一.TEMPORARY|TEMP TABLE
会话级或事务级的临时表，临时表在会话结束或事物结束自动删除，任何在临时表上创建的索引也会被自动删除。除非用模式修饰的名字引用，否则现有的同名永久表在临时表存在期间，在本会话或事务中是不可见的。另外临时表对其他会话也是不可见的，但是会话级的临时表也可以使用临时表所在模式修饰的名字引用。
创建临时表的语法：
CREATE TEMP tbl_name()ON COMMIT{PRESERVE ROWS|DELETE ROWS|DROP};
PRESERVE ROWS：默认值，事务提交后保留临时表和数据
DELETE ROWS：事务提交后删除数据，保留临时表
DROP：事务提交后删除表
示例1
会话A:
创建临时表
test=# create temp table tbl_temp(a int); CREATE TABLE
会话B：
1.在会话B查询临时表tbl_temp,提示表不存在
test=# select * from tbl_temp; ERROR: relation "tbl_temp" does not exist LINE 1: select * from tbl_temp;
2.但是在会话B查询pg_class中可以查到tbl_temp的记录
test=# select relname,relnamespace from pg_class where relname = 'tbl_temp'; relname | relnamespace ----------+-------------- tbl_temp | 16488 (1 row)
3.从上述查询结果中可以看到临时表tbl_temp属于16488的模式
test=# select nspname from pg_namespace where oid = 16488; nspname ----------- pg_temp_3 (1 row)
4.直接使用模式修饰的表名访问成功
test=# select * from pg_temp_3.tbl_temp ; a --- (0 rows)
会话A:
退出会话A
会话B：
再次查询tbl_temp时提示不存在
test=# select * from pg_temp_3.tbl_temp ; ERROR: relation "pg_temp_3.tbl_temp" does not exist LINE 1: select * from pg_temp_3.tbl_temp ; ^
示例2.创建ON COMMIT DELETE ROWS的临时表
test=# begin ; BEGIN test=# create temp table tbl_temp(a int) on commit delete rows; CREATE TABLE test=# insert into tbl_temp values (1); INSERT 0 1 test=# select * from tbl_temp ; a --- 1 (1 row) test=# commit ; COMMIT test=# select * from tbl_temp ; a --- (0 rows)
示例3.创建ON COMMIT DROP临时表
test=# begin ; BEGIN test=# create temp table tbl_temp(a int) on commit drop; CREATE TABLE test=# commit ; COMMIT test=# select * from tbl_temp; ERROR: relation "tbl_temp" does not exist LINE 1: select * from tbl_temp; ^
示例4.查询数据库中所有临时表
test=# select relname,nspname from pg_class join pg_namespace on(relnamespace=pg_namespace.oid) where pg_is_other_temp_schema(relnamespace); relname | nspname ----------+----------- tbl_test | pg_temp_2 (1 row)
二.UNLOGGED TABLE
unlogged table是为临时数据设计的，写入性能较高，但是当postgresql进程崩溃时会丢失数据。
创建一张普通表test和一张unlogged表test，测试性能情况
普通表：
test=# create table test(a int); CREATE TABLE test=# /timing Timing is on. test=# insert into test select generate_series(1,1000000); INSERT 0 1000000 Time: 3603.715 ms
unlogged表
test=# create unlogged table testu(a int); CREATE TABLE Time: 12.920 ms test=# insert into testu select generate_series(1,1000000); INSERT 0 1000000 Time: 801.376 ms
比较以上两个结果，unlogged表的写性能是普通表的4.5倍。
杀死postgresql的主进程，重启DB服务
[[email protected] ~]# ps -elf | grep postgres 0 S postgres 2129 1 0 80 0 - 66830 poll_s 04:24 ? 00:00:00 /opt/pg9.6/bin/postgres -D /mnt/pgdata 1 S postgres 2130 2129 0 80 0 - 29645 ep_pol 04:24 ? 00:00:00 postgres: logger process 1 S postgres 2132 2129 0 80 0 - 66898 poll_s 04:24 ? 00:00:00 postgres: checkpointer process 1 S postgres 2133 2129 0 80 0 - 66830 ep_pol 04:24 ? 00:00:00 postgres: writer process 1 S postgres 2134 2129 0 80 0 - 66871 ep_pol 04:24 ? 00:00:00 postgres: wal writer process 1 S postgres 2135 2129 0 80 0 - 66954 ep_pol 04:24 ? 00:00:00 postgres: autovacuum launcher process 1 S postgres 2136 2129 0 80 0 - 29677 ep_pol 04:24 ? 00:00:00 postgres: stats collector process 0 S root 2262 2099 0 80 0 - 28768 n_tty_ 04:52 pts/1 00:00:00 /opt/pg9.6/bin/psql -d test -U postgres 1 S postgres 2264 2129 0 80 0 - 67351 ep_pol 04:52 ? 00:00:02 postgres: postgres test [local] idle 0 S root 2334 2198 0 80 0 - 25813 pipe_w 05:15 pts/2 00:00:00 grep postgres [[email protected]-R1CL-srv ~]# kill -9 2129 [[email protected]-R1CL-srv ~]# rm -rf /mnt/pgdata/postmaster.pid [[email protected]-R1CL-srv ~]# su -l postgres -c '/opt/pg9.6/bin/pg_ctl -D /mnt/pgdata start' server starting [[email protected]-R1CL-srv ~]# 2016-06-22 05:16:04.399 CST 2372 LOG: redirecting log output to logging collector process 2016-06-22 05:16:04.399 CST 2372 HINT: Future log output will appear in directory "/var/log/pg_log".
再次查询unlogged表testu，发现数据已丢失
test=# select * from testu ; a --- (0 rows)
原创文章，作者：ItWorker，如若转载，请注明出处：https://blog.ytso.com/tech/database/4882.html