[swift-evolution] SE-165: Dictionary & Set Enhancements

[Nate Cook]

> On Apr 5, 2017, at 9:43 PM, Brent Royal-Gordon via swift-evolution <swift-evolution at swift.org> wrote:
> 
>> On Apr 5, 2017, at 5:45 PM, Ben Cohen via swift-evolution <swift-evolution at swift.org <mailto:swift-evolution at swift.org>> wrote:
>> 
>> 	• What is your evaluation of the proposal?
> 
> (As a meta issue, I'm not sure I like the grab-bag review style; I'm finding this proposal a little bit difficult to navigate.)
> 
> Sequence-based initializer and merging initializer
> 
> Good idea, but I think these two are redundant with each other, and I don't think "merging" is an accurate way to describe what the second one does. (`merging` would suggest to me that it was combining several dictionaries or lists, not combining conflicting elements.) I'd suggest a single initializer along the lines of:
> 
> 	init<S: Sequence>(_ keysAndValues: S, correctingConflictsBy resolveConflict: (Value, Value) throws -> Value = { fatalError("Duplicate values \($0) and \($1)") }) rethrows
> 		where S.Iterator.Element == (key: Key, value: Value)

Thanks for all your feedback, Brent! One note on this item in particular—if you specify a default argument for a throws/rethrows closure, you have to use "try" at the call site even if the default closure argument doesn't throw. Modules currently don't promise that default closure arguments don't throw, and a default argument could change from non-throwing to throwing in a later version of a library.

There's a bug tracking the issue here: https://bugs.swift.org/browse/SR-2979

> Merging methods
> 
> Good idea, but I'm not a fan of the `mergingValues:` label. I would suggest the same `correctingConflictsBy resolveConflict:` label I suggested for the previous method—possibly including the default value. I also think `merge(_:correctingConflictsBy:)`'s first parameter should be labeled `with`, just as the `merged` variant is.
> 
> I wonder if we might also want a method that copies the Dictionary, but with a single key added/removed/changed:
> 
> 	func withValue(_ value: Value?, forKey key: Key) -> [Key: Value]
> 
> Key-based subscript with default value
> 
> I like the functionality, but not way this proposal does it. I don't like having the default value be a labeled parameter to the subscript, because it isn't used to locate the value. However, I can't come up with a better syntax without adding language features. What I'd like to do is make it possible to assign through `??`:
> 
> 	frequencies[c] ?? 0 += 1
> 
> But that would require either that we support `inout` functions, or that `??` become magic syntax instead of a standard library feature. The former is not coming in Swift 4 and the latter is less than ideal.
> 
> Still, if we would rather have that syntax and we think we'll soon have the language improvements needed to pull it off, I'd suggest rejecting this portion of the proposal.
> 
> Dictionary-specific map and filter
> 
> I am +114 on this. I say that because I have received 114 upvotes on my Stack Overflow answer explaining how to write a `Dictionary.map` method: <http://stackoverflow.com/questions/24116271/whats-the-cleanest-way-of-applying-map-to-a-dictionary-in-swift/24219069#24219069 <http://stackoverflow.com/questions/24116271/whats-the-cleanest-way-of-applying-map-to-a-dictionary-in-swift/24219069#24219069>>
> 
> I agree with the decision not to pass keys to the closures in these methods; that keeps them simple and focused, and ensures they stay parallel with ordinary `map` and `filter`. I also agree with the decision to not build in a form of `map` which allows key remapping; you can always do that with the sequence-based initializer, which would let you add conflict-handling logic more elegantly than a key-value `map` could.
> 
> Visible dictionary capacity
> 
> I doubt I'll ever use the `capacity` property, but I'm not opposed to it. Adding a `reserveCapacity(_:)` method is a good idea.
> 
> A grouped(by:) method for sequences
> 
> Yes, please.
> 
> Apply relevant changes to Set
> 
> These make sense. (I considered suggesting the `filter` method be called `intersection(where:)`, but on second thought, I don't think that conveys the actual use cases for the method very well.)
> 
> I wonder if we might want to conform `Dictionary` to `SetAlgebra`. They have compatible semantics, and I've occasionally wanted to use them in the same places. On the other hand, some of the methods might form attractive nuisances; perhaps I should just write a SetAlgebra-conforming view when I want to do that.
> 
> General notes
> 
> If SE-0161 is accepted, we may want to support key path variants of some of these APIs (like `grouped(by:)`, `map(_:)`, and `filter(_:)`). On the other hand, we may want to defer that so we can consider that issue holistically, including with existing Collection APIs.
> 
>> 	• Is the problem being addressed significant enough to warrant a change to Swift?
> 
> Yes. These are all common needs; when working with dictionaries, I find myself writing `for` loops with `var`s far more often than I'd like.
> 
>> 	• Does this proposal fit well with the feel and direction of Swift?
> 
> Yes, these are all pretty straightforward additions.
> 
>> 	• If you have used other languages or libraries with a similar feature, how do you feel that this proposal compares to those?
> 
> Ruby's `Hash` has many of these features and I appreciate them there; `NSDictionary` does not and it suffers for it. Perl hashes have a flattening behavior that tends to be amenable to editing them in various ways, but that's not really suitable for Swift.
> 
>> 	• How much effort did you put into your review? A glance, a quick reading, or an in-depth study?
> 
> 
> Quick reading.
> 
> -- 
> Brent Royal-Gordon
> Architechies
> 
> _______________________________________________
> swift-evolution mailing list
> swift-evolution at swift.org
> https://lists.swift.org/mailman/listinfo/swift-evolution

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-evolution/attachments/20170406/f26ae505/attachment.html>