[superseded] new pointers.toUncheckedArray

[timotheecour]

alternative to #15490

adds toUncheckedArray

# before
let px = cast[ptr UncheckedArray[type(x)]](x)
# after
let px = x.toUncheckedArray

it provides a safer and more readable construct than the explicit cast (which is error prone since the cast will happily accept any type without giving any compiler error, eg in case of let px = cast[ptr UncheckedArray[int]](pointerToInt8)

[juancarlospaco]

Documentation needs a warning that is unsafe,
and I think you can not guarantee is not "error prone" so maybe remove that comment.

[Araq]

This is going in the right direction but addr(a[i]).toUncheckedArray is messier than e.g. a.toUncheckedArray(i). Don't take the address of a single element and then cast it to a fullblown array. It was an array, it is turned into a pointer to an UncheckedArray.

[timotheecour]

addr(a[i]).toUncheckedArray is messier than e.g. a.toUncheckedArray(i). Don't take the address of a single element and then cast it to a fullblown array. It was an array, it is turned into a pointer to an UncheckedArray.

your assumption is too restrictive, toUncheckedArray is meant to be usable for dealing with any ptr[T] wherever they come from (eg FFI or addr of something), where you can't assume it's an array[T] or seq[T] etc; it's the most general case

[timotheecour]

PTAL

[Araq]

This is a new std module that nobody uses, with no RFC process to back it up, without use cases to justify it and not useful for the low level code I typically write (I need "pointer +- offset"). I don't understand why we need more obviously not battle-tested standard modules. Why not add it to "fusion" instead?

[disruptek]

How is that any different from fusion?

[dom96]

What Araq said. Additions to the stdlib should be well thought through, discussed and hopefully tested ahead of time as a Nimble package

[timotheecour]

@Araq

Why not add it to "fusion" instead?

=> nim-lang/fusion#20
(fusion is a better fit than a separate nimble package)

I'll close this PR if the fusion one gets merged

[timotheecour]

=> nim-lang/fusion#20