Extract non-existent values based on previous months? - sql

Question

I HAVE tb1
code Name sal_month
==== ===== ========
101 john 02/2017
102 mathe 02/2017
103 yara 02/2017
104 sara 02/2017
101 john 03/2017
102 mathe 03/2017
103 yara 03/2017
104 sara 03/2017
101 john 04/2017
103 yara 04/2017
In February all of them received salaries as well as March
How do I extract non-existent values based on previous months?
the result should be come
code sal_month
==== =======
102 04/2017
104 04/2017
Thank in advance

Answer 1

First I created this table:
create table #T(code int, sal_month varchar(10))
insert into #T values(101,'2/2017'),(102,'2/2017'),(103,'2/2017'),(104,'2/2017'),
(101,'3/2017'),(102,'3/2017'),(104,'3/2017'),(101,'4/2017'),(103,'4/2017')
Second, I executed this query:
SELECT code, Max(sal_Month)
From #T
Where code not in (select code from #T where sal_Month = (select Max(sal_Month) from #T))
Group by code
Then I got the following results:
Note: I am using SQL SERVER 2012

Answer 2

I think you can count salary_month grouped by id, something like this, and select the rows that shows less than 3 times.
select code, count (sal_month) from tb1
group by code
having count (sal_month) < 3
After that you join with initial table (just to filter the full rows which you need) on code.
So the final query will look like his:
select code, sal_month
from tb1 a
join (select code, count (sal_month) from tb1
group by code
having code < 3) X on a.code = X.code

Answer 3

Something like this:
DECLARE #DataSource TABLE
(
[code] INT
,[sal_month] VARCHAR(12)
);
INSERT #DataSource ([code], [sal_month])
VALUES (101, '2/2017')
,(102, '2/2017')
,(103, '2/2017')
,(104, '2/2017')
,(101, '3/2017')
,(102, '3/2017')
,(104, '3/2017')
,(101, '4/2017')
,(103, '4/2017');
WITH DataSource AS
(
SELECT *
,DENSE_RANK() OVER (ORDER BY [sal_month]) AS [MonthID]
,MAX([sal_month]) OVER () AS [MaxMonth]
FROM #DataSource DS1
)
SELECT DS1.[code]
,DS1.[sal_month]
FROM DataSource DS1
LEFT JOIN DataSource DS2
ON DS1.[code] = DS2.[code]
AND DS1.[MonthID] = DS2.[MonthID] - 1
LEFT JOIN DataSource DS3
ON DS1.[code] = DS3.[code]
AND DS1.[MonthID] = DS3.[MonthID] + 1
WHERE DS2.[code] IS NULL
AND DS3.[code] IS NOT NULL
AND DS1.[sal_month] <> DS1.[MaxMonth];
Some notes:
we need a way to sort the months and it is not easy as you are storing them in very unpractical way; you are not using a date/datetime column and your string is not a valid date; also, the string you are using is not good at all, because if you have [sal_month] from different years, we will not be able to sort them; you should think about this - one alternative is to use this format:
201512
201701
201702
201711
In this way we can sort by string.
in the core I am using ROW_NUMBER and sorting months as strings;
the idea is to look for all records, that have not exists in the next month, but have a record in the previous; at the same time, excluded records which are from the last month, as it's not possible for them to have record in the next month;

Answer 4

Try this:
select tb2.code, tb2.sal_month from tb
right join (
select code, sal_month, datepart(month, sal_month) + 1 as next_sal_month from tb) as tb2
on (tb.code = tb2.code and datepart(month, tb.sal_month) = tb2.next_sal_month)
where tb2.next_sal_month < 5 and tb.sal_month is null
In the result set there's one additional record: code 103 didn't receive salary in March, but did so in February, so it is included as well.
Here's SQL fiddle, to try :)

Answer 5

In the absence of more facts about your tables, create a cartesian product of the 2 axes of month & code, then left join the stored data. Then it is easy to identify missing items when no stored data exists when compared to every possible combination.
You might already have master tables of sal_month and/or code to use, if you do use those, but if not you can dynamically create them using select distinct as seen below.
create table tbl1 (code int, sal_month varchar(10))
insert into tbl1 values(101,'2/2017'),(102,'2/2017'),(103,'2/2017'),(104,'2/2017'),
(101,'3/2017'),(102,'3/2017'),(104,'3/2017'),(101,'4/2017'),(103,'4/2017')
select c.code, m.sal_month
from ( select distinct sal_month from tbl1 ) m
cross join ( select distinct code from tbl1 ) c
Left join tbl1 t on m.sal_month = t.sal_month and c.code = t.code
Where t.sal_month IS NULL
code | sal_month
---: | :--------
103 | 3/2017
102 | 4/2017
104 | 4/2017
dbfiddle here