棘手的SQL查询涉及连续值

Question

I need to perform a relatively easy to explain but (given my somewhat limited skills) hard to write SQL query.

Assume we have a table similar to this one:

 exam_no | name | surname | result | date
---------+------+---------+--------+------------
 1       | John | Doe     | PASS   | 2012-01-01
 1       | Ryan | Smith   | FAIL   | 2012-01-02 <--
 1       | Ann  | Evans   | PASS   | 2012-01-03
 1       | Mary | Lee     | FAIL   | 2012-01-04
 ...     | ...  | ...     | ...    | ...
 2       | John | Doe     | FAIL   | 2012-02-01 <--
 2       | Ryan | Smith   | FAIL   | 2012-02-02
 2       | Ann  | Evans   | FAIL   | 2012-02-03
 2       | Mary | Lee     | PASS   | 2012-02-04
 ...     | ...  | ...     | ...    | ...
 3       | John | Doe     | FAIL   | 2012-03-01
 3       | Ryan | Smith   | FAIL   | 2012-03-02
 3       | Ann  | Evans   | PASS   | 2012-03-03
 3       | Mary | Lee     | FAIL   | 2012-03-04 <--

Note that exam_no and date aren't necessarily related as one might expect from the kind of example I chose.

Now, the query that I need to do is as follows:

• From the latest exam ( exam_no = 3) find all the students that have failed ( John Doe , Ryan Smith and Mary Lee ).
• For each of these students find the date of the first of the batch of consecutively failing exams. Another way to put it would be: for each of these students find the date of the first failing exam that comes after their last passing exam. (Look at the arrows in the table).

The resulting table should be something like this:

 name | surname | date_since_failing
------+---------+--------------------
 John | Doe     | 2012-02-01
 Ryan | Smith   | 2012-01-02
 Mary | Lee     | 2012-03-04

How can I perform such a query?

Thank you for your time.

Answer 1

You can take advantage of the fact that if someone passed the most recent exam, then they have not failed any exams since their most recent pass: therefore the problem reduces to finding the first exam failed since the most recent pass:

SELECT   name, surname, MIN(date) date_since_fail
FROM     results NATURAL LEFT JOIN (
  SELECT   name, surname, MAX(date) lastpass
  FROM     results
  WHERE    result = 'PASS'
  GROUP BY name, surname
) t
WHERE    result = 'FAIL' AND date > IFNULL(lastpass,0)
GROUP BY name, surname

See it on sqlfiddle .

Answer 2

I should use a subquery that fetch last passed exam, somthing like:

SET @query_exam_no = 3;
SELECT
 name,
 surname,
 MIN(IF(date > last_passed_exam, date, NULL)) AS date_failing_since
FROM
 exam_results
 LEFT JOIN (
  SELECT
   name,
   surname, 
   MAX(date) AS last_passed_exam 
  FROM exam_results 
  WHERE result = 'PASS' 
  GROUP BY name, surname
 ) AS last_passed_exams USING (name, surname)
HAVING
 MAX(IF(exam_no = @query_exam_no, result, NULL)) = 'FAIL'
GROUP BY name, surname

Answer 3

This is enough:

select t.name,
       t.surname,
       t.date as 'date_since_failing'
from tablename t
inner join
(
    select name,
           surname,
           max(exam_no) as exam_no
    from tablename
    group by name, surname
    having min(result) = 'FAIL'
) aux on t.name = aux.name and t.surname = aux.surname and t.exam_no = aux.exam_no

Answer 4

The condition you are asking for is good for nothing you can do it without it. Here is the working example.

select
  e.name,
  e.sur_name,
  min(e.date) as `LastFailed`
from exams as e
where e.result = 'Fail'
group by e.name
    order by e.name 

This produces this result

name            sur_name    LastFailed
Ann         Evans       2012-02-03
John        Doe         2012-02-01
Mary        Lee         2012-01-04
Ryan        Smith       2012-01-02