From PostgreSQL wiki
Authors: Michael Fuhr, Erwin Brandstetter
Comment by: Emanuel Calvo Franco
The basic SQL standard query to count the rows in a table is:
SELECT COUNT(*) FROM tbl;
This can be rather slow because PostgreSQL has to check visibility for all rows, due to the MVCC model. (There have been improvements in PostgreSQL 9.2.)
If you don't need an exact count, the current statistic from the catalog table
pg_class might be good enough and is much faster to retrieve for big tables.
SELECT reltuples FROM pg_class WHERE relname = 'tbl'; reltuples ----------- 100
"tbl" can live in multiple schemas of a database, in which case you get multiple rows for this query. To avoid ambiguity:
SELECT * FROM pg_class WHERE oid = 'schema_name.tbl'::regclass;
Number of rows in the table. This is only an estimate used by the planner. It is updated by
ANALYZE, and a few DDL commands such as
For more sophisticated queries (other than counting all rows from a table), or if you cannot
SELECT from the catalog table
pg_class (which every user can by default), consider this plpgsql function by Michael Fuhr that collects information from
CREATE FUNCTION count_estimate(query text) RETURNS INTEGER AS $func$ DECLARE rec record; ROWS INTEGER; BEGIN FOR rec IN EXECUTE 'EXPLAIN ' || query LOOP ROWS := SUBSTRING(rec."QUERY PLAN" FROM ' rows=([[:digit:]]+)'); EXIT WHEN ROWS IS NOT NULL; END LOOP; RETURN ROWS; END $func$ LANGUAGE plpgsql;
CREATE TEMP TABLE tbl AS SELECT * FROM generate_series(1, 1000) AS t; ANALYZE tbl; SELECT count_estimate('SELECT * FROM tbl WHERE t < 100'); count_estimate ---------------- 100 EXPLAIN SELECT * FROM tbl WHERE t < 100; QUERY PLAN ------------------------------------------------------ Seq Scan on tbl (cost=0.00..35.00 rows=100 width=4) Filter: (t < 100)
As you can see, it's an estimate - actual count would be 99.
If you didn't
ANALYZE recently (after last changes), the estimates can be off more or less.
If you are running the autovacuum daemon as is the default for modern PostgreSQL,
ANALYZE is run automatically, too (except for temporary tables which need manual attention). So the estimates should be good unless you had major changes very recently.