[Liblas-devel] problems with lasblock and big datasets

[Howard Butler]

On Sep 10, 2010, at 12:05 PM, Hugo Ledoux wrote:

> thanks everyone for your help. I use a 32bit machine and my swap is the "standard" 2X the memory. I could indeed increase it but then it would only be good until I try a bigger dataset...
> 
> I do not need a temporary workaround, I was just testing the capabilities of lasblock and wanted to understand what you've done, and ran into that problem.
> 
> Splitting with the scan order is indeed likely to fail, but I'm curious about lasindex. What index are you planning to use? And when do you reckon it'll be ready?

libLAS has recently gained a generic octree implementation [1] that can be used for windowed queries.  With some work it could be adapted to give you "blocks" that you could then use to inform splitting the files.  Unlike the chipper in lasblock, the octree doesn't need a couple of vectors of 48bytes * the size of the file to be created, but it doesn't give the organizational benefits (capacity, build speed, shape) that lasblock works to achieve either.  

"Spatial splitting" would need to add some code to las2las2 to build an index, iterate the liblas::Index octree cells directly, and determine how to split based on the data in those nodes.  It'd be fun project for someone to tinker with, or a nice, paid project for me to do ;)

Howard


[1] http://lists.osgeo.org/pipermail/liblas-devel/2010-August/000921.html