Re: [eigen] Scalar vs. int / double in SelfAdjointEigenSolver.h

On Fri, Aug 12, 2016 at 2:24 PM, Christoph Hertzberg <chtz@xxxxxxxxxxxxxxxxxxxxxxxx> wrote:

On 2016-08-11 08:40, Pavel Holoborodko wrote:

*...which may not the case for user defined type.*

From my experience of creating custom scalar type (mpreal) -
I can say that keeping code without explicit conversions to Scalar is very
beneficial.

I generally agree on that -- and this is also true for AutoDiff-types. However we traditionally did convert all POD scalars to "Scalar" before using it, in order to be compatible with custom scalars which do not support operators with mixed types.

I've added the required explicit conversions: https://bitbucket.org/eigen/eigen/commits/b8b291f89653/

I guess an optimal solution could be add a templated convert method, into NumTraits<Scalar> like:

That's an interesting idea. With C++11 we could even make use of its return type to automatically promote scalar factors.

gael

template<class Scalar>
struct NumTraits {
// other NumTraits stuff ...
template<class X>
static Scalar convert(const X& x) { return Scalar(x); }
};

which, if Scalar supports operations with PODs, can be specialized to:
template<class X>
static const X& convert(const X& x) { return x; }

Unfortunately, this would require massive refactoring. Maybe we just add the Scalar() conversion for now, and consider changing this for 3.3.x or 3.4?

Christoph

--
Dipl. Inf., Dipl. Math. Christoph Hertzberg

Universität Bremen
FB 3 - Mathematik und Informatik
AG Robotik
Robert-Hooke-Straße 1
28359 Bremen, Germany

Zentrale: +49 421 178 45-6611

Besuchsadresse der Nebengeschäftsstelle:
Robert-Hooke-Straße 5
28359 Bremen, Germany

Tel.: +49 421 178 45-4021
Empfang: +49 421 178 45-6600
Fax: +49 421 178 45-4150
E-Mail: chtz@xxxxxxxxxxxxxxxxxxxxxxxx

Weitere Informationen: http://www.informatik.uni-bremen.de/robotik