ClickHouse / Hive 数组交并差计算

Posted 2022-03-22 东海陈光剑

tags:

篇首语：本文由小常识网(cha138.com)小编为大家整理，主要介绍了ClickHouse / Hive 数组交并差计算相关的知识，希望对你有一定的参考价值。

Hive

Hive 计算数组交并差函数

select
    array_intersect(array(1, 2), array(2, 3)) i,
    array_union(array(1, 2), array(2, 3)) u,
    array_except(array(1, 2), array(2, 3)) e;

SQL 实例：

select
    size(t.res) as cnt
from
    (
        select
            array_intersect(
                (
                    select
                        collect_set(user_id)
                    from
                        ecom_kunlun.crm_s_u
                    where
                        date = '20220315'
                        and shop_id = 7571122
                        and pay_90d > '0'
                ),
                (
                    select
                        collect_set(user_id)
                    from
                        ecom_kunlun.crm_s_u
                    where
                        date = '20220315'
                        and shop_id = 7571122
                        and member_level in ('1', '2', '3', '4', '5', '6')
                )
            ) as res
    ) t;

ClickHouse

select
    length(
        arrayDistinct(arrayFilter(x -> x is not null, t.res))
    ) as cnt
from
    (
        select
            arrayIntersect(
                (
                    select
                        groupUniqArray(user_id)
                    from
                        ecom_dmp_shop_user.crm_s_u
                    where
                        p_date = '2022-03-15'
                        and shop_id = 7571122
                        and pay_90d > '0'
                ),
                (
                    select
                        groupUniqArray(user_id)
                    from
                        ecom_dmp_shop_user.crm_s_u
                    where
                        p_date = '2022-03-15'
                        and shop_id = 7571122
                        and member_level in ('1', '2', '3', '4', '5', '6')
                )
            ) as res
    ) t;

ClickHouse 实现数组交并差计算

计算数组交集

函数：arrayIntersect(arr1,arr2,...)
功能说明：计算 arr1,arr2等数组元素交集。
返回值：交集元素子数组。
SQL 实例：

SELECT
    arrayIntersect([1, 2, 3], [4, 5, 6]) AS noIntersect,
    arrayIntersect([1, 2, 3], [2, 3, 4, 5, 6]) AS hasIntersect
FORMAT Vertical

输出：
noIntersect: []
hasIntersect: [3,2]

计算数组并集

组合使用函数来实现arrayDistinct(arrayConcat(a, b))。
SQL 实例：

SELECT
    [1, 2] AS a,
    [2, 3] AS b,
    arrayDistinct(arrayConcat(a, b)) AS res
FORMAT Vertical

输出：
a: [1,2]
b: [2,3]
res: [1,2,3]

计算数组差集

差集的实现要有一些技术含量了(感觉 ClickHouse 后面应该内置数组差集计算函数，实现类似arrayExcept() 函数)，需要使用数组交集函数arrayIntersect() 结合高阶函数 arrayMap()和 arrayFilter()来组合实现。
SQL 实例：

SELECT
    arrayIntersect([1, 2, 3], [4, 5, 6]) AS noIntersect,
    arrayIntersect([1, 2, 3], [2, 3, 4, 5, 6]) AS hasIntersect
FORMAT Vertical
SELECT
    [1, 2] AS a,
    [2, 3] AS b,
    arrayFilter(x -> (x IS NOT NULL), arrayMap(x -> multiIf(x NOT IN arrayIntersect(a, b), x, NULL), a)) AS res
FORMAT Vertical

输出：
a: [1,2]
b: [2,3]
res: [1]

使用交(INTERSECT)、并(UNION)、差(EXCEPT)的SQL子句关键字实现

另外， ClickHouse 中有集合交(INTERSECT)、并(UNION)、差(EXCEPT)的SQL子句关键字，可以实现数组的交并差运算。实例 SQL 如下。

交集SQL：

SELECT a.i
FROM
(
    SELECT arrayJoin([1, 2]) AS i
) AS a
INTERSECT
SELECT b.i
FROM
(
    SELECT arrayJoin([2, 3]) AS i
) AS b

输出：2

并集 SQL：

SET union_default_mode = 'ALL';
SELECT DISTINCT t.i
FROM
(
    SELECT a.i
    FROM
    (
        SELECT arrayJoin([1, 2]) AS i
    ) AS a
    UNION
    SELECT b.i
    FROM
    (
        SELECT arrayJoin([2, 3]) AS i
    ) AS b
) AS t

输出：
1
2
3

差集 SQL：

SELECT a.i
FROM
(
    SELECT arrayJoin([1, 2]) AS i
) AS a
EXCEPT
SELECT b.i
FROM
(
    SELECT arrayJoin([2, 3]) AS i
) AS b

输出：1

上述[1,2]和[2,3]分别表示要求差集的数组

数据函数解释：
arrayIntersect表示多个数组求交集
arrayConcat表示连接多个数组为一个数组
arrayDistinct表示将多个数组元素去重组合为一个数组
arrayMap表示使用lambda函数转换
arrayFilter表示使用lambda函数过滤

参考资料：

https://clickhouse.tech/docs/zh/sql-reference/functions/higher-order-functions/

以上是关于ClickHouse / Hive 数组交并差计算的主要内容，如果未能解决你的问题，请参考以下文章