about rand and reflect

select
    regexp_replace(reflect("java.util.UUID", "randomUUID"), "-", "") as uuid
    ,rand()      -- rand_num
    ,rand(100)   -- rand_num_seed
;

SELECT * FROM <Table_Name> DISTRIBUTE BY RAND() SORT BY RAND()  LIMIT <N rows to sample>;

CREATE TABLE lxw1234 AS SELECT * FROM lxw1 TABLESAMPLE (50 PERCENT); -- 取原表中50%的数据
CREATE TABLE lxw1234_2 AS SELECT * FROM lxw1 TABLESAMPLE (30M);      -- 取原表中30M大的数据
SELECT COUNT(1) FROM (SELECT * FROM lxw1 TABLESAMPLE (200 ROWS)) x;  -- 取原表中每个map的200行
SELECT COUNT(1) FROM lxw1 TABLESAMPLE (BUCKET 1 OUT OF 10 ON rand()); -- 将表随机分成10个桶，抽样第一个桶的数据；

CREATE TABLE lxw1_bucketed (pcid STRING) CLUSTERED BY(pcid) INTO 10 BUCKETS; -- 创建一个分桶表
INSERT overwrite TABLE lxw1_bucketed SELECT pcid FROM lxw1;                  -- 插入数据
SELECT COUNT(1) FROM lxw1_bucketed TABLESAMPLE(BUCKET 1 OUT OF 10 ON pcid);  -- 从10个桶中抽样第一个桶的数据
SELECT COUNT(1) FROM lxw1_bucketed TABLESAMPLE(BUCKET 1 OUT OF 20 ON pcid)   -- 在第一个桶中抽样一半的数据
SELECT COUNT(1) FROM lxw1 TABLESAMPLE(BUCKET 1 OUT OF 20 ON pcid);           -- 从源表中直接分桶抽样，也能达到一样的效果

-- Hive实现从表中随机抽样得到一个不重复的数据样本
select * from table_a order by rand() limit 100;
select * from (select e.*, cast(rand() * 100000 as int) as vidx from e) vt order by vt.vidx limit 100;

select
     id
    ,name
    ,age
    ,rank
from (
    select
         id
        ,name
        ,age
        ,rank
        ,row_number()over(partition by rank order by rand()) as rn
    from a
) t
where t.rn <=2
;