oracle sql 优化

1.Sql优化：explain

分析

• seelect * from tf_f_user;--全表扫描
• select t.user_id from tf_f_user t;--idex_fast_full_sacn

• select * from tf_f_user t where t.user_id is null;--全表扫描
• select * from tf_f_user t where t.user_id is not null;--全表扫描
• select * from tf_f_user t where t.user_id = '123';--无索引全表扫描，有索引查索引

• select count(1) from tf_f_user t;--index_fast_full_sacan
• select * from tf_f_user t , tf_f_user_member tt where t.user_id = tt.member_role_id;--两张表全表扫描
• select t.* from tf_f_user t , tf_f_user_member tt where t.user_id = tt.member_role_id;--tf_f_user_member表index_fast_full_scan
• select * from tf_f_user t left join tf_f_user_member tt on t.user_id = tt.member_role_id;--全表扫描
• select t.* from tf_f_user t left join tf_f_user_member tt on t.user_id = tt.member_role_id;--tf_f_user_member表index_fast_full_scan

总结

1. Select * 或者无索引的字段，即使表有索引，也需全表查询
2. Where条件无索引字段，is null ,null,’’，全表查询
3. 左连接或者右连接或者内连接的使用等不会缩短查询时间，只是一种匹配方式
4. 关联查询若只是查其中一张表数据，不要去查另一张表数据，否则变成两张表全表查询
5. 关联查询条件需走索引，不然走全表查询

 

2. 几种insert数据的效率比较

原始sql，数据量为9640188，一个左连接

insert   into serv

  (select USER_ID SERV_ID,

          0 AGREEMENT_ID,

          CUST_ID CUST_ID,

          PRODUCT_ID PRODUCT_ID,

          0 ACCT_ID,

          0 BILLING_CYCLE_TYPE_ID,

          in_date CREATE_DATE,

          serial_number ACC_NBR,

          open_date EFF_DATE,

          to_date('20501231235959','YYYYMMDDHH24MISS') EXP_DATE,

          eparchy_code REGION_ID,

          brand_code BAND_ID,

          decode(mm.member_role_code, 9225, 1, 9226, 2, 0) ATTR_CODE,

          0 CREDIT_VALUE,

          0 BILL_TYPE,

          SERVICE_STATE_CODE STATE,

          '' STATE_DATE,

          '' ENCRYPT_INFO,

          '' STOP_ATTR,

          PREPAY_TAG SERV_PASSWD,

          '0' CYCLE_STATE,

          in_date CYCLE_DATE,

          user_type_code USER_TYPE,

          CITY_CODE CITY_CODE,

          sysdate UPDATE_DATE,

          '' RLB_DISP,

          '' ROLL_FLAG,

          NET_TYPE_CODE NET_TYPE_CODE,

          FIRST_CALL_DATE FIRST_USER_DATE,

          LAST_STOP_DATE LAST_STOP_DATE

     from i_user t left join

(select distinct m.member_role_id, m.member_role_code

               from tf_f_user_member m

              where m.member_role_code in ('9225', '9226') and m.relation_type_code='ZF'

                and m.end_date >= trunc(sysdate, 'mm')) mm

 on t.user_id = mm.member_role_id

 )

 

 

几种运行结果

1. -- insert 有索引:194s
2. --inseert 无索引：68
3. --insert /*+parallel(16)*/ 有索引:175
4. alter table serv nologging;--加上后isnert:187s
5. --create as： 53s + 31s建索引

总结

若是新表，则用create table [tablename] as 最快

1. 若是老表，则加上nologging在insert
2. 若是老表数据不多，先让索引失效，在重建索引
3. parallel(16)并发有效果，感觉不明显

 几种join的总结

若有两张表：t1,t2

T1 Right join t2:至少count(t1)条记录，t1,t2中有两条匹配数据，则多count(1)-2+2*2条记录，

右连接一样

内连接：条数为符合两边条件的乘机，即若两边都有2条数据满足，则2*2条，若无 则2*0条