SQL count changes in column - sql

Question

I need to count the changes in assigned group on a ticket. The problem is my log also count changes in assignee that are in the same group.
Here is some sample data
ticket_id | assigned_group | assignee | date
----------------------------------------------------
1001 | group A | john | 1-1-15
1001 | group A | michael | 1-2-15
1001 | group A | jacob | 1-3-15
1001 | group B | eddie | 1-4-15
1002 | group A | john | 1-1-15
1002 | group B | eddie | 1-2-15
1002 | group A | john | 1-3-15
1002 | group B | eddie | 1-4-15
1002 | group A | john | 1-5-15
I need this to return
ticket_id | count
--------------------
10001 | 2
10002 | 4
My query is like this
select ticket_id, assigned_group, count(*) from mytable group by ticket_id, assigned_group
But that gives me
ticket_id | count
--------------------
10001 | 4
10002 | 5
edit:
Also if I use
select ticket_id, count(Distinct assigned_group) as [Count] from mytable group by ticket_id
I only get
ticket_id | count
--------------------
10001 | 2
10002 | 2
Any advice?

Answer 1

Use Distinct Count to get the result
select ticket_id, count(Distinct assigned_group) as [Count]
from mytable
group by ticket_id

Answer 2

try this..
with temp as
(
select ticket_id, assigned_group, count(*) as count,date from mytable group by ticket_id, assigned_group,date
)
select ticket_id, count from temp

Answer 3

You can use Row_number() function to look into the next record's value.
with tbl as (select *, row_number() over(partition by ticket_id order by 1) from table)
select a.ticket_id, a.assigned_group, a.assignee_name, a.date,
count(case when a.assigned_group <> b.assigned_group then 1 else 0 end) as No_of_change
from tbl as a
left join tbl as b
on a.rn = b.rn + 1

Answer 4

If you are using SQL Server 2012, then you can use the LAG function to determine the previous assigned group easily. Then, if the previous assigned group is different from the current assigned group, you can increment the count, as below:
WITH previous_groups AS
(
SELECT
ticket_id,
assign_date,
assigned_group,
LAG(assigned_group, 1, NULL) OVER (PARTITION BY ticket_id ORDER BY assign_date) AS prev_assign_group
FROM mytable
)
SELECT
ticket_id,
SUM(CASE
WHEN assigned_group <> prev_assign_group THEN 1
ELSE 0
END) AS count
FROM previous_groups
WHERE prev_assign_group IS NOT NULL
GROUP BY ticket_id
ORDER BY ticket_id;
If you are using SQL Server 2008 or earlier versions, then you need an extra step to determine the previous assigned group, as below:
WITH previous_assign_dates AS
(
SELECT
mt1.ticket_id,
mt1.assign_date,
MAX(mt2.assign_date) AS prev_assign_date
FROM mytable mt1
LEFT JOIN mytable mt2
ON mt1.ticket_id = mt2.ticket_id
AND mt2.assign_date < mt1.assign_date
GROUP BY
mt1.ticket_id,
mt1.assign_date
),
previous_groups AS
(
SELECT
mt1.*,
mt2.assigned_group AS prev_assign_group
FROM mytable mt1
INNER JOIN previous_assign_dates pad
ON mt1.ticket_id = pad.ticket_id
AND mt1.assign_date = pad.assign_date
LEFT JOIN mytable mt2
ON pad.ticket_id = mt2.ticket_id
AND pad.prev_assign_date = mt2.assign_date
)
SELECT
ticket_id,
SUM(CASE
WHEN assigned_group <> prev_assign_group THEN 1
ELSE 0
END) AS count
FROM previous_groups
WHERE prev_assign_group IS NOT NULL
GROUP BY ticket_id
ORDER BY ticket_id;
SQL Fiddle demo
References:
The LAG function on MSDN

Answer 5

Adding an ordinal number within the ticket, then a self join where the group is different and consecutive ordinals, should work:
SELECT t1.ticket_id, COUNT(*) FROM
(SELECT *, ROW_NUMBER() OVER(PARTITION BY ticket_id ORDER BY date) ordinal
FROM mytable) t1
JOIN
(SELECT *, ROW_NUMBER() OVER(PARTITION BY ticket_id ORDER BY date) ordinal FROM nytable) t2
ON t1.ticket_id=t2.ticket_id AND t1.assigned_group<>t2.assigned_group AND t1.ordinal+1=t2.ordinal
GROUP BY t1.ticket_id