Question

I have an SQL query that joins few tables and I want to filter records using same LIKE condition for many columns of result set.
For example, I have columns t1.Name, t1.FullName, t1.Comment, t2.Name, t3.Description, etc in my query (t1, t2 and t3 are joined tables names) and I want to check whether t1.Name or t1.FullName or t1.Comment or t2.Name or t3.Description is LIKE ‘%sometext%’.
I’m just interested what SQL will be faster?

WHERE t1.FullName LIKE '%sometext%' OR t1.Comment LIKE '%sometext%' OR
      t2.Name LIKE '%sometext%' OR t3.Description  LIKE '%sometext%'

or

WHERE ISNULL(t1.Name,'') + '|' + ISNULL(t1.FullName,'') + '|' + ISNULL(t1.Comment,'') + '|' + ISNULL(t2.Name,'') + '|' + ISNULL(t3.Description,'') LIKE '%sometext%'

Or may be there is some even faster way?
I’m using MS SQL Server 2008 R2.

UPD: I’ve edited my second query to handle situations when some fields are NULL and when concatenation may contain pattern but fields separately don’t.

Answer 1

I would keep it written like this:

WHERE t1.FullName LIKE '%sometext%' OR t1.Comment LIKE '%sometext%' OR
      t2.Name LIKE '%sometext%' OR t3.Description  LIKE '%sometext%'

Several reasons for not using the latter, where faster doesn’t even come into play:

1. String summary statistics may be used to service the query, and it may combine different indexes on different columns under exceptional circumstances. This is not possible with the concatenated text
2. t1.FullName + .... may result in NULL – you have to handle this
3. t1.FullName + .... may result in Text being found between t1.Comment and t2.Name, e.g.
   • t1.Comment Welcome to the hub
   • t2.Name = Blenheim
   • sometext = ‘hubble’
   • .. t1.Comment + t2.Name … = … hubBlenheim matches hubble