count existing buyer with mysql - mysql

Question

i have table like this with mysql version 5.7
CREATE TABLE order_match (
ID INT,
user_id INT,
createdAt DATE,
status_id INT,
quantity INT
);
INSERT INTO order_match VALUES
(1, 12, '2020-01-01', 4, 1),
(2, 12, '2020-01-03', 7, 1),
(3, 12, '2020-01-06', 7, 2),
(4, 13, '2020-01-02', 5, 2),
(5, 13, '2020-01-03', 6, 1),
(6, 14, '2020-03-03', 8, 0.5),
(7, 13, '2020-03-04', 4, 1),
(8, 15, '2020-04-04', 7, 3),
(9, 14, '2020-03-02', 7, 2),
(10, 14, '2020-03-10', 5, 4),
(11, 13, '2020-04-10', 8, 3),
(12, 13, '2020-04-11', 8, 2),
(13, 16, '2020-04-15', 8, 3);
select * from order_match
order by createdAt;
the output just like this
+---------+---------+------------+-----------+----------+
| ID | user_id | createdAt | status_id | quantity |
+---------+---------+------------+-----------+----------+
| 1 | 12 | 2020-01-01 | 4 | 1 |
| 4 | 13 | 2020-01-02 | 5 | 2 |
| 2 | 12 | 2020-01-03 | 7 | 1 |
| 5 | 13 | 2020-01-03 | 6 | 1 |
| 3 | 12 | 2020-01-06 | 7 | 2 |
| 9 | 14 | 2020-03-02 | 7 | 2 |
| 6 | 14 | 2020-03-03 | 8 | 1 |
| 7 | 13 | 2020-03-04 | 4 | 1 |
| 10 | 14 | 2020-03-10 | 5 | 4 |
| 8 | 15 | 2020-04-04 | 7 | 3 |
| 11 | 13 | 2020-04-10 | 8 | 3 |
| 12 | 13 | 2020-04-11 | 8 | 2 |
| 13 | 16 | 2020-04-15 | 8 | 3 |
| 13 rows | | | | |
+---------+---------+------------+-----------+----------+
with ID as the id of transaction, user_id as the buyer who doing transaction, createdAt as the date transaction happen, status_id as the status of transaction (which 4, 5, 6, 8 as the approval transaction) and quantity as the amount of quantity of every transaction
this is the fiddle
so i want to find out the statistic of how many transaction, total amount of quantity, and total frequency of unique user between 2020-03-01 until 2020-04-01, unique user is the user who doing his first approval transaction before 2020-03-01 and at least doing 1 approval transaction in between 2020-03-01 until 2020-04-01, based on the table i made the expected result just like this
+------------+------------------+-----------------+
| count user | total_order (kg) | total_order (x) |
+------------+------------------+-----------------+
| 1 | 1 | 1 |
+------------+------------------+-----------------+
explanation : as we know the user who become unique user in between 2020-03-01 until 2020-04-01 are user_id 13, because he doing his first approval transaction on 2020-01-02 (before 2020-03-01) and then doing his approval transaction at least one time on 2020-03-01 until 2020-04-01, on time range, user_id 13 (count user) doing 1 transaction (total_order (x)) and the amount are 1 kg (total_order (kg )
i've doing this syntax
select
count(distinct om.user_id) as count,
sum(om.quantity) as total_order_kg,
count(om.id) as order_x
from (select count(xx.count_) as count_
from (select count(user_id) as count_ from order_match
where status_Id in (4, 5, 6, 8)
group by user_id
) xx
) x1,
(select user_id
from order_match
group by user_id
) yy,
order_match om
where yy.user_id = om.user_id and
status_id in (4, 5, 6, 8)
and om.createdAt < '2020-03-01'
and EXISTS (select 1 from order_match om2
where om.user_id = om2.user_id
and status_id in (4, 5, 6, 8)
and om2.createdAt >= '2020-03-01'
and om2.createdAt <= '2020-04-01');
but idk why the result like this
+------------+------------------+-----------------+
| count user | total_order (kg) | total_order (x) |
+------------+------------------+-----------------+
| 1 | 3 | 2 |
+------------+------------------+-----------------+
THE FIDDLE

Answer 1

-- separate users statistic
SELECT user_id,
SUM(quantity * (createdAt >= #start)) total_order_kg,
SUM(createdAt >= #start) order_x
FROM order_match
WHERE createdAt <= #finish
GROUP BY user_id
HAVING SUM(createdAt >= #start)
AND SUM(createdAt >= #start) < COUNT(createdAt);
-- overall statistic
SELECT COUNT(*) users_count,
SUM(order_kg) total_order_kg,
SUM(order_count) total_order_count
FROM ( SELECT user_id,
SUM(quantity * (createdAt >= #start)) order_kg,
SUM(createdAt >= #start) order_count
FROM order_match
WHERE createdAt <= #finish
GROUP BY user_id
HAVING SUM(createdAt >= #start)
AND SUM(createdAt >= #start) < COUNT(createdAt) ) totals;
fiddle

Answer 2

'why the result like this' - you are using comma joins so are starting from a cartesian product you can see what is happening if you substitute the aggregations for actual values for example
select
om.user_id,
om.quantity,
om.id,
x1.count_,
yy.user_id
from (select count(xx.count_) as count_
from (select count(user_id) as count_ from t
where status_Id in (4, 5, 6, 8)
group by user_id
) xx
) x1,
(select user_id
from t
group by user_id
) yy,
t om
where yy.user_id = om.user_id and
status_id in (4, 5, 6, 8)
and om.createdAt < '2020-03-01'
and EXISTS (select 1 from t om2
where om.user_id = om2.user_id
and status_id in (4, 5, 6, 8)
and om2.createdAt >= '2020-03-01'
and om2.createdAt <= '2020-04-01');
Where t is my table name and a copy of order_match.
If you run this query without the where clause then you get 65 rows returned, if you run it with the where clause but not the exists check you get 3 rows returned if you run it in it's entirety you get
---------+----------+------+--------+---------+
| user_id | quantity | id | count_ | user_id |
+---------+----------+------+--------+---------+
| 13 | 2 | 4 | 4 | 13 |
| 13 | 1 | 5 | 4 | 13 |
+---------+----------+------+--------+---------+
2 rows in set (0.002 sec)
Which when aggregated produces the result you get from your query.
NB group by without any aggregation functions is just wrong.