Count Similar Substrings SQL query - sql

Question

I've tried a few scenarios and googled a lot, but still can't find a solution.
I have a table of user names with entries something like the below:
UserName
Cakes420
18Jack01
18Jack04
16Jack22
22Jack16
Mapple7609
Chrom44
chrom22
chrom77
013Cake
016Cake
122Cake
123Cake87
So I need a query that checks for all records that share 4 or more (in sequence) characters in the table.
So I need to return something like :
Characters
Times Used
Names Sharing
Cake
5
Cakes420, 013Cake, 016Cake, 122Cake, 123Cake87
Chro
3
Chrom44, chrom22, chrom77
or anything similar as I'd prefer not to repeat patterns, but hey, at this stage if it returns the values properly, I don't mind.
The shared characters can naturally appear in any place in the string, which is what makes this so difficult.

Answer 1

Should you do this in T-SQL? Probably not.
Can you do this in T-SQL? Yes.
Sample data
create table Names
(
Name nvarchar(20)
);
insert into Names (Name) values
('Cakes420'),
('18Jack01'),
('18Jack04'),
('16Jack22'),
('22Jack16'),
('Mapple7609'),
('Chrom44'),
('chrom22'),
('chrom77'),
('013Cake'),
('016Cake'),
('122Cake'),
('123Cake87');
Solution
Using STRING_AGG() for easy concatenation. Available from SQL Server 2017. Alternatives available for older SQL versions (use the search box on this site, there are many examples).
with rcte as
(
select n.Name,
convert(nvarchar(4), substring(n.Name, 1, 4)) as Part,
1 as PartFrom
from Names n
where len(n.Name) >= 4
union all
select r.Name,
convert(nvarchar(4), substring(r.Name, r.PartFrom+1, r.PartFrom+4)),
r.PartFrom+1
from rcte r
where len(r.Name) >= r.PartFrom+4
),
cte_count as
(
select r.Part,
count(1) as PartCount
from rcte r
where r.Part not like '%[0-9]%' -- exclude parts with numbers in them
group by r.Part
having count(1) > 1
)
select c.Part,
c.PartCount,
string_agg(r.Name, ', ') as Names
from cte_count c
join rcte r
on r.Part = c.Part
group by c.Part,
c.PartCount
order by c.Part;
Result
Part PartCount Names
---- --------- ----------------------------------------------
Cake 5 Cakes420, 123Cake87, 122Cake, 016Cake, 013Cake
Chro 3 Chrom44, chrom22, chrom77
hrom 3 chrom77, chrom22, Chrom44
Jack 4 22Jack16, 16Jack22, 18Jack04, 18Jack01
Fiddle to see it in action with the intermediate CTE results.

Answer 2

Let's use Itzik Ben-Gan's Tally Function to break out a list of substrings, then group them. This is called N-Gram, after the more common Trigram which is 3-character substrings.
I've removed one extra cross-join from the function to speed it up slightly, it's now good for up to varchar(65536):
CREATE OR ALTER FUNCTION dbo.GetNums(#num AS BIGINT)
RETURNS TABLE
AS
RETURN
WITH
L0 AS ( SELECT 1 AS c
FROM (VALUES(1),(1),(1),(1),(1),(1),(1),(1),
(1),(1),(1),(1),(1),(1),(1),(1)) AS D(c) ),
L1 AS ( SELECT 1 AS c FROM L0 AS A CROSS JOIN L0 AS B ),
L2 AS ( SELECT 1 AS c FROM L1 AS A CROSS JOIN L1 AS B ),
Nums AS ( SELECT ROW_NUMBER() OVER(ORDER BY (SELECT NULL)) AS rownum
FROM L2 )
SELECT TOP(#num)
rownum AS rn
FROM Nums
ORDER BY rownum;
GO
DECLARE #substringLen int = 4;
SELECT
Characters,
[Times Used] = COUNT(*),
[Names Sharing] = STRING_AGG(Username, ', ')
FROM (
SELECT DISTINCT
-- remove DISTINCT if you want to know about multiple in a single username
t.Username,
Characters = SUBSTRING(t.Username, n.rn, #substringLen)
FROM myTable t
CROSS APPLY dbo.GetNums (LEN(t.UserName) - #substringLen + 1) n
) t
GROUP BY t.Characters
HAVING COUNT(*) > 1