> I fled from C++ back to C exactly because of templates
Obviously I sympathize as someone who spent the front half of their career writing C and has no interest in writing C++ beyond toy examples. Nevertheless, IMO you're cutting off a choice here, forcing yourself to make potentially sub-optimal choices.
Sometimes the templates would have been better and writing X-macros is just worse, ergonomically at least. Perhaps C++ people use templates more often than they should, but I'm sure "never" is not the correct amount.
Still, if C gets fat pointers that's at least a step in the right direction. So I must wish you the best with that part.
I used to write expression template libraries, so I know what C++ can do. I just figured out I am a lot more productive in C than in C++ and that in the time I waste on C++ I can do a lot more in C. I also do rarely use macros in C, so what C++ people get wrong is assuming that C programmer would do the same thing just using macros. This is not the case. One has to realize that compile-time programming is 95% of times a waste of time and effort and the same thing can be achieved in an easier and better way.
I also recently removed some code written with templates in C++ from one of my projects, because this single file caused compilation times for a ca. 450 file C project to increasse by 50% from 20s to ca. 30s.
> Still, if C gets fat pointers that's at least a step in the right direction.
AFAIK the C standard doesn't prevent an implementation from using fat pointers, this is one of the reasons why the conversion from pointer to integer only works in one direction. This is actually necessary for segmented memory. The compiler is allowed to optimize based on the assumption, which allocation a pointer comes from (including the allocations boundaries, i.e. size), even if bytewise the pointers would be equal, so you could argue, that C abstract machine already has fat pointers.
Indeed. C implementation could just use fat pointers, but it is unrealistic that they now break their ABI. Also the cost of fat pointers is higher when not needed. So the plan is to give them a new wide pointer type.
Obviously I sympathize as someone who spent the front half of their career writing C and has no interest in writing C++ beyond toy examples. Nevertheless, IMO you're cutting off a choice here, forcing yourself to make potentially sub-optimal choices.
Sometimes the templates would have been better and writing X-macros is just worse, ergonomically at least. Perhaps C++ people use templates more often than they should, but I'm sure "never" is not the correct amount.
Still, if C gets fat pointers that's at least a step in the right direction. So I must wish you the best with that part.