Generalized Existentials / Type Erasers

Chris_Eidhof · January 31, 2017, 10:07am

Hi swift-evolution,

I was wondering if anything more detailed has been written about
generalized existentials / type erasers. I noticed that AnyCollection only
delegates a number of methods to its wrapped value, but often doesn't. For
example, all the conditionally inherited items aren't being delegated (e.g.
min()). This behavior was surprising to me (but makes sense once I thought
about it longer).

How would this work with generalized existentials? Are generalized
existentials in scope at all for Swift 4?

···

--
Chris Eidhof

hartbit · January 31, 2017, 12:27pm

Hi Chris,

The topic has been heavily discussed last year, culminating into a massing proposal by Austin Zheng:

github.com

austinzheng/swift-evolution/blob/az-existentials/proposals/XXXX-enhanced-existentials.md#nested-typealias-existential

# Enhanced Existential Support

* Proposal: [SE-NNNN](https://github.com/apple/swift-evolution/blob/master/proposals/NNNN-name.md)
* Authors: [Douglas Gregor](https://github.com/DougGregor), [Joe Groff](https://twitter.com/jckarter), Austin Zheng
* Status: **[Awaiting review](#rationale)**
* Review manager: TBD

## Introduction

Swift's support for existential types is currently quite limited: two or more protocols can be composed using the `P1 & P2 & P3` syntax. We propose, in the spirit of [*Completing Generics*](https://github.com/apple/swift/blob/master/docs/GenericsManifesto.md), to add support to Swift for describing more complex existential types, including those involving protocols with associated types.

## Acknowledgements

Many thanks to [Matthew Johnson](https://github.com/anandabits), Thorsten Seitz, [David Smith](https://twitter.com/Catfish_Man), [Adrian Zubarev](https://github.com/DevAndArtist), Dave Abrahams, and TBD, whose feedback and contributions were instrumental in composing this proposal.

## Motivation

There are a variety of useful types which cannot be expressed currently within Swift's type system. Here are two examples:

1. **Types involving protocols with associated types, and constraints upon those associated types**.

This file has been truncated. show original

The proposal was never merged and postponed for a future Swift release. This feature is very important for me so I reopened it earlier this month:

https://lists.swift.org/pipermail/swift-evolution/Week-of-Mon-20170102/029890.html

Douglas Gregor replied to this thread saying that part of the proposal (superclass and protocol constraints) should be proposed and implemented for Swift 4, but that the remaining more generalised has little chance to make it in time for Swift 4.

In consequence, I started a new thread with the smaller feature of superclass and protocol constraints:

https://lists.swift.org/pipermail/swift-evolution/Week-of-Mon-20170123/031066.html

Quite a few comments have already been made about the proposal so I’ll post a new draft soon. But please let me know if you have more comments!

David.

···

On 31 Jan 2017, at 11:07, Chris Eidhof via swift-evolution <swift-evolution@swift.org> wrote:

Hi swift-evolution,

I was wondering if anything more detailed has been written about generalized existentials / type erasers. I noticed that AnyCollection only delegates a number of methods to its wrapped value, but often doesn't. For example, all the conditionally inherited items aren't being delegated (e.g. min()). This behavior was surprising to me (but makes sense once I thought about it longer).

How would this work with generalized existentials? Are generalized existentials in scope at all for Swift 4?

--
Chris Eidhof
_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution