<div dir="ltr">I'm a rather enthusiastic +1 on this, given a sufficiently long time table for the switch!<div><div><br></div><div>We ARE asking a lot of people to break code here.</div><div><br></div><div>It does drastically reduce the amount of "magic" involved in the ApplicativeDo story and avoids enshrining the current historical artifact in a future Haskell report that includes the AMP.</div><div><br></div><div>I'd also like to consider employing the exact same migration plan for (>>) at the same time. It is redundant with (*>), and the choice of which gets implemented in terms of the other introduces precisely the same set of issues. If we can do that then this would fix the remaining performance issues for mapM, letting us eventually remove that from Traversable as well, moving it to a top level alias in the same way -- maybe not until 8.8 or something, but it'd at least give us a roadmap to get there eventually.</div><div><br></div><div>-Edward</div></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Sep 24, 2015 at 5:43 PM, Herbert Valerio Riedel <span dir="ltr"><<a href="mailto:hvr@gnu.org" target="_blank">hvr@gnu.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hello *,<br>
<br>
Concluding AMP and MFP, We (David and I) proudly present you the final<br>
installment of the Monad trilogy:<br>
<br>
<br>
Monad of no `return` Proposal<br>
=============================<br>
<br>
TLDR: To complete the AMP, turn `Monad(return)` method into a<br>
top-level binding aliasing `Applicative(pure)`.<br>
<br>
<br>
Current Situation<br>
-----------------<br>
<br>
With the implementation of Functor-Applicative-Monad Proposal (AMP)[1] and<br>
(at some point) the MonadFail proposal (MFP)[2] the AMP class hierarchy<br>
becomes<br>
<br>
<br>
class Functor f where<br>
fmap :: (a -> b) -> f a -> f b<br>
<br>
<br>
class Functor f => Applicative f where<br>
pure :: a -> f a<br>
(<*>) :: f (a -> b) -> f a -> f b<br>
<br>
(*>) :: f a -> f b -> f b<br>
u *> v = …<br>
<br>
(<*) :: f a -> f b -> f a<br>
u <* v = …<br>
<br>
<br>
class Applicative m => Monad m where<br>
(>>=) :: m a -> (a -> m b) -> m b<br>
<br>
return :: a -> m a<br>
return = pure<br>
<br>
(>>) :: m a -> m b -> m b<br>
m >> k = …<br>
<br>
<br>
class Monad m => MonadFail m where<br>
fail :: String -> m a<br>
<br>
<br>
Consequently, the `Monad` class is left with a now redundant `return`<br>
method as a historic artifact, as there's no compelling reason to<br>
have `pure` and `return` implemented differently.<br>
<br>
Traditionally, `return` is often used where `pure` would suffice<br>
today, forcing a `Monad` constraint even if a weaker `Applicative`<br>
would have sufficed.<br>
<br>
As a result, language extensions like `ApplicativeDo`[3] have to<br>
rewrite `return` to weaken its `Monad m =>` constraint to<br>
`Applicative m =>` in order to benefit existing code at the cost<br>
of introducing magic behavior at the type level.<br>
<br>
Finally, this redundancy becomes even more significant when viewed in<br>
light of the renewed Haskell standardisation process[7]: The next<br>
Haskell Report will almost certainly incorporate the AMP (and MFP)<br>
changes, and there's no justification for the Report to retain<br>
`return` as a method of `Monad`. A good reason would have been to<br>
retain backward compatibility with Haskell 2010. However, as the AMP<br>
superclass hierarchy requires `Monad` instances to be accompanied by<br>
`Applicative` instances (which aren't part of Haskell 2010, c.f. [6]),<br>
backward compatibility with Haskell 2010 goes out the window when it<br>
comes to defining `Monad` instances (unless via use of `-XCPP` or<br>
similar). Consequently, meeting the high bar for a formal document<br>
such as the Haskell Report demands that `Monad` shall not carry a<br>
redundant `return` method that serves no purpose anymore. Moreover,<br>
getting `return` out of the way is desirable to facilitate<br>
standardising potential candidates such as the earlier mentioned<br>
`ApplicativeDo` in the future and avoids the technical debt incurred<br>
by keeping around this language wart.<br>
<br>
<br>
Proposed Change<br>
---------------<br>
<br>
Remove `return` as a method from the `Monad` class and in its place<br>
define a top-level binding with the weaker `Applicative` typeclass<br>
constraint:<br>
<br>
<br>
-- | Legacy alias for 'pure'<br>
return :: Applicative f => a -> f a<br>
return = pure<br>
<br>
<br>
This allows existing code using `return` to benefit from a weaker<br>
typeclass constraint as well as cleaning the `Monad` class from a<br>
redundant method in the post-AMP world.<br>
<br>
A possible migration strategy is described further below.<br>
<br>
<br>
Compatibility Considerations<br>
----------------------------<br>
<br>
Generalizing the type signature of a function from a `Monad`<br>
constraint to its superclass `Applicative` doesn't cause new<br>
type-errors in existing code.<br>
<br>
However, moving a method to a top-level binding obviously breaks code<br>
that assumes `return` to be a class method. Foremost, code that<br>
defines `Monad` instances it at risk:<br>
<br>
### Instance Definitions<br>
<br>
Code defining `return` as part of an instance definition<br>
breaks. However, we had the foresight to provide a default<br>
implementation in `base-4.8` for `return` so that the following<br>
represents a proper minimal instance definition post-AMP:<br>
<br>
<br>
instance Functor Foo where<br>
fmap g foo = …<br>
<br>
instance Applicative Foo where<br>
pure x = …<br>
a1 <*> a2 = …<br>
<br>
instance Monad Foo where<br>
m >>= f = …<br>
<br>
-- NB: No mention of `return`<br>
<br>
<br>
Consequently, it is possible to write forward-compatible instances<br>
that are valid under this proposal starting with GHC 7.10/`base-4.8`.<br>
<br>
Heuristically `grep`ing through Hackage source-code reveals a<br>
non-negligible number of packages defining `Monad` instances with<br>
explicit `return` definitions[4]. This has a comparable impact to the<br>
AMP, and similarly will require a transition scheme aided by compiler<br>
warnings.<br>
<br>
### Module Import/Export Specifications<br>
<br>
A second source of incompatibility may be due to<br>
`import`s. Specifically module import that assert `return` to be a<br>
method of `Monad`, e.g.:<br>
<br>
import Control.Monad (Monad ((>>=), return))<br>
<br>
or<br>
<br>
import Prelude hiding (Monad(..))<br>
import Control.Monad (Monad(..)) as Monad<br>
<br>
f = Monad.return ()<br>
<br>
The dual situation can occur when re-exporting `return` via module<br>
export specifications.<br>
<br>
However, given that `return` is exported by `Prelude` and the examples<br>
above are rather artificial, we don't expect this to be a major source<br>
of breakage in the case of `return`. In fact, a heuristic grep[5] over<br>
Hackage source-code revealed only 21 packages affected.<br>
<br>
### Example for writing compatible code<br>
<br>
<br>
instance Functor Foo where<br>
fmap g foo = …<br>
<br>
instance Applicative Foo where<br>
pure x = …<br>
a1 <*> a2 = …<br>
<br>
instance Monad Foo where<br>
m >>= f = …<br>
<br>
#if !(MIN_VERSION_base(4,8,0))<br>
return = pure<br>
#endif<br>
<br>
<br>
Migration Strategy<br>
------------------<br>
<br>
The migration strategy is straightforward:<br>
<br>
**Phase 1** *(GHC 8.0)*: Implement new warning in GHC which gets<br>
triggered when `Monad` instances explicitly override the<br>
default `return` method implementation.<br>
<br>
**Phase 2** *(GHC 8.2 or later)*: When we're confident that the<br>
majority of Hackage has reacted to the warning (with the help of<br>
Stackage actively pursuing maintainers to update their packages) we<br>
turn the `return` method into a top-level binding and remove the<br>
warning implemented in Phase 1 from GHC again.<br>
<br>
<br>
Discussion period<br>
-----------------<br>
<br>
A discussion period of three weeks (until 2015-10-15) should be enough<br>
to allow everyone to chime in as well as leave enough time to make the<br>
required preparations for GHC 8.0 should this proposal pass as we hope.<br>
<br>
----<br>
<br>
[1]: <a href="https://wiki.haskell.org/Functor-Applicative-Monad_Proposal" rel="noreferrer" target="_blank">https://wiki.haskell.org/Functor-Applicative-Monad_Proposal</a><br>
[2]: <a href="https://wiki.haskell.org/MonadFail_Proposal" rel="noreferrer" target="_blank">https://wiki.haskell.org/MonadFail_Proposal</a><br>
[3]: <a href="https://ghc.haskell.org/trac/ghc/wiki/ApplicativeDo" rel="noreferrer" target="_blank">https://ghc.haskell.org/trac/ghc/wiki/ApplicativeDo</a><br>
[4]: <a href="https://gist.github.com/hvr/b0e34463d85b58f169d9" rel="noreferrer" target="_blank">https://gist.github.com/hvr/b0e34463d85b58f169d9</a><br>
[5]: <a href="https://gist.github.com/hvr/afcd040783d980594883" rel="noreferrer" target="_blank">https://gist.github.com/hvr/afcd040783d980594883</a><br>
[6]: <a href="https://ghc.haskell.org/trac/ghc/ticket/9590" rel="noreferrer" target="_blank">https://ghc.haskell.org/trac/ghc/ticket/9590</a><br>
[7]: <a href="https://mail.haskell.org/pipermail/haskell-prime/2015-September/003936.html" rel="noreferrer" target="_blank">https://mail.haskell.org/pipermail/haskell-prime/2015-September/003936.html</a><br>
<br>
--<br>
<br>_______________________________________________<br>
Libraries mailing list<br>
<a href="mailto:Libraries@haskell.org">Libraries@haskell.org</a><br>
<a href="http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries" rel="noreferrer" target="_blank">http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries</a><br>
<br></blockquote></div><br></div>