-- Function that will create a partition CREATE OR REPLACE FUNCTION historize_check_partition( schema_dest name, table_source name, delta integer default 1) RETURNS integer LANGUAGE plpgsql AS $$ DECLARE dateStr varchar; dateUpStr varchar; partition varchar; table_log varchar; BEGIN table_log := table_source || '_log'; IF NOT EXISTS (SELECT 1 FROM information_schema.tables WHERE table_schema=schema_dest AND table_name=table_log) THEN RETURN 1; END IF; SELECT to_char(DATE 'today' + make_interval(days => delta), 'YYYYMMDD') INTO dateStr; SELECT to_char(DATE 'tomorrow' + make_interval(days => delta), 'YYYYMMDD') INTO dateUpStr; partition := table_log || '_' || dateStr::text; IF EXISTS (SELECT 1 FROM information_schema.tables WHERE table_schema=schema_dest AND table_name=partition) THEN RETURN 0; ELSE RETURN 2; END IF; END; $$; CREATE OR REPLACE FUNCTION historize_create_partition( schema_dest name, table_source name, delta integer default 1) RETURNS integer LANGUAGE plpgsql AS $$ DECLARE dateStr varchar; dateUpStr varchar; partition varchar; table_log varchar; BEGIN table_log := table_source || '_log'; IF NOT EXISTS (SELECT 1 FROM information_schema.tables WHERE table_schema=schema_dest AND table_name=table_log) THEN RETURN 1; END IF; SELECT to_char(DATE 'today' + make_interval(days => delta), 'YYYYMMDD') INTO dateStr; SELECT to_char(DATE 'tomorrow' + make_interval(days => delta), 'YYYYMMDD') INTO dateUpStr; partition := table_log || '_' || dateStr::text; IF NOT EXISTS (SELECT 1 FROM information_schema.tables WHERE table_schema=schema_dest AND table_name=partition) THEN EXECUTE format('CREATE TABLE %s (LIKE %s INCLUDING INDEXES)', partition, table_log); EXECUTE format('ALTER TABLE %s ATTACH PARTITION %s FOR VALUES FROM (%L) TO (%L)', table_log, partition, dateStr, dateUpStr); RETURN 0; ELSE RETURN 2; END IF; END; $$; CREATE OR REPLACE FUNCTION historize_create_partition(table_source name, delta integer default 1) RETURNS integer LANGUAGE plpgsql AS $$ DECLARE result integer; BEGIN SELECT historize_create_partition('public', table_source, delta) INTO result; RETURN result; END; $$; -- Drop a partion -- -- CREATE OR REPLACE FUNCTION historize_drop_partition( schema_dest name, table_source name, delta integer default 1) RETURNS integer LANGUAGE plpgsql AS $$ DECLARE dateStr varchar; dateUpStr varchar; partition varchar; table_log varchar; BEGIN table_log := table_source || '_log'; SELECT to_char(DATE 'today' + make_interval(days => delta), 'YYYYMMDD') INTO dateStr; SELECT to_char(DATE 'tomorrow' + make_interval(days => delta), 'YYYYMMDD') INTO dateUpStr; partition := table_log || '_' || dateStr::text; IF EXISTS (SELECT relname FROM pg_class WHERE relname=partition) THEN EXECUTE format('DROP TABLE %s.%s', schema_dest, partition); RETURN 1; ELSE RETURN 0; END IF; END; $$; CREATE OR REPLACE FUNCTION historize_drop_partition(table_source name, delta integer default 1) RETURNS integer LANGUAGE plpgsql AS $$ DECLARE result integer; BEGIN SELECT historize_drop_partition('public', table_source, delta) INTO result; RETURN result; END; $$; -- This function is used to initialize the data historization -- CREATE OR REPLACE FUNCTION historize_get_logname( schema_source NAME, table_source NAME) RETURNS text LANGUAGE plpgsql AS $$ BEGIN RETURN schema_source || '.' || table_source || '_log'; END; $$; -- This function is used to initialize the data historization -- -- Drop the partitionned tables CREATE OR REPLACE FUNCTION historize_table_clean( schema_source NAME, table_source NAME) RETURNS void LANGUAGE plpgsql AS $$ DECLARE partition varchar; BEGIN IF NOT EXISTS (SELECT 1 FROM information_schema.tables WHERE table_schema=schema_source AND table_name=table_source) THEN RAISE EXCEPTION 'table %.% does not exists', schema_source, table_source USING HINT = 'Check the table name and schema, fix the search_path is case of needed', ERRCODE = '42P01'; END IF; -- Stop the historization to ensure to not block users doing an error PERFORM historize_table_stop(schema_source, table_source); EXECUTE format('DROP TABLE %s ', historize_get_logname(schema_source, table_source )); END; $$; -- -- Implicit schema public -- CREATE OR REPLACE FUNCTION historize_table_clean(table_source NAME) RETURNS void LANGUAGE plpgsql AS $$ BEGIN PERFORM historize_table_clean('public'::name, table_source); END; $$; -- This function is used to define conr entries in another database -- -- schema_dest : -- table_source : -- days_in_advance : the number of partition to create in advance -- foreign_server : the name of the ofreign server defined with CREATE SERVER statement -- cron_foreign_schema : the name of the scheme where the pg_cron extesion is installed CREATE OR REPLACE FUNCTION historize_cron_define( schema_dest NAME, table_source NAME, days_in_advance integer DEFAULT 4, days_to_keep integer DEFAULT 7, schedule text DEFAULT '00 08 * * *', foreign_server varchar DEFAULT 'historize_foreign_cron' ) RETURNS integer LANGUAGE plpgsql AS $$ DECLARE qry_c text; qry_d text; drop_start_from integer; BEGIN qry_c := format('SELECT schedule_in_database(%L, %L, $eof$SELECT historize_create_partition(%L, %L, generate_series(1, %s) )$eof$, %L) ', 'histo_create_part_' || schema_dest || '_' || table_source, schedule, schema_dest, table_source, days_in_advance, current_database()); -- Drop one week more to ensure that no partitions are missed drop_start_from := 0 - (7 + days_to_keep); qry_d := format('SELECT schedule_in_database(%L, %L, $eof$SELECT historize_drop_partition(%L, %L, generate_series(%s, -%s) )$eof$, %L) ', 'histo_drop_part_' || schema_dest || '_' || table_source, schedule, schema_dest, table_source, drop_start_from, days_to_keep, current_database()) ; EXECUTE format(' SELECT * FROM dblink(%L, %L ) AS t1 (schedule_in_database bigint)', foreign_server, qry_c); EXECUTE format(' SELECT * FROM dblink(%L, %L ) AS t1 (schedule_in_database bigint)', foreign_server, qry_d); RETURN 0; END; $$; CREATE OR REPLACE FUNCTION historize_cron_remove( schema_dest NAME, table_source NAME, foreign_server varchar DEFAULT 'historize_foreign_cron' ) RETURNS integer LANGUAGE plpgsql AS $$ DECLARE qry_c text; qry_d text; BEGIN qry_c := format('SELECT unschedule(%L)', 'histo_create_part_' || schema_dest || '_' || table_source); qry_d := format('SELECT unschedule(%L)', 'histo_drop_part_' || schema_dest || '_' || table_source); EXECUTE format(' SELECT * FROM dblink(%L, %L ) AS t1 (result boolean)', foreign_server, qry_c); EXECUTE format(' SELECT * FROM dblink(%L, %L ) AS t1 (result boolean)', foreign_server, qry_d); RETURN 0; END; $$; -- List all cron entries for the current database -- -- CREATE OR REPLACE FUNCTION historize_cron_list( foreign_server varchar DEFAULT 'historize_foreign_cron' ) RETURNS TABLE (jobid bigint, schedule text, command text, nodename text, nodeport integer, username text, active boolean, jobname text) LANGUAGE plpgsql AS $$ DECLARE qry text; BEGIN qry := format('SELECT jobid, schedule, command, nodename, nodeport, database, username, active,jobname FROM job' ); RETURN QUERY SELECT t1.jobid, t1.schedule, t1.command, t1.nodename, t1.nodeport, t1.username, t1.active, t1.jobname FROM dblink(foreign_server, qry ) AS t1 (jobid bigint,schedule text, command text, nodename text, nodeport integer, database text, username text, active boolean, jobname text) WHERE t1.database = current_database(); END; $$; -- This function is used to initialize the data historization -- -- It creates multiple objects -- - a table with the name of the table to historize adding a suffix _log -- - an index -- - a column histo_version on the table source -- - a column histo_sys_period on the table source CREATE OR REPLACE FUNCTION historize_table_init( schema_dest NAME, table_source NAME) RETURNS void LANGUAGE plpgsql AS $$ DECLARE partition varchar; BEGIN -- check if the table source exists IF NOT EXISTS (SELECT true FROM information_schema.tables WHERE table_schema = schema_dest AND table_name = table_source) THEN RAISE EXCEPTION 'table %.% does not exists', schema_dest, table_source USING ERRCODE = '42P01'; END IF; -- check if the columns does not already exists EXECUTE format(' CREATE TABLE IF NOT EXISTS %s ( id int, eventtime timestamp with time zone, txid bigint, sys_period tstzrange, data jsonb ) PARTITION BY RANGE (eventtime)', schema_dest || '.' || table_source || '_log'); -- Create an index on the id the easily regroup all the tuple for the same initial one -- EXECUTE format(' CREATE INDEX %s_log_id_idx ON %s_log(id)', table_source, schema_dest || '.' || table_source); -- Add a new column on the source table to keep the version directly in the row -- EXECUTE format(' ALTER TABLE %s ADD COLUMN histo_version int default 0', table_source); EXECUTE format(' ALTER TABLE %s ADD COLUMN histo_sys_period tstzrange NOT NULL DEFAULT tstzrange(current_timestamp, null)', table_source); -- Create 7 first partition from today -- EXECUTE format(' SELECT historize_create_partition(%L, generate_series(0,6) )', table_source ); -- If a foreign server exists and named as default, define the cron entries -- IF EXISTS (SELECT 1 FROM pg_foreign_server WHERE srvname='historize_foreign_cron') THEN EXECUTE format(' SELECT historize_cron_define(%L, %L)', schema_dest, table_source ); END IF; END; $$; -- -- Implicit schema public -- CREATE OR REPLACE FUNCTION historize_table_init(table_source NAME) RETURNS void LANGUAGE plpgsql AS $$ BEGIN PERFORM historize_table_init('public'::name, table_source); END; $$; -- This function is used to initialize the data historization -- -- Reset the related objects likned to the historization -- -- - stop the historization -- - drop columns created on source table -- - remove the cron commands CREATE OR REPLACE FUNCTION historize_table_reset( schema_source NAME, table_source NAME) RETURNS void LANGUAGE plpgsql AS $$ DECLARE partition varchar; BEGIN IF NOT EXISTS (SELECT 1 FROM information_schema.tables WHERE table_schema=schema_source AND table_name=table_source) THEN RAISE EXCEPTION 'table %.% does not exists', schema_source, table_source USING HINT = 'Check the table name and schema, fix the search_path is case of needed', ERRCODE = '42P01'; END IF; -- Stop the historization to ensure there is no trigger left EXECUTE format(' SELECT historize_table_stop(%L, %L)', schema_source, table_source ); EXECUTE format(' ALTER TABLE %s DROP COLUMN histo_version', table_source); EXECUTE format(' ALTER TABLE %s DROP COLUMN histo_sys_period', table_source); -- If a foreign server exists and named as default, define the cron entries -- IF EXISTS (SELECT 1 FROM pg_foreign_server WHERE srvname='historize_foreign_cron') THEN EXECUTE format(' SELECT historize_cron_remove(%L, %L)', schema_source, table_source ); END IF; END; $$; -- -- Implicit schema public -- CREATE OR REPLACE FUNCTION historize_table_reset(table_source NAME) RETURNS void LANGUAGE plpgsql AS $$ BEGIN PERFORM historize_table_reset('public'::name, table_source); END; $$; -- This function is used to initialize the data historization -- -- It creates multiple objects -- - a table with the name of the table to historize adding a suffix _log -- - an index -- - a new column on the table source CREATE OR REPLACE FUNCTION historize_table_start(schema_dest NAME, table_source NAME) RETURNS void LANGUAGE plpgsql AS $EOF$ DECLARE check_part boolean; BEGIN SELECT historize_check_partition(schema_dest, table_source, 0) = 0 INTO check_part; IF NOT check_part THEN RAISE EXCEPTION 'no available partition in log table' USING HINT = 'check if you init the historization with historize_table_init function'; END IF; -- -- Function to manage UPDATE statements -- EXECUTE format('CREATE OR REPLACE FUNCTION %s_historization_update_trg() RETURNS trigger LANGUAGE plpgsql AS $$ BEGIN NEW.histo_version = OLD.histo_version + 1; NEW.histo_sys_period = tstzrange(CURRENT_TIMESTAMP,null); INSERT INTO %s.%s_log (id, txid, eventtime, sys_period, data) VALUES (OLD.id, txid_current(), now(), tstzrange(lower(OLD.histo_sys_period), CURRENT_TIMESTAMP), to_jsonb(OLD) - ''histo_sys_period''); RETURN NEW; END; $$', table_source, schema_dest, table_source); -- -- Function to manage INSERT statements -- EXECUTE format('CREATE OR REPLACE FUNCTION %s_historization_insert_trg() RETURNS trigger LANGUAGE plpgsql AS $$ BEGIN NEW.histo_version = 1; NEW.histo_sys_period = tstzrange(CURRENT_TIMESTAMP,null); RETURN NEW; END; $$', table_source, schema_dest, table_source); -- -- Create two triggers, one for UPDATE and one for INSERT -- EXECUTE format(' CREATE TRIGGER %s_historization_update_trg BEFORE UPDATE ON %s.%s FOR EACH ROW WHEN (NEW <> OLD) EXECUTE PROCEDURE %s_historization_update_trg()', table_source, schema_dest, table_source, table_source); EXECUTE format(' CREATE TRIGGER %s_historization_insert_trg BEFORE INSERT ON %s.%s FOR EACH ROW EXECUTE PROCEDURE %s_historization_insert_trg()', table_source, schema_dest, table_source, table_source); END; $EOF$; -- -- Implicit schema public -- CREATE OR REPLACE FUNCTION historize_table_start(table_source NAME) RETURNS void LANGUAGE plpgsql AS $$ BEGIN PERFORM historize_table_start('public'::name, table_source); END; $$; -- This function is used to stop the historization -- CREATE OR REPLACE FUNCTION historize_table_stop( schema_dest NAME, table_source NAME) RETURNS void LANGUAGE plpgsql AS $EOF$ BEGIN IF NOT EXISTS (SELECT 1 FROM information_schema.tables WHERE table_schema=schema_dest AND table_name=table_source) THEN RAISE EXCEPTION 'table %.% does not exists', schema_dest, table_source USING HINT = 'Check the table name and schema, fix the search_path in case of needed', ERRCODE = '42P01'; END IF; EXECUTE format('DROP TRIGGER IF EXISTS %s_historization_update_trg ON %s.%s', table_source, schema_dest, table_source); EXECUTE format('DROP TRIGGER IF EXISTS %s_historization_insert_trg ON %s.%s', table_source, schema_dest, table_source); -- -- Function that manage UPDATE statements -- EXECUTE format('DROP FUNCTION IF EXISTS %s_historization_update_trg()', table_source); -- -- Function that manage INSERT statements -- EXECUTE format('DROP FUNCTION IF EXISTS %s_historization_insert_trg()', table_source); END; $EOF$; CREATE OR REPLACE FUNCTION historize_table_stop(table_source NAME) RETURNS void LANGUAGE plpgsql AS $$ BEGIN PERFORM historize_table_stop('public'::name, table_source); END; $$;