Hive中的explode函数、posexplode函数与later view函数

1.概述

在离线数仓处理通过HQL业务数据时，经常会遇到行转列或者列转行之类的操作，就像concat_ws之类的函数被广泛使用，今天这个也是经常要使用的拓展方法。

2.explode函数

2.1 函数语法

sql 复制代码

-- explode(a) - separates the elements of array a into multiple rows, or the elements of a map into multiple rows and columns 
Function class:org.apache.hadoop.hive.ql.udf.generic.GenericUDTFExplode
Function type:BUILTIN
-- explode()用于array的语法如下
select explode(arraycol) as newcol from tablename;
-- explode()用于map的语法如下：
select explode(mapcol) as (keyname,valuename) from tablename;

2.2 函数说明

explode 函数是UDTF函数，将hive一列中复杂的array或者map结构拆分成多行。
Explode函数是不允许在select再有其他字段,
- explode(ARRAY) 列表中的每个元素生成一行。
- explode(MAP) map中每个key-value对，生成一行，key为一列，value为一列。

2.3 使用案例

sql 复制代码

-- explode (array)
select explode(array('A','B','C'));
select explode(array('A','B','C')) as col;
select tf.* from (select 0) t lateral view explode(array('A','B','C')) tf;
select tf.* from (select 0) t lateral view explode(array('A','B','C')) tf as col;
-- 结果
col
A
B
C
-- explode (map)
select explode(map('A',10,'B',20,'C',30));
select explode(map('A',10,'B',20,'C',30)) as (key,value);
select tf.* from (select 0) t lateral view explode(map('A',10,'B',20,'C',30)) tf;
select tf.* from (select 0) t lateral view explode(map('A',10,'B',20,'C',30)) tf as key,value;
-- 结果
key value
A	10
B	20
C	30

3.posexplode函数

2.1 函数语法

sql 复制代码

-- posexplode(a) - behaves like explode for arrays, but includes the position of items in the original array
Function class:org.apache.hadoop.hive.ql.udf.generic.GenericUDTFPosExplode
Function type:BUILTIN
select posexplode(ARRAY<T> a)
-- Explodes an array to multiple rows with additional positional column of int type (position of items in the original array, starting with 0). Returns a row-set with two columns (pos,val), one row for each element from the array.

2.2 函数说明

posexplode 函数 ，将ARRAY数组a展开，每个Value一行，每行两列分别对应数组从0开始的下标和数组元素。

2.3 使用案例

sql 复制代码

-- posexplode (array)
select posexplode(array('A','B','C'));
select posexplode(array('A','B','C')) as (pos,val);
select tf.* from (select 0) t lateral view posexplode(array('A','B','C')) tf;
select tf.* from (select 0) t lateral view posexplode(array('A','B','C')) tf as pos,val;
-- 结果
pos val
0	A
1	B
2	C

4.later view

4.1 语法

sql 复制代码

lateralView: LATERAL VIEW udtf(expression) tableAlias AS columnAlias (',' columnAlias)*
fromClause: FROM baseTable (lateralView)*
-- columnAlias是给udtf(expression)列起的别名。
-- tableAlias 虚拟表的别名。

4.2 用法描述

lateral view为侧视图,意义是为了配合UDTF来使用,把某一行数据拆分成多行数据。
不加lateral view的UDTF只能提取单个字段拆分,并不能塞回原来数据表中。
加上lateral view就可以将拆分的单个字段数据与原始表数据关联上。
lateral view函数会将UDTF生成的结果放到一个虚拟表中，然后虚拟表中的数据和输入行进行join来达到连接UDTF外的select字段的目的。(本质是笛卡尔积)

4.3 使用案例

4.3.1 准备数据

下表 pageAds. 它有两个字段: pageid (页码) and adid_list (页面上的adid):

Column name	Column type
pageid	STRING
adid_list	Array

表中数据如下：

pageid	adid_list
front_page	[1, 2, 3]
contact_page	[3, 4, 5]

需求：统计各个页面出现的广告的次数

4.3.2 代码实现

第一步： 使用 lateral view 和 explore() 函数将 adid_list 列的 list 拆分，sql代码如下：

sql 复制代码

select pageid, adid
FROM pageAds lateral view explode(adid_list) ad_view as adid;

可的如下结果

pageid	adid
front_page	1
front_page	2
front_page	3
contact_page	4
contact_page	5

第二步： 使用 count/group by 语句统计出每个adid出现的次数：

sql 复制代码

select adid,count(1) as cnt
FROM pageAds lateral view explode(adid_list) ad_view as adid
group by adid;

adid	cnt
1	1
2	1
3	2
4	1
5	1

4.4 Multiple Lateral Views

FROM子句可以有多个LATERAL VIEW子句。后面的LATERAL VIEWS子句可以引用出现在LATERAL VIEWS左侧表的任何列。

例如，如下查询：

sql 复制代码

SELECT * FROM exampleTable
LATERAL VIEW explode(col1) myTable1 AS myCol1
LATERAL VIEW explode(col2) myTable2 AS myCol2;

例如使用以下基表：

Array pageid_list	Array adid_list
[1, 2, 3]	["a", "b", "c"]
[3, 4]	["c", "d"]

单个Lateral View查询:

sql 复制代码

SELECT pageid_list, adid
FROM pageAds_1
         LATERAL VIEW explode(adid_list) adTable AS adid;
[1,2,3]	a
[1,2,3]	b
[1,2,3]	c
[4,5]	c
[4,5]	d

多个Lateral View查询：

sql 复制代码

select pageid,adid FROM pageAds_1
lateral view explode(pageid_list) adTable as pageid
lateral view explode(adid_list) adTable as adid;
1,a
1,b
1,c
2,a
2,b
2,c
3,a
3,b
3,c
3,c
3,d
4,c
4,d

4.5 later view json_tuple()

4.5.1 准备数据

sql 复制代码

create table lateral_tal_3
(
    id   int,
    col1 string,
    col2 string
);

 insert into lateral_tal_3 values(1234,'{"part1" : "61", "total" : "623", "part2" : "560", "part3" : "1", "part4" : "1"}','	{"to_part2" : "0", "to_part4" : "0", "to_up" : "0", "to_part3" : "0", "to_part34" : "0"}'),
(4567,'{"part1" : "451", "total" : "89928", "part2" : "88653", "part3" : "789", "part4" : "35"}','{"to_part2" : "54", "to_part4" : "6", "to_up" : "65", "to_part3" : "2", "to_part34" : "3"}'),
(7890,'{"part1" : "142", "total" : "351808", "part2" : "346778", "part3" : "4321", "part4" : "567"}','{"to_part2" : "76", "to_part4" : "23", "to_up" : "65", "to_part3" : "14", "to_part34" : "53"}');

id	col1	col2
1234	{"part1" : "61", "total" : "623", "part2" : "560", "part3" : "1", "part4" : "1"}	{"to_part2" : "0", "to_part4" : "0", "to_up" : "0", "to_part3" : "0", "to_part34" : "0"}
4567	{"part1" : "451", "total" : "89928", "part2" : "88653", "part3" : "789", "part4" : "35"}	{"to_part2" : "54", "to_part4" : "6", "to_up" : "65", "to_part3" : "2", "to_part34" : "3"}
7890	{"part1" : "142", "total" : "351808", "part2" : "346778", "part3" : "4321", "part4" : "567"}	{"to_part2" : "76", "to_part4" : "23", "to_up" : "65", "to_part3" : "14", "to_part34" : "53"}

需求： 解析非结构化的json数据类型

"json_tuple(jsonStr, p1, p2, ..., pn) - like get_json_object, but it takes multiple names and return a tuple. All the input parameters and output column types are string."

Function class:org.apache.hadoop.hive.ql.udf.generic.GenericUDTFJSONTuple

Function type:BUILTIN

json_tuple : 第一个参数是json 字符串所在的列名,其它参数是获取 json 字符串中的哪些key值;

4.5.2 代码实现

sql 复制代码

SELECT id,
       part1,
       part3,
       part4,
       to_part2,
       to_part3,
       to_part4,
       IF(part3 = 0, 0.0, to_part3 / part3) as ratio3,
       IF(part4 = 0, 0.0, to_part4 / part4) as ratio4
FROM lateral_tal_3
         lateral VIEW json_tuple(col1, 'part3', 'part4', 'part1') json1 AS part3, part4, part1
         lateral VIEW json_tuple(col2, 'to_part2','to_part3', 'to_part4') json2 AS to_part2, to_part3, to_part4
;

1234,61,1,1,0,0,0,0,0
4567,451,789,35,54,2,6,0.0025348542458808617,0.17142857142857143
7890,142,4321,567,76,14,23,0.0032399907428835918,0.04056437389770723

5.使用案例

需求1： 如何产生1-100的连续的数字？

sql 复制代码

--方式1： 结合space函数与split函数，posexplode函数，lateral view函数获得
select id_start + pos as id
from (
         select 1   as id_start,
                100 as id_end
     ) m lateral view posexplode(split(space(id_end - id_start), '')) t as pos, val;
  
-- 方式2：结合space函数与split函数，explode函数，lateral view函数+窗口函数获得
select row_number() over () as id
from (select split(space(99), '') as x) t
         lateral view
             explode(x) ex;
-- 方式2：结合space函数与split函数，posexplode函数，lateral view函数获取
from (select split(space(99), ' ') as x) t
         lateral view
             posexplode(x) ex as pos,val;

需求2： 获取2024-07-15至2024-07-29间所有的日期

sql 复制代码

SELECT pos,
       date_add(start_date, pos) dd
FROM (SELECT '2024-07-15' AS start_date, '2024-07-29' AS end_date) temp
         lateral VIEW
             posexplode(split(space(datediff(end_date, start_date)), '')) t
         AS pos, val;