Optimisation possibilities: current, future and enhancements

[Patrick Schluter via Digitalmars-d]

On Friday, 26 August 2016 at 17:52:36 UTC, Meta wrote:
> On Friday, 26 August 2016 at 14:12:24 UTC, Basile B. wrote:
>> That's another story. Of course the optimization pass for this 
>> should check that **ALL** the calls to Test in a sub program 
>> (or in this scope if you want) are const... Which is not the 
>> case here.
>
> My point is that getN is strongly pure but is not referentially 
> transparent. Purity is broken in a strange way that you 
> wouldn't expect, and you don't even have to use an escape hatch 
> like debug.

getN() is not pure at all, it refers to an object (in C term) 
outside of its scope not depending on its parameter. It is 
irrelevant if the outside object is a global variable or a 
member. Access to a memory defined outside of its scope breaks 
pureness.