Hi there, 
Sorry for the delayed response but this went to my spam.  Are you planning on 
open sourcing this UDF? I’d be happy to do a code review and give you some 
suggestions.   Could you explain more about the use case however as I am 
totally unfamiliar with this. 
Thanks!
— Charles

> On Oct 4, 2026, at 14:45, richard van der koppen <[email protected]> 
> wrote:
> 
> Hi Drill Developers/experts,
> 
> I hope you are all doing well.
> 
> I have been working on a geospatial architecture use case for a while now,
> researching the best way to handle large-scale Dutch coordinate
> transformations directly inside Apache Drill. Next week, I am planning to
> test my first working prototype / MVP for a custom Java Scalar UDF
> (rd_to_wgs84).
> The goal of this UDF is to perform high-performance, on-the-fly
> transformations from the Dutch Rijksdriehoekscoördinaten (RD New /
> Amersfoort, EPSG:28992) to global WGS84 coordinates (Latitude/Longitude,
> EPSG:4326).
> 
> To maximize performance and avoid external GIS library overhead during
> runtime code generation, the MVP will implement the official
> Bessel-to-WGS84 polynomial regression formulas directly into the eval()
> method using primitive doubles, Netty DrillBuf writes, and a VarCharHolder
> output.
> 
> Since Drill UDFs operate on a very specific subset of Java and rely heavily
> on internal scalar replacement, I want to make sure my implementation
> strictly adheres to Drill's compilation mechanics and memory management
> rules.
> 
> Once the MVP and the Maven project are up and running next week, would
> anyone with experience in Drill's query engine internals and code
> generation be open to reviewing the Java code? I am planning to publish the
> final project as an open-source repository on GitHub to help the geospatial
> data engineering community.
> 
> Any guidance or interest in doing a quick code review would be highly
> appreciated! Feel free to send me a private message at this email address,
> and we can discuss things privately between ourselves.
> 
> Best regards,
> 
> M.D

Reply via email to