Looking for increasing performance in SpatialRelator flow with large datasets

gio wrote:

A tiling strategy would help. (tile all sets involved)

Tile by block or municipality.

Or run per tile.

Workspacerunner.

Portproces for border objects.

"Any source feature could possibly end of in any number of municipalities."

How? ;)

I'm considering tiling, but one based on aggregating municipalities into regions first. I really don't want to have to break my workspace into multiple.

 

 

You ask "How?". What I meant was that there is no way (except by spatial comparison) to pre-determine which municipalities a given feature might end up in. Note that I've added a km wide buffer around each of the municipalities to account for border issues, so each feature may end up in 1-3 or even more municipalities. It's a requirement from the customer.

 

 

gio wrote:

A tiling strategy would help. (tile all sets involved)

Tile by block or municipality.

Or run per tile.

Workspacerunner.

Portproces for border objects.

"Any source feature could possibly end of in any number of municipalities."

How? ;)

Hi gio.

 

Well, I reconsidered your grid suggestion, and found a way to utilize it. I first created a grid of 1 km cells, and sorted out all cells uniquely within a single municipality. I then preemptively matched my features with these grid cells, only performing the "expensive" matching on municipality polygons with the features not entirely within these grid cells. It seems that about 2/3 can be sorted this way.

 

I'm running my two 5 million+ feature sets with this improved workflow (which crashed before), hoping it'll speed up things. I'll let y'all know how it went.

 

Cheers