Feedback on Átila's Vision for D

[Paul Backus]

On Wednesday, 16 October 2019 at 06:53:37 UTC, Sebastiaan Koppe 
wrote:
> Surely it would be possible to define a partial ordering over 
> our template constraints.
>
> Since isRandomAccessRange!R includes isForwardRange!R (which 
> includes isInputRange!R), the compiler should be able to 
> determine that isRandomAccessRange!R is more strict than 
> isInputRange!R.
>
> If anything, you could simply count the && and the ||'s. That 
> is, A && B > B. That ought to be a simple AST+lookup operation. 
> Right?

Template bodies may be arbitrarily complex, and determining 
whether one template predicate implies another is 
halting-equivalent in the general case.

The only way you can do this kind of analysis is if you define, 
in the language spec, a restricted subset of templates to apply 
it to--for example, enum templates whose body consists of a 
single boolean expression.

Personally, I find that kind of special-case handling a bit 
unprincipled, so it wouldn't be my first choice, but it's 
probably the shortest path from where we are now to partial 
ordering for template constraints.