Problem with ST_Intersection

[ezimanyi]

test=# select st_intersects(geometry 'Linestring(1 1,1 1)', geometry 'Linestring(1 1,3 3)');
 st_intersects
---------------
 t
(1 row)

test=# select st_astext(st_intersection(geometry 'Linestring(1 1,1 1)', geometry 'Linestring(1 1,3 3)'));
    st_astext
------------------
 LINESTRING EMPTY
(1 row)

[robe]

I have replicated the issue. I'm guessing the issue is because linestring 1 is a degenerate linestring. I'm honestly not sure what the answer should be to the intersection question.

[ezimanyi]

The problem is in the GEOS function GEOSIntersection

[mdavis]

This is a design choice made in the JTS/GEOS intersection algorithm, about how to handle invalid degenerate lines.

The overlay algorithm is designed to never provide invalid output. Rightly or wrongly, zero-length lines are considered to be invalid. But perhaps it would be better to treat them as valid (or mostly valid - i.e. only if they occur in the input.)

I'm not sure how hard it would be to provide this semantic.

[robe]

Replying to mdavis:

This is a design choice made in the JTS/GEOS intersection algorithm, about how to handle invalid degenerate lines.

The overlay algorithm is designed to never provide invalid output. Rightly or wrongly, zero-length lines are considered to be invalid. But perhaps it would be better to treat them as valid (or mostly valid - i.e. only if they occur in the input.)

I'm not sure how hard it would be to provide this semantic.

I would say probably not worth changing.

@ezimanyi is this issue inconveniencing you a lot?