🙅🏼‍♂️ [stdlib] Add method for accessing full storage capacity of array

[natecook1000]

This is an implementation of a withUnsafeMutableBufferPointerToStorage method and init(unsafeUninitializedCapacity:initializingWith:) for arrays. Under review here: https://forums.swift.org/t/se-0223-accessing-an-arrays-uninitialized-buffer/15194

To do:

• Check inlining annotations
• Add @_semantics attributes?
• Revise documentation
• Duplicate onto ContiguousArray

Resolves #45677.

[natecook1000]

@swift-ci Please smoke test

[natecook1000]

Ugh — it looks like adding init(minimumCapacity: Int) to RangeReplaceableCollection changes the overload selection in [1: 10, 2: 20].mapValues(String.init) so that the new initializer is being chosen instead of the integer-to-string-converting initializer. That's really unfortunate.

let problem = [1: 10, 2: 20].mapValues(String.init)
// old: [1: "10", 2: "20"]
// new: [1: "", 2: ""]

[natecook1000]

@swift-ci Please smoke test

[jrose-apple]

Then why is it being passed inout?

[natecook1000]

It looks like I'm cargo-culting the rationale for the buffer being inout in withUnsafeMutableBufferPointer, which is a very old behavior (introduced in c5c17e2). It looks like that was probably because the type used to access the elements then (UnsafeMutableArray) needed to be mutable to access its subscript setter, unlike UnsafeMutableBufferPointer.

[natecook1000]

This should probably prefer the correct semantics (non-inout) over matching the existing method, esp. since you can't pass the same closure to both anyway. Thanks!

[jrose-apple]

Ahhh right, you do need mutation if you want to do collectiony things on the UnsafeMutableBufferPointer. But in this case it's not safe to do collectiony things unless you take a slice anyway, at which point you can use var.

[natecook1000]

@swift-ci Please smoke test

[natecook1000]

Closing until there's a new SE pitch.