MySQL query to show running total per week - mysql

Question

I have a table where a close date will be updated upon close...until that time the field is null. I need to show a close trend grouped every 7 days from the start. So for example if there are 10 rows and 2 were closed in the first week then the total for that week should be 8. I have been able to create a query that works to show how many were closed in each week but I am struggling to find a way to account for the previous weeks totals.
Table
SET FOREIGN_KEY_CHECKS=0;
DROP TABLE IF EXISTS `test_data`;
CREATE TABLE `test_data` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`close_date` date DEFAULT NULL,
`location` varchar(255) DEFAULT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB AUTO_INCREMENT=23 DEFAULT CHARSET=utf8;
INSERT INTO `test_data` VALUES ('1', '2015-02-02', 'one');
INSERT INTO `test_data` VALUES ('2', '2015-02-02', 'one');
INSERT INTO `test_data` VALUES ('3', '2015-02-09', 'one');
INSERT INTO `test_data` VALUES ('4', '2015-02-09', 'one');
INSERT INTO `test_data` VALUES ('5', '2015-02-09', 'one');
INSERT INTO `test_data` VALUES ('6', '2015-02-16', 'one');
INSERT INTO `test_data` VALUES ('7', '2015-02-16', 'one');
INSERT INTO `test_data` VALUES ('8', '2015-02-16', 'one');
INSERT INTO `test_data` VALUES ('9', '2015-02-16', 'one');
INSERT INTO `test_data` VALUES ('10', '2015-02-16', 'one');
INSERT INTO `test_data` VALUES ('11', '2015-02-02', 'two');
INSERT INTO `test_data` VALUES ('12', '2015-02-02', 'two');
INSERT INTO `test_data` VALUES ('13', '2015-02-09', 'two');
INSERT INTO `test_data` VALUES ('14', '2015-02-09', 'two');
INSERT INTO `test_data` VALUES ('15', '2015-02-09', 'two');
INSERT INTO `test_data` VALUES ('16', '2015-02-16', 'two');
INSERT INTO `test_data` VALUES ('17', '2015-02-16', 'two');
INSERT INTO `test_data` VALUES ('18', '2015-02-16', 'two');
INSERT INTO `test_data` VALUES ('19', '2015-02-16', 'two');
INSERT INTO `test_data` VALUES ('20', '2015-02-16', 'two');
Query so far
select
'2015-02-02' + INTERVAL (DATEDIFF(test_data.close_date, '2015-02-02') DIV 7) WEEK as start_week,
(SELECT COUNT(*) FROM test_data WHERE location = 'one') - COUNT(a.id) AS one,
(SELECT COUNT(*) FROM test_data WHERE location = 'two') - COUNT(b.id) AS two
from test_data
left join test_data as a on a.id = test_data.id and a.location = 'one'
left join test_data as b on b.id = test_data.id and b.location = 'two'
where test_data.close_date >= '2015-02-02'
group by DATEDIFF(test_data.close_date,'2015-02-02') DIV 7
Output
start_week one two
2015-02-02 8 8
2015-02-09 7 7
2015-02-16 5 5
Output I am trying to achieve
start_week one two
2015-02-02 8 8
2015-02-09 5 5
2015-02-16 0 0
A push in the right direction is appreciated as I continue to work on this.
EDIT: More explanation of expected results.
As you can see in "Output" it is calculating the number of closed subtracted from the total. For location "one" there are 10 entries and in week 1 there were 2 closed (so 8 were still open in week 1), in week 2 3 more were closed giving the total closed since the beginning as 5 to the total in week 2 should be 5 instead.
So for my desired output.
start_week one
2015-02-02 8 <-- Total 10 - 2 closed in this week = 8
2015-02-09 5 <-- Total 10 - (3 closed in this week + the 2 in week 1) = 5
2015-02-16 0 <-- Total 10 - (5 closed in this week + the previous weeks) = 0

Answer 1

Columns a1 and a2 are the answers you are looking for. I tried this without the derived table did not get the correct answer. There is either a mysql bug or a misunderstanding on my part. But with the derived table it seems to work fine.
SELECT
sums.start_week,sums.*,
sums.tot1-sums.cur1-sums.back1 a1,sums.tot2-sums.cur1-sums.back2 a2
FROM
( /* Derived table cf counts */
SELECT DATEDIFF(test_data.close_date,'2015-02-02') DIV 7 AS sd,
'2015-02-02' + INTERVAL (DATEDIFF(test_data.close_date, '2015-02-02') DIV 7) WEEK as start_week,
(select COUNT(*) from test_data d1 where d1.location='one') tot1,
(select COUNT(*) from test_data d2 where d2.location='two') tot2,
count(CASE WHEN test_data.location='one' then 1 else null end ) cur1,
count(CASE WHEN test_data.location='two' then 1 else null end
) cur2,
(select count(*) from test_data d1 where d1.location='one' AND d1.close_date<
('2015-02-02' + INTERVAL (DATEDIFF(test_data.close_date, '2015-02-02') DIV 7) WEEK) ) as back1,
(select count(*) from test_data d2 where d2.location='two' AND d2.close_date<
('2015-02-02' + INTERVAL (DATEDIFF(test_data.close_date, '2015-02-02') DIV 7) WEEK) ) as back2
from test_data
where test_data.close_date >= '2015-02-02'
group by DATEDIFF(test_data.close_date,'2015-02-02') DIV 7 ,
'2015-02-02' + INTERVAL (DATEDIFF(test_data.close_date, '2015-02-02') DIV 7) WEEK
) sums