Support covered_by() for box, geometry combinations #1046
Conversation
| box_type box_areal; | ||
| geometry::envelope(geometry, box_areal, strategy); | ||
| return strategy.covered_by(box_areal, box).apply(box_areal, box); |
There was a problem hiding this comment.
This approach while giving the correct result is suboptimal because it cannot be interrupted if e.g. one of the points/segments of a geometry is outside the box. The whole envelope is calculated each time. If that's acceptable then ok. But if it isn't but for whatever reason you decide to not implement this now maybe you could put TODO here and create another issue about this.
| using is_cartesian = std::is_same | ||
| < | ||
| typename traits::coordinate_system<typename traits::point_type<Box>::type>::type, | ||
| cs::cartesian | ||
| >; |
There was a problem hiding this comment.
You should probably use Strategy::cs_tag to distinguish between CSes. It's because the CS of a geometry may be undefined and it's the Strategy that defines CS-related operations.
| BOOST_GEOMETRY_STATIC_ASSERT_FALSE( | ||
| "Not implemented for this coordinate system.", | ||
| typename traits::coordinate_system<typename traits::point_type<Box>::type>::type); |
There was a problem hiding this comment.
Is this a breaking change? AFAIU previously box was silently converted to ring in non-cartesian CSes, correct?
There was a problem hiding this comment.
I think we are OK since this operation was not supported in the past, i.e. covered_by(box, polygon) do not compile in any cs before this PR.
|
@awulkiew thanks for your comments I addressed them by pushing new commits. |
|
@barendgehrels are you OK with this PR? |
| return strategy.covered_by(box1, box2).apply(box1, box2); | ||
| box_type box_areal; | ||
| geometry::envelope(geometry, box_areal, strategy); | ||
| return strategy.covered_by(box_areal, box).apply(box_areal, box); |
There was a problem hiding this comment.
Is this sufficient for the generic case? It looks surprisingly simple.
But probably yes.
There was a problem hiding this comment.
Fox box/box , calculating the envelope is actually not necessary.
For box/point neither.
Maybe the replaced code should still stay, and this new case to support other combinations.
You covered that already below, fine.
| } | ||
| }; | ||
| using point_type = typename point_type<Geometry>::type; | ||
| using box_type = model::box<point_type>; |
There was a problem hiding this comment.
We probably should use a helper geometry here, in case point_type is not mutable
See my review remark, basically yes, but can you look at my comments I made earlier? |
Thanks for the review! Somehow your comments appeared to me yesterday. I addressed them and now I can merge this PR. |
This PR implements support for
covered_by()for the following combinations of a box and a geometry:convered_by(box, areal)(only cartesian CS)convered_by(geometry, box)The first case resolves issue #604