Question

i have a table which has many fields but i want to get count of every word in any three fields of that table

find all title in a table that exist more than once…so for that i can issue this statement

SELECT title, COUNT(title) AS NumOccurrences FROM users
GROUP BY titleHAVING ( COUNT(title) > 1 )

suppose my table has three fields called title,url,description.
basically i do not know which word has been stored in which 3 fields in that table maximum time.

i want to issue a sql statement which can show me which word found maximum time…like

word-name          occurance
---------          -------
sqlserver           300
jquery              120
ajax                110

please guide me with sample sql for sql server 2000/2005 thanks

Here is my updated full code…..please have look

    IF OBJECT_ID('tempdb..#tempSearch') IS NOT NULL
BEGIN
    DROP TABLE #tempSearch
END

CREATE TABLE #tempSearch(
    ID INT,
    Title nvarchar(4000),
    Description ntext,
    Url nvarchar(4000),
    Type char(1))

INSERT INTO #tempSearch 
    SELECT * from vwProductSearch

INSERT INTO #tempSearch 
    SELECT * from vwContentSearch

    SELECT  Word, 
    COUNT(Word) AS TotalOccurrences,
    COUNT(CASE WHEN Field = 'Title' THEN Word END) AS OccurancesInTitle,
    COUNT(CASE WHEN Field = 'URL' THEN Word END) AS OccurancesInURL,
    COUNT(CASE WHEN Field = 'Description' THEN Word END) AS OccurancesInDescription
    FROM    (   SELECT  CONVERT(NTEXT, Title) AS Word, 'Title' AS Field
        FROM    #tempSearch
        UNION ALL
        SELECT  CONVERT(NTEXT, URL), 'URL' AS Field
        FROM    #tempSearch
        UNION ALL
        SELECT  CONVERT(NTEXT, Description), 'Description' AS Field
        FROM    #tempSearch
    ) As Fields
    GROUP BY Word
    HAVING  COUNT(Word) > 1

   DROP TABLE #tempSearch

Answer 1

You need to use UNION to combine your 3 fields into a single column so you can use this to group by. I’ve also added a few more counts in case you need to drill down as to where the word occurs the most.

SELECT  Word, 
        COUNT(Word) AS TotalOccurrences,
        COUNT(CASE WHEN Field = 'Title' THEN Word END) AS OccurancesInTitle,
        COUNT(CASE WHEN Field = 'URL' THEN Word END) AS OccurancesInURL,
        COUNT(CASE WHEN Field = 'Description' THEN Word END) AS OccurancesInDescription
FROM    (   SELECT  Title AS Word, 'Title' AS Field
            FROM    Users
            UNION ALL
            SELECT  URL, 'URL' AS Field
            FROM    Users
            UNION ALL
            SELECT  Description, 'Description' AS Field
            FROM    Users
        ) As Fields
GROUP BY Word
HAVING  COUNT(Word) > 1

EDIT

I know you have asked about SQL_Server 2005 and 2000, but if you were ever to upgrade to 2008 or later there is a much cleaner solution:

SELECT  Word, 
        COUNT(Word) AS TotalOccurrences,
        COUNT(CASE WHEN Field = 'Title' THEN Word END) AS OccurancesInTitle,
        COUNT(CASE WHEN Field = 'URL' THEN Word END) AS OccurancesInURL,
        COUNT(CASE WHEN Field = 'Description' THEN Word END) AS OccurancesInDescription
FROM    Users
        CROSS APPLY 
        (   VALUES 
                (Title, 'Title'), 
                (URL, 'URL'), 
                (Description, 'Description')
        ) AS T (Word, Field)
GROUP BY Word
HAVING  COUNT(Word) > 1

EDIT 2

If all your columns are different datatypes you will need to explicitly convert them:

SELECT  Word, 
        COUNT(Word) AS TotalOccurrences,
        COUNT(CASE WHEN Field = 'Title' THEN Word END) AS OccurancesInTitle,
        COUNT(CASE WHEN Field = 'URL' THEN Word END) AS OccurancesInURL,
        COUNT(CASE WHEN Field = 'Description' THEN Word END) AS OccurancesInDescription
FROM    (   SELECT  CONVERT(NTEXT, Title) AS Word, 'Title' AS Field
            FROM    Users
            UNION ALL
            SELECT  CONVERT(NTEXT, URL), 'URL' AS Field
            FROM    Users
            UNION ALL
            SELECT  CONVERT(NTEXT, Description), 'Description' AS Field
            FROM    Users
        ) As Fields
GROUP BY Word
HAVING  COUNT(Word) > 1

EDIT 3

There is no way around the error you are getting, you cannot group by NTEXT. The best solution I can come up with feels very dirty, and I’m not particularly happy with it…

SELECT  COALESCE(Title, URL, Description) AS Word,
        COALESCE(Title.Occurances, 0) + COALESCE(URL.Occurances, 0) + COALESCE(Description.Occurances, 0) AS TotalOccurances,
        COALESCE(Title.Occurances, 0) AS TitleOccurances,
        COALESCE(URL.Occurances, 0) AS URLOccurances,
        COALESCE(Description.Occurances, 0) AS DescriptionOccurances
FROM    (   SELECT  CONVERT(NTEXT, Title) AS Title, COUNT(*) AS Occurances
            FROM    #tempSearch
            GROUP BY Title
        ) AS Title
        FULL JOIN 
        (   SELECT  CONVERT(NTEXT, URL) AS URL, COUNT(*) AS Occurances
            FROM    #tempSearch
            GROUP BY URL
        ) AS URL
            ON URL LIKE Title
        FULL JOIN
        (   SELECT  Description, 1 AS Occurances
            FROM    #tempSearch
        ) AS Description
            ON Description LIKE Title

This works, but like I said, it isn’t perfect and probably won’t perform very well. Strongly consider upgrading to a later version of SQL-Server!