[Dune-devel] StoreGeometryReference

[Martin Nolte]

Hi Oli,

On 02/14/2014 03:46 PM, Oliver Sander wrote:
> Am 14.02.2014 15:32, schrieb Martin Nolte:
>> Hi Oli,
>
> Hi Martin,
> excuse the harsh tone of my previous mail.
>
>>
>> well, it is, of course, possible that something got messed up with my copy of the master branch. I'll investigate on that later. The lack of traffic on the mailing list indicates that I'm the only one
>> with problems, so let's ignore it for now.
>
> Well, even if you are the only one who is experiencing problems, it is still an oversight
> on my part.
>
I think we should lay low here until someone else confirms my problems. Simply 
configuring without Alberta solves the problem for now.

>>
>> Now the file has been removed from the master branch, I don't think it makes much sense to update the release notes accordingly.
>>
>
> Unless you insist that the file should be put back in.  I don't think it should, but others may well disagree with me.
>
As indicated in my previous post: I do not insist on the file being put back 
into dune/grid/common. However, that means copies might reappear in different 
places (e.g., dune/grid/albertagrid).

>> As to the question whether this should be an official feature of dune-grid: I don't care either way. Personally, I'm perfectly fine with duplicating the code in each grid requiring it.
>
> I think it is too complicated to be available as a general feature.
> I suppose you'll want to disagree with me on this :-)
>
Complicated is not the word I would have used. GeometryReference is a nearly 
exact copy of Dune::Geometry, storing a pointer to the implementation instead 
of the implementation itself. The other two helper classes just adapt it the 
fixed template argument list of Dune::Geometry.

Personally, I find it hard to make "the right decision". Since you seem to 
have a strong opinion on this, let us keep things as they are until further 
discussion is requested.

Best,

Martin

> Cheers,
> Oliver
>
>>
>> Best,
>>
>> Martin
>>
>>
>> On 02/14/2014 02:17 PM, Oliver Sander wrote:
>>> Hi Martin,
>>>
>>>>
>>>> yes, removing GeometryReference break AlbertaGrid (and SPGrid, but who cares).
>>>>
>>>
>>> not me, to be honest.  It is not in dune-grid, in fact I can't even find it mentioned
>>> anywhere on our homepage.  Certainly not on the list of external modules
>>> http://www.dune-project.org/downloadext.html
>>>
>>>> The GeometryReference class was explicitly introduced (by me) to allow returning references to geometries actually stored on the grid. For example, there are only two local geometries for
>>>> geometryInFather in AlbertaGrid. Why bother to recompute them every time the user asks for them? Notice that this does not violate the new lifetime guarantees. The "reference" is valid as long as the
>>>> grid exists.
>>>>
>>>> Actually, the idea of the facade options was to avoid this code duplication in the first place. Since my fellow developers voted for deprecation of the reference version, I needed a replacement. But I
>>>> will happily duplicate this piece of code into every single grid that wants to return a reference to its local geometries. Originally it was an implementation detail of SPGrid anyway.
>>>>
>>>
>>> Is that documented anywhere?  If it is an official new feature in dune-grid, why doesn't
>>> it appear in any release notes?  I'm sorry but I can't read your mind.  All I remember
>>> is that it was requested that geometries be returned by value now, and that that StoreReference
>>> shim was introduced to ease the transition.
>>>
>>>> As a final comment: Since when do we ask such questions merely three hours before removing the code?
>>>
>>> I actually did ask the same thing a while ago (can't find it now, though), with no reaction at all.
>>>
>>> Cheers,
>>> Oliver
>>>
>>>>
>>>> Best,
>>>>
>>>> Martin
>>>>
>>>>
>>>>
>>>> On 02/14/2014 09:09 AM, Oliver Sander wrote:
>>>>> Dear Dune team,
>>>>> IIRC the GeometryReference class and the corresponding traits class StoreGeometryReference
>>>>> were supposed to exist only for a transitional period.  Are there objections to their
>>>>> removal now that 2.3 has been released?
>>>>> Cheers,
>>>>> Oliver
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Dune-devel mailing list
>>>>> Dune-devel at dune-project.org
>>>>> http://lists.dune-project.org/mailman/listinfo/dune-devel
>>>>>
>>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Dune-devel mailing list
>>> Dune-devel at dune-project.org
>>> http://lists.dune-project.org/mailman/listinfo/dune-devel
>>>
>>
>
>
>
>
>
>
>
>
> _______________________________________________
> Dune-devel mailing list
> Dune-devel at dune-project.org
> http://lists.dune-project.org/mailman/listinfo/dune-devel
>

-- 
Dr. Martin Nolte <nolte at mathematik.uni-freiburg.de>

Universität Freiburg                                   phone: +49-761-203-5630
Abteilung für angewandte Mathematik                    fax:   +49-761-203-5632
Hermann-Herder-Straße 10
79104 Freiburg, Germany