SQL：从历史记录表中获取最新条目

Question

I have 3 tables

person (id, name)
area (id, number)
history (id, person_id, area_id, type, datetime)

In this tables I store the info which person had which area at a specific time. It is like a salesman travels in an area for a while and then he gets another area. He can also have multiple areas at a time.

history type = 'I' for CheckIn or 'O' for Checkout. Example:

id    person_id    area_id    type    datetime
1     2            5          'O'     '2011-12-01' 
2     2            5          'I'     '2011-12-31' 
A person started traveling in area 5 at 2011-12-01 and gave it back on 2011-12-31.

Now I want to have a list of all the areas all persons have right now.

person1.name, area1.number, area2.number, area6.name
person2.name, area5.number, area9.number
....

The output could be like this too (it doesn't matter):

person1.name, area1.number
person1.name, area2.number
person1.name, area6.number
person2.name, area5.number
....

How can I do that?

Answer 1

This question is, indeed, quite tricky. You need a list of the entries in history where, for a given user and area, there is an 'O' record with no subsequent 'I' record. Working with just the history table, that translates to:

SELECT ho.person_id, ho.area_id, ho.type, MAX(ho.datetime)
  FROM History AS ho
 WHERE ho.type = 'O'
   AND NOT EXISTS(SELECT *
                    FROM History AS hi
                   WHERE hi.person_id = ho.person_id
                     AND hi.area_id   = ho.area_id
                     AND hi.type = 'I'
                     AND hi.datetime > ho.datetime
                 )
 GROUP BY ho.person_id, ho.area_id, ho.type;

Then, since you're really only after the person's name and the area's number (though why the area number can't be the same as its ID I am not sure), you need to adapt slightly, joining with the extra two tables:

SELECT p.name, a.number
  FROM History AS ho
  JOIN Person  AS p  ON ho.person_id = p.id
  JOIN Area    AS a  ON ho.area_id   = a.id
 WHERE ho.type = 'O'
   AND NOT EXISTS(SELECT *
                    FROM History AS hi
                   WHERE hi.person_id = ho.person_id
                     AND hi.area_id   = ho.area_id
                     AND hi.type = 'I'
                     AND hi.datetime > ho.datetime
                 );

The NOT EXISTS clause is a correlated sub-query; that tends to be inefficient. You might be able to recast it as a LEFT OUTER JOIN with appropriate join and filter conditions:

SELECT p.name, a.number
  FROM History AS ho
  JOIN Person  AS p  ON ho.person_id = p.id
  JOIN Area    AS a  ON ho.area_id   = a.id
  LEFT OUTER JOIN History AS hi
    ON hi.person_id = ho.person_id
   AND hi.area_id   = ho.area_id
   AND hi.type = 'I'
   AND hi.datetime > ho.datetime
 WHERE ho.type = 'O'
   AND hi.person_id IS NULL;

All SQL unverified.

Answer 2

You're looking for results where each row may have a different number of columns? I think you may want to look into GROUP_CONCAT()

SELECT p.`id`, GROUP_CONCAT(a.`number`, ',') AS `areas` FROM `person` a LEFT JOIN `history` h ON h.`person_id` = p.`id` LEFT JOIN `area` a ON a.`id` = h.`area_id`

I haven't tested this query, but I have used group concat in similar ways before. Naturally, you will want to tailor this to fit your needs. Of course, group concat will return a string so it will require post processing to use the data.

EDIT I thikn your question has been edited since I began responding. My query does not really fit your request anymore...

Answer 3

Try this:

select *
from person p
inner join history h on h.person_id = p.id
left outer join history h2 on h2.person_id = p.id and h2.area_id = h.area_id and h2.type = 'O'
inner join areas on a.id = h.area_id
where h2.person_id is null and h.type = 'I'