Count 2 different columns in one result set - sql

Question

I am facing a problem and getting incorrect results. I'd like to count records that have leavedates and deceaseddates for each year and display them in one table.
This is my query so far:
select year(leavedate) as leave_year, count(year(leavedate)) as cnt_leave , count(year(DeceasedDate)) as cnt_dod
from clients
where leavedate is not null and DeceasedDate is not null
group by year(leavedate),year(DeceasedDate)
order by 1 desc
This is the goal I'd like like to have:
+------+-----------+---------+
| Year | Cnt_leave | Cnt_dod |
+------+-----------+---------+
| 2018 | 542 | 5685 |
| 2017 | 8744 | 5622 |
| 2016 | 556 | 325 |
| etc | | |
+------+-----------+---------+
Should I do it in two separate select statements or is it possible to do it in one?
I'd appreciate a feedback!

Answer 1

I would do:
with
l as (
select year(leavedate) as y, count(*) as c
from clients
group by year(leavedate)
),
d as (
select year(deceaseddate) as y, count(*) as c
from clients
group by year(deceaseddate)
)
select
coalesce(l.y, d.y) as year,
coalesce(l.c, 0) as cnt_leave,
coalesce(d.c, 0) as cnt_dod
from l
full outer join d on l.y = d.y
order by coalesce(l.y, d.y) desc

Answer 2

If i understand, you are looking for something like this :
select year(leavedate) as leave_year, sum(case leavedate when null then 0 else 1 end) as cnt_leave,
sum(case DeceasedDate when null then 0 else 1 end) as cnt_dod
from clients
where leavedate is not null or DeceasedDate is not null
group by year(leavedate)
order by 1 desc

Answer 3

You'll need to count in two steps as Year(LeaveDate) is not related to Year(DeceasedDate); but you combine those SELECTs via some CTE construction.
In fact, there are (at least) two ways you can do this, not sure which one will be the better performer; you'll need to try to figure out what works best on your data.
SELECT client_id = 1, LeaveDate = Convert(datetime, '20080118'), DeceasedDate = Convert(datetime, '20090323')
INTO #test
UNION ALL SELECT client_id = 2, LeaveDate = '20100118', DeceasedDate = '20100323'
UNION ALL SELECT client_id = 3, LeaveDate = '20100118', DeceasedDate = '20120323'
;WITH src
AS (SELECT *
FROM #test
WHERE LeaveDate IS NOT NULL
AND DeceasedDate IS NOT NULL),
leave
AS (SELECT yr = Year(LeaveDate), cnt = COUNT(*) FROM src GROUP BY Year(LeaveDate)),
deceased
AS (SELECT yr = Year(DeceasedDate), cnt = COUNT(*) FROM src GROUP BY Year(DeceasedDate))
SELECT yr = ISNULL(l.yr, d.yr),
cntLeaveDate = ISNULL(l.cnt, 0),
cntDeceasedDate = ISNULL(d.cnt, 0)
FROM leave l
FULL OUTER JOIN deceased d
ON d.yr = l.yr
ORDER BY 1 DESC
;WITH src
AS (SELECT *
FROM #test
WHERE LeaveDate IS NOT NULL
AND DeceasedDate IS NOT NULL),
years
AS (SELECT DISTINCT yr = Year(LeaveDate) FROM src
UNION -- Do not UNION ALL here, you want to avoid doubles!
SELECT DISTINCT yr = Year(DeceasedDate) FROM src)
SELECT yr,
cntLeaveDate = SUM ( CASE WHEN Year(LeaveDate) = yr THEN 1 ELSE 0 END),
cntDeceasedDate = SUM ( CASE WHEN Year(DeceasedDate) = yr THEN 1 ELSE 0 END)
FROM years, #test
WHERE LeaveDate IS NOT NULL
AND DeceasedDate IS NOT NULL
GROUP BY yr
ORDER BY 1 DESC
DROP TABLE #test

Answer 4

I would use apply for this . . . I think it is the simplest solution:
select year(dte) as year, sum(isleave) as num_leaves, sum(isdeceased) as numdeceased
from clients c cross apply
(values (leavedate, 1, 0),
(deceaseddate, 0, 1)
) v(dte, isleave, isdeceased)
where dte is not null
group by year(dte)
order by 1 desc;

Answer 5

SQL FIDDLE
Based on your comments, all you need is to remove the where clause and limit the group by.
select
year(leavedate) as leave_year,
count(year(leavedate)) as cnt_leave ,
count(year(DeceasedDate)) as cnt_dod
from clients
--where leavedate is not null and DeceasedDate is not null
group by year(leavedate) --,year(DeceasedDate)
--order by 1 desc
Or, if your data could have NULL's in either column, which I assume above and thus omitted the where clause you'd want:
select
coalesce(year(leavedate),year(DeceasedDate)) as the_year,
count(year(leavedate)) as cnt_leave ,
count(year(DeceasedDate)) as cnt_dod
from clients
--where leavedate is not null and DeceasedDate is not null
group by coalesce(year(leavedate),year(DeceasedDate))
--order by 1 desc
Here's the example:
declare #clients table (leavedate date null, DeceasedDate date null)
insert into #clients
values
('20180101',NULL)
,(null,'20170101')
,('20150101','20150101')
select
year(leavedate) as leave_year,
count(year(leavedate)) as cnt_leave ,
count(year(DeceasedDate)) as cnt_dod
from #clients
--where leavedate is not null and DeceasedDate is not null
group by year(leavedate) --,year(DeceasedDate)
--order by 1 desc
select
coalesce(year(leavedate),year(DeceasedDate)) as the_year,
count(year(leavedate)) as cnt_leave ,
count(year(DeceasedDate)) as cnt_dod
from #clients
--where leavedate is not null and DeceasedDate is not null
group by coalesce(year(leavedate),year(DeceasedDate))
--order by 1 desc