Tip #2: Spatial Indexing and Primary Keys
For fast and efficient search and retrieval all columns containing spatial data should have a spatial index created on them.
For Oracle an index can be created on any SDO_GEOMETRY column regardless as to whether the table has a primary key or not. This is because Oracle uses an internal, unique, ROWID pseudo-column in its indexes: the ROWID links a leaf in the spatial index to the table’s actual row on disk (in the table’s tablespace).
However, SQL Server 2008 “Katmai” requires that all tables must have a primary key before a spatial index can be constructed. The primary key can be based on one or more columns in the table with those columns being of any data type.
The table I loaded for Tip #1, TAS_LGA, can have its primary key constructed as follows:
1. Make the column that will become the primary key “NOT NULL”
alter table dbo.tas_lga alter column lga_pid char(15) not null;
2. Declaratively define the primary key on our not null column lga_pid as follows:
alter table dbo.tas_lga add constraint tas_lga_pk primary key (lga_pid);
We are now in a position to create our spatial index. Microsoft’s spatial index requires you define the bounding box of the spatial index. For our TAS_LGA table this will be the complete extent of the data. This can be discovered as follows:
select min( a.geog.STEnvelope().STPointN(1).STX) as minx, min( a.geog.STEnvelope().STPointN(1).STY) as miny, max( a.geog.STEnvelope().STPointN(3).STX) as maxx, max( a.geog.STEnvelope().STPointN(3).STY) as maxy from gisdb.dbo.tas_lga a;
minx miny maxx maxy
143.813480017165 -43.8603741008445 148.503628803775 -39.1918359319521
Now that we know the extent of our data we can create a spatial index as follows:
CREATE SPATIAL INDEX tas_lga_geog_idx ON tas_lga (geog) USING GEOMETRY_GRID WITH ( BOUNDING_BOX = ( 143, -44, 149, -39), GRIDS = (LEVEL_1 = HIGH, LEVEL_2 = HIGH, LEVEL_3 = HIGH, LEVEL_4 = HIGH), CELLS_PER_OBJECT = 16);