[Dune] [Fwd: Re: [Dune-Commit] dune-istl r929 - trunk/istl]

[Carsten Graeser]

Although I don't recognize why non-existence of a feature is better
than a slow implementation of it I would suggest to remove the
cast to avoid repeating the same arguments again and again.

Probably we're the only ones using this class, but if we really need
the cast we can keep in in our working copies.

But before removing it I'd like to test if it really is that slow.
I will do this after next week.

Best regards
Carsten

Markus Blatt schrieb:
>>> But this cast might have side effects which surprise people and are
>>> really hard to find.
>>>
>>> If FieldMatrix and ScaledIDMatrix are so much correlated. Then
>>> ScaledIDMAtrix should be moved to common and assignment and explicit
>>> type conversion constructors should be implemented in FieldMatrix.
>>>   
>> I do not think moving this class to dune-common is a good idea.
>> It is very special will not be used frequently.
>>
> 
> And I dont like the side effects of implicit type conversion. IMHO
> moving is the best to keep the code clear.
> 
> Another possiblity (with other side effect, I admit) is providing a
> constructor and assignment operator in FieldMatrix that has a template
> parameter for the matrix.

-- 
----------------------------------------------------------------------
Carsten Gräser           | phone: +49-30 / 838-75349
Freie Universität Berlin | fax  : +49-30 / 838-54977
Institut für Mathematik  | email: graeser at math.fu-berlin.de
Arnimallee 6             |
14195 Berlin, Germany    | URL  : http://page.mi.fu-berlin.de/graeser
----------------------------------------------------------------------