SQL Server : remove duplicates from count() - sql

Question

I'm creating a report in a SQL Server database. I will show it's code first and then describe what it does and where is problem.
SELECT
COUNT(e.flowid) AS [count],
t.name AS [process],
CAST(DATEPART(YEAR, e.dtcr) AS VARCHAR) + '-' + CAST(RIGHT('0' + RTRIM(DATEPART(MONTH, e.dtcr)), 2) AS VARCHAR) + '-' + CAST(RIGHT('0' + RTRIM(DATEPART(DAY, e.dtcr)), 2) AS VARCHAR) AS [day]
FROM
dbo.[Event] e
JOIN
dbo.Flow f ON e.flowid = f.id
JOIN
dbo.WorkOrder o ON f.workorderno = o.number
AND o.treenodeid IN (26067, 26152, 2469, 1815, 1913) -- only from requested processes
JOIN
dbo.TreeNode t ON o.treenodeid = t.id -- for process name in select statement
JOIN
dbo.Product p ON f.productid = p.id
AND p.materialid NOT IN (26094, 27262, 27515, 27264, 28192, 28195, 26090, 26092, 26093, 27065, 26969, 27471, 28351, 28353, 28356, 28976, 27486, 29345, 29346, 27069, 28653, 28654, 26735, 26745, 28686) -- exclude unwanted family codes
WHERE
e.pass = 1 -- only passed units
AND e.treenodeid IN (9036, 9037, 9038, 9039, 12594, 26330) -- only from requested events
AND e.dtcr BETWEEN '2015-12-01 00:00:00.000' AND '2016-05-31 23:59:59.999' -- only from requested time interval
GROUP BY
DATEPART(YEAR, e.dtcr), DATEPART(MONTH, e.dtcr), DATEPART(DAY, e.dtcr), t.name
ORDER BY
[day]
What query does is count units that passed specific events in a time periods (with some filters).
Important tables are:
Event - basically log for units passing specific events.
Product - list of units.
Output is something like this:
COUNT PROCESS DAY
71 Process-1 2015-12-01
1067 Process-2 2015-12-01
8 Process-3 2015-12-01
3 Process-4 2015-12-01
15 Process-1 2015-12-02
276 Process-2 2015-12-02
47 Process-3 2015-12-02
54 Process-4 2015-12-02
It does well but there is an issue. In some specific cases unit can pass same event several times and this query counts every such passing. I need to count every unit only once.
"Duplicated" records are in Event table. They have different dates and ids. Same for all records I need to count only once is flowid. Is there any simple way to achieve this?
Thank you for your time and answers!

Answer 1

To count each flowid only once, do count(distinct flowid), i.e.
SELECT
COUNT(distinct e.flowid) AS [count],
t.name AS [process],
CAST(DATEPART(YEAR, e.dtcr) AS VARCHAR) + '-' + CAST(RIGHT('0' + RTRIM(DATEPART(MONTH, e.dtcr)), 2) AS VARCHAR) + '-' + CAST(RIGHT('0' + RTRIM(DATEPART(DAY, e.dtcr)), 2) AS VARCHAR) AS [day]
FROM
...

Answer 2

It sounds like you need the first time that something passes the threshold. You can get the first time using row_number(). This can be tricky with the additional conditions on the query. This modification might work for you:
select sum(case when seqnum = 1 then 1 else 0 end) as cnt,
. . .
from (select e.*,
row_number() over (partition by eventid order by e.dtcr) as seqnum
from event e
where e.pass = 1 and -- only passed units
e.treenodeid IN (9036, 9037, 9038, 9039, 12594, 26330) and
e.dtcr >= '2015-12-01' AND e.dtcr < '2016-06-01'
) e join
. . .
You don't specify how the same event is identified for the duplicates. The above uses eventid for this purpose.