Re: [eigen] Strange operator behaviour

[Christoph Hertzberg <chtz@xxxxxxxxxxxxxxxxxxxxxxxx>]

On 19.09.2014 15:01, mwb@xxxxxxxxxxxx wrote:
> Personally, I would prefer to make expression ctors explicit, since the
> current behavior is quite unintuitive and potentially dangerous, though
> you probably get used to it.

I totally agree on making any constructor explicit when in:
  X x;  // some kind of expression or scalar
  A a1(x);
  A a2; a2=x;
The last line is either not possible or leads to different behavior than 
the line before. In other words, only make it implicit if
  A a3=x;
has the same result as the a2 case.
I can't think of any case, where we would break intended behavior with 
changing that -- I guess it will rather make some show some unintended 
misuses visible.

I agree that usage will be a bit more painful, e.g.

  TransposeType transpose() const {
    return derived();
  }

and related, would need to be replaced by (the IMO easier to understand):

  TransposeType transpose() const {
    return TransposeType(derived());
  }

Actually, this would also motivate to consequently add typedefs for all 
return types.


> > Does Eigen3.2.2 somewhere overrides the ?: operator with some strange side
> > effects?

Overloading ?: is not possible:
http://en.wikipedia.org/wiki/Operators_in_C_and_C++#Other_operators

Christoph



--
----------------------------------------------
Dipl.-Inf., Dipl.-Math. Christoph Hertzberg
Cartesium 0.049
Universität Bremen
Enrique-Schmidt-Straße 5
28359 Bremen

Tel: +49 (421) 218-64252
----------------------------------------------