Patch for clang-format to allow references and pointers to bind differently

[Alp Toker]

On 05/12/2013 10:10, Daniel Jasper wrote:
> As I have mentioned before, I don't think this is a useful feature 
> (let alone being a very intrusive change on the current codebase). 
> Whenever we have done anything into this direction, we have been 
> getting the feedback that it is not helpful. People want automatically 
> enforceable styles and the consistency it brings them. Bailing out of 
> doing specific changes is not the solution.

Then what is the solution?

Skipping an unenforceable format style seems to offer 99% of the 
benefits of the tool.

The user can then potentially do remaining polishing with a perl script 
or something if they want complete enforceability of an unusual style.

If they want full enforceability then I think it's reasonable to expect 
users to stay close to the currently supported styles. Not convinced 
that offering yet another way of formatting pointer formatting is the 
way to go.

The strategy also makes sense for column width violations when 
formatting comments. Right now clang-format pushes a dangling word the 
next line, which is enforcing something, but not what anyone wants.

If clang-format were to leave the line it didn't understand how to 
reflow untouched (maybe with a note to stderr), it'd be easier to get to 
a finished result with another tool or manually.

Of course if this is work to implement, then the other option might be 
easier. It's just I was under the impression ignoring an edit would be 
straightforward, which might not be the case in TokenAnnotator.cpp :-)

Alp.



>
> On Thu, Dec 5, 2013 at 10:39 AM, Alp Toker <alp at nuanti.com 
> <mailto:alp at nuanti.com>> wrote:
>
>
>     On 04/12/2013 22:45, Daniel Jasper wrote:
>
>         --detect-style will be nice, but it is going to be
>         significantly more effort than you might think.
>
>
>     Returning to my idea to "leave X formatting untouched", I imagine
>     this might give Ben a chance to get usage out of the tool while
>     also becoming a first step towards --detect-style.
>
>     The algorithm would be roughly:
>
>     for each formatting criterion X, apply X alone to the in-memory
>     buffer and count deltas. Report ambiguity if Xn == 0, otherwise
>     select X with lowest number of deltas as the detected style variable.
>
>     I'm sure there are other ways to do it, so just throwing about the
>     idea if only to avoid the perennial debate on what constitutes
>     acceptable corner-case styles :-)
>
>     Alp.
>
>
>     -- 
>     http://www.nuanti.com
>     the browser experts
>
>

-- 
http://www.nuanti.com
the browser experts