从SQL中的位置表获取最新价格

Question

I have a table called location which has one to many relations with rates table. The following is a description of the rate table:

+-------------+--------------+------+-----+---------+----------------+
| Field       | Type         | Null | Key | Default | Extra          |
+-------------+--------------+------+-----+---------+----------------+
| id          | bigint(20)   | NO   | PRI | NULL    | auto_increment |
| location_id | bigint(20)   | YES  | MUL | NULL    |                |
| price       | double       | NO   |     | NULL    |                |
| start_date  | datetime     | NO   |     | NULL    |                |
+-------------+--------------+------+-----+---------+----------------+

How could I get the latest price which has most recent start_date? I tried many queries the following query is one of them, but I still can't get the correct result:

SELECT *
FROM rate r
     INNER JOIN (SELECT r2.start_date,r2.id AS id,
                        r2.location_id,
                        MIN(ABS(DATEDIFF(r2.start_date, NOW()))) AS version
                FROM rate r2
                GROUP BY r2.location_id) o2
     ON (o2.id = r.id AND r.start_date = o2.start_date);

What am I doing wrong? how can I fix it?

Answer 1

I'll take a guess at your data structure. I will guess that you have pricing for many locations. I will guess that you have both past and future pricing for each location.

So, if you want the so-called "current" price for each location, this query gets it. This will give back one row for each distinct location with the current price.

SELECT a.id, a.price, a.location, a.start_date
  FROM rate a
  JOIN (
          SELECT location, MAX(start_date) start_date
            FROM rate 
           WHERE start_date <= NOW()
           GROUP BY location
       ) b USING(location, start_date)

An index on (start_date, location, price, id) will accelerate this query nicely.

It may seem strange to have to join the subquery back to the original table. But, if you want the price and id columns, that's necessary. Why? Because the subquery is an aggregate ( GROUP BY ) query that can only return a single start_date for each location. All it can do is identify the appropriate date. The JOIN recovers the other columns from the table.

Answer 2

This question appears about once a week, just with different data. My latest answer is here .

Your query will look like this:

select  l.*, r.price
from    location l
join    rate     r
    on  r.Location_ID = l.ID
    and r.Start_Date =(
        select  Max( Start_Date )
        from    rate
        where   Location_ID = r.Location_ID );

You could allow future dates for Start_Date, for time when a rate change is scheduled ahead of time. In that case, the subquery would need a slight change to pull out the current rate which is the latest rate still in the past:

        ...
        where   Location_ID = r.Location_ID
            and Start_Date <= CurDate() );

Assuming the PK of the Rate table is (Location_ID, Start_Date), this index will provide very fast lookup of the exact row you're looking for.

Yes, I can see the PK for the Rate table is a surrogate key. Bad idea. Have you ever or do you think you will ever access the Rate table using the surrogate key value? Look at your own query. You have the Location_ID to access the Rate table -- and now the Start_Date value. Those two fields form a natural unique key and they are also the values you will have on hand when you access the table. There is your best primary key.

Answer 3

SELECT *
FROM rate r
GROUP BY r.location_id
ORDER BY r.start_date DESC

I believe this would work without a join.

You might need to wrap this, I can't remember if MySQL allows an ORDER BY after a GROUP BY. If it doesn't then this should work:

SELECT r.* FROM (SELECT * FROM rate ORDER BY start_date DESC) r GROUP BY r.location_id

Answer 4

I may be misunderstanding the question but it looks like you are trying to get the most recent start date for each location_id. If that is the case then I believe the following should work by using a subquery to get the maximum (most recent) date for each location_id and then joining it on itself to get the rest of the columns (id, price). Your query does something similar but it includes the price and id in the subquery. Since you're only looking for the most recent date by location_id the rest of the columns must be left out of the subquery and brought back in after you determine the most recent date.

SELECT 
    r.id,
    r.location_id,
    r.price,
    r.start_date
FROM (SELECT 
            location_id,
            MAX(start_date) as MaxStartDate
        FROM rate 
        GROUP BY location_id
     ) mr
JOIN rate r
    ON r.location_id = mr.location_id
    AND r.start_date = mr.MaxStartDate

Answer 5

How could I get the latest price which has most recent start_date?

Try this, if you want to fetch any of price with recent start_date

SELECT price
FROM location
ORDER BY start_date DESC
LIMIT 1

Try this, if you want to fetch all prices with recent start_date, separated with commas and sorted by their values

SELECT GROUP_CONCAT(price ORDER BY price)
FROM location
GROUP BY start_date
ORDER BY start_date DESC
LIMIT 1

It's not tested, but you may use it as foundation for futher researches