geometric object crashes server with Out Of Memory

[cousinm]

The attached object crashes the PostgreSQL server after allocating more than 20 Gigabytes of ram (sometimes bringing the server with it), when using any GIS function even st_isvalid().

I don't know if it is valid, it was provided «as is» by a user, and I can't run st_isvalid() on it.

SELECT postgis_full_version();

postgis_full_version

---

POSTGIS="1.5.2" GEOS="3.2.2-CAPI-1.6.2" PROJ="Rel. 4.7.1, 23 September 2009" LIBXML="2.7.8" USE_STATS

(1 row)

The attached script creates a «test» table, and puts this object inside it.

Please tell me if I can provide anything to help on this problem.

[cousinm]

script creating test data

[robe]

I suspect you are experiencing the GEOS bug described in #1007. The fix is to upgrade to GEOS 3.3.0 or GEOS 3.2.3.

Please give that a try and let us know how that works.

[robe]

It would also help if you printed out the

SELECT version();

[cousinm]

SELECT version();

version

---

PostgreSQL 9.0.4 on x86_64-unknown-linux-gnu, compiled by GCC gcc (GCC) 4.6.0 20110415 (prerelease), 64-bit

(1 row)

I'm going to try with a newer GEOS.

[robe]

Sorry I forgot 3.3.0 is already released so the fixed GEOS are 3.2.3SVN and 3.3.1SVN

[cousinm]

I have the same problem using 3.3.1SVN.

[pramsey]

Confirmed, same behaviour on GEOS 3.2.3 under OS/X. Other pure-PostGIS functions work, but building the edge graph in GEOS leads to memory explosion.

[pramsey]

From Martin:

This geometry causes problems in JTS as well (all versions). The problem is that there is an area of the geometry where there are about 2000 circular polygons (totalling about 71000 pts) which are all *almost* identical. This causes the noding algorithm to have a hissy fit.

It should be possible to get validation working at least, since the geometry is obviously invalid (due to the numerous overlaps).

Also, Unary Union works fine, with respectable performance. So if it's necessary to reduce the polygon to something workable, that's the way to do it.