change mySQL subquery to a join for efficiency - mysql

Question

Is it possible to change the following mySQL query to use a join instead of a subquery for efficiency (or another way to increase efficiency)? I have a table with patient visits to an emergency department. The table lists arrival and departure time. I need the query to return the total number of patients that were already present in the emergency department (the "census") when the patient arrived.
My table looks something like this:
+------+------+---------------------+---------------------+
| id | name | arrival | departure |
+------+------+---------------------+---------------------+
| 1 | Joe | 2010-01-01 00:00:00 | 2010-01-01 02:00:00 |
| 2 | John | 2010-01-01 00:05:00 | 2010-01-01 03:00:00 |
| 3 | Jane | 2010-01-01 01:00:00 | 2010-01-01 04:00:00 |
...
With a desired result like this:
+------+--------+
| name | census |
+------+--------+
| Joe | 0 |
| John | 1 |
| Jane | 2 |
...
The following query works, but is quite slow (about 3.5 seconds on 180,000 rows). Is there a way to increase the efficiency of this query (with some sort of join, or other method)?
select name, arrival,
(SELECT count(*)
FROM patient_arrivals as b
WHERE b.arrival <= a.arrival and b.departure >= a.departure) as census
FROM patient_arrivals as a

Answer 1

I don't think a join will help. Instead, you need to restructure the query. The following gets the number of patients in the room at any particular time:
select t, sum(num) as num, #total := #total + num as total
from (select arrival as t, 1 as num
from patient_arrivals
union all
select departure, -1
from patient_arrivals
) t cross join
(select #total := 0) vars
group by t
order by t
Then, you can use this as a subquery for the join:
select pa.*, t.total as census
from patient_arrivals pa join
(select t, sum(num) as num, #total := #total + num as total
from (select arrival as t, 1 as num
from patient_arrivals
union all
select departure, -1
from patient_arrivals
) t cross join
(select #total := 0) vars
group by t
order by t
) tnum
on pa.arrival = tnum.t;
This gives the number when the patient arrives. For the total that overlap:
select pa.*, max(t.total) as census
from patient_arrivals pa join
(select t, sum(num) as num, #total := #total + num as total
from (select arrival as t, 1 as num
from patient_arrivals
union all
select departure, -1
from patient_arrivals
) t cross join
(select #total := 0) vars
group by t
order by t
) tnum
on tnum.t between pa.arrival and pa.departure
group by pa.id