Question

我需要一个SQL查询，它模仿表格的if-then-else语句：

if (query1 != null)
  return query1
else
  return query2

由于COALESCE不能使用结果集，我创建了一个联合查询来完成这项工作：

SELECT * FROM obs WHERE cond1  --query1
UNION
SELECT * FROM obs WHERE (NOT EXISTS(query1)) AND cond2

在SQL中：

  ( SELECT * FROM obs WHERE src = @id AND tstart <= @instant AND tend >= @instant )
  UNION
  ( SELECT * FROM obs WHERE (NOT EXISTS (SELECT 1 FROM obs WHERE src = @id AND tstart <= @instant AND tend >= @instant )) AND src = @id AND tstart <= @instant ORDER BY tend DESC LIMIT 1);

表obs有字段（src | tstart | tend | ...）。我想选择那些与@instant重叠的行。如果未找到重叠的行，则应返回@instant之前的最近一行。

SQL UNION语句有效，但它非常笨拙，我正在寻找一个更短更清晰的语句。 COALESCE（query1，query2）精神的东西会很好。我的数据库是Postgresql。

Answer 1

首先，在这种情况下，union all可能比union更合适。

其次，您可以使用with来简化查询：

WITH t1 as (
      SELECT *
      FROM obs
      WHERE src = @id AND tstart <= @instant AND tend >= @instant
     )
SELECT t1.*
FROM t1
UNION ALL
(SELECT *
 FROM obs
 WHERE NOT EXISTS (SELECT 1 FROM t1) AND
       src = @id AND tstart <= @instant
 ORDER BY tend DESC
 LIMIT 1
);

但是，如果您正在寻找单行，这更简单：

 SELECT *
 FROM obs
 WHERE src = @id
 ORDER BY (CASE WHEN  tstart <= @instant AND tend >= @instant THEN 1
                ELSE 2
           END),
          tend DESC
 LIMIT 1;

而且，如果不是单行，则也可以使用窗口函数：

SELECT o.*
FROM (SELECT o.*,
             DENSE_RANK() OVER (PARTITION BY src
                                ORDER BY (CASE WHEN tstart <= @instant AND tend >= @instant THEN 1
                                               ELSE 2
                                          END),
                                         (CASE WHEN tstart <= @instant AND tend >= @instant THEN NULL
                                               ELSE tend
                                          END) DESC
                               ) as seqnum
      FROM obs o
      WHERE src = @id
     ) o
WHERE seqnum = 1;

合并两个SQL查询

1 个答案: