[cfe-dev] clang-format behavior for Obj-C methods with block parameters

Daniel Jasper via cfe-dev cfe-dev at lists.llvm.org
Thu Dec 24 02:19:05 PST 2015


I don't know anything about that, don't use ObjC myself. The main problem
will be finding someone who can maintain this in the long(er) run. Adding
Nico for additional thoughts.

On Thu, Dec 24, 2015 at 9:25 AM, Yavuz Nuzumlalı <manuyavuz at gmail.com>
wrote:

> Hi Daniel,
>
> Are you OK with an option like "ObjCColonAlignMultipleBlocks" which aligns
> multiple block arguments from colon. I think this is needed because
> aligning multiple block arguments from colon character is so common in
> Objective-C community. It's also the default indentation behavior of Xcode.
>
> We can still set this to false by default, but could provide an option to
> disable that behavior?
>
> I've done an implementation internally, if you're OK and suggest a proper
> key, I can send a patch for it.
>
> On Mon, Dec 14, 2015 at 1:16 PM Daniel Jasper via cfe-dev <
> cfe-dev at lists.llvm.org> wrote:
>
>> On Mon, Dec 14, 2015 at 9:43 AM, manuyavuz <yavuz.nuzumlali at gmail.com>
>> wrote:
>>
>>> On 13 Dec 2015, at 20:15, Daniel Jasper <djasper at google.com> wrote:
>>>
>>>
>>>
>>> On Sun, Dec 13, 2015 at 6:53 PM, Yavuz Nuzumlalı <manuyavuz at gmail.com>
>>> wrote:
>>>
>>>> Thanks for the response Daniel,
>>>>
>>>> Could you guide me on where can I look inside clang-format
>>>> implementation to see how clang-format currently handles lambdas in other
>>>> languages, so that I may try to implement the same functionality for
>>>> Objective-C case?
>>>>
>>>
>>> Not really, it is all in ContinuationIndenter.cpp. However, this is
>>> among the most complex code of clang-format and giving you details on how
>>> to do this probably would take me longer than looking at this myself. If
>>> you still want to try and take a look, feel free. I am happy to answer any
>>> specific questions you might have.
>>>
>>>
>>>>
>>>> Besides that, I have one more question, which is somewhat related to
>>>> topic.
>>>>
>>>> When I try to use clang-format to format the following simple code:
>>>>
>>>> - (void)asdf {
>>>> [UIView animateasdfasdf:0 animations:^{ }];
>>>> [UIView animations:^{ } done:YES];
>>>> [UIView animations:^{ } completion:^{ }];
>>>> [UIView testasdfasdfsd:YES animations:^{ } completion:^{ }];
>>>> [UIView animations:^{ } tessd:YES completion:^{ }];
>>>> }
>>>>
>>>> Output is the following:
>>>>
>>>> - (void)asdf {
>>>>   [UIView animateasdfasdf:0
>>>>                animations:^{
>>>>                }];
>>>>   [UIView animations:^{
>>>>   }
>>>>                 done:YES];
>>>>   [UIView animations:^{
>>>>   }
>>>>       completion:^{
>>>>       }];
>>>>   [UIView testasdfasdfsd:YES
>>>>       animations:^{
>>>>       }
>>>>       completion:^{
>>>>       }];
>>>>   [UIView animations:^{
>>>>   }
>>>>       tessd:YES
>>>>       completion:^{
>>>>       }];
>>>> }
>>>>
>>>> I think clang-format somehow decides to not use colon alignment but
>>>> align using `ContinuationIndentWidth` where there are more than one
>>>> block parameters.
>>>> Is this an intended behavior? Seems like a bug to me.
>>>>
>>>
>>> It is intentional. Column alignment with more than one internal block
>>> doesn't really work. And this is also explicitly written in Google's style
>>> guide: https://google.github.io/styleguide/objcguide.xml#Blocks
>>>
>>> Independent of that some of you examples look like there are still other
>>> bugs in there.
>>>
>>> I carefully read the google style guide, and behavior about multiple
>>> inline blocks seems reasonable now.
>>> However, I still feel it's not fully consistent because of the third
>>> method call example above.
>>> I think that case should have something like the following format in
>>> order to be consistent across the code:
>>>
>>> [UIView
>>>     animations:^{
>>>     }
>>>     completion:^{
>>>     }];
>>>
>>
>> As I mentioned in my previous email, there are cases in your example that
>> look buggy to me. I was just stating that in general, this is intended.
>>
>> Also there is another case in which formatting messes completely:
>>>
>>>   [UIView animateWithDuration:123 delay:123 usingSpringWithDamping:123
>>> initialSpringVelocity:123 options:123 animations:^{ } completion:^(BOOL
>>> finished){ }];
>>>
>>> becomes this:
>>>
>>>   [UIView animateWithDuration:123
>>>       delay:123
>>>       usingSpringWithDamping:123
>>>       initialSpringVelocity:123
>>>       options:123
>>>       animations:^{
>>>       }
>>>       completion:^(BOOL finished){
>>>       }];
>>>
>>> Is this (cancelling colon alignment also for non-block parameters ) also
>>> intentional?
>>>
>>
>> It is intended and I think it makes sense. However, I have never worked
>> on ObjC code, so my understanding of the pros and cons is *very* limited.
>>
>>
>>> If so, I understand that clang-format will apply standard
>>> `ContinuationIndentWidth` value whenever a method call includes more than
>>> one inline blocks.
>>>
>>> I question this scenario too much because the above code is a frequent
>>> code block to animate views on screen, and having a consistent behavior is
>>> important in order to use clang-format’s automatic formatting without any
>>> hesitancy
>>>
>>
>> Sure, I understand. But as I mentioned above, I might actually not be the
>> best person to ask. As the main author of clang-format, I try to support
>> ObjC on a best-effort basis, without understanding much of it. There is
>> also limited time I can spend on this at least in the foreseeable future.
>>
>> Cheers
>> Daniel
>>
>>
>>>
>>> PS : I'm using the default LLVM configuration for the above case.
>>>>
>>>> On 13 Dec 2015, at 16:42, Daniel Jasper <djasper at google.com> wrote:
>>>>
>>>> Nope, this behavior isn't currently implemented. I agree that it might
>>>> make sense, as it is similar to how clang-format treats lambdas at the end
>>>> of several other parameters.
>>>>
>>>> On Fri, Dec 11, 2015 at 4:08 PM, Yavuz Nuzumlalı <
>>>> cfe-dev at lists.llvm.org> wrote:
>>>> Hi everyone,
>>>>
>>>> I'm experiencing a problem with clang-format for my code.
>>>>
>>>> I'm using ColumnLimit as 120. When I write some code like this:
>>>>
>>>> [self testMethod:15 withBlock:^{
>>>> NSLog(@"asdfasdf");
>>>> NSLog(@"asdfasdf");
>>>> NSLog(@"asdfasdf");
>>>> NSLog(@"asdfasdf");
>>>> }];
>>>>
>>>> This code would be converted to this by clang-format:
>>>>
>>>> [self testMethod:15
>>>>        withBlock:^{
>>>>        NSLog(@"asdfasdf");
>>>>        NSLog(@"asdfasdf");
>>>>        NSLog(@"asdfasdf");
>>>>        NSLog(@"asdfasdf");
>>>>    }];
>>>>
>>>> That seems wrong to me because the method call before block variable
>>>> actually fits in width of 120. I guess this is because clang-format behaves
>>>> above code as a single method call and assumes it as one line. But I really
>>>> want to not split the method like this because it feels weird.
>>>>
>>>> Is there a way to accomplish this with current system?
>>>>
>>>> I put my configuration file as an attachment.
>>>>
>>>> Best
>>>>
>>>>
>>>> _______________________________________________
>>>> cfe-dev mailing list
>>>> cfe-dev at lists.llvm.org
>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>>>>
>>>>
>>>>
>>> _______________________________________________
>> cfe-dev mailing list
>> cfe-dev at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20151224/c868ee38/attachment-0001.html>


More information about the cfe-dev mailing list