[swift-evolution] [Pitch] Retiring `where` from for-in loops

[Thorsten Seitz]

> Am 11.06.2016 um 21:57 schrieb Xiaodi Wu <xiaodi.wu at gmail.com>:
> 
>> On Sat, Jun 11, 2016 at 2:50 PM, Thorsten Seitz <tseitz42 at icloud.com> wrote:
>> 
>> 
>>> Am 10.06.2016 um 17:22 schrieb Erica Sadun via swift-evolution <swift-evolution at swift.org>:
>>> 
>>> 
>>>> On Jun 10, 2016, at 8:02 AM, Xiaodi Wu via swift-evolution <swift-evolution at swift.org> wrote:
>>>> 
>>>> On Fri, Jun 10, 2016 at 7:18 AM, Haravikk <swift-evolution at haravikk.me> wrote:
>>>>>> * The word "where" does not consistently imply `break` or `continue`. In current Swift, `where` implies `break` in the context of a `while` loop and `continue` in the context of a `for` loop. Some users intuitively guess the correct meaning in each context, while others guess the wrong meaning. Therefore, the only way to learn for sure what `where` means in any context is to read the rulebook. That, by definition, means that this is unintuitive.
>>>>> 
>>>>> This is an argument for renaming the where keyword on for loops to be more clear, or to somehow integrate continue/break to be more explicit about what the developer intends for it to do.
>>>> 
>>>> Sure: I conclude that the keyword should be *either* removed *or* reformed; both outcomes could address the issue.
>>> 
>>> This is my stance as well and I reserve the right to flit between both choices until we've fully talked it through.
>>> 
>>> One more data point. 
>>> 
>>> In the standard library there are just under 950 uses of "for in loops". There are 3 uses of "for in while" :
>> 
>> How many of the 950 "for in loops" use a guard with continue? Only these can be compard with the "for in where loops".
> 
> That is a different argument, I think, from the one Erica is making. The data you're asking about would answer the question, "How often do stdlib authors prefer `guard` over `where`?" That is an interesting question, certainly.

It is the relevant question to ask when deciding whether `where` should be better replaced by `guard`. And it should be qualified by asking whether the respective author did know about `where`. I wouldn't want a language to be designed by looking at usage examples of inexperienced users...

> 
> I think Erica's point is that what Vladimir earlier called the "simple, common" case may be simple but isn't actually common. I did a quick GitHub search earlier today (and the tools aren't there, afaik, for an accurate regexp search), but what I noticed on a random, not-statistically-sound sampling was that uses of `continue` and `break`, when they do occur inside a for loop (which isn't always), actually tend to happen after some work has been done at the top of the loop. These wouldn't be replaceable by a `where` clause. Moreover, I noticed an appreciable share of `return` and `fatalError()` calls from inside the loop, which surprised me; I had assumed I'd find mostly `break` or `continue`, but even when I did find those it was as likely as not to be accompanied by logging. These again preclude refactoring into `where`.

Logging hints at logic used for filtering out data errors. That's a good usage for `guard` as opposed to `where`. The latter is about logic, not about error handling. 

-Thorsten 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-evolution/attachments/20160611/c6911753/attachment.html>