加入来自同一个表的 2 个选择查询

Question

我有来自同一个表的两个 SQL 查询，如下所示。

Query1 -> select * from (select (start_date::text), (end_date::text), count("NEW") "NEW" from table_name where refnum in (select refnum from table_name where start_date in (SELECT start_date FROM table_name ORDER BY ID DESC LIMIT 1)) and "NEW">0 group by start_date, end_date order by end_date DESC limit 10) top_10 order by end_date ASC

Query2 -> select * from (select (start_date::text), (end_date::text), count("DELETED") "DELETED" from table_name where refnum in (select refnum from table_name where start_date in (SELECT start_date FROM table_name ORDER BY ID DESC LIMIT 1)) and "DELETED">0 group by start_date, end_date order by end_date DESC limit 10) top_10 order by end_date ASC

请建议一个最终查询以合并这两个子查询，从而生成一个表

我想要包含列 start_date、end_date、NEW、DELETED

的最终表格

Answer 1

您可以使用FULL JOIN，如：

select 
  coalesce(a.start_date, b.start_date) as start_date,
  coalesce(a.end_date, b.end_date) as end_date,
  a.new,
  b.deleted
from (
  -- query #1 here; exclude the ORDER BY clause
) a
full join (
  -- query #2 here; exclude the ORDER BY clause
) b on b.start_date = a.start_date and b.end_date = a.end_date
order by coalesce(a.end_date, b.end_date) ASC

Answer 2

您可以使用FULL [OUTER] JOIN 的建议。 但请注意，NULL 值上的行不匹配！如果您的结果行之一在start_date 或end_date 中有NULL，则不会合并“匹配”行。有一些解决方法，但我不打算这样做，因为我不希望它适用。此外，您的查询已经在 order by end_date DESC 中断，它首先对 NULL 值进行排序 - 可能是无意的......

无论哪种方式，如果您选择FULL JOIN，请使用USING 子句：

SELECT start_date, end_date, a.new, b.deleted
FROM        ( <"top_10" subquery of Query1 here> ) a
FULL   JOIN ( <"top_10" subquery of Query2 here> ) b USING (start_date, end_date)  -- !
ORDER  BY end_date;

The manual:

USING ( a, b, ... ) 形式的子句是ON left_table.a = right_table.a AND left_table.b = right_table.b .... 的简写 此外，USING 意味着每对等效列中只有一个 将包含在连接输出中，而不是两者都包含。

不再需要COALESCE(a.start_date, b.start_date) 等。更短、更简单、更快。

你可能甚至去NATURAL加入：

SELECT start_date, end_date, a.new, b.deleted
FROM              a
NATURAL FULL JOIN b  -- !
ORDER   BY end_date;

The manual:

NATURAL 是 USING 列表的简写，它提到了 两个具有匹配名称的表。如果没有公共列 名称，NATURAL 等价于 ON TRUE。

仅当您的子查询严格符合描述时才有意义，并且不会改变。我只见过极少数情况下建议这样做。

然而，这一切很可能是给猪涂口红。通常，您应该能够将两个查询合并为一个 - 简化（并修复）您当前拥有的查询。

向我们展示准确的表定义和准确的职位描述，我很有可能可以用更好的查询替换这句话。

Answer 3

您可以使用union 运算符来堆叠表格：

select * from (select (start_date::text), (end_date::text), count("NEW") "NEW" from table_name where refnum in (select refnum from table_name where start_date in (SELECT start_date FROM table_name ORDER BY ID DESC LIMIT 1)) and "NEW">0 group by start_date, end_date order by end_date DESC limit 10) top_10 order by end_date ASC
UNION
select * from (select (start_date::text), (end_date::text), count("DELETED") "DELETED" from table_name where refnum in (select refnum from table_name where start_date in (SELECT start_date FROM table_name ORDER BY ID DESC LIMIT 1)) and "DELETED">0 group by start_date, end_date order by end_date DESC limit 10) top_10 order by end_date ASC