Question

The following only returns surnames that are one character long.
Can I add a further condition so that it only returns Latin alphabet characters i.e. A-Z (plus a-z)

SELECT Surname
FROM WHData.dbo.vw_DimUser
WHERE 
    LEN(Surname) =1
    AND <extra condition required>
GROUP BY Surname

Answer 1

SELECT Surname
FROM WHData.dbo.vw_DimUser
WHERE 
    LEN(Surname) = 1
    AND Surname like '[a-Z]'
GROUP BY Surname

Wildcard matching

Edit after comment below:

Let’s test:

create table t1 (c1 char(1) collate Latin1_General_CS_AS
                 , c2 char(1) collate Latin1_General_CS_AS
                 , wild1 varchar(10) collate Latin1_General_CS_AS
                , wild2 varchar(10) collate Latin1_General_CS_AS)

insert into t1 values ('A', 'a', '[A-z]', '[a-Z]')

select match1 = case when c1 like wild1 then 'Matched' else 'Unmatched' end
  , match2 = case when c1 like wild2 then 'Matched' else 'Unmatched' end
  , match3 = case when c2 like wild1 then 'Matched' else 'Unmatched' end
  , match4 = case when c2 like wild2 then 'Matched' else 'Unmatched' end
from t1

You can see that both A and a are only matched correctly when [a-Z] is used as the match pattern.

SQL Fiddle for example.

This can be explained in the answer to this previous SO question. Basically, with Latin1_General_CS_AS, SQL server will sort characters like:

a
A
b
B

However, to steal from the question again, Books Online states:

In range searches, the characters included in the range may vary depending on the sorting rules of the collation.

So I guess it’s COLLATION dependent to some extent, so really a case of testing in any particular environment where something liken this is deployed.