Question

I’ve already solved this issue using a CASE statement, but I’m wondering why, when using ISNULL, can I concatenate a value to the check_expression but not the replacement_value. Here’s an example:

create table name_test 
(
    nm          varchar(25)
    ,mid_init   varchar(1)
    ,name_type  int
)

insert into name_test values('Joe',NULL,1)
insert into name_test values('Joe','X',2)

I have this query.

SELECT
    n1.nm, isnull(n2.mid_init+'b',n1.mid_init) as m1,
    isnull(n1.mid_init,n2.mid_init+'b') as m2
FROM 
    name_test n1
    JOIN name_test n2 
    on n1.nm = n2.nm and n1.name_type = 1 and n2.name_type = 2

In the first statement (m1), I use the NOT-NULL value first, concatenated with a ‘b’. In the second statement (m2), I use the NULL value first, so the query should pass through to the replacement_value. It does, but only takes the first value, not the concatenated value.

Result:

nm     m1   m2
Joe    Xb   X

With a little more investigation, I see that this query:

select 
    n1.nm, isnull(n2.mid_init+'b',n1.mid_init) as m1,
    isnull(n1.mid_init,'1'+ n2.mid_init+'b') as m2
from 
    name_test n1
    join name_test n2 
    on n1.nm = n2.nm and n1.name_type = 1 and n2.name_type = 2

returns

nm  m1  m2
Joe Xb  1

So, the ISNULL function will take the entire check_expression, but if the check_expression is null, it will only take the first part of the replacement_value.

Does anyone know why there is a discrepancy. Also, I solved it with a case – is there a better solution?

Answer 1

ISNULL inherits the data type of the first expression, while COALESCE inherits the data type according to data type precedence. (What is the point of varchar(1) anyway?)

I would do:

select n1.nm, 
COALESCE(CONVERT(VARCHAR(25), n1.mid_init),'1'+ n2.mid_init+'b') as m2
from name_test n1
join name_test n2 on n1.nm = n2.nm and n1.name_type = 1 and n2.name_type = 2;