MPP（高斯DB）常用语句（不定时更新添加）

更新时间20210923

目录

1.获取分布键

2. 获取主键

3.查看视图注释

4.改变表所属schema

5.查找某个schema里有什么个表

6.一个merge的举例

7.根据mpp表生成hive的建表语句（表名加_di后缀）

8.分区操作

①.查看表分区

②设置每月一个分区（区间）

③设置上界

④删除分区

⑤新增分区

⑥分区合并

⑦分区分割

⑧重命名分区

9.列操作

①查询列信息

②拼接所有列

10.数据去重（未优化版本）

11.建立与使用序列

12.查看表数据在DN上的分布倾斜情况

---

1.获取分布键

select

attname

from pg_attribute

where attrelid = '表schema.表名称'::regclass

and not attisdropped

and attnum in (SELECT unnest(pcattnum) from PGXC_CLASS where pcrelid = '表schema.表名称'::regclass)

2. 获取主键
 

select

a.attname as pk_name

from

pg_index i

join pg_attribute a on a.attrelid = i.indrelid and a.attnum = any(i.indkey)

where

i.indrelid = format('%s.%s', ?, ?)::regclass

and i.indisprimary

3.查看视图注释

-- 方法一视图字段注释
select

pa.attnum,

pa.attname, quote_literal(col_description('视图schema.视图名称'::regclass::oid,pa.attnum)) attdesc

from

pg_attribute pa

where

pa.attnum > 0 and not pa.attisdropped

and pa.attrelid = '视图schema.视图名称'::regclass::oid and attdescis not null

-- 方法二获取视图注释
select

string_agg(

format('COMMENT ON COLUMN "%s"."%s"."%s" IS %s;',

table_schema,

table_name,

column_name,

quote_literal(col_description((table_schema || '.' || table_name)::regclass::oid, ordinal_position)))

, E'\n'

) as comment_ddl

from

information_schema.columns

where

table_schema = 'jwy_odm' and table_name = 'myv'

4.改变表所属schema

alter table 表 set schema 表schema

5.查找某个schema里有什么个表

SELECT

'表schema.'||table_name table_name

FROM

information_schema.tables

WHERE

table_schema = '表schema'

-- and table_name ~ 's_[^_]+_t_.*'

-- and table_name not like 's_code%'

and table_type = 'BASE TABLE'



-- 带注释

select
distinct
	  pv.table_name,
		cast(obj_description(cl.relfilenode,'pg_class') as varchar)as comment
	from 
		pg_class pc
		left join pg_namespace ns on pc.relnamespace = ns.oid
		left join information_schema.table_privileges pv on pv.table_schema = ns.nspname and pv.table_name = pc.relname
		left join pg_class cl on  cl.relname = pv.table_name
	where
		pv.table_schema is not null
		and pv.table_schema = 'odm'
		and pv.table_name  ~ 'o_.*'
		order by table_name

		-- and pv.grantee <> 'odm'
		-- pc.relkind = 'v' and  pc.oid = oid and 

6.一个merge的举例

MERGE INTO 表sche.主表名称1 AS s

USING 表sche.次表名称2 as o
-- 条件可以有多个
ON ( s.id = o.id AND o.xxrksj > to_date('19001010'))

WHEN MATCHED THEN

UPDATE SET
-- 符合条件就更新
s.rybh = o.rybh,

s.zpxh = o.zpxh

WHEN NOT MATCHED THEN
-- 不符合条件就插入
INSERT ( id, rybh)

VALUES (o.id, o.rybh) ;

7.根据mpp表生成hive的建表语句（表名加_di后缀）

with cols as (

select

table_schema || '.' || table_name as sat,

column_name as cn,

case

when data_type in ('text') then 'string'

when data_type in ('nvarchar2', 'character varying', 'character', '"char"') and character_maximum_length is null then 'string'

when data_type in ('nvarchar2', 'character varying', 'character', '"char"') then 'varchar(' || character_maximum_length || ')'

when data_type in ('time without time zone', 'time with time zone') then 'string'

when data_type in ('timestamp with time zone', 'timestamp without time zone') then 'timestamp'

when data_type in ('tinyint', 'smallint', 'integer', 'bigint', 'interval') then data_type

when data_type in ('double precision') then 'double'

when data_type = 'bytea' then 'string'

when data_type = 'numeric' and nvl(numeric_scale,0) <> 0 then 'double'

when data_type = 'numeric' then 'integer'

else 'null-' || data_type

end as dt,

QUOTE_LITERAL(col_description(sat::regclass::oid, ordinal_position) )as cm

from

information_schema.columns

where

table_schema = '表schema'

-- 过滤条件

-- and table_name ~ 's_[^_]+_t_.*'

and table_name not like 's_code%'

-- 特出处理

-- and table_name in ('表名称1', '表名称2')

order by ordinal_position

)


select

quote_literal(obj_description(sat::regclass::oid)) tm,

E'-- ' || sat || E'_di \n'

|| 'create table if not exists ' || sat || '_di' || E'(\n'

|| string_agg(' `' || cols.cn || '` ' || cols.dt ||

case when cols.cm is null then null else ' comment ' || cols.cm end

, E',\n')

|| E')\n'

|| case when tm is null then null else E'comment ' || tm || E'\n' end

|| E'partitioned by (etl_rkrq string)\n'

|| E'stored as orc;\n'

|| E'grant select, update, insert, delete on table '|| sat || '_di to user 授权用户名;'

|| E'\n\n' as ddl

from cols

group by sat

order by sat

8.分区操作

自动建立的分区的命名应该是P_1 , P_2 , P_3 ...

CREATE TABLE "odm"."test_partition" (

"id" int4 NOT NULL,

"sj" timestamp(6),

"ms" varchar(255)

)

PARTITION by RANGE(sj) (

PARTITION P_MIN VALUES LESS THAN(DATE '2019-01-01')

);


-- 或者开始：

WITH (ORIENTATION = COLUMN, COMPRESSION = LOW)

DISTRIBUTE BY HASH(id)

PARTITION by RANGE(sj) (

PARTITION P START(DATE '2018-01-01') END(DATE '2040-01-01') EVERY( INTERVAL '1' MONTH)

)
-- 或者结束



-- 插入数据示例
INSERT INTO test_partition_pd("id", "sj", "ms") VALUES (1, '1970-01-01 12:12:12', '最小');

INSERT INTO test_partition_pd("id", "sj", "ms") VALUES (2, '2020-01-02 01:02:03', '一月');

INSERT INTO test_partition_pd("id", "sj", "ms") VALUES (3, '2019-12-31 09:11:33', '十二月');

INSERT INTO test_partition_pd("id", "sj", "ms") VALUES (4, '2020-06-17 09:48:12', '六月');

①.查看表分区

-- 方法一
select c.relname,p.relname as part_relname,p.parttype,p.partkey,p.relfilenode,p.boundaries,p.reloptions

from pg_partition p

left join pg_tablespace t on p.reltablespace=t.oid

left join pg_class c on p.parentid=c.relfilenode

where c.relname = '表名称（不加schema）'

order by c.relname,p.relfilenode

-- 方法二
SELECT

relname, boundaries, spcname

FROM

pg_partition p

JOIN pg_tablespace t ON p.reltablespace=t.oid and p.parentid='标schema.表名称'::regclass

ORDER BY 1;

-- 方法三
SELECT * from pg_partition where relname = '表名称（不加schema）'

②设置每月一个分区（区间）

ALTER TABLE "odm"."test_partition"

ADD PARTITION P START(DATE '2019-01-01') END(DATE '2040-01-01') EVERY( INTERVAL '1' MONTH);

③设置上界

ALTER TABLE "odm"."test_partition"

ADD PARTITION P_MAX VALUES LESS THAN(MAXVALUE);

④删除分区

ALTER TABLE "odm"."test_partition_pd" drop PARTITION P_MAX;

⑤新增分区

ALTER TABLE "odm"."test_partition_pd"

ADD PARTITION p_123 start(date '2020-05-01') end (date '2020-06-01 ');

⑥分区合并

分区合并，分割均为新分区，非已存在分区

ALTER table 表schema.表名称 merge partitions

p1, p2

into PARTITION P_MIN;

⑦分区分割

-- 分割1

ALTER table 表schema.表名称

SPLIT partition p_min at (date '2005-01-01') into ( partition P_MIN_S, partition S)

-- 分割2

ALTER table 表schema.表名称

SPLIT partition s into ( partition s START(DATE '2005-01-01') END(DATE '2014-01-01') EVERY( INTERVAL '1' MONTH))

⑧重命名分区

ALTER TABLE 表schema.表名称 RENAME PARTITION P8 TO P_9;

9.列操作

①查询列信息

SELECT

*

from

information_schema.columns

where

table_schema = '表schema' and table_name = '表名称'

-- and column_name = 'xxrksj'

②拼接所有列

SELECT string_agg(column_name, ', ') val

from information_schema.columns

where table_schema = '表schema' and table_name = '表名称'

and column_name not in ('aaa', 'bbb', 'xxx')

10.数据去重（未优化版本）

WITH ccc AS (
	WITH bbb AS (
		SELECT
			时间或其他排序字段,
			主键,
			ROW_NUMBER ( ) OVER ( PARTITION BY 主键 ) AS rn 
		FROM
			表schema.表名称 
		WHERE
			主键 IN (
				WITH aaa AS ( SELECT 主键, COUNT ( 主键 ) AS ccc FROM 表schema.表名称 GROUP BY 主键 HAVING ccc > 1 ) SELECT
				主键 
			FROM
				aaa 
			) 
		ORDER BY
			主键,
			时间或其他排序字段 DESC 
		) SELECT
		主键 || 时间或其他排序字段 
	FROM
		bbb 
	WHERE
		rn = 1 
	) SELECT *
	-- delete 
FROM
	表schema.表名称 
WHERE
	主键 || 时间或其他排序字段 IN ( SELECT * FROM ccc )

11.建立与使用序列

-- 建立：
 DROP SEQUENCE IF EXISTS "test"."p_rxzt_t_zd_dmb_id_seq";

 CREATE SEQUENCE "test"."p_rxzt_t_zd_dmb_id_seq" 
   INCREMENT 1
   MINVALUE  1
   MAXVALUE 9223372036854775807
   START 1
   CACHE 1;

--使用
 CREATE TABLE "test"."p_rxzt_t_zd_dmb" (
   "zdbh" varchar(30) COLLATE "pg_catalog"."default" NOT NULL,
   "zdmc" nvarchar2 COLLATE "pg_catalog"."default",
   "dmbh" varchar(30) COLLATE "pg_catalog"."default",
   "dmmc" nvarchar2 COLLATE "pg_catalog"."default",
   "dmms" nvarchar2 COLLATE "pg_catalog"."default",
   "dmsx" numeric(8),
   "sjdmbh" varchar(30) COLLATE "pg_catalog"."default",
   "yxbz" char(1) COLLATE "pg_catalog"."default",
   "sxrq" timestamp(0),
   "xxrksj" timestamp(3),
   "xxgxsj" timestamp(3),
   "id" int8 NOT NULL DEFAULT nextval('"test".p_rxzt_t_zd_dmb_id_seq'::regclass)
 )
 ;
 
 或者这样使用：
   select id,nextval('test.p_rxzt_t_zd_dmb_id_seq') as ccc from jwy_sdm.s_1400_t_kkjbxx limit 10

12.查看表数据在DN上的分布倾斜情况

SELECT
    a.count,b.node_name
    FROM 
        (
             SELECT COUNT(*) AS count,xc_node_id FROM table_name GROUP BY xc_node_id
        ) a,
        pgxc_node b 
WHERE
    a.xc_node_id=b.node_id
ORDER BY 
    a.count desc;