Postgres Count Function for monthly transactions, counting distinct only - postgresql

Question

I am counting the size of my teams within a department. All employees have an employee ID beginning with "E" and then a designating number (i.e. "0", "1", etc) to denote which team they are on.
I have the following query in Postgres to count the size of the teams, but the problem is that with this query, I get a lot of rows that are empty, because some months are duplicated. For example, the row containing "May/2016" may be duplicated 3 times, with only 1 row containing the actual team counts.
select to_char("Date", 'Mon/YYYY') as "Date",
sum(case when l_part LIKE 'E0%%' then count end) as "ACCOUNTING",
sum(case when l_part LIKE 'E1%%' then count end) as "SW",
sum(case when l_part LIKE 'E2%%' then count end) as "SUPPORT",
sum(case when l_part LIKE 'E3%%' then count end) as "CALLCENTER",
sum(case when l_part LIKE 'E4%%' then count end) as "ADMIN",
sum(case when l_part LIKE 'E5%%' then count end) as "MARKETING",
sum(case when l_part LIKE 'E9%%' then count end) as "MANAGEMENT"
from (
select left("Type",4)as l_part, count(*),"Date" from
"Transactions" group by "Date",l_part
) p group by "Date"
order by min("Date");
If I can just get the count down to one row per month/yyyy, and order by the date that would be helpful and less confusing. Any tweaks to my attempt appreciated.
Here is what populates, as an example using September 2015:
This is what I get:
DATE | ACCOUNTING | SW | SUPPORT | CALLCENTER | ADMIN | MARKETING |
Sep/15| | | | | | |
Sep/15| | | | | | |
Sep/15| 1 | 2 | 1 | 5 | 5 | 3 |

Answer 1

I suspect your issue is the GROUP BY clause, which I think is solved using DATE_TRUNC(). Not sure if you need the where clause.
SELECT
to_char(DATE_TRUNC('month',"Date"), 'Mon/YYYY') as "Date"
, SUM(CASE WHEN left("Type",4) LIKE 'E0%%' THEN 1 END) AS "ACCOUNTING"
, SUM(CASE WHEN left("Type",4) LIKE 'E1%%' THEN 1 END) AS "SW"
, SUM(CASE WHEN left("Type",4) LIKE 'E2%%' THEN 1 END) AS "SUPPORT"
, SUM(CASE WHEN left("Type",4) LIKE 'E3%%' THEN 1 END) AS "CALLCENTER"
, SUM(CASE WHEN left("Type",4) LIKE 'E4%%' THEN 1 END) AS "ADMIN"
, SUM(CASE WHEN left("Type",4) LIKE 'E5%%' THEN 1 END) AS "MARKETING"
, SUM(CASE WHEN left("Type",4) LIKE 'E9%%' THEN 1 END) AS "MANAGEMENT"
FROM "Transactions"
WHERE "Date" IS NOT NULL
GROUP BY
DATE_TRUNC('month',"Date")
ORDER BY
DATE_TRUNC('month',"Date")
btw: Instead of SUM() an alternatve using COUNT() would be:
SELECT
to_char(DATE_TRUNC('month',"Date"), 'Mon/YYYY') as "Date"
, COUNT(CASE WHEN left("Type",4) LIKE 'E0%%' THEN 1 END) AS "ACCOUNTING"
, COUNT(CASE WHEN left("Type",4) LIKE 'E1%%' THEN 1 END) AS "SW"
, COUNT(CASE WHEN left("Type",4) LIKE 'E2%%' THEN 1 END) AS "SUPPORT"
, COUNT(CASE WHEN left("Type",4) LIKE 'E3%%' THEN 1 END) AS "CALLCENTER"
, COUNT(CASE WHEN left("Type",4) LIKE 'E4%%' THEN 1 END) AS "ADMIN"
, COUNT(CASE WHEN left("Type",4) LIKE 'E5%%' THEN 1 END) AS "MARKETING"
, COUNT(CASE WHEN left("Type",4) LIKE 'E9%%' THEN 1 END) AS "MANAGEMENT"
COUNT() increments by one for any NON-NULL value it encounters.