Question

虽然我的帖子类似于this，但我仍然认为这与此不同。

我有2个CSV文件。

File A                                       File B
+-------------------------------------------------------------------+
| Name         | Country                     Name         | Country |
+-------------------------------------------------------------------+
| Ferrari      | Italy                       Jaguar       | British |
| Mercedes     | Germany                     Chevrolet    | America |
| Jaguar       | British                     Bugatti      | Italy   |
| Nissan       | Japan                       Tata         | India   |
| Chevrolet    | USA                         Nissan       | Japan   |
+-------------------------------------------------------------------+

以上内容仅供参考。一般来说，我在两个文件中都有更多的行和列，但它们的结构是相同的。

我被要求有效地对所有列进行行级别比较。因此，我建议使用HSQLDB而不是以编程方式进行 CREATE TEXT TABLE和SET SOURCE分别包含文件，然后在文件之间执行EXCEPT操作。我做了它的编码，它就像一个魅力。下面是我为实现相同目的而编写的SQL部分。

CREATE TABLE COMPARE_TABLE AS (SELECT SRC.*, 'SRC-TGT' compare_order FROM TABLEA SRC EXCEPT SELECT TGT.*, 'SRC-TGT' compare_order FROM TABLEB TGT) WITH DATA;
INSERT INTO COMPARE_TABLE SELECT TGT.*, 'TGT-SRC' compare_order FROM TABLEB TGT EXCEPT SELECT SRC.*, 'TGT-SRC' compare_order FROM TABLEA SRC;

这给了我一个结果表，如下所示（考虑上面的样本数据）：

比较表

+------------+----------+---------------+
|   Name     | Country  | Compare_order |
+------------+----------+---------------+
| Ferrari    | Italy    | SRC-TGT       |
| Mercedes   | Germany  | SRC-TGT       |
| Chevrolet  | USA      | SRC-TGT       |
| Chevrolet  | America  | TGT-SRC       |
| Bugatti    | Italy    | TGT-SRC       |
| Tata       | India    | TGT-SRC       |
+------------+----------+---------------+

从这里开始，我需要确定每一行的原因是什么不匹配？至少，我希望将其归为三类：

Source at Source
目标
值已更改（如果可能，哪些列？）

最后，我希望我的表格如下所示：

COMPARE_TABLE

+------------+----------+---------------+------------------------+
|   Name     | Country  | Compare_order |     Failure_Reason     |
+------------+----------+---------------+------------------------+
| Ferrari    | Italy    | SRC-TGT       | New at Source          |
| Mercedes   | Germany  | SRC-TGT       | New at Source          |
| Chevrolet  | USA      | SRC-TGT       | Country value mismatch |
| Chevrolet  | America  | TGT-SRC       | Country value mismatch |
| Bugatti    | Italy    | TGT-SRC       | New at Target          |
| Tata       | India    | TGT-SRC       | New at Target          |
+------------+----------+---------------+------------------------+

我该怎么做呢？我们甚至可以在SQL中执行此操作吗？

非常感谢任何帮助。

Answer 1

您可以进行基本比较：

select name, country,
       (case when sum(which = 'src') > 0 and sum(which = 'tgt') then 'DROPPED'
             when sum(which = 'src') = 0 and sum(which = 'tgt') then 'NEW'
        end) as OP
from ((select 'src' as which, name, country
       from tableA
      ) union all
      (select 'tgt', name, country
       from tableB
      )
     ) ab
group by name, country;

但这并没有给你逐列比较。这有点困难。我假设Name是唯一的，因此它可以用作密钥。以下是比较，但每个名称产生一行：

select name,
       (case when src.country is null then dest.country
             when tgt.country is null then tgt.country
             when src.country = tgt.country then dest.country
             else (src.country, '-->', tgt.country)
        end) as country,
       (case when src.country is null then 'new'
             when tgt.country is null then 'dropped'
             when src.country = tgt.country then 'same'
             else 'changed'
        end) as country,
from (select name from TableA union select name from TableB
     ) names left outer join
     TableA src
     on names.name = src.name left outer join
     TableB tgt
     on names.name = tgt.name;

当列中的值发生变化时，为每个名称获取多行似乎更加困难，尽管它也是可能的。

SQL EXCEPT：如何识别新记录VS更改记录

1 个答案: