pgSphere uses GiST and Block Range INdexing (BRIN) algorithms to create spherical indices. GiST indexes utilize an R-tree implementation for spherical objects, while BRIN indexes are based on the "summarization" of data blocks (pages) on physical storage in order to organize data searches on ranges of summarized data that can be easily skipped on the base of search filters (see PostgreSQL documentation for further details on BRIN indexes). As a consequence, BRIN indexes are very small indexes (up to 1000 times smaller than GiST ones), generally with lower performance compared with a GiST one, but up to 100 times faster than a full sequential scan of a table performed without any index. So BRIN indexes are particularly suitable in a big data context. An index speeds up the execution time of searches based on operators <@, @, &&, #, =, and !=.

You can create a GiST index with the following spherical data types:

A GiST index can be also used for quickly finding the points closest to the given one when ordering by an expression with the <-> operator, as shown in an example below.

BRIN indexing supports just spherical points (spoint) and spherical coordinates range (sbox) at the moment.

CREATE TABLE test (
  pos spoint NOT NULL
);
-- Put in data now
CREATE INDEX test_pos_idx ON test USING GIST (pos);
VACUUM ANALYZE test;
          

SELECT * FROM test ORDER BY pos <-> spoint (0.2, 0.3) LIMIT 10 
          

CREATE INDEX test_pos_idx USING BRIN ON test (pos);
          

CREATE INDEX test_pos_idx USING BRIN ON test (pos) WITH (pages_per_range = 16);