Re: psycopg2.errors.InternalError_: GEOSTopologyPreserveSimplify: IllegalArgumentException: CGAlgorithmsDD::orientationIndex encountered NaN/Inf numbers

This entry from the 3.12.3 changelog seems pertinent 

  - DouglasPeuckerLineSimplifier, avoid crash with Point input and NaN tolerance (GH-1078, Dan Baston)

Running the latest patch releases is always the best practice.

P.

> On Jul 29, 2025, at 12:25 PM, Bharath Tadikonda <[email protected]> wrote:
> 
> Team, 
> 
> Dear PostGIS Team,
> We’re encountering the following exception during execution of a spatial operation in our Python application using PostGIS:
> 
> 
> 
> psycopg2.errors.InternalError_: GEOSTopologyPreserveSimplify: IllegalArgumentException: CGAlgorithmsDD::orientationIndex encountered NaN/Inf numbers
> This error arises from a call to ST_TopologyPreserveSimplify() in our function that processes and simplifies geometry.
> 
> After some investigation, our development team suspects the issue is related to the PostGIS and GEOS versions installed in different environments:
> 
> 🔴 Validation Environment — Not Working
> 
> 
> SELECT postgis_full_version();
> -- POSTGIS="3.4.0 0874ea3" [EXTENSION] PGSQL="160" GEOS="3.12.2-CAPI-1.18.2" ...
> ✅ Production (AWS RDS) — Working Fine
> 
> 
> SELECT postgis_full_version();
> -- POSTGIS="3.4.3 e365945" [EXTENSION] PGSQL="160" GEOS="3.13.0-CAPI-1.19.0" ...
> It appears that the error occurs only on PostGIS 3.4.0 with GEOS 3.12.2, while the same code works as expected in environments with PostGIS 3.4.3 and GEOS 3.13.0.
> 
> Currently, we are unable to find and install GEOS 3.13.0 on-prem and only see GEOS 3.13.1 as available.
> We’d appreciate your guidance on the following:
> 
> Has this issue (NaN/Inf handling in ST_TopologyPreserveSimplify) been addressed between PostGIS 3.4.0 and 3.4.3 or GEOS 3.12 → 3.13?
> 
> Would upgrading to GEOS 3.13.1 with PostGIS 3.4.3 on-prem likely resolve this issue?
> 
> What is the best practice to keep PostGIS + GEOS versions in sync between on-prem and cloud (AWS RDS) environments?
> 
> Thank you for your support and any insights you can share.
> 
> Best regards,
> Bharath
> 
>