我们将XML数据作为名为XML的单个字符串列加载到Hadoop中.我们试图检索到数据级别进行规范化或将其分解为单行进行处理(你知道,就像一个表!)尝试过爆炸功能,但没有得到我们想要的.
<Reports>
<Report ID="1">
<Locations>
<Location ID="20001">
<LocationName>Irvine Animal Shelter</LocationName>
</Location>
<Location ID="20002">
<LocationName>Irvine City Hall</LocationName>
</Location>
</Locations>
</Report>
<Report ID="2">
<Locations>
<Location ID="10001">
<LocationName>California Fish Grill</LocationName>
</Location>
<Location ID="10002">
<LocationName>Fukada</LocationName>
</Location>
</Locations>
</Report>
</Reports>
Run Code Online (Sandbox Code Playgroud)
我们正在查询更高级别的Report.Id,然后是孩子的ID和名称(位置/位置).以下基本上给出了所有可能组合的笛卡尔积(在这个例子中,8行而不是我们希望的4行).
SELECT xpath_int(xml, '/Reports/Report/@ID') AS id, location_id, location_name
FROM xmlreports
LATERAL VIEW explode(xpath(xml, '/Reports/Report/Locations/Location/@ID')) myTable1 AS location_id
LATERAL VIEW explode(xpath(xml, '/Reports/Report/Locations/Location/LocationName/text()')) myTable2 AS location_name;
Run Code Online (Sandbox Code Playgroud)
试图分组成一个结构然后爆炸,但这会返回两行和两个数组.
SELECT id, loc.col1, loc.col2
FROM (
SELECT xpath_int(xml, '/Reports/Report/@ID') AS id,
array(struct(xpath(xml, '/Reports/Report/Locations/Location/@ID'), xpath(xml, '/Reports/Report/Locations/Location/LocationName/text()'))) As foo
FROM xmlreports) x
LATERAL VIEW explode(foo) exploded_table as loc;
Run Code Online (Sandbox Code Playgroud)
结果
1 ["20001","20002"] ["Irvine Animal Shelter","Irvine City Hall"]
2 ["10001","10002"] ["California Fish Grill","Irvine Spectrum"]
Run Code Online (Sandbox Code Playgroud)
我们想要的是什么
1 "20001" "Irvine Animal Shelter"
1 "20002" "Irvine City Hall"
2 "10001" "California Fish Grill"
2 "10002" "Irvine Spectrum"
Run Code Online (Sandbox Code Playgroud)
看起来像是一件常见的事情,但却找不到任何例子.任何帮助是极大的赞赏.
我看到有两种方法可以解决这个问题。
创建自定义 UDF,它将解析一个 XML 元素并返回您需要的数组。之后爆炸阵。
使用子选择。
我使用子选择实现了解决方案 2。即使在使用子选择时,Hive 也“足够智能”,只为此创建一个映射缩减作业,所以我认为您不会遇到性能问题。
SELECT
l2.key,
l2.rid,
l2.location_id,
location_name
FROM (
SELECT
l1.key as key,
l1.rid as rid,
location_id as location_id,
l1.xml as xml
FROM (
SELECT key, xml, rid
FROM xmlreports
LATERAL VIEW explode(xpath(xml, '/Reports/Report/@ID')) rids as rid
) l1
LATERAL VIEW explode(xpath(l1.xml, concat('/Reports/Report[@ID = ',l1.rid, ']/Locations/Location/@ID'))) locids as location_id
) l2
LATERAL VIEW explode(xpath(l2.xml, concat('/Reports/Report[@ID = ',l2.rid, ']/Locations/Location[@ID = ', l2.location_id ,' ]/LocationName/text()'))) locnames as location_name;
Run Code Online (Sandbox Code Playgroud)
对您提供的 XML 文件运行此查询后,我得到了您正在搜索的结果
1 1 20001 Irvine Animal Shelter
1 1 20002 Irvine City Hall
1 2 10001 California Fish Grill
1 2 10002 Fukada
Run Code Online (Sandbox Code Playgroud)
希望这能解决您的问题。
问候,迪诺