Re: [Geotools-devel] Future if H2 module in GeoTools. Observations..

[Jody G. <jod...@gm...>]

JTS is LGPL, and is being re licensed to EPL+BSD.
On Wed, May 27, 2015 at 6:02 PM Vitali Dyachkov <vi...@ar...> wrote:

> Hello .
>
> We completed big project on uDig, GeoTools, PostGIS, H2 and
> SymmetricDS.  In context  of this project we invested into adopting H2
> 1.4 through GeoTools interfaces (H2DataStore) , basically by tuning
> gt-jdbc-h2 module , geodb and hatbox libraries.  I want to share some
> observations first.
>
> H2 is a great small database which shows excellent performance. Being
> integrated into uDig and after all optimization we did it makes
> rendering of dozens of thousands of features in multiple layers at the
> same time very fast.
>
> There are  several changes that we have done:
>
> - Usage of VARBINARY data type instead of BLOB. For many reasons BLOB
> was not very efficient and became even less efficient in latest H2.
> - Changes in SQL generation by FilterToSQL-derived classes and
> optimizations for prepared statements
> - Declaring most Hatbox and geodb functions as DETERMINISTIC which in
> certain scenarios helps to H2 to make right execution plan and avoid
> calling functions unnecessarily many times with the same parameters.
> - Optimizing hatbox for handling MultiPolygons by splitting them into
> parts, requesting spatial index separately and merging results - boost
> in performance for MPs  in spatial predicates that use spatial index of
> Hatbox.
> - some minor fixes  everywhere.
>
>
> Also H2 database nowadays  has built-in GEOMETRY type (based on WKB as a
> storage format and JTS for handling it in Java) and even spatial RTree
> based index.   http://www.h2gis.org  guys  are developing  a wrapping
> tier over H2, its built-in geometry type and JTS.  In the end JTS is a
> driving engine for everything there.  But it's  GPL and I assume can not
> be used as a part of GeoTools library being distributed as GeoTools module.
>
> I would like to  discuss the future  of H2 module in GeoTools.  I see 2
> options.
>
> 1) First option is to establish probably a separate module in GT for H2
> 1.4  and   using current infrastructure of geodb and hatbox which works
> just fine. I would contribute all our results to this module and would
> be a maintainer  e.g.
>
> 2) Implement from scratch something similar as H2GIS as , again, a new
> GT module with usage of built-in GEOMETRY type and built-in RTree index.
> This option should give even better performance because of internal
> index utilization comparing with Hatbox (which is a simulation of an
> index using normal tables).
>
>
> Regards,
> Vitaly.
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> GeoTools-Devel mailing list
> Geo...@li...
> https://lists.sourceforge.net/lists/listinfo/geotools-devel
>