Re: [boost] Another GGL review

Subject: Re: [boost] Another GGL review
From: Barend Gehrels (barend_at_[hidden])
Date: 2009-11-17 10:47:06

Hi Jose,

> Design
> ---------
>
> Spatial indexes are very important. What is the plan ? (I saw the
> brief mention in your paper).
The plan is still there. We have an implementation, as you know, and 
want to revisit it a little more than we did until now. So it is 
currently not included, also because it is not yet used in the 
algorithms, so not yet essential.

As Mateusz said, we want to have an implementation based on templates; 
the SOC version did use virtual methods and we wish to avoid them.

> I also saw complaints about the bad quality of the SOC project code.
>   
People should not complain the way they did. We will take care for good 
quality.

> I found an r-tree implementation that is part of a thesis and asked
> the author if he would be willing to donate the code (if it made
> sense!)
>   
The implementation we have is r-tree. If the implementation you refer to 
is completely template based and has no virtual methods, we're certainly 
interested to compare or even include it. Our actual wish is a Priority 
R-Tree because it is worst-case optimal.

> [quoted from more recent mail] I recall long ago checking an r-tree algorithm (maybe from
> rtreeportal) that used persistent r-trees
We're currently not planning persistent trees inside the library (though 
it is interesting of course). We could do research to combine them (e.g. 
based on our concepts), so interfacing indeed, but that might more have 
the form of an external-sample than real integration.

Regards, Barend