[QGIS-Developer] Heads up on GEOS Overlay New Generation

[Régis Haubourg]

Hi

Le jeu. 17 sept. 2020 à 11:31, Andreas Neumann <a.neumann at carto.net> a
écrit :

> Hi Régis,
>
> Indeed interesting.
>
> Is there evidence that it will also be faster than previous versions? If
> yes - for certain operations or generally faster?
>
Well not yet, and this will depend a lot on the type of features, vertex
density and so on. I we trust some unit tests, and compare with ESRI on big
datasets (France 4G antennas intersect with municipalities), we could
expect something between twice and four times faster. That's a hope a that
point not a measured fact.

> We had issues in the past with the "Geometry checks" that you could define
> on a layer (Layer Properties --> Digitizing settings) - tests about
> overlays and gaps. It was always slow and not so reliable. Once you tried
> to fix one error you would get even more errors through the fix and then
> you would end up with even more errors when trying to fix the error ...
> Hopefully, this will work better then ...
>
I think this should help indeep.
Bye

> Greetings,
>
> Andreas
>
> On 2020-09-17 09:33, Régis Haubourg wrote:
>
> Hi there,
> I'm forwarding a call [0] from Paul Ramsey and the JTS-GEOS teams.
> I've been following this closely for 3 years now and we have been
> providing testing and test datasets to try to help here.
>
>
>
> In short, GEOS provides most of the geometry computations tools to PostGIS
> QGIS & al. The intersection, difference etc.. (overlay operations) suffered
> from using full numeric precision and lacked snapping and rounding
> operations.
>
> Thus our libraries were slow, and prone to topology errors due to those
> numeric precision issues. Most of us have seen intersections or union
> operations failing from time to time.
>
> The new overlay operation implements snap rounding operations (which is
> implemented in ESRI tools for instance) and will provide fast and robust
> operations.
>
> Now it is available in the 3.9 branch of GEOS.
>
> As QGIS uses GEOS extensively in editing, displaying and algorithms, we
> have to take a look seriously at the positive impacts and potential
> regressions.
>
> Paul warns that they spotted no regression in Z handling, which probably
> means the test coverage is very low in GEOS CI, and we might expect
> regressions on the Z handling here, especially for snapping I bet.
>
> Anyhow this is GREAT news for the OSGEO community because this was one
> major glass ceiling for us. We sometimes had no choice than switching to
> GRASS topological operation to do some very basic operations. Hopefully
> these times are gone :)
>
> And combined with subdividing features, we should gain a massive order of
> magnitude in speed for overlay.
>
> Time to test this big win!
>
> Best regards
>
> Régis
>
> [0]
> https://lists.osgeo.org/pipermail/geos-devel/2020-September/009670.html
>
> _______________________________________________
> QGIS-Developer mailing list
> QGIS-Developer at lists.osgeo.org
> List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20200917/48d48cf3/attachment.html>