A minor syntax request

Daniel van Vugt daniel.van.vugt at canonical.com
Wed Apr 17 01:28:28 UTC 2013

Actually I was not requesting "/3/ Allowing multiple variables to be 
defined in a declaration". Just using it as an example. Although I would 
like to be able to do so, use cases for it are less common so am happy 
to say one variable per declaration if that is the style.

And I totally agree that */& being part of the variable was a poor 
language design decision. But that was probably 40 years ago. It's 
established and that's the way it works.

Back on the const'ness issue... Another reason for const prefixing is 
this case:
     const T * const x = y;
If you don't prefix const then it is significantly more confusing:
     T const * const x = y;

On 16/04/13 18:32, Alan Griffiths wrote:
> On 16/04/13 11:06, Daniel van Vugt wrote:
>> Can we please start to allow the style:
>>      const T x
>>      const T &x
>>      T *x
>> I believe it is probably the most common syntax in general industry
>> use, and also one that many of us are used to and most comfortable
>> with. Other arguments for this syntax are:
>>    1. It's part of the C/C++ language definition that &/* belongs to
>> the variable and not the type...
>>         T x, *y = &x;
>>    2. It feels more natural for many readers to think "constant T" when
>> reading "const T", than when reading "T const".
>>    3. It's more obvious that your variable is a variable if you say:
>>         const T *a;
>>         a = b;
>>       However writing it this way, it's less obvious that a is not
>> constant itself:
>>         T const* a;
>>         a = b;
> C/C++ declaration syntax is famously "an experiment that failed", but
> using the well known "right to left" heuristic gives:
> "a is a pointer to constant T" vs "a is a pointer to T constant"
>> Do we really need to be so strict about using the Mir convention:
>>      T const x
>>      T cons& x
>>      T* x
>> ?
> There are thee related requests hidden here.
> /1/ Putting cv-qualifiers to the left of the type in the few (but
> common) cases it is allowed syntactically.
> Most cases have to be to the right anyway, so I've a mild preference to
> "on the right" for consistency, and a strong preference for using what
> the rest of the code does.
> /2/ Putting * or & with the name, not the type.
> /3/ Allowing multiple variables to be defined in a declaration
> Putting * or & with the type has a long standing tradition within the
> C++ culture - it started with the first edition of "The C++ Programming
> Language" and has been common practice ever since.
> There is an associated tradition of declaring each variable separately.
> (These two go together as "pretending" the type includes * or & doesn't
> work if multiple variables are declared.)
> This is very different to the C tradition.

More information about the Mir-devel mailing list