This workspace takes a set of road centrelines and a set of building centroids and determines which building belongs to which road, based on proximity and relative bearing.

I envision this would be used for map data that you are trying to create intelligence for.

This workspace actually generates its own sample data, applying a random 'wobble' to each feature to mimic bad digitising. The building points are generated as text features with a rotation set so that they face the road they belong to. The idea is that a building will be assigned to a road not just because that road line is closest.

Although the workspace may look complicated, a large portion of it relates to the sample data generation - the actual processing part is fairly minimal. The data has an 'overall' rotation so it can be tested at angles other than a N-S/E-W alignment - it seems to work OK until items are rotated 90 degrees or more (but at 90 degrees N-S would be running E-W and you'd need to be a really bad digitizer to manage that).