SQL Server, combining % and [] wildcards in LIKE statement

Question

In t-sql, is there a way to do pattern matching in a like statement such that you can search for 1 or more characters in a given set?

To be specific, I'm trying to come up with a LIKE statement that will return strings that begin with 1 or more letters and end in 1 or more numbers.

So these strings should match:

• abcd1234
• a1
• abcdef2
• ab123456

And these strings should not match:

• abcd
• 1234
• abcd1234abcd
• 1abcd1

I know you can use the % wildcard to match a string of 0 or more characters, and you can use brackets[] to match a single character in a given set. But is there any way to combine those so that I can match on 1 or more characters in a given set?

Something like this would be nice, but of course doesn't work:

WHERE ColumnName LIKE '[%a-z][%0-9]'

Does anyone know of a solution to this problem? Or is it just not possible in SQL Server?

Thanks, Jim

Solution 1

Like pattern matching is very limited, it does not allow for normal regular expressions.

See here for details.

For your needs use:

WHERE ColumnName LIKE '[a-z]%[0-9]'

This will match any letter followed by anything, followed by a number. SQL will enforce that the letter and number are at the two ends of the string because there is no pattern or literal character before or after our [] match set.

Solution 2

If you want to use Regex and have permissions, you can write a user defined function to give yourself access to the Regular Expressions parser on the SQL server's .NET CLR:

http://msdn.microsoft.com/en-us/magazine/cc163473.aspx (SQL 2005 and higher)

Comments

[Jim]

Yeah I've come to the realization that it can't be done without CLR. Thanks for the suggestion, but that doesn't do what I need. It will make something like "a1234bcde1" match, when it shouldn't.

[Martin Smith]

@Jim WHERE ColumnName LIKE '[a-z]%[0-9]' AND ColumnName NOT LIKE '%[0-9][a-z]%'?

[Jim]

Yes @Martin, that's actually what I was looking for and failing to come up with. I guess I needed 2 separate statements to accomplish it. Thank you.

[Martin Smith]

@Jim - Though I suppose you actually need a third to exclude strings containing characters that are neither alpha nor numeric AND ColumnName NOT LIKE '%[^0-9A-Z]%'

[Jim]

Yes @Martin, you're right with that as well, and I plan on doing that