Alternative Approaches to Parallel GIS Processing

HadoopGIS-create.java

if(key == 1008130) {
        GIS created = new GIS();
        GIS.update("\"97123897\",\"POLYGON((-112.0859375 33.4349975585938,-112.0859375 33.4675445556641,-112.059799194336 33.4675445556641,-112.059799194336 33.4349975585938,-112.0859375 33.4349975585938))\",\"C\"");
        output.collect(new LongWritable(new Integer(97123897)), created);
}
output.collect(key, value);

HadoopGIS-filter.java

f(value.geometry.intersects(box)) {
        output.collect(key, value);
}

HadoopGIS-nearest.java

double minDistance = Double.MAX_VALUE, currDistance;
int closestParcel = -1;

Iterator it = parcels.iterator();
while (it.hasNext())
{
        GIS parcel = (GIS) it.next();

        currDistance = value.geometry.distance(parcel.geometry);
        if (currDistance < minDistance)
        {
                minDistance = currDistance;
                closestParcel = new Integer(parcel.attributes.get("id"));
        }
}

LongWritable lngClosestParcel = new LongWritable (closestParcel);
output.collect(key, lngClosestParcel);

HadoopGIS-update.java

if(key.equals(new LongWritable(1008130))) {
        value.attributes.put("devtype", "C");
}
output.collect(key, value);

PostGIS-chained.sql

begin;

-- Figure out the distance to all applicable parcels
create index jobs_devtype_index on jobs using btree ("devtype");
create index parcels_devtype_index on parcels using btree ("devtype");
create temp table distances on commit drop as
select jobs.id as job_id, parcels.id as parcel_id, distance(jobs.the_geom, parcels.the_geom)
from jobs, parcels
where jobs.devtype = parcels.devtype;

-- Find the distance to the closest parcel
create temp table min on commit drop as
select job_id, min(distance) as distance
from distances
group by job_id;

-- Match the job to the closest parcels
create index distances_distance_index on distances using btree ("distance");
create index min_distance_index on min using btree ("distance");
create temp table nearest on commit drop as
select distances.job_id, distances.parcel_id
from distances, min
where distances.job_id = min.job_id
and distances.distance = min.distance;

-- Only use unique job_id's
create index nearest_job_id_index on nearest using btree ("job_id");
create table "jobs_parcels" as
select nearest.job_id, nearest.parcel_id
from nearest
where nearest.parcel_id = (select n.parcel_id from nearest as n where n.job_id = nearest.job_id limit 1);

commit;

PostGIS-create.sql

insert into parcels values (97123897, ST_GeomFromText('POLYGON((-112.0859375 33.4349975585938,-112.0859375 33.4675445556641,-112.059799194336 33.4675445556641,-112.059799194336 33.4349975585938,-112.0859375 33.4349975585938))', 4326), 'C');

PostGIS-filter.sql

create table output as
select parcels.* from parcels, (select ST_GeomFromText('POLYGON((-112.0859375 33.4349975585938,-112.0859375 33.4675445556641,-112.059799194336 33.4675445556641,-112.059799194336 33.4349975585938,-112.0859375 33.4349975585938))', 4326) as the_geom) as box
where ST_Intersects(parcels.the_geom, box.the_geom);

PostGIS-update.sql

update parcels set devtype = 'C' where id = 1008130;