Standardize Validation #228
Conversation
- Add builds for libgeos 3.4, 3.5, 3.6, 3.7, 3.8 (via ubuntu versions) - Add osx build (with latest geos installed via homebrew) - Add ruby 2.7 - Use jruby 9.2.13.0
Broaden travis ci build matrix
…int error on buffer_with_style test
…or. Removed self-intersecting is_simple? test on CAPIPolygons because this will be version dependent.
|
After trying again, I was able to get tests to fail locally by switching libgeos versions. The issue is that in older versions of libgeos (<3.8), only The simplicity check that the This will be a breaking change since users with older versions of libgeos will now have more strict validation requirements on projected factories. |
|
The reason for overriding The only issue with this could be if there's a case where a |
| # having issues with floating point errors on some systems | ||
| # 4.3 -> 4.29999999999999, for example, and throws an error | ||
| # iterating through points and using assert_in_delta instead | ||
| # of assert_equal | ||
| b_coords = buffered_line_string.exterior_ring.coordinates | ||
| p_coords = polygon.exterior_ring.coordinates | ||
| p_coords.zip(b_coords).each do |p_coord, b_coord| | ||
| p_coord.zip(b_coord).each do |pt1, pt2| | ||
| assert_in_delta(pt1, pt2, 1e-7) | ||
| end | ||
| end |
There was a problem hiding this comment.
This one is not satisfying, but this PR is not the place to fix floating point errors!
test/simple_mercator/polygon_test.rb
Outdated
| @@ -15,6 +15,14 @@ def setup | |||
| @factory = RGeo::Geographic.simple_mercator_factory | |||
| end | |||
|
|
|||
| def test_is_simple_validation_behavior | |||
| # issue 218 | |||
There was a problem hiding this comment.
I always find it easier to show an url in that case:
| # issue 218 | |
| # See https://github.com/rgeo/rgeo/issues/218 |
Summary
PR to fix #218 where
is_simple?returns inconsistent results across libgeos versions.Other Information
I personally tried using geos versions 3.5.2, 3.6.4, 3.7.3, and 3.8.1 across multiple Rubies and did not have a failing test case. I'm using Ubuntu Bionic so that may have something to do with it, but this is inconsistent with what is described in #218 .
I also looked at the libgeos docs and there is a note in there that suggests
GeosIsSimplewill always returntruefor polygons and thatGeosIsValidwould be the appropriate function to check for self-intersections. This is also described here https://postgis.net/docs/using_postgis_dbmanagement.html#OGC_Validity where it says "By definition, a POLYGON is always simple."