[swift-evolution] [Pitch] Remove destructive consumption from Sequence

Matthew Johnson matthew at anandabits.com
Mon Jun 27 09:40:19 CDT 2016

> On Jun 26, 2016, at 10:56 PM, Jonathan Hull via swift-evolution <swift-evolution at swift.org> wrote:
> Can’t a Sequence be potentially infinite, whereas a collection has a defined count/endIndex?  Other than that, I agree with your statement.
> Here is what I see as the appropriate structure:
> Iterator: Single destructive pass, potentially infinite, (should be for-in able)
> Sequence: Guaranteed non-destructive multi-pass (vends Iterators), potentially infinite, (should be subscript-able, gain most of collection, but lose anything that relies on it ending)
> Collection: Multi-pass, guaranteed finite, (no changes from current form, except extra inits from Iterator/Sequence with end conditions)
> Right now we are allowed to have an infinite sequence, but calling dropLast or non-lazy map will cause an infinite loop.  These cases could be made much safer by considering the potentially infinite and finite cases separately…

I think this is pointing in the right general direction.  It would make working with `Sequence` much more straightforward and allow us to depend on the multi-pass property that is true in practice of the most common models of `Sequence`.

But I agree that we should give much more careful consideration to finite / infinite generally and for..in specifically.

Now that I have been thinking about the finite / infinite distinction more closely I have begun to notice a lot of code that is written generically using `Sequence` where a for..in loop is really what is required, however the “finite sequence” precondition is not explicitly stated.  Interestingly, this is the case with the standard library’s eager `map` (but not the case with `dropLast` which explicitly notes the precondition).  I have been somewhat surprised to realize how common this “bug” is (i.e. not stating a precondition).  I think we have gotten away with it thus far because the sequences most people use most of the time in practice are finite.  But that doesn’t mean we should accept this as good enough - IMO it is way to easy to forget to document this precondition (and obviously easier for users to overlook than preconditions that are actually encoded in the type system, violations of which are caught at compile time).

The fact that this pattern is so pervasive is what I meant when I said for..in “naturally” requires a finite sequence. 

IMO it’s better to encode preconditions in the type system when that is practical, and especially when the precondition is shared by a vast majority of code written using a particular construct (in this case a for..in loop written using the most generic for..in-able protocol).

I think the safest solution is to take the position that writing an infinite loop is relatively uncommon and is a more “advanced” technique, and thus should be done explicitly.  Do people really write infinite loops often enough that the convenience of using for..in when writing infinite loops outweighs the safety benefit of preventing accidental infinite loops?  I haven’t seen a compelling argument for this.  

If we adopt that position then for..in would need to be built on top of a guaranteed finite construct.  This would allow programmers to continue writing generic code agains the most generic for..in-able construct while eliminating a precondition that is often (usually?) unstated and likely unconsidered.  

If we do decide to move forward with infinite for..in loops I think we need to establish strong guidance around how to properly write generic code with these protocols.  Should such code really be constrained to `Collection` rather than `Sequence` (i.e. should a potentially infinite `Sequence` have an eager map)?  If this is the guidance, should it be paired with guidance that all finite sequences should conform to `Collection`?  Or is it sufficient to just educate developers about this issue and expect people to document the “finite Sequence” precondition when the constraint is `Sequence` rather than `Collection`?  

I hope we will give serious consideration to these questions while this topic is open for discussion.


> Thanks,
> Jon
>> on Wed Jun 22 2016, David Waite <david-AT-alkaline-solutions.com <http://david-at-alkaline-solutions.com/>> wrote:
>> >> On Jun 22, 2016, at 2:57 PM, Dave Abrahams via swift-evolution <swift-evolution at swift.org <https://lists.swift.org/mailman/listinfo/swift-evolution>> wrote:
>> >> 
>> >> <Ahem> “Iterators,” please.
>> >
>> > That makes me happy - for some reason I thought it was still GeneratorProtocol
>> >
>> >>> destructively, but such Generators would not conform to the needs of
>> >>> Sequence. As such, the most significant impact would be the inability
>> >>> to use such Generators in a for..in loop, 
>> >> 
>> >> Trying to evaluate this statement, it's clear we're missing lots of
>> >> detail here:
>> >> 
>> >> * Would you remove Sequence?
>> >> * If so, what Protocol would embody “for...in-able?”
>> > No, I would just remove the allowance in the documentation and API
>> > design for a destructive/consuming iteration. Sequence would be the
>> > interface to getting access to repeatable iteration, without the need
>> > for meeting the other requirements for Collection.
>> That would be wrong unless there exist substantial examples of a
>> multipass Sequence that *can't* meet the other requirements of
>> Collection without loss of efficiency.  And since I can write an adaptor
>> that turns any multipass sequence into a Collection, I think it's
>> trivial to prove that no such examples exist.
>> -- 
>> -Dave
> _______________________________________________
> 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/20160627/73aacecc/attachment.html>

More information about the swift-evolution mailing list