Subject: [ggl] rtree query interface
From: Barend Gehrels (barend)
Date: 2011-08-28 23:57:25
On 22-8-2011 23:39, Adam Wulkiewicz wrote:
> Barend Gehrels wrote:
>> Hi Adam,
>>> I now realize that this may be confusing since everybody thinks that
>>> predicates works exactly like filters from Boost.Range but they
>>> shouldn't. They should work more like Expressions.
>> Yes, actually I did as well. What is exactly the difference?
> Well it isn't commutative and transparent like filters which just
> operates on iterators. index::filter (or a sequence of index::
> predicates) must be passed just after the tree object and it doesn't
> operate on iterators but calls some searching routine, stores elements
> inside some container (e.g. std::deque) and exposes this container as
> a range.
Is this not what we want? The filter should be passed just after the
tree object, seems OK to me, and exposes the output as a range.
I understand that after (and before) this there were proposals to pass
other range adaptoprs after, trying to combine them, which makes the
implementation (and interface) probably much more troublesome. We must
be careful to keep it simple. I agree with Luke's remark about lets not
create a DSEL, and see also the answer of feverzsj, (9-8-2011 15:57) . I
will reply on that now.
Geometry list run by mateusz at loskot.net