Utku Demir

> The coercions from non-linear functions to linear functions may be blocking optimisations as well. I think this is the current major issue. As an example, this simple code: ```haskell...

After #329 and #330; `Data.Array.Mutable.Linear.get` is only three times slower than `Data.Vector` counterpart on GHC HEAD. I don't see a way to work around the the `unsafeCoerce`'s blocking optimisations. I...

Sorry for the radio silence. For now I think we should hold off merging this PR (alongside with other new collection PR's #293 and #295), because we already have some...

> This is incorrect but gets at the rough idea I don't get it :). `flip use` seems like just `id`, and I don't see how that `Owned` would witness...

> What would be the type of dupVM? Very good point; it can not be `a #1-> VM n a` because then we can get an unrestricted `VM`. So it...

@tbagrel1 Do you think this can still be useful, especially after #360? Please feel free to close this otherwise :).

> Could you tell me what does `VM` would bring over `Data.LArray.Mutable.Unlifted.Linear`? `VM` is length-indexed, `LArray`'s are not. As far as I can remember this is the only difference. Whether...

+1, it seems like this situation is becoming pretty common. I had this issue before while writing lenses for mutable collections, where the collection itself is linear but the elements...

> You give me tips on how to resolve the issues Ormolu has with my code. (Or you tell me it's not possible!) CPP support is almost impossible to get...

I implemented a solution to this at [`ud/grouping-complete-pragmas` branch](https://github.com/tweag/ormolu/compare/ud/grouping-complete-pragmas); simply by handling `COMPLETE` pragma as we handle others, where we omit the newline if a pragma refers to a name...