Chris Wong

It looks like this has been implemented by @kallentu (behind an experimental flag): https://medium.com/dartlang/dart-declaration-site-variance-5c0e9c5f18a5

@Levi-Lesches I can't speak for the Dart team, but I don't think that's practical with the sheer amount of code that's already written. At the scale we're working at –...

`&[u8]`, `&[Option]`, and `&[u16]` can be decoded in place as well, since every element of a JSON array takes up at least two bytes. With some effort this can apply...

@webmaster128 this StackOverflow answer is more up-to-date: https://stackoverflow.com/a/14608300/617159

Note that it's impossible in general to tell whether whitespace is significant or not, as [you can control whitespace behavior in CSS](https://developer.mozilla.org/en-US/docs/Web/CSS/white-space).

Looks like this bug depends on haskell/cabal#1189.

[This reddit thread](https://www.reddit.com/r/haskell/comments/2abi06/hackage_update_part_3/citetak) outlines a possible solution.

Would it help if the macro were named `with_clone!` instead? That would better emphasize that it's some parameter being cloned, not the whole thing.

How does this break referential transparency? I'd like to see that elaborated upon.

> Wouldn't this lead to inference errors? Is this something that we can figure out with a crater run?