Percentage of Values for Top 3 from a Character Field - tsql

Question

I have an unusual situation. Please consider the following code:
IF OBJECT_ID('tempdb..#CharacterTest') IS NOT NULL
DROP TABLE #CharacterTest
CREATE TABLE #CharacterTest
(
[ID] int IDENTITY(1, 1) NOT NULL,
[CharField] varchar(50) NULL
)
INSERT INTO #CharacterTest (CharField)
VALUES ('A')
, ('A')
, ('A')
, ('A')
, ('B')
, ('B')
, ('B')
, ('C')
, ('C')
, ('D')
, ('D')
, ('F')
, ('G')
, ('H')
, ('I')
, ('J')
, ('K')
, ('L')
, ('M')
, ('N')
, (' ')
, (' ')
, (' ')
, (NULL)
, ('');
I would like a query which gives me a character string like this:
A (16%), B (12%), C(8%)
Please notice the following:
I don't want to have empty strings, strings with all blanks, or nulls listed in the top 3, but I do want the percentage of values calculated using the entire record count for the table.
Ties can be ignored, so if there were 22 values in the list with 8% frequency, it's alright to simply return whichever one is first.
Percentages can be rounded to whole numbers.
I'd like to find the easiest way to write this query while still retaining T-SQL compatibility back to SQL Server 2005. What is the best way to do this? Window Functions?

Answer 1

I'd go for.
WITH T1
AS (SELECT [CharField],
100.0 * COUNT(*) OVER (PARTITION BY [CharField]) /
COUNT(*) OVER () AS Pct
FROM #CharacterTest),
T2
AS (SELECT DISTINCT TOP 3 *
FROM T1
WHERE [CharField] <> '' --Excludes all blank or NULL as well
ORDER BY Pct DESC)
SELECT STUFF((SELECT ',' + [CharField] + ' (' + CAST(CAST(ROUND(Pct,1) AS INT) AS VARCHAR(3)) + ')'
FROM T2
ORDER BY Pct DESC
FOR XML PATH('')), 1, 1, '') AS Result

Answer 2

My first attempt would probably be this. Not saying that it's the best way to handle it, but that it would work.
DECLARE #TotalCount INT
SELECT #TotalCount = COUNT(*) FROM #CharacterTest AS ct
SELECT TOP(3) CharField, COUNT(*) * 1.0 / #TotalCount AS OverallPercentage
FROM #CharacterTest AS ct
WHERE CharField IS NOT NULL AND REPLACE(CharField, ' ', '') <> ''
GROUP BY CharField
ORDER BY COUNT(*) desc
DROP TABLE #CharacterTest

Answer 3

This should get the character string you need:
declare #output varchar(200);
with cte as (
select CharField
, (count(*) * 100) / (select count(*) from #CharacterTest) as CharPct
, row_number() over (order by count(*) desc, CharField) as RowNum
from #CharacterTest
where replace(CharField, ' ', '') not like ''
group by CharField
)
select #output = coalesce(#output + ', ', '') + CharField + ' (' + cast(CharPct as varchar(11)) + '%)'
from cte
where RowNum <= 3
order by RowNum;
select #output;
-- Returns:
-- A (16%), B (12%), C (8%)
I would draw attention to storing a single character in a varchar(50) column, however.