与Oracle不同。PostgreSQL须要手动控制分区规则触发器。
步骤一：创建分区
CREATE TABLE table_partition_1( CHECK partition_column criteria) INHENRITS (table)
步骤二：为分区表创建PK跟index，这里使用btree
ALTER TABLE ONLY table_partition_1 ADD CONSTRAINT table_partition_1_pkey PRIMARY KEY (key_column);
CREATE INDEX index_table_partition_1 ON table_partition_1 USING btree(column);步骤三：手动创建触发器
CREATE OR REPLACE FUNCTION before_insert_table() RETURNS trigger AS$BODY$ DECLARE BEGINif criteria then EXECUTE 'insert into appropriate table ...' SELECT ($1).* ' USING NEW;end if;return null;END;$BODY$ LANGUAGE plpgsql VOLATILE COST 100;ALTER FUNCTION before_insert_table() OWNER TO db;rule在批量操作时更合适，可是对于单独操作会占用较大的开销。
CREATE RULE table_partition_1 _insert AS(criteria ...)DO INSTEAD INSERT INTO table_partition_1 VALUES (NEW.*)步骤四：触发器方式hibernate向分区插入数据时。获得的result count为0，会导致推断失败而回滚。
解决方法是使用rule，或者声明分区插入时不进行result检查。
@SQLInsert(sql = "INSERT INTO " + "table(column,...)" + " VALUES(?,...)", check = ResultCheckStyle.NONE)
在Java项目中。考虑到分区创建会採用job方式自己主动创建，能够通过function完毕创建。
CREATE OR REPLACE FUNCTION "public"."function"() RETURNS void AS $BODY$ DECLARE _tablename text ; quarter integer; record1 record; BEGIN select function(now()) as quarter into record1; quarter := record1.quarter; _tablename := 'partition_name';PERFORM 1FROM pg_catalog.pg_class cWHERE c.relname = _tablename; IF FOUND <> TRUE THEN EXECUTE 'CREATE TABLE ' || _tablename || ' ( CHECK ( criteria) ) INHERITS (table)'; EXECUTE 'ALTER TABLE ' || _tablename || ' OWNER TO db '; EXECUTE ' alter table ' || _tablename || ' add CONSTRAINT ' || _tablename||'_pkey PRIMARY key (column) ' ; EXECUTE ' CREATE INDEX ' || _tablename|| '_index ON '|| _tablename ||' USING btree (column)'; END IF;END$BODY$ LANGUAGE 'plpgsql' VOLATILE COST 100;ALTER FUNCTION "public"."function"() OWNER TO "db";