Question

In our case, we’re restricted to SQL Server 2000.

Say we have a table “Articles” with a field “ArticleText”. How do we search a string consisted of many words, and return the number of found occurrences.

An example of a search string: John is a nice boy (so it contains 5 words)

ID  ArticleText                         Result
1   John is going to learn              2
2   John is doing his homework nice-ly  3
3   John is a nice boy                  5

Up to now I’ve found a split function
for sql server 2000 to delimit the search string. Now I’m trying to iterate through the records and display the count of found words in one field, but I got stuck. Any help?

Answer 1

SQL Server 2000 solution

New idea. Try splitting the search string first into 1 temp table. Then joining back to the articles using wild cards.

DECLARE @Search nvarchar(200); 
SET @Search = 'John is a nice boy'; 
IF NOT OBJECT_ID('tempdb..#S') IS NULL
DROP TABLE #S; 

SELECT * 
INTO #S 
FROM dbo.Split(@Search, ' '); 

SELECT a.ID, a.ArticleText, COUNT(*) as [Result] 
FROM #S as s 
JOIN Articles as a on a.ArticleText like s.Data + ' %' 
OR a.ArticleText like '% ' + s.Data + '%' 
GROUP BY a.ID, a.ArticleText

I tested this on SQL Server 2008 R2 with 2000 compatibility mode.

The 2 join conditions ensure that it matches the first word or any beginning of words (to get ‘nice-‘ ).
You should tweak this to your requirements and consider replace functions for the hyphen and other punctuation.