Google bigquery Bigquery:将SPLIT()输出行分解为多列
我在一列中有一个长字符串,需要将其分解为多行,然后拆分为多列。数据如下所示:((a:10,b:20,c:test1)(a:40,b:50,c:test2)(a:60,b:70,c:test3))。当我应用split和regexp_replace时,得到的结果如下 从[mydataset.mytable]中选择SPLIT(REGEXP_REPLACE(REGEXP_REPLACE(message,r'))'、''、r'(('、''))('))作为消息 输出: 味精Google bigquery Bigquery:将SPLIT()输出行分解为多列,google-bigquery,Google Bigquery,我在一列中有一个长字符串,需要将其分解为多行,然后拆分为多列。数据如下所示:((a:10,b:20,c:test1)(a:40,b:50,c:test2)(a:60,b:70,c:test3))。当我应用split和regexp_replace时,得到的结果如下 从[mydataset.mytable]中选择SPLIT(REGEXP_REPLACE(REGEXP_REPLACE(message,r'))'、''、r'(('、''))('))作为消息 输出: 味精 a:10,b:20,c:test
a:10,b:20,c:test1
a:40,b:50,c:test2
a:60,b:70,c:test3
我要找的是:
a b c
10 20测试1
40 50测试2
60 70测试3
我再次使用split按(,)分割行,但它只给我一行,而不是3行。非常感谢您的帮助。试试下面的例子
SELECT
MIN(CASE WHEN name = 'a' THEN value END) AS a,
MIN(CASE WHEN name = 'b' THEN value END) AS b,
MIN(CASE WHEN name = 'c' THEN value END) AS c
FROM (
SELECT
message, msg,
REGEXP_EXTRACT(pair, r'(\w*):') AS name,
REGEXP_EXTRACT(pair, r':(\w*)') AS value
FROM (
SELECT message, msg,
SPLIT(msg) AS pair
FROM (
SELECT message,
SPLIT(REPLACE(REPLACE(message, '))',''), '((','') ,')(') AS msg
FROM
(SELECT '((a:10,b:20,c:test1)(a:40,b:50,c:test2)(a:60,b:70,c:test3))' AS message),
(SELECT '((a:12,b:22,c:test4)(a:42,b:52,c:test5)(a:62,b:72,c:test6))' AS message),
)
)
)
GROUP BY message, msg
试试下面的例子
SELECT
MIN(CASE WHEN name = 'a' THEN value END) AS a,
MIN(CASE WHEN name = 'b' THEN value END) AS b,
MIN(CASE WHEN name = 'c' THEN value END) AS c
FROM (
SELECT
message, msg,
REGEXP_EXTRACT(pair, r'(\w*):') AS name,
REGEXP_EXTRACT(pair, r':(\w*)') AS value
FROM (
SELECT message, msg,
SPLIT(msg) AS pair
FROM (
SELECT message,
SPLIT(REPLACE(REPLACE(message, '))',''), '((','') ,')(') AS msg
FROM
(SELECT '((a:10,b:20,c:test1)(a:40,b:50,c:test2)(a:60,b:70,c:test3))' AS message),
(SELECT '((a:12,b:22,c:test4)(a:42,b:52,c:test5)(a:62,b:72,c:test6))' AS message),
)
)
)
GROUP BY message, msg
这里有一个替代解决方案,使用(取消选中“显示选项”下的“使用旧SQL”框),该解决方案仍然比较详细,但需要较少的文本操作:
WITH MyTable AS (
SELECT messages
FROM UNNEST(['((a:10,b:20,c:test1)(a:40,b:50,c:test2)(a:60,b:70,c:test3))',
'((a:12,b:22,c:test4)(a:42,b:52,c:test5)(a:62,b:72,c:test6))'])
AS messages)
SELECT
(SELECT value FROM UNNEST(message_parts) WHERE name = 'a') AS a,
(SELECT value FROM UNNEST(message_parts) WHERE name = 'b') AS b,
(SELECT value FROM UNNEST(message_parts) WHERE name = 'c') AS c
FROM (
SELECT ARRAY(SELECT AS STRUCT
SPLIT(part, ':')[OFFSET(0)] AS name,
SPLIT(part, ':')[OFFSET(1)] AS value
FROM UNNEST(SPLIT(message, ',')) AS part) AS message_parts
FROM (SELECT message FROM MyTable,
UNNEST(REGEXP_EXTRACT_ALL(messages, r'\(([^\(\)]+)\)')) AS message)
);
这里有一个替代解决方案,使用(取消选中“显示选项”下的“使用旧SQL”框),该解决方案仍然比较详细,但需要较少的文本操作:
WITH MyTable AS (
SELECT messages
FROM UNNEST(['((a:10,b:20,c:test1)(a:40,b:50,c:test2)(a:60,b:70,c:test3))',
'((a:12,b:22,c:test4)(a:42,b:52,c:test5)(a:62,b:72,c:test6))'])
AS messages)
SELECT
(SELECT value FROM UNNEST(message_parts) WHERE name = 'a') AS a,
(SELECT value FROM UNNEST(message_parts) WHERE name = 'b') AS b,
(SELECT value FROM UNNEST(message_parts) WHERE name = 'c') AS c
FROM (
SELECT ARRAY(SELECT AS STRUCT
SPLIT(part, ':')[OFFSET(0)] AS name,
SPLIT(part, ':')[OFFSET(1)] AS value
FROM UNNEST(SPLIT(message, ',')) AS part) AS message_parts
FROM (SELECT message FROM MyTable,
UNNEST(REGEXP_EXTRACT_ALL(messages, r'\(([^\(\)]+)\)')) AS message)
);
实际数据是长字符串,大约分为40行,行数不固定。实际数据是长字符串,大约分为40行,行数不固定。非常感谢Mikhail,此解决方案对我有效:)太好了,你可以考虑接受它:o)你可以在投票的下方用记号左边的记号写<代码>标记接受答案< /代码>。看看为什么它很重要!非常感谢米哈伊尔,这个解决方案对我来说很有用:“好的,你可以考虑接受它:O,你可以通过使用在投票下面的答案的左边的记号写<代码>标记接受的答案< /代码>。看看为什么它很重要!我希望BigQuery标准SQL文档能够和您最近的相关答案一样好!我真的很喜欢这些例子!谢谢你,艾略特!!是的,没问题!您似乎也很快就了解了标准SQL:)我们将继续改进文档,因为我们正在向公共测试版迈进。我在这个答案中看到的唯一问题是,初始数据没有问题!如果您重写它来处理非常初始的数据,那就太好了。我想你最终会有更沉重的额外转变。因此,在我看来,在这种情况下,标准sql“需要更少的文本操作”是适用的,或者至少不是那么明显。我觉得有更好的方法用标准sql实现。只是感觉:哦,哎呀,再试一次;我已经解决了这个问题的一个简单版本。遗留的BigQuery SQL和标准SQL解决方案都比较冗长;)非常感谢艾略特的帮助。解决方案运行良好,唯一的问题是我需要在标准SQL中运行它。我希望BigQuery标准SQL文档将和您最近的相关答案一样好!我真的很喜欢这些例子!谢谢你,艾略特!!是的,没问题!您似乎也很快就了解了标准SQL:)我们将继续改进文档,因为我们正在向公共测试版迈进。我在这个答案中看到的唯一问题是,初始数据没有问题!如果您重写它来处理非常初始的数据,那就太好了。我想你最终会有更沉重的额外转变。因此,在我看来,在这种情况下,标准sql“需要更少的文本操作”是适用的,或者至少不是那么明显。我觉得有更好的方法用标准sql实现。只是感觉:哦,哎呀,再试一次;我已经解决了这个问题的一个简单版本。遗留的BigQuery SQL和标准SQL解决方案都比较冗长;)非常感谢艾略特的帮助。解决方案运行良好,唯一的问题是我需要在标准SQL中运行它。