Grouping by date and combining data from 3 tables - mysql

Question

The following are the three tables I have where session.id = signup.session_id AND session.loc_id = location.id. The max override is as the name suggest override the default max capacity for the location hence IFNULL(session.max_override, location.max_cap).
mysql> SELECT * FROM session;
+----+---------------------+---------------+--------+
| id | date_time | max_override | loc_id |
+----+---------------------+---------------+--------+
| 1 | 2014-02-04 10:30:00 | 35 | 2 |
| 2 | 2014-02-04 17:00:00 | | 2 |
| 3 | 2014-02-06 11:30:00 | 50 | 2 |
| 4 | 2014-02-09 13:30:00 | | 1 |
+----+---------------------+---------------+--------+
mysql> SELECT * FROM location;
+-----------------+---------+
| id | location | max_cap |
+-----------------+---------+
| 1 | up | 20 |
| 2 | down | 103 |
| 3 | right | 50 |
| 4 | left | 50 |
+-----------------+---------+
mysql> SELECT * FROM signups;
+-----------------+------------+
| id | name | session_id |
+-----------------+------------+
| 1 | test | 3 |
| 2 | admin | 1 |
| 3 | meme | 2 |
| 4 | anna | 4 |
+-----------------+------------+
The report I am trying to create looks simple but I am not sure how to approach the problem. The following is how I would like the report/output to look like..
mysql> query ouput;
+------------+----------+-----------+----------+----------+-----------+----------+
| date | am_time | am_ses_id | am_spots | pm_time | pm_ses_id | pm_spots |
+------------+----------+-----------+----------+----------+-----------+----------+
| 2014-02-04 | 10:30 AM | 1 | 34 | 05:00 PM | 2 | 102 |
| 2014-02-06 | 11:30 AM | 3 | 49 | | | |
| 2014-02-09 | | | | 01:30 PM | 4 | 49 |
+------------+----------+-----------+----------+----------+-----------+----------+
I can group the date and time correctly and also managed to get the session_id to match since it is all within one table but to calculate the am/pm spots which is nothing but counting the records in signups table for a particular session and deducting the value from either the max_cap or max_override depending on the situation.
THIS is what I tried
Using the following query
SELECT
DATE_FORMAT(a.date_time,'%m/%d/%Y') AS ses_date,
DATE_FORMAT(a.date_time,'%r') AS ses_time,
a.id,
COUNT(b.id) as signed_up,
IFNULL(a.max_override,c.max_cap) AS cap
FROM
test.session a
LEFT JOIN
test.signups b
ON (b.session_id = a.id)
LEFT JOIN
test.location c
ON (c.id = a.loc_id)
GROUP BY b.session_id
I get the following output
+------------+----------+--------+-----------+------+
| date | ses_time | ses_id | signed_up | cap |
+------------+----------+--------+-----------+------+
| 2014-02-04 | 10:30 AM | 1 | 1 | 35 |
| 2014-02-04 | 05:00 PM | 2 | 1 | 103 |
| 2014-02-06 | 10:30 AM | 3 | 1 | 50 |
| 2014-02-09 | 10:30 AM | 4 | 1 | 50 |
+------------+----------+--------+-----------+------+
But I cannot seem to find a way to group it only by the date so the output would appear like desired! I don't know if I should union two queries either.

Answer 1

Here is a very convoluted way of doing it...
sqlfiddle: http://sqlfiddle.com/#!2/d85ca/11
select c.ses_date `date`, a.ses_time am_time, a.id am_ses_id, a.cap-a.signed_up am_spots,
b.ses_time pm_time, b.id pm_ses_id, b.cap-b.signed_up pm_spots
from (
select distinct DATE_FORMAT(a.date_time,'%m/%d/%Y') ses_date
from session a) c
left join (
SELECT
DATE_FORMAT(a.date_time,'%m/%d/%Y') AS ses_date,
DATE_FORMAT(a.date_time,'%r') AS ses_time,
a.id,
COUNT(b.id) as signed_up,
IFNULL(a.max_override,c.max_cap) AS cap
FROM
session a
LEFT JOIN
signups b
ON (b.session_id = a.id)
LEFT JOIN
location c
ON (c.id = a.loc_id)
where date_format(a.date_time, '%p') = 'AM'
GROUP BY b.session_id) a on c.ses_date = a.ses_date
left join (
SELECT
DATE_FORMAT(a.date_time,'%m/%d/%Y') AS ses_date,
DATE_FORMAT(a.date_time,'%r') AS ses_time,
a.id,
COUNT(b.id) as signed_up,
IFNULL(a.max_override,c.max_cap) AS cap
FROM
session a
LEFT JOIN
signups b
ON (b.session_id = a.id)
LEFT JOIN
location c
ON (c.id = a.loc_id)
where date_format(a.date_time, '%p') = 'PM'
GROUP BY b.session_id) b on c.ses_date = b.ses_date;

Answer 2

You need to use the JOIN operator to let the SQL DB know the relationship between the tables.
In this case also, it may be easier to do a subquery to get the count (to avoid GROUP BY). I've not separated out AM and PM by day but you could do that.
SELECT session.date_time,
IFNULL(session.max_override,location.max_cap)-(
SELECT COUNT(signups.id)
FROM signups
WHERE signups.session_id = session.id) as avail_spots
FROM session LEFT JOIN location ON session.loc_id = location.id;
Note the LEFT JOIN will include 2014-02-04 17:00:00 with a NULL avail_spots since neither max_override nor max_cap have a value, whereas INNER JOIN would not report that session at all.
fiddle
EDIT: once you have the information by day you can use it on output. Trying to pivot out the times but grouping on the date adds a lot of complexity to the query that could be solved a lot more simply by whatever program you are using for your UI.