Jean Klingler

Hi! I had a look into the test repo, and tried to code an [alternative implementation](https://github.com/mtrudel/iolist_test/compare/master...sabiwara:alternative-impl?expand=1) that "just" recursively walks the IO list and doesn't build intermediate tuples (and probably...

> The intent here is to allow bandit's http2 implementation to 'peel off' some leading number of bytes from a Plug-generated iolist response Oh OK, I was missing some context....

> It is probably not that relevant to most people, especially with the new type system coming up, but I think this probably breaks Dialyzer's "opaqueness" access rules and will...

I understand this concern, my thinking was that filtering is still possible by returning the noop value 0 or 1. Maybe we can explicitly provide an example doing so in...

I've been thinking about this one, and `map_sum/2` would actually make a lot of sense if we think in terms of types. It would be consistent with `map_join/2`, which has...

> I’m not convinced we need this in the first place, though. I think it would be good from the perspective of encouraging users to pick efficiency without sacrificing readability....

Closing for now since the appropriate solution still needs discussion.

> @sabiwara you said closing as it still needs discussion but immediately reopened - accidental double click? 😅 Nope, Jose asked me to keep it open 😉

@josevalim after thinking about this for a while, I was more leaning towards `map_sum/2` due to the very close situation with `map_join`. It could be caught by a linter rule...

- Rebased - Renamed to `sum_by` for now - Added an entry to the cheatsheet and reworked the docs a bit Once this is merged, will send an equivalent for...