Re: [eigen] Re: meeting in February?

[Gael Guennebaud <gael.guennebaud@xxxxxxxxx>]

On Mon, Jan 11, 2010 at 10:23 PM, Benoit Jacob <jacob.benoit.1@xxxxxxxxx> wrote:
> 2010/1/11 Benoit Jacob <jacob.benoit.1@xxxxxxxxx>:
>> 2010/1/11 Thomas Capricelli <orzel@xxxxxxxxxxxxxxx>:
>>>
>>> On Monday 11 January 2010 19:42:23 Benoit Jacob wrote:
>>>> The biggest purpose of ei_traits<...> is forward-declarations.
>>>> Actually, at some point, I named it ForwardDecl<...>. The idea is that
>>>> with our CRTPs, when we want to access Derived::SomeTypedef from
>>>> MatrixBase<Derived>, we can't, because Derived isn't fully declared at
>>>> that point, so we put SomeTypedef instead in ei_traits<Derived>. Does
>>>> that help? Should ei_traits be renamed?
>>>
>>> Let me drop my thoughts now.
>>> You see, all what you are talking about is rather obscure to me. I more or less know what traits are used for, but this is really blur to me. When i read such mails, i can't help but thinking that all this valuable information should be put somewhere (doxygen, wiki?).
>>
>> Yes, that's true, we need more design documentation. But a better time
>> will be just after the release, as internals are still evoluting right
>> now. For example, if we had documented how cwise() works, it would now
>> be obsolete. So for now, I'm actually better off explaining stuff by
>> e-mail if you would like to know more about something in particular!
>> But I would like very much to take the meeting also as an opportunity
>> for you to learn more about Eigen internals.
>>
>> At the very least, if you want, you could prepare a list of "things
>> you always wanted to know in Eigen internals" and we'll go over it.
>>
>> I also proposed the idea of masterclasses at the meeting but there
>> doesn't seem to be any demand so far, see the wiki
>> http://eigen.tuxfamily.org/index.php?title=Paris_2010_Meeting
>>
>>>
>>> I'm really sure i can't give any useful talk at the eigen meeting, but I'm avid of knowledge of this kind, and if i manage to understand enough, I'll try to complete some documentation.
>>>
>>> Don't get me wrong : the documentation for eigen is really cool. I could do whatever i want within 5 minutes and that's one of the points that convinced me of using eigen at the beginning (I did a quite complete study of what exist, Eigen won!). What I'm talking about here is design doc: documentation for those working on eigen internals, not for users.
>>>
>>> Anyway, I'm looking forward for the meeting, and I'm sorry I can't provide 'content' to it :/
>>
>> Actually, if we had a bit more time, it would be useful to discuss the
>> NonlinearOptimization module, but it's true that in 3 days and with
>> Eigen 3.0 to prepare, it's difficult to dedicate much time to it. If
>> others don't mind, then I think it might be useful if you gave a short
>> presentation on it, focusing on its design and API, so we can review
>> it together...? Or at the very least, you could prepare a list of
>> points that need to be discussed together...
>
> Sorry, I've been an asshole again.
>
> It's actually obvious that out of 3 days it's not too much to take 1
> hour for the NonlinearOptimization module:
>  - we could listen to a short talk
>  - we could discuss the API, the design, how it to make it best fit
> with the rest of Eigen...
>
> Rationale:
>  - we want it to eventually become a supported module, right? So we
> should be concerned about it.
>  - there's a whole class of users for whom this will be one of the
> most prominent features.
>
> Opinions?

The non linear optimization module is definitely a hot feature, and
I'd like we spend some time on its API design.

gael

> Benoit
>
>
>>
>> Benoit
>>
>>
>>>
>>> Thomas
>>> --
>>> Thomas Capricelli <orzel@xxxxxxxxxxxxxxx>
>>> http://www.freehackers.org/thomas
>>>
>>>
>>>
>>
>
>
>