Contains issue for complex LineStrings

[sgillies]

Hi All,

For:
from shapely.geometry import LineString
good1 = LineString([(0,2),(0,0),(2,2)])
good2 = LineString([(2,0),(0,2),(0,0),(2,2)])
broken = LineString([(1,0),(0,2),(0,0),(2,2)])
contained = LineString([(0,0),(2,2)])
print good1.contains(contained), good2.contains(contained), broken.contains(contained)

I get:
True True False

where I would expect:
True True True

I am running: Shapely-1.2.14-py2.6.

Am I overlooking something?

Thanks for your help,
Julia

[mwtoews]

I'm seeing odd behaviour elsewhere, where the DE-9IMs are completely different between Shapely and JTS. If broken = A, and contained = B:

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

JTS (1.12.0):
AB : 001F001F2
BA : 0F100F102

Furthermore, Within and Contains are all false for all combinations. What's really odd with the JTS result is that the interior ∩ interior result (first position) is '0', meaning only a point interior intersection. This should be a '1' for line, like GEOS/Shapely has it.

The expected intersection matrix should be:
AB : 101F00FF2
BA : 1FF00F102

[sgillies]

What's the GEOS version?

[mwtoews]

The behaviour is consistent with Shapely versions 1.2.14 to the latest, which have shapely.geos.geos_capi_version (1, 6, 2) to (1, 7, 1) (I'm not sure how to squeeze the GEOS version out of Shapely, only the C API version).

Also tested this case with POSTGIS="2.0.0 r9605" / GEOS="3.3.3-CAPI-1.7.4" with same results:

with data as (select 'LINESTRING (0 2, 0 0, 2 2)'::geometry AS good1,
    'LINESTRING (2 0, 0 2, 0 0, 2 2)'::geometry AS good2,
    'LINESTRING (1 0, 0 2, 0 0, 2 2)'::geometry AS broken,
    'LINESTRING (0 0, 2 2)'::geometry AS contained)
select ST_Contains(good1, contained),
    ST_Contains(good2, contained),
    ST_Contains(broken, contained)
from data;

returns t t f

[sgillies]

Do you mean PostGIS has the same results as JTS, or same results as Shapely? If the latter, we should forward this on to the GEOS tracker (I'm subscribed to that too).

[mwtoews]

All tested versions of JTS, PostGIS and Shapely have the same results for "Contains". I recall Shapely and PostGIS have the same DE-9IM, thus this suggests issues with GEOS.

This issue has been passed to GEOS http://trac.osgeo.org/geos/ticket/572

[mwtoews]

FYI, this issue has been filed further upstream with JTS: locationtech/jts#396

[dr-jts]

This is now fixed by libgeos/geos#937