creal.re and creal.im are not lvalues: is that intentional?

[Lars T. Kyllingstad]

Denis Koroskin wrote:
> For some reason compiler rewrites someComplexVar.re and 
> someComplexVar.im into cast(real)someComplexVar and 
> cast(ireal)someComplexVar respectively, which results in those variables 
> being non-lvalues and me being unable to update either of a properties 
> independently of each other. As one of the consequences, I can't set 
> imaginary part of a complex variable to signalling nan:
> 
> real snan = real.init; // reals are signalling nans by default
> creal c = ...;
> //c.im = snan; // doesn't work
> c = snan + snan * 1i;
> 
> assert(isSignallingNan(c.re)); // pass
> assert(isSignallingNan(c.im)); // fail
> 
> c = snan + cast(ireal)snan; // doesn't work either
> 
> I know complex types are scheduled for deprecation but I'm still confused.


I don't know whether the current behaviour is intentional, but in 
std.complex.Complex, which is supposed to replace the built-in types, 
Complex.re and Complex.im are ordinary public member variables.

-Lars