state_agg is subject to the same memory hunger described in issue #475 possibly making it unusable for large data sets.
It seems states should be enumerated and state_agg should deal with only integer states anyway, obviating the problem.
If we do need to operate directly on strings, the simplest fix may be to tell postgres not to parallelize it and then use a HashMap to avoid duplicating strings as we go.
state_aggis subject to the same memory hunger described in issue #475 possibly making it unusable for large data sets.It seems states should be enumerated and state_agg should deal with only integer states anyway, obviating the problem.
If we do need to operate directly on strings, the simplest fix may be to tell postgres not to parallelize it and then use a
HashMapto avoid duplicating strings as we go.