SQL: self JOIN top 1 from table SELECTs? (or better way of doing it…)

Question

This has probably been discussed before (or it's too simple), but I can't find an easy answer: How should this be done: select the (single) earliest future date and most recent date per id?

t1:             ==?==>         (earliest)  (most recent)
|id|date(DESC)|          |id|future_date  |   past_date   |
+==+==========+          +==+=============+===============+
|1 |  d1      |          | 1|       d1    |        d3     |
|2 |  d2      |          | 2|       d2    |        d6     |
           (<==now)      | 3|             |        d4     |
|1 |  d3      |
|3 |  d4      |
|1 |  d5      |
|2 |  d6      |

I was thinking along the lines of the following, but I have the feeling that this is rather complicated/bad syntax, and I haven't figured out how to limit to single results (i.e. most recent/nearest in future). Any suggestions?

 SELECT t_1.id,t_1.date AS future_date,t_2.date AS past_date 
    FROM (SELECT * FROM t1 WHERE t1.date>CURRENT_TIMESTAMP) t_1
    LEFT OUTER JOIN 
    (SELECT TOP 1 * FROM t1 WHERE t1.date<CURRENT_TIMESTAMP) t_2 
    ON t_1.id=t_2.id

Answer 1

I think you could use:

   SELECT x.id,
          MIN(y.date) AS future_date,
          MAX(z.date) AS past_date
     FROM (SELECT DISTINCT t.id
             FROM YOUR_TABLE t) x
LEFT JOIN YOUR_TABLE y ON y.id = x.id
                      AND y.date > CURRENT_TIMESTAMP
LEFT JOIN YOUR_TABLE z ON z.id = x.id
                      AND z.date < CURRENT_TIMESTAMP
 GROUP BY x.id

It's unclear what database you're working with - TOP is TSQL/SQL Server syntax only (2000+), while NOW as NOW() is supported by MySQL and PostgreSQL... CURRENT_TIMESTAMP is ANSI, and supported by all.

Answer 2

why wouldn't this work?

select min(date) past_date,max(DATE)future_date
from t1
HAVING MIN(date) < CURRENT_TIMESTAMP AND MAX(DATE) < CURRENT_TIMESTAMP
group by id