
Hi Emily,
As David suggested, removal of `mapM` will be problematic for boxed `Vector`: https://github.com/haskell/vector/blob/a15a52155f281ca6753ecef1247f84a0579e3...
The reason behind that fact is that streams in `vector` are monadic, so `traverse` implementation has to go through a list, while `mapM` can be implemented in a more efficient manner.
My personal opinion on this is that such slimming of Traversable is counterproductive, the only thing we incur is a bunch of broken libraries and redundant work on maintainers part.
Is there anything that can be won from a user's perspective from such a change? A potential performance or usability benefit that I am missing?
Sincerely,
Alexey.
‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
On Saturday, November 7, 2020 11:03 PM, David Feuer
I'm generally in favor of removing `sequence`, and probably also `sequenceA`. Can you demonstrate that there can be no instance for which `mapM` is much more efficient than `traverse`? Consider especially types based on monadic streams.
On Sat, Nov 7, 2020 at 2:56 PM Emily Pillmore
wrote: Hi All,
I have a proposal for the `base` library: I would like to move `mapM` and `sequence` out of the class definition for `Traversable`, redefining them as toplevel aliases:
```haskell mapM :: Traversable t ⇒ Monad m ⇒ (a -> m b) -> t a -> m (t b) mapM = traverse
sequence :: Traversable t ⇒ Monad m=> t (m a) -> m (t a) sequence = sequenceA ```
This slims `Traversable` by 50%. This would be a very small breaking change, which is completely tractable, but a great improvement for the ecosystem imo. Thoughts? What timeline should we shoot for with a change like this?
Cheers, Emily
_______________________________________________ Libraries mailing list Libraries@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries