Can I select distinct/group by characters in a field in SQL? - sql

Question

I'd like to run a query which for a given field will count the instances of a particular character.
For example if I had a table called 'Friends' with a 'Names' field containing rows: Edward, James, Mike. I'd get the output:
A 2,
D 2,
E 3,
I 1,
and so on...

Answer 1

The generic answer is that you need to split each name into a table of constituent characters, then count those.
You don't mention which RDBMS you're using. The answer varies a bit by database engine.
For example, this will work on SQL 2005+:
DECLARE #friends TABLE (NAMES VARCHAR(30))
DECLARE #maxLen INT
INSERT #friends (NAMES)
SELECT 'Edward'
UNION SELECT 'James'
UNION SELECT 'Mike'
SELECT #maxLen = MAX(LEN(NAMES)) FROM #friends
;WITH numsCte
AS
(
--dynamic numbers table. If you have a numbers table in your database
--use that instead, as it will be more efficient.
SELECT 1 AS n
UNION ALL
SELECT n+1 FROM numsCte
WHERE n < #maxLen
)
,charCTE
AS
(
--split the string into a dataset
SELECT *
FROM numsCte AS nm
CROSS APPLY (SELECT NAMES, SUBSTRING(NAMES, n, 1) AS splitChar
FROM #friends
) AS st
WHERE splitChar > ''
)
SELECT UPPER(splitChar) AS letter
,COUNT(1) AS cnt
FROM charCTE
GROUP BY splitChar
ORDER BY splitChar
But almost certainly won't work on any other database engines.

Answer 2

One way is to use a temporary table, and populate it in a WHILE loop:
declare #letters table (letter varchar(1))
declare #pos int
set #pos = 1
while 1=1
begin
insert into #letters
select substring(name,#pos,1)
from #names
where len(name) >= #pos
if ##rowcount = 0
break
set #pos = #pos + 1
end
select letter, count(*)
from #letters
group by letter
Another way is to create a list of valid character positions in a temporary table, or as in this example, with a recursive common table expression (CTE):
declare #maxLen int
select #maxLen = max(len(name)) from #names
;WITH CharPositions (i) AS (
select 1
union all
select i+1
from CharPositions
where i < #maxLen
)
select substring(n.name,cp.i,1), count(*)
from #names n
inner join CharPositions cp on cp.i <= len(n.name)
group by substring(n.name,cp.i,1)
I've tested the code samples against this dataset:
declare #names table (name varchar(max))
insert into #names values ('abc')
insert into #names values ('def')
insert into #names values ('def')
insert into #names values ('g')
insert into #names values ('g')
insert into #names values ('g')