Joins & relationships

A join answers "which of these goes with which of those", and there are only two ways to answer it: read a key somebody already wrote down, or compute the relationship from the geometry. The first is faster and authoritative. The second works when no key exists, which is most of the time.

The surprise in spatial joins is arithmetic rather than spatial. An attribute join on a unique key is a function — each left row matches at most one right row. A spatial join is a relation, and a relation can be true several times, so a left feature that falls inside two overlapping zones comes back twice. The row count growing is not a fault; it is the join reporting every relationship it found, and the fix is deciding what one output row should mean.

The pages below run from what a join is doing, through the joins themselves — point-in-polygon, polygon overlay, nearest neighbour — to the specific ways one returns nothing, or far too much.

Learning paths run through this topic too — see the end of this page.