[swift-dev] Fixit for trailing closures

[Xi Ge]

> On Jul 5, 2016, at 8:31 PM, Jordan Rose <jordan_rose at apple.com> wrote:
> 
>> 
>> On Jul 5, 2016, at 17:19, Ben Langmuir via swift-dev <swift-dev at swift.org <mailto:swift-dev at swift.org>> wrote:
>> 
>>> 
>>> On Jul 5, 2016, at 4:34 PM, Xi Ge via swift-dev <swift-dev at swift.org <mailto:swift-dev at swift.org>> wrote:
>>> 
>>> Hi Swift-devs,
>>> I have tried to add a fixit to help developers using trailing closures more, motivated by my observation during WWDC that some developers 
>>> do not even realize that we have such a feature. In my opinion, trailing closures are more concise, and once you get used to it, more readable; 
>>> therefore users should adopt trailing closures whenever doing so introduces no ambiguity.
>>> 
>>> Fixits can enhance the discoverability of trailing closures by identifying misuses and by transforming users’ code automatically. However, adding the fixit introduces new issues:
>>> 
>>> Issue 1: The fixit has to be associated with a warning. Adding the warning means we declare wars against convertible non-trailing closures, which is a valid syntax choice by users.
>>> 
>>> Issue 2: Ambiguity checking should be exhaustive. We have several known situations when non-trailing closures cannot be convert to trailing closures, including:
>>> 	
>>> Trailing closures are followed by other brackets, e.g., “if foo({}) {}” cannot be converted to “if foo {} {}”.
>>> Removing the label of the last closure causes ambiguous function references, e.g. “foo(v: {})” cannot be converted to “foo {}” when “foo(v1: {})” also exists.
>>> 
>>> So Swift-devs, is the warning worth adding? If yes, are there other situations of ambiguity that are not covered?
>>> 
>>> Thanks for your feedback!
>>> Xi
>>> 
>> 
>> -1, this feels like a really good tool to have in a code-linter, but not something that we should put in the compiler and prescribe to all our users.
>> 
>> My biggest concern is that it’s not always obvious what the closure is used for.  The argument label can help with this:
>> 
>> [1].lexicographicallyPrecedes([0]) { <#code#> }
>> [1].lexicographicallyPrecedes([0], isOrderedBefore: { <#code#> })
> 
> I strongly agree with Ben. This is a style choice and highly context-dependent. We should not warn about it or produce a fix-it unless specifically requested, and we don’t have any place for such requests today.
> 
> (Motivation: Another case where I prefer not using a trailing closure is when there are other closure arguments in the call.)
> 

I agree with you if by “context-sensitive”, you mean “function-signature-sensitive”; and we can solve that by adding a new attribute on function decls to categorize a function into either “trailing closure preferred” or “inline closure preferred”.
“lexicographicallyPrecedes” is a good example of the later while "DispatchQueue.async(execute:)” is of the former. Any other contexts in your mind?
Xi

> Jordan

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-dev/attachments/20160705/bad711ad/attachment.html>