Should negating an unsigned integral be an error?

[Michael]

On Tuesday, 30 January 2018 at 14:08:41 UTC, bauss wrote:
> On Tuesday, 30 January 2018 at 11:09:08 UTC, Michael wrote:
>> On Monday, 29 January 2018 at 10:06:41 UTC, Jonathan M Davis 
>> wrote:
>>> On Monday, January 29, 2018 09:58:00 Dave Jones via 
>>> Digitalmars-d wrote:
>>>> Given
>>>>
>>>> uint i = 12345;
>>>>
>>>> should
>>>>
>>>> writeln(-i)
>>>>
>>>> be an error? or maybe i should be automatically cast to a 
>>>> larger signed type?
>>>
>>> It arguably should be, but it's pretty common in C/C++ to get 
>>> uint.max with -1, and I suspect that it's fairly common in D 
>>> as well even though you could just use uint.max. So, it could 
>>> be argued either way.
>>>
>>> - Jonathan M Davis
>>
>> I'd like to see these kinds of potential mistakes to be forced 
>> to use something like uint.max. Of course, at this point it 
>> might be too late to introduce something as breaking as this, 
>> but these kinds of things should definitely be forced by the 
>> language, or made a warning at least, pointing to the use of 
>> .max instead.
>
> A normal deprecation process should be enough, since it'll give 
> time to update code that uses it.
>
> I don't think there is a lot of idiomatic D code that uses this 
> feature.

But it's one of those bad habits lots of people have probably 
gotten into because it was the way to do things in C/C++ or 
whatever. Good point, though. I think it would be worth catching 
all of these small potential errors that can be replaced with 
code that is much more clear as to its intent.