将列中的值拆分为 BigQuery 中一行的所有列答案

【问题标题】：Splitting value in a column into all columns of a row in BigQuery将列中的值拆分为 BigQuery 中一行的所有列
【发布时间】：2017-03-14 21:47:37
【问题描述】：

我有一个包含 7 列的表格，如下所示：

date                         org   cus_id   prod_id   sales_qty   sales_amount   profit_amount
30-AUG-14 55 12 34 56 78 99  null   null      null     null           null        null
31-AUG-14 22 32 43 65 76 88  null   null      null     null           null        null

确实，第一列中的值由每一行中所有列的值连接。我想通过将第一列中的值拆分到所有列来修复它。预期的输出应该如下

date        org   cus_id   prod_id   sales_qty   sales_amount   profit_amount
 30-AUG-14   55       12        34          56             78              99  
 31-AUG-14   22       32        43          65             76              88

我认为拆分这样的字符串值是适用的，但我不熟悉拆分它并放入现有列中。我可以有你的建议吗？提前谢谢你。

【问题讨论】：

标签： sql google-bigquery

【解决方案1】：

您可以使用用户定义函数将值扩展到现有或新列。

#standardSQL
CREATE TEMPORARY FUNCTION AddField(s STRUCT<tdate STRING, org INT64,cus_id INT64,prod_id INT64,sales_qty INT64,sales_amount INT64,profit_amount INT64>)
  RETURNS STRUCT<tdate STRING, org INT64,cus_id INT64,prod_id INT64,sales_qty INT64,sales_amount INT64,profit_amount INT64> LANGUAGE js AS """
var fields = s.tdate.split(' ');
  s.org=fields[1];
  s.cus_id=fields[2];
  s.prod_id=fields[3];
  s.sales_qty=fields[4];
  s.sales_amount=fields[5];
  s.profit_amount=fields[6];
  return s;
""";
with mytable as (
select 
"30-AUG-14 55 12 34 56 78 99" as tdate,  null as org,   null as cus_id,    null as prod_id , null as sales_qty ,null as  sales_amount ,null as  profit_amount
union all
select "31-AUG-14 22 32 43 65 76 88" as tdate,  null as org,   null as cus_id,    null as prod_id , null as sales_qty ,null as  sales_amount ,null as  profit_amount
)
SELECT AddField(t).*
FROM mytable AS t;

要使用标准 SQL 将行值传递给 JavaScript 函数，请定义一个函数，该函数采用与表相同的行类型的结构。

例如：

s STRUCT<tdate STRING, org INT64,cus_id INT64,prod_id INT64,sales_qty INT64,sales_amount INT64,profit_amount INT64>

然后使用 Javascript 代码转换您现有的值

 var fields = s.tdate.split(' ');
 s.org=fields[1];

您可以添加逻辑，如果存在则不要覆盖，或创建为新列，然后对整行运行这样的查询

SELECT AddField(t).*
FROM mytable AS t;

您可以在migration guide 和UDF docs 中找到多个复杂的UDF。

【讨论】：

【解决方案2】：

下面试试

#standardSQL
SELECT 
  SPLIT(date, ' ')[OFFSET(0)] AS date,        
  SPLIT(date, ' ')[OFFSET(1)] AS org,   
  SPLIT(date, ' ')[OFFSET(2)] AS cus_id,   
  SPLIT(date, ' ')[OFFSET(3)] AS prod_id,   
  SPLIT(date, ' ')[OFFSET(4)] AS sales_qty,   
  SPLIT(date, ' ')[OFFSET(5)] AS sales_amount,   
  SPLIT(date, ' ')[OFFSET(6)] AS profit_amount  
FROM yourTable

您可以使用以下示例中的虚拟数据对其进行测试

#standardSQL
WITH yourTable AS (
  SELECT '30-AUG-14 55 12 34 56 78 99' AS date, NULL AS org, NULL AS cus_id, NULL AS prod_id, NULL AS sales_qty, NULL AS sales_amount, NULL AS profit_amount UNION ALL
  SELECT '31-AUG-14 22 32 43 65 76 88', NULL, NULL, NULL, NULL, NULL, NULL
)
SELECT 
  SPLIT(date, ' ')[OFFSET(0)] AS date,        
  SPLIT(date, ' ')[OFFSET(1)] AS org,   
  SPLIT(date, ' ')[OFFSET(2)] AS cus_id,   
  SPLIT(date, ' ')[OFFSET(3)] AS prod_id,   
  SPLIT(date, ' ')[OFFSET(4)] AS sales_qty,   
  SPLIT(date, ' ')[OFFSET(5)] AS sales_amount,   
  SPLIT(date, ' ')[OFFSET(6)] AS profit_amount  
FROM yourTable

如果您需要将字段转换为 INT - 请在下面使用

#standardSQL
SELECT 
  SPLIT(date, ' ')[OFFSET(0)] AS date,        
  CAST(SPLIT(date, ' ')[OFFSET(1)] AS INT64) AS org,   
  CAST(SPLIT(date, ' ')[OFFSET(2)] AS INT64) AS cus_id,   
  CAST(SPLIT(date, ' ')[OFFSET(3)] AS INT64) AS prod_id,   
  CAST(SPLIT(date, ' ')[OFFSET(4)] AS INT64) AS sales_qty,   
  CAST(SPLIT(date, ' ')[OFFSET(5)] AS INT64) AS sales_amount,   
  CAST(SPLIT(date, ' ')[OFFSET(6)] AS INT64) AS profit_amount  
FROM yourTable

【讨论】：