8

Im currently trying to convert my scripts from ArcGIS to QGIS but one aspect that is extremely slow is trying to intersect a polygon with a line feature. I feel as though there must be a better way to accomplish trying to clip the voronoi polylines with the original polygon itself i.e. :

Voronoi_Lines

by

Original Polygon

These are some of the bench tests that i've done with these two layers.

ArcGIS 10.1 Intersect 2min 24sec

Grass 7 v.Overlay quit after ~1hour (~70% Selecting lines...) - Currently not supported in processing (Sextante)

Grass 7 v.split + v.Overlay ~2min + 45min

QGIS 1.9 quit after ~3hours

ogr2ogr -clipsrc quit after > 1hour

SAGA GIS Intersect Line-Polygon ~15min -- acceptable but still 7x slower than ArcGIS and produces wrong results? The images below show SAGA GIS Intersect(with dangles removed) and same area with ArcGIS Intersect - Clearly SAGA did not intersect the line and polygon correctly as there should be no connection of the lines. These errors are scattered throughout the new output!!! In addition Line-Polygon does not retain the polygon intersect attributes to the line unlike the other methods mentioned above...

SAGA GIS Intersect

I have placed the original dataset in a GitHub repository here so hopefully others can make their own benchmarks and thereby provide suggestions?

I notice that ArcGIS Desktop splits (cracking) the featureclass so I do not know if that is another possibility to improve intersection speed within one of the opensource alternatives mentioned above?

***Note that the reason I mention QGIS here is that I have made some scripts within the processing toolbox (formally known as Sextante) and i'd like to remain in that environment if at all possible.

System - Intel Core i7-2760QM CPU @2.40GHz and 8GB ram

  • you should try a benchmark using OGR directly: darrencope.com/2011/03/31/clipping-large-shapefiles-using-ogr - you can use that in the Processing toolbox as well. – Darren Cope Sep 16 '13 at 19:07
  • 1
    How about the SAGA function from the toolbox? They seem to be very fast and reliable. The default intersect and other default tools in QGIS are a real pain in the a.. – Bernd V. Sep 16 '13 at 19:32
  • 3
    @BJEBN: Can you put the data set online so that we can benchmark ourselves? – markusN Sep 16 '13 at 20:16
  • @Darren/@Bernd/@markusN see updated post – BJEBN Sep 18 '13 at 8:33
  • 1
    @BJEBN; unfortunately I don't have time right now to try a benchmark, but another thought may be to try the 64bit version of QGIS available as an experimental build from trac.osgeo.org/osgeo4w (assuming you're on a 64bit machine) and see if the processing time is any faster. Otherwise, you could add the GRASS v.split and v.overlay to the Processing toolbox - not that hard to do following the instructions here: fossies.org/linux/misc/qgis-2.0.1.tar.gz:a/qgis-2.0.1/python/… – Darren Cope Sep 18 '13 at 12:05
1

What I did to speed up this process was to load SHP files in PostgreSQL and do clipping there. Not a perfect solution but faster for me since I quite powerful server behind it. And yes, this operations are slow.

| improve this answer | |

Not the answer you're looking for? Browse other questions tagged or ask your own question.