Buffer creates artefact with flat endcap [JTS fails too]

[barendgehrels]

Hi,

A buffer of a quite simple linestring, using endcap=flat, creates two artefacts at the top.

The PostGIS SQL command to reproduce this is:

select ST_AsText(ST_Buffer(ST_GeomFromText('LINESTRING(2 7,3 5,5 4,7 5,8 7)'), 5, 'endcap=flat'));

I've not tested this particular case in GEOS, but have seen similar artefacts originating from GEOS, so suspect GEOS will generate this.

How a buffer with endcap=flat would ideally look like, I'm not sure of, because as soon as a buffer is not having round endcaps and joins, it is not mathematically defined, AFAIK, so the behaviour is more or less up to the implementation. In this case the buffer distance is larger as the vertex distances, so this is probably the cause of this artefact. Anyway, I'm calling it artefact because it's looking strange.

I'm implementing a buffer as well (Boost.Geometry), and especially currently need an asymmetric buffer distance (left/right different) AND end-cap, and the results are different (but also not yet satisfactory).

Regards, Barend

[strk]

This would be best discussed with Martin Davis, as JTS is what drives GEOS.

[robe]

I'm going to assume JTS still fails since I don't have JTS to test with. Push back if this is an easy fix.

[robe]

Still an issue

[strk]

Ticket retargeted after milestone closed

[mdavis]

As the OP observes, the "expected" value of a buffer with flat endcaps and a large buffer distance is not really well-defined. In the absence of a definition, it's not possible to determine how to improve the current result.

[mdavis]

What do other computational geometry libraries do in this case?