Inefficient retrival of elements within an area about librecad_3

[feragon]

That's right, this test fails:

TEST(Test, Test) {
    lc::QuadTree<lc::entity::CADEntity_CSPtr> qt;
    qt.insert(std::make_shared<lc::entity::Line>(lc::geo::Coordinate(-5000, -5000),
                                                 lc::geo::Coordinate(5000, 5000),
                                                 std::make_shared<lc::Layer>())
    );

    ASSERT_EQ(0, qt.retrieve(lc::geo::Area(lc::geo::Coordinate(1, -1), lc::geo::Coordinate(5000, -5000))).size()); //Returns 1 element
}

[rvt]

In context of quad tree this is correct. The quad tree doesn't care about exact coordinates but it just finds the area's that include (even partially) in the requested area.
Since the bounding box of the lines falls over the boundig box of the requested area the line is included.

For painting it matters such that element's really outside of a quad will not be retreived, but even taht depends.

For example if you have 10 elements in a quad tree, no matter what you do, you will always retrieve them all. If you have 1000 they will be pushed down the quad tree levels and store them at finer granularity.

This could even optmise painting during canvas moving. Because you could easely request all entities up to some level X and you will only get a maximum number of entities but nothing below that (if it applies).

For the painter is doesn't matter much in some conditions. We need to be carefull not to optimise to much because sometimes it can be faster to just paint it rather than to find out if we need to paint it. The quad tree just helps with it... We do have some more exact routines that can find all entities within a bounding box, thet just take a lot more time to calculate.

For OpenGL it could even slow things down because with OpenGL you can keep everything in memory (I hope :) ) and as such it's properly faster to let the GPU handle it.
In addition, both the painter and kernel can use there own implementation to speed up operations if they wish...

Hope that makes sense...