MySQL百分位排名

Question

I have a table containing date, id, and value, with about 1000 id rows per date. I need to calculate the percentile rank of each row, by date. I am using the following code for percentile rank for a single date, but with over 10 years of daily data this is very inefficient to run date-by-date. Seems that it should be able to be formulated in MySQL but I've not been able to make it work.

Date   ID    Value
date1  01    -7.2
date1  02     0.6
date2  01     1.2
date2  02     3.8

SELECT c.id, c.value, ROUND( (
(@rank - rank) / @rank ) *100, 2) AS rank
FROM (
SELECT * , @prev := @curr , @curr := a.value, 
@nxtRnk := @nxtRnk + 1,
@rank := IF( @prev = @curr , @rank , @nxtRnk ) AS rank
FROM (
SELECT id, value
FROM temp
WHERE date = '2013-06-28'
) AS a, (

SELECT @curr := NULL , @prev := NULL , @rank :=0, @nxtRnk :=0
) AS b
ORDER BY value DESC
) AS c

So basically I want to SELECT DISTINCT(date), and then for each date perform the above SELECT, which is preceeded by INSERT INTO table2( ... ) to write the results to table2.

Thanks for any help, Hugh

Answer 1

I finally developed an acceptable solution by using a temporary table. Maybe not the optimum solution, but it works in about 5 sec on a million + record table.

My temporary table (t1) contains date and the count of rows for date.

The third select above is changed to SELECT t1.date, t1.cnt, id, value FROM t1 LEFT JOIN temp ON(t1.date = temp.date)

Also, the calculations in the first SELECT above were changed to use c.cnt rather than @rank, and an @prevDate variable was created to reset the rank count on date changes.

Thanks to anyone who looked at this and tried to work up a solution.

Answer 2

I was trying to solve this for quite some time and then I found the following answer. Honestly brilliant. Also quite fast even for big tables (the table where I used it contained approx 5 mil records and needed a couple of seconds).

SELECT 
    CAST(SUBSTRING_INDEX(SUBSTRING_INDEX( GROUP_CONCAT(field_name ORDER BY 
    field_name SEPARATOR ','), ',', 95/100 * COUNT(*) + 1), ',', -1) AS DECIMAL) 
    AS 95th Per 
FROM table_name;

As you can imagine just replace table_name and field_name with your table's and column's names.

For further information check Roland Bouman 's original post