add DE-9IM predicates and test

[rafaqz]

@evetion these might help with JuliaGeo/GeoInterface.jl#106

There are a few more to add, we have intersects, within, contains, union and intersection so far.

In GeoInterface we can just do things like

within(geom, ext::Extent) = Extents.within(GI.extent(geom), ext)

[codecov-commenter]

Codecov Report

Merging #13 (a743a1b) into main (b76338a) will increase coverage by 3.36%.
The diff coverage is 87.87%.

@@            Coverage Diff             @@
##             main      #13      +/-   ##
==========================================
+ Coverage   84.40%   87.76%   +3.36%     
==========================================
  Files           1        1              
  Lines         109      139      +30     
==========================================
+ Hits           92      122      +30     
  Misses         17       17              

Files	Coverage Δ
src/Extents.jl	87.76% <87.87%> (+3.36%)	⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[evetion]

Sorry to take this long in reviewing this, these things are hard to quickly review (on mobile), it requires a bit of careful thought.

Nice to expand the spatial predicates. You could add disjoint quite easily, as it's the inverse of intersects.

Any plans on adding equals (should be Base.isequal, except for strict ofcourse), and or others (touches is mina==maxb or maxa==minb), and there is overlaps (intersects with > instead of >=, but not equal)?

Some comments.

[evetion]

Still commented out code here to remove that should be included? Doesn't seem to be complete yet.

Might be good to have a comment like
# abab or baba (which also matches abba and baab, but not aabb or bbaa)
and/or even destructure the tuples to (mina, maxa), (minb, maxb) for readability.

Note that this is sometimes written as the inverse (maxa < minb) || (mina > maxb)

[evetion]

I love the extensive tests ❤️ . However, we might need to add even more, seeing the missed cases for intersects, and also to test the equal in the <=?

It might be useful to comment specific cases where you test something very specific (like an equality edgecase)?

[rafaqz]

Yes to disjoint and overlaps, and I guess equals.

Also to very clear tests and docs. This package should be bomb-proof.

But not sure when I will be able to lift it to that standard.

[rafaqz]

@evetion I went all out ISO and implemented DE-9IM

Seems to work well but some feedback would be good. I'm not sure about the strict keyword and generally what to do as the default when there is an extra dimension in one of the extents (like there is a time or Z dimension in a spatial raster as well as X/Y - what do when checking if it intersects with a regular X/Y raster.

They all seem to take a few nanosecond. I'm thinking we can use them in fast loops over many geometries, matching some predicate like everything that might overlap.

[evetion]

Nice work! It's a pain to get all these right. 👍