Truncating all tables in a Postgres database

Question

I regularly need to delete all the data from my PostgreSQL database before a rebuild. How would I do this directly in SQL?

At the moment I've managed to come up with a SQL statement that returns all the commands I need to execute:

SELECT 'TRUNCATE TABLE ' ||  tablename || ';' FROM pg_tables WHERE tableowner='MYUSER';

But I can't see a way to execute them programmatically once I have them.

Answer 1

FrustratedWithFormsDesigner is correct, PL/pgSQL can do this. Here's the script:

CREATE OR REPLACE FUNCTION truncate_tables(username IN VARCHAR) RETURNS void AS $$
DECLARE
    statements CURSOR FOR
        SELECT tablename FROM pg_tables
        WHERE tableowner = username AND schemaname = 'public';
BEGIN
    FOR stmt IN statements LOOP
        EXECUTE 'TRUNCATE TABLE ' || quote_ident(stmt.tablename) || ' CASCADE;';
    END LOOP;
END;
$$ LANGUAGE plpgsql;

This creates a stored function (you need to do this just once) which you can afterwards use like this:

SELECT truncate_tables('MYUSER');

Answer 2

Could you use dynamic SQL to execute each statement in turn? You would probably have to write a PL/pgSQL script to do this.

http://www.postgresql.org/docs/8.3/static/plpgsql-statements.html (section 38.5.4. Executing Dynamic Commands)

Answer 3

If I have to do this, I will simply create a schema sql of current db, then drop & create db, then load db with schema sql.

Below are the steps involved:

1) Create Schema dump of database (--schema-only)

pg_dump mydb -s > schema.sql

2) Drop database

drop database mydb;

3) Create Database

create database mydb;

4) Import Schema

psql mydb < schema.sql

Answer 4

Cleaning AUTO_INCREMENT version:

CREATE OR REPLACE FUNCTION truncate_tables(username IN VARCHAR) RETURNS void AS $$
DECLARE
    statements CURSOR FOR
        SELECT tablename FROM pg_tables
        WHERE tableowner = username AND schemaname = 'public';
BEGIN
    FOR stmt IN statements LOOP
        EXECUTE 'TRUNCATE TABLE ' || quote_ident(stmt.tablename) || ' CASCADE;';

        IF EXISTS (
            SELECT column_name 
            FROM information_schema.columns 
            WHERE table_name=quote_ident(stmt.tablename) and column_name='id'
        ) THEN
           EXECUTE 'ALTER SEQUENCE ' || quote_ident(stmt.tablename) || '_id_seq RESTART WITH 1';
        END IF;

    END LOOP;
END;
$$ LANGUAGE plpgsql;

Answer 5

Explicit cursors are rarely needed in plpgsql. Just use the simpler and faster implicit cursor of a FOR loop:

Note: Since table names are not unique per database, you have to schema-qualify table names to be sure. Also, I limit the function to the default schema 'public'. Adapt to your needs, but be sure to exclude the system schemas pg_* and information_schema.

Be very careful with these functions. They nuke your database. I added a child safety device. Comment the RAISE NOTICE line and uncomment EXECUTE to prime the bomb ...

CREATE OR REPLACE FUNCTION f_truncate_tables(_username text)
  RETURNS void AS
$func$
DECLARE
   _tbl text;
   _sch text;
BEGIN
   FOR _sch, _tbl IN 
      SELECT schemaname, tablename
      FROM   pg_tables
      WHERE  tableowner = _username
      AND    schemaname = 'public'
   LOOP
      RAISE NOTICE '%',
      -- EXECUTE  -- dangerous, test before you execute!
         format('TRUNCATE TABLE %I.%I CASCADE', _sch, _tbl);
   END LOOP;
END
$func$ LANGUAGE plpgsql;

format() requires Postgres 9.1 or later. In older versions concatenate the query string like this:

'TRUNCATE TABLE ' || quote_ident(_sch) || '.' || quote_ident(_tbl)  || ' CASCADE';

Single command, no loop

Since we can TRUNCATE multiple tables at once we don't need any cursor or loop at all:

• Passing table names in an array

Aggregate all table names and execute a single statement. Simpler, faster:

CREATE OR REPLACE FUNCTION f_truncate_tables(_username text)
  RETURNS void AS
$func$
BEGIN
   RAISE NOTICE '%', 
   -- EXECUTE  -- dangerous, test before you execute!
  (SELECT 'TRUNCATE TABLE '
       || string_agg(format('%I.%I', schemaname, tablename), ', ')
       || ' CASCADE'
   FROM   pg_tables
   WHERE  tableowner = _username
   AND    schemaname = 'public'
   );
END
$func$ LANGUAGE plpgsql;

Call:

SELECT truncate_tables('postgres');

Refined query

You don't even need a function. In Postgres 9.0+ you can execute dynamic commands in a DO statement. And in Postgres 9.5+ the syntax can be even simpler:

DO
$func$
BEGIN
   RAISE NOTICE '%', 
   -- EXECUTE
   (SELECT 'TRUNCATE TABLE ' || string_agg(oid::regclass::text, ', ') || ' CASCADE'
    FROM   pg_class
    WHERE  relkind = 'r'  -- only tables
    AND    relnamespace = 'public'::regnamespace
   );
END
$func$;

About the difference between pg_class, pg_tables and information_schema.tables:

• How to check if a table exists in a given schema

About regclass and quoted table names:

• Table name as a PostgreSQL function parameter

For repeated use

It might be simpler and (much) faster to create a "template" database (let's name it my_template) with your vanilla structure and all empty tables. Then go through a DROP / CREATE DATABASE cycle:

DROP DATABASE mydb;
CREATE DATABASE mydb TEMPLATE my_template;

This is extremely fast, because Postgres copies the whole structure on the file level. No concurrency issues or other overhead slowing you down.

Answer 6

In this case it would probably be better to just have an empty database that you use as a template and when you need to refresh, drop the existing database and create a new one from the template.