scala	99.12% <0%> (-0.02%)	⬇️
core/src/main/scala/cats/instances/ordering.scala	100% <0%> (ø)	⬆️
core/src/main/scala/cats/data/Ior.scala	100% <0%> (ø)	⬆️
... and 16 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 40465ac...d7d347d. Read the comment docs.

[SystemFw]

Sure, I agree that the basic problem of underlying methods taking precedence over typeclass ones is there.
I guess the question is how many instances of this are out there for the map case? And the other question is how big of a problem it actually is. Unless people always call fmap , you still need to know whether the map you're calling is a good one, or a bad one, in which case you can do Functor[F].map.

In any case this is making me sound way more opposed to the proposal than I actually am. My only concern is that an advanced user knows that Functor[F].map will work, and a beginner might be confused as to why fmap is similar to flatMap, especially given that the name fmap is not nice in Haskell either. Unfortunately I can't think of a better name that isn't map or lift, both already taken.

[LukaJCB]

This is a great addition, thank you very much @fosskers!

[fosskers]

Lol wuh github? I didn't dismiss a review.

[LukaJCB]

It automatically dismissed reviews when you push, don't worry :)

[kailuowang]

thanks!