Get time duration per each available day - mysql

Question

I need to get users visits duration for each day in MySQL.
I have table like:
user_id,date,time_start, time_end
1, 2018-09-01, 09:00:00, 12:30:00
2, 2018-09-01, 13:00:00, 15:10:00
1, 2018-09-03, 09:30:00, 12:30:00
2, 2018-09-03, 13:00:00, 15:10:00
and need to get:
user_id,2018-09-01_duration,2018-09-03_duration
1,03:30:00,03:00:00
2,02:10:00,02:10:00
So columns need to be dynamic as some dates can be missed (2018-09-02).
Is it possible to do with one query without explicit joins per each day (as some days can be null)?
Update #1
Yes, I can generate columns in application side, But I still have terrible query like
SELECT user_id, d1.dt AS "2018-08-01_duration", d2.dt AS "2018-08-03_duration"...
FROM (SELECT
user_id,
time_format(TIMEDIFF(TIMEDIFF(time_out,time_in),time_norm),"%H:%i") AS dt
FROM visits
WHERE date = "2018-09-01") d1
LEFT JOIN(
SELECT
user_id,
time_format(TIMEDIFF(TIMEDIFF(time_out,time_in),time_norm),"%H:%i") AS dt
FROM visits
WHERE date = "2018-09-03") d3
ON users.id = d3.user_id...
Update #2
Yes, data like
select user_id, date, SEC_TO_TIME(SUM(TIME_TO_SEC(time_out) - TIME_TO_SEC(time_in))) as total
from visits
group by user_id, date;
is correct, but in this case data for users goes consistently. And I hope there's the way when I have rows with users and columns with dates (like in example above)

Answer 1

Try something like this:
select user_id, date, sum(time_end - time_start)
from table
group by user_id, date;
You will need to do some tweaking, as you didn't mention the RDBMS provider, but it should give you a clear idea on how to do it.

Answer 2

There's no dynamic way to use pivotting in MySQL but you might use the following for your case :
create table t(user_id int, time_start timestamp, time_end timestamp);
insert into t values(1,'2018-09-01 09:00:00', '2018-09-01 12:30:00');
insert into t values(2,'2018-09-01 13:00:00', '2018-09-01 15:10:00');
insert into t values(1,'2018-09-03 09:30:00', '2018-09-03 12:30:00');
insert into t values(2,'2018-09-03 13:00:00', '2018-09-03 15:10:00');
select min(q.user_id) as user_id,
min(CASE WHEN (q.date='2018-09-01') THEN q.time_diff END) as '2018-09-01_duration',
min(CASE WHEN (q.date='2018-09-03') THEN q.time_diff END) as '2018-09-03_duration'
from
(
select user_id, date(time_start) date,
concat(concat(lpad(hour(timediff(time_start, time_end)),2,'0'),':'),
concat(lpad(minute(timediff(time_start, time_end)),2,'0'),':'),
lpad(second(timediff(time_start, time_end)),2,'0')) as time_diff
from t
) q
group by user_id;

Answer 3

If you know the dates that you want in the result set, you don't need a dynamic query. You can just use conditional aggregation:
select user_id,
SEC_TO_TIME(SUM(CASE WHEN date = '2018-09-01' THEN TIME_TO_SEC(time_out) - TIME_TO_SEC(time_in))) as total_20180901,
SEC_TO_TIME(SUM(CASE WHEN date = '2018-09-02' THEN TIME_TO_SEC(time_out) - TIME_TO_SEC(time_in))) as total_20180902,
SEC_TO_TIME(SUM(CASE WHEN date = '2018-09-03' THEN TIME_TO_SEC(time_out) - TIME_TO_SEC(time_in))) as total_20180903
from visits
group by user_id;
You only need dynamic SQL if you don't know the dates you want in the result set. In that case, I would suggest following the same structure with the dates that you do want.

Answer 4

By the query you can solve your problem. the query is dynamic and you can improve it.
i use TSQL for the query, you can use the idea in MySQL.
declare
#columns as nvarchar(max),
#query as nvarchar(max)
select
#columns =
stuff
((
select
distinct
',' + quotename([date])
from
table_test
for xml path(''), type
).value('.', 'nvarchar(max)'), 1, 1, '')
--select #columns
set #query =
'with
cte_result
as
(
select
[user_id] ,
[date] ,
time_start ,
time_end ,
datediff(minute, time_start, time_end) as duration
from
table_test
)
select
[user_id], ' + #columns + '
from
(
select
[user_id] ,
[date] ,
duration
from
cte_result
)
sourceTable
pivot
(
sum(duration)
for [date] in (' + #columns + ')
)
pivotTable'
execute(#query)