Contains issue for complex LineStrings [JTS fails too]

[Mike Taves]

This issues was first identified by Julia via ​Shapely #17.

With non-simple A and simple B:

A: LINESTRING (1 0, 0 2, 0 0, 2 2)
B: LINESTRING (0 0, 2 2)

Compare DE-9IM results. First with Shapely (1.2.14, with GEOS C API 1.7.1; also verified with PostGIS 2.0 with 3.3.3-CAPI-1.7.4):

A.relate(B) : 101F001F2
B.relate(A) : 1F100F102

JTS (1.12.0):

AB : 001F001F2
BA : 0F100F102

The expected intersection matrix should be (based on LINESTRING (2 0, 0 2, 0 0, 2 2)):

AB : 101F00FF2
BA : 1FF00F102

Related issues are apparent with "Contains" (T*****FF*), which return False for AB and BA combinations for both JTS and GEOS. This result is consistent with the DE-9IM relate patterns. The expected result for AB is True (and False for BA).

[robe]

Still an issue

[strk]

Ticket retargeted after milestone closed

[mdavis]

See also ​https://github.com/locationtech/jts/issues/270.

This branch may have a fix: ​https://github.com/locationtech/jts/tree/bug-test-intersects

Needs some solid testing though! (Although all JTS unit tests pass, FWIW).

[mdavis]

Upstream ticket is ​https://github.com/locationtech/jts/issues/396

Still no solution for this.

[mdavis]

Fixed by ​https://github.com/libgeos/geos/pull/937