Subject: [ggl] spacial index
From: Bruno Lalande (bruno.lalande)
Date: 2011-03-27 18:12:13
I have read more now, here are some comments.
New design: Very good overall, I think the migration to Boost.Variant is a
success. Thanks Barend and Luke for pushing in that direction. The only
thing I would like you to do is: confirm the performance difference with an
even bigger tree (would like to see times greater than 1 second), if
possible. And also analyse real memory consumption (objects size don't tell
me enough). This can simply be done by checking the total memory occupied by
a program with a fully loaded huge tree...
I can see that you're using the rtree_ prefix everywhere. This should
probably better be a namespace, just under "index". So
One point I wanted to raise before but I thought there was enough
discussions to address already. It might surprise you, but I think you
shouldn't propose the user to choose which box type your tree should use
internally. I.e. the Box template argument shouldn't exist at all. This is
an implementation detail, which has only to do with the guts of your tree.
It should be up to you to decide what's the optimal choice, and it doesn't
seem to me that the optimal choice depends on any user-specific context. And
it doesn't interact at all with the types used by the user. So what's the
real justification behind that?
I'll probably add a comment later about the Tag structure, I have to read
more to be sure...
-------------- next part --------------
An HTML attachment was scrubbed...
Geometry list run by mateusz at loskot.net