过滤所有行的特定条件，并为具有相同值的每一列仅返回一行

Question

I have house_leases and house_lease_terms (see table schemas below). A house_lease can have multiple house_lease_terms however there can only be one "active" term at a time.

Table Definitions:

CREATE TABLE `house_leases` (
  `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `house_id` int(10) unsigned NOT NULL,
  PRIMARY KEY (`id`)
);

CREATE TABLE `house_lease_terms` (
  `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `house_lease_id` int(10) unsigned NOT NULL,
  `date_start` datetime NOT NULL,
  `date_end` datetime DEFAULT NULL,
  PRIMARY KEY (`id`),
  KEY `house_lease_terms_house_lease_id_foreign` (`house_lease_id`),
  CONSTRAINT `house_lease_terms_house_lease_id_foreign` FOREIGN KEY (`house_lease_id`) REFERENCES `house_leases` (`id`) ON DELETE CASCADE
);

As you can see the house_lease_terms.house_lease_id corresponds to a specific house_lease , however there can be multiple rows with the same house_lease_id .

The rules for determining the "active" terms are:

date_start <= NOW() AND date_end > NOW() OR date_end IS NULL

If no rows are returned then the "active" terms must be in the future, so then the rules change to be:

date_start > NOW()

We order by date_start DESC if the terms are not in the future since multiple rows could be returned we want the latest date_start at the top of the results. Otherwise we sort by date_start ASC because we want the closest date_start to now to be at the top.

I then limit by 1 to get only one result and that row is considered the "active" terms. If no results come back, then there are no "active" terms.

I have a SQL statement that has this logic for getting a specific house_lease_id . That looks like this:

SELECT * FROM house_lease_terms
WHERE 
CASE 
    WHEN 
        (SELECT COUNT(*) FROM house_lease_terms WHERE date_start <= NOW() AND (date_end > NOW() OR date_end IS NULL) AND house_lease_id = 1)
    THEN 
        date_start <= NOW() AND (date_end > NOW() OR date_end IS NULL)
    ELSE
        date_start > NOW()
END
AND house_lease_id = 1
ORDER BY
    IF(
        (SELECT COUNT(*) FROM house_lease_terms WHERE date_start <= NOW() AND (date_end > NOW() OR date_end IS NULL) AND house_lease_id = 1), 
        unix_timestamp(date_start), 
        -unix_timestamp(date_start)
    ) DESC
LIMIT 1;

This statement works, but I wish there was a better way (more efficient) of fetching the "active" terms for a specific house_lease_id (If you know a better solution please share).

Now I need to have a query that will fetch the "active" terms for all the different house_lease_id 's.

I don't want any type of custom MySQL function or stored procedure to do this. I don't know where to start to create this query. I figure I can use the query from above in some sub select or join, but am not sure how I would do so.

Any help will be appreciated!

SQLFiddle with data: http://sqlfiddle.com/#!9/cab159/2/0

Answer 1

Look for:

SELECT *
FROM house_lease_terms
WHERE house_lease_id = 1
  AND (   (   (    date_start <= NOW() 
               AND date_end > NOW()
              ) 
           OR date_end IS NULL
          )
       OR (date_start > NOW()
          )
      )
ORDER BY CASE WHEN ((date_start <= NOW() AND date_end > NOW()) OR date_end IS NULL)
              THEN DATEDIFF(NOW(), date_start)
              ELSE DATEDIFF(date_start, NOW()) + 1000000
              END

---

SELECT id,
       CASE WHEN @var2=house_lease_id 
            THEN @var1:=house_lease_id 
            ELSE 1
            END row_number_in_house_lease_id,
       @var2:=house_lease_id house_lease_id,
       date_start,
       date_end,
       created_at,
       updated_at,
       deleted_at
FROM house_lease_terms, (SELECT @var1:=0, @var2:=0) variables
WHERE house_lease_id = 1
  AND (   (   (    date_start <= NOW() 
               AND date_end > NOW()
              ) 
           OR date_end IS NULL
          )
       OR (date_start > NOW()
          )
      )
ORDER BY house_lease_id,
         CASE WHEN ((date_start <= NOW() AND date_end > NOW()) OR date_end IS NULL)
              THEN DATEDIFF(NOW(), date_start)
              ELSE DATEDIFF(date_start, NOW()) + 1000000
              END

---

Next generation.

SELECT id, house_lease_id, date_start, date_end
FROM (
SELECT id,
       CASE WHEN @var2=house_lease_id 
            THEN @var1:=@var1+1 
            ELSE @var1:=1
            END row_number_in_house_lease_id,
       @var2:=house_lease_id house_lease_id,
       date_start,
       date_end
FROM house_lease_terms, (SELECT @var1:=0, @var2:=0) variables 
WHERE (   (   (    date_start <= NOW() 
               AND date_end > NOW()
              ) 
           OR date_end IS NULL
          )
       OR (date_start > NOW()
          )
      )
ORDER BY house_lease_id,
         CASE WHEN ((date_start <= NOW() AND date_end > NOW()) OR date_end IS NULL)
              THEN DATEDIFF(NOW(), date_start)
              ELSE DATEDIFF(date_start, NOW()) + 1000000
              END
) AS cte
WHERE row_number_in_house_lease_id = 1;

fiddle

Answer 2

I'm not sure this is what you would consider clean, but it seems like any approach to this is going to result in a fairly complex and messy query.

Based on your original query:

SELECT house_lease_id, 
       SUBSTRING_INDEX(GROUP_CONCAT(id ORDER BY
                                    IF(
                                        (SELECT COUNT(*) FROM house_lease_terms AS hlt2 
                                         WHERE date_start <= NOW() 
                                         AND (date_end > NOW() OR date_end IS NULL) 
                                         AND hlt2.house_lease_id = hlt.house_lease_id), 
                                         unix_timestamp(date_start), 
                                         -unix_timestamp(date_start)
                                    ) DESC),',',1) AS id,
       SUBSTRING_INDEX(GROUP_CONCAT(date_start ORDER BY
                                    IF(
                                        (SELECT COUNT(*) FROM house_lease_terms AS hlt3 
                                         WHERE date_start <= NOW() 
                                         AND (date_end > NOW() OR date_end IS NULL) 
                                         AND hlt3.house_lease_id = hlt.house_lease_id), 
                                         unix_timestamp(date_start), 
                                         -unix_timestamp(date_start)
                                    ) DESC),',',1) AS date_start,
       SUBSTRING_INDEX(GROUP_CONCAT(date_end ORDER BY
                                    IF(
                                        (SELECT COUNT(*) FROM house_lease_terms AS hlt4 
                                         WHERE date_start <= NOW() 
                                         AND (date_end > NOW() OR date_end IS NULL) 
                                         AND hlt4.house_lease_id = hlt.house_lease_id), 
                                         unix_timestamp(date_start), 
                                         -unix_timestamp(date_start)
                                    ) DESC),',',1) AS date_end 
FROM house_lease_terms AS hlt
WHERE 
CASE 
    WHEN 
        (SELECT COUNT(*) FROM house_lease_terms AS hlt5 WHERE date_start <= NOW() 
         AND (date_end > NOW() OR date_end IS NULL) 
         AND hlt5.house_lease_id = hlt.house_lease_id)
    THEN 
        date_start <= NOW() AND (date_end > NOW() OR date_end IS NULL)
    ELSE
        date_start > NOW()
END
GROUP BY hlt.house_lease_id;

That should work.