CascadedPolygonUnion performance

[Marco]

Hi

In my current project I have to merge (union) a few thousands
polygons. As this is not a trivial task the process needs usually
around 2 minutes for the whole union. Inquiries to speed up the
processing have led me to Paul Ramseys blog:
http://blog.cleverelephant.ca/2009/01/must-faster-unions-in-postgis-14.html

He is describing there a fast way to union polygons with the help of
an STR-Tree. Is the current implementation of CascadedPolygonUnion in
NTS 1.x and 2.x based on a similar algorithm or is there a chance to
improve this function?

Im currently using NTS 1.x. Please let me know if there would be
already a performance increase in CascadedPolygonUnion of NTS 2.x

Thanks a lot,
Marco

[FObermaier]

That is the implementation. IIRC there is an option to use multiple
threads, you may want to try that.
I personally doubt that the v2 implementation is any faster than the
v1 implementation, since it works on interfaces only which prevents
inlining.

Hth FObermaier

On Nov 27, 2:53 pm, Marco <ruf...@gmail.com> wrote:
> Hi
>
> In my current project I have to merge (union) a few thousands
> polygons. As this is not a trivial task the process needs usually
> around 2 minutes for the whole union. Inquiries to speed up the

> processing have led me to Paul Ramseys blog:http://blog.cleverelephant.ca/2009/01/must-faster-unions-in-postgis-1...

[Diego Guidi]

> That is the implementation. IIRC there is an option to use multiple
> threads, you may want to try that.

here the issue list:
http://code.google.com/p/nettopologysuite/issues/detail?id=44
maybe can be useful to retrieve other info

[Marco Birchler]

Thanks for your response.

FObermaier and Diego, you wrote that there is an OPTION to use
multiple threads. Is the code not using multiple threads by default? I
did a review of your code but could not figure it out. Can you please
tell me how to enable multithreading if it isn't already?

2011/11/28 Diego Guidi <diego...@gmail.com>

> --
> You received this message because you are subscribed to the Google Groups "NetTopologySuite" group.
> To post to this group, send email to nettopol...@googlegroups.com.
> To unsubscribe from this group, send email to nettopologysui...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/nettopologysuite?hl=en.
>

[Diego Guidi]

> how to enable multithreading if it isn't already?

talking about the trunk (NTS 1.X), the code isn't multithreading "as
is": it's up to you to split your code in "pieces" to call
asynchronously, actually, not an easy task to do.
Beware: NTS "should be" thread-safe, but as example a bug was found
yesterday: http://code.google.com/p/nettopologysuite/issues/detail?id=102

[FObermaier]

You are right Marco, the algorithm creates worker threads for the
actual unioning of the geometries.
If you feel the algorithm is too slow, please raise an issue and
provide your sample geometries, so we can profile,
FObermaier

On Nov 30, 2:55 pm, Marco Birchler <ruf...@gmail.com> wrote:
> Thanks for your response.
>
> FObermaier and Diego, you wrote that there is an OPTION to use
> multiple threads. Is the code not using multiple threads by default? I
> did a review of your code but could not figure it out. Can you please
> tell me how to enable multithreading if it isn't already?
>

> 2011/11/28 Diego Guidi <diegogu...@gmail.com>

[Diego Guidi]

> talking about the trunk (NTS 1.X), the code isn't multithreading "as
> is": it's up to you to split your code in "pieces" to call
> asynchronously, actually, not an easy task to do.

Of course, CascadedPolygonUnion does actually this thing: split the
code in "pieces" that are called in different threads. Maybe it's time
to know better the code that I suppose to already know.
Sorry fot the mistake, forget my answer.