Define ranges to cover gaps in a number sequence (T-SQL) - sql

Question

Simplifying my problem down - I have 6-digit field which assigns numbers to customers starting from 1 and ending to 999999. Most numbers are sequentially assigned, but numbers can be assigned manually by users, and this feature has been used in an unpredicatable pattern throughout the range.
We now need to identify numbers that have not been assigned (easy) - and then convert this into a number of ranges (seems complex).
For example given the following numbers have been assigned
1,2,3,4,5,
1001,1002,1003,1004,1005,
999101,999102,999103,999104,999105
I need a resulting set of ranges like
Start End
6 1000
1006 999100
999106 999999
My thinking so far is this is probably too complex to write in queries - and best achieved by looping from 1 to 999999, and adding ranges to a temp table.
Interested to hear ideas as I can imagine there are a few approaches. I'm using SQL Server 2008 R2. This is a one-off exercise so even a non-SQL solution might be appropriate, if this were for example easily done in Excel.

Answer 1

Try this
declare #t table (num int)
insert #t values (2),(3),(6),(7),(9),(10),(11)
select
MIN(number) as rangestart,
MAX(number) as rangeend
from
(
select *,
ROW_NUMBER() over (order by number) -
ROW_NUMBER() over (order by num,number) grp
from
(
select number from master..spt_values where type='p' and number between 1 and 15
) numbers
left join #t t
on numbers.number = t.num
) v
where num is null
group by grp
Reference : gaps and islands by itzik ben-gan
To create a numbers query upto 999999
select p1.number + p2.number * 2048 as number
from
(select * from master..spt_values where type='p' ) p1,
(select * from master..spt_values where type='p' and number<489) p2
where p1.number + p2.number * 2048 <=999999

Answer 2

declare #t table (num int)
insert #t values
(2),(3),(4),(5),
(1001),(1002),(1003),(1004),(1005),
(999101),(999102),(999103),(999104),(999105)
;with cte as
(
select num,(ROW_NUMBER() OVER(ORDER BY num)) + 1 as idx from #t
union
select 0 [num],1 [idx] --start boundary
union
select 1000000 [num],COUNT(num) + 2 [idx] from #t --end boundary
)
select c1.num + 1 [Start], c2.num - 1 [End]
from cte c1
inner join cte c2 on c2.idx = c1.idx + 1
where c2.num != c1.num + 1

Answer 3

create table #temp (id int)
insert into #temp (id)
values (1),(2),(3),(1000),(1001),(1002),(2000)
--drop table #temp
with cte as
(
select *, ROW_NUMBER() over(order by id) as rn
from #temp a
)
select a.id + 1, b.id - 1
from cte a join cte b on a.rn = b.rn - 1 and a.id <> b.id -1
it wont include tail ranges, like 2001-9999

Answer 4

Here is SQLFiddle demo
select
case when max(n1)=0 then 1 else max(n1)end,
case when max(n2)=0 then 999999 else max(n2)end
from
(
select t.n+1 as n1,0 n2,
row_number() over(order by t.n)
+isnull((select 0 from t where n=1),1)
rn
from t
left join t t2 on t.n+1=t2.n
where t2.n is null
union all
select 0 n1, t.n-1 as n2 ,
row_number() over(order by t.n) rn
from t
left join t t2 on t.n-1=t2.n
where t2.n is null
and t.n>1
) t3
group by rn

Answer 5

declare #t table(id int)
insert #t values
(1),(2),(3),(4),(5),(1001),(1002),(1003),(1004),(1005),
(999101),(999102),(999103),(999104),(999105)
select t1.id+1 [start], coalesce(t3.[end], 999999) [end]
from #t t1
left join #t t2 on t1.id +1 = t2.id
cross apply
(select min(id)-1 [end] from #t where t1.id < id
) t3
where t2.id is null

Answer 6

if you have a table called "kh" for example with a column "myval" which is your list of integers you could try this SELECT.
SELECT MAX(t1.myval+1) AS 'StartRange',t3.myval-1 AS 'EndRange'
FROM kh t1, kh t3
WHERE t1.myval+1 NOT IN (SELECT myval FROM kh t2 ORDER BY myval)
AND t3.myval-1 NOT IN (SELECT myval FROM kh t4 ORDER BY myval)
AND t1.myval < t3.myval
GROUP BY t3.myval
ORDER BY StartRange