Question

I have a table availablities which holds availabilities of users:

row_id | user_id  | available_from      | available_to
-------------------------------------------------------------
1      | 1        | 2012-02-01 08:00:00 | 2012-02-01 09:00:00
2      | 1        | 2012-02-01 09:00:00 | 2012-02-01 10:00:00
3      | 2        | 2012-02-01 08:00:00 | 2012-02-01 10:00:00
4      | 3        | 2012-02-01 07:00:00 | 2012-02-01 12:00:00

I need to get all available users from this table for a certain event.

Scenario: get all available users for an event that starts 2012-02-01 08:00:00 and ends 2012-02-01 10:00:00.

I have no problem getting user_ids (2,3):

SELECT `user_id` FROM `availablities`
WHERE `available_from` <= "2012-02-01 08:00:00"
AND `available_to` >= "2012-02-01 10:00:00"

I am having a hard time though to find a query that will also return user_id (1). For that the query must somehown combine the two rows (1) and (2) of user_id (1) because only the sum of those two availabilities fits the event.

In a perfect world the adjacent availabilities on the rows (1,2) would be in just one row. There are reasons why it is saved this ways and I can’t just “optimize” the data in the tables to combine adjacent availabilities in one row.

So what I need is a way in mysql to return user_ids (1,2,3) in one query for the given timeframe – possible or do I have to find another approach?

Answer 1

If you need to do this in one query you could try combining MySQL User Defined variables and some inline view trickery:

select  d.*
from
(
select u.user_id,u.consec as available_from,max(u.available_to) as available_to
from
(
select a.*,
case when @lastEndDate != a.available_from then @curStartDate := a.available_from else @curStartDate end as consec,
@lastEndDate := a.available_to as c
from availabilities a
inner join (select @curStartDate := null,@lastEndDate := "1970-01-01 00:00:00") as t
order by a.user_id,a.available_from
) u
group by u.user_id,u.consec
) d
where d.available_from <= "2012-02-01 08:00:00"
AND d.available_to >= "2012-02-01 10:00:00";

I took your test data on a bit and included a user (user_id = 4) who was available between 2012-02-01 08:00:00 and 2012-02-01 08:30:00 and then again between 2012-02-01 09:00:00 and 2012-02-01 10:00:00. So he was not available for the entire period between 2012-02-01 08:00:00 and 2012-02-01 10:00:00 (coffee break?!).

So the expected result is that user_id = 1 is returned for the time period but user_id =4 is not.

Here is the test I used:

drop table if exists availabilities;

create table availabilities
(row_id integer unsigned not null primary key,
user_id integer not null,
available_from datetime not null,
available_to datetime not null
);

insert into availabilities values (1,1,'2012-02-01 08:00:00','2012-02-01 09:00:00');
insert into availabilities values (2,1,'2012-02-01 09:00:00','2012-02-01 10:00:00');
insert into availabilities values (3,2,'2012-02-01 08:00:00','2012-02-01 10:00:00');
insert into availabilities values (4,3,'2012-02-01 07:00:00','2012-02-01 12:00:00');
insert into availabilities values (5,4,'2012-02-01 08:00:00','2012-02-01 08:30:00');
insert into availabilities values (6,4,'2012-02-01 09:00:00','2012-02-01 10:00:00');