Why can't i refer to a column alias in the ORDER BY using CASE?

Question

Sorry if this a duplicate, but i haven't found one. Why can't i use my column alias defined in the SELECT from the ORDER BY when i use CASE?

Consider this simple query:

SELECT NewValue=CASE WHEN Value IS NULL THEN '<Null-Value>' ELSE Value END
FROM dbo.TableA
ORDER BY CASE WHEN NewValue='<Null-Value>' THEN 1 ELSE 0 END

The result is an error:

Invalid column name 'NewValue'

Here's a sql-fiddle. (Replace the ORDER BY NewValue with the CASE WHEN... that´'s commented out)

I know i can use ORDER BY CASE WHEN Value IS NULL THEN 1 ELSE 0 END like here in this case but actually the query is more complex and i want to keep it as readable as possible. Do i have to use a sub-query or CTE instead, if so why is that so?

Update as Mikael Eriksson has commented any expression in combination with an alias is not allowed. So even this (pointless query) fails for the same reason:

SELECT '' As Empty
FROM dbo.TableA
ORDER BY Empty + ''

Result:

Invalid column name 'Empty'.

So an alias is allowed in an ORDER BY and also an expression but not both. Why, is it too difficult to implement? Since i'm mainly a programmer i think of aliases as variables which could simple be used in an expression.

Answer 1

This has to do with how a SQL dbms resolves ambiguous names.

I haven't yet tracked down this behavior in the SQL standards, but it seems to be consistent across platforms. Here's what's happening.

create table test (
  col_1 integer,
  col_2 integer
);

insert into test (col_1, col_2) values 
(1, 3), 
(2, 2), 
(3, 1);

Alias "col_1" as "col_2", and use the alias in the ORDER BY clause. The dbms resolves "col_2" in the ORDER BY as an alias for "col_1", and sorts by the values in "test"."col_1".

select col_1 as col_2
from test
order by col_2;

col_2
--
1
2
3

Again, alias "col_1" as "col_2", but use an expression in the ORDER BY clause. The dbms resolves "col_2" not as an alias for "col_1", but as the column "test"."col_2". It sorts by the values in "test"."col_2".

select col_1 as col_2
from test
order by (col_2 || '');

col_2
--
3
2
1

So in your case, your query fails because the dbms wants to resolve "NewValue" in the expression as a column name in a base table. But it's not; it's a column alias.

PostgreSQL

This behavior is documented in PostgreSQL in the section Sorting Rows. Their stated rationale is to reduce ambiguity.

Note that an output column name has to stand alone, that is, it cannot be used in an expression — for example, this is not correct:

SELECT a + b AS sum, c FROM table1 ORDER BY sum + c;          -- wrong

This restriction is made to reduce ambiguity. There is still ambiguity if an ORDER BY item is a simple name that could match either an output column name or a column from the table expression. The output column is used in such cases. This would only cause confusion if you use AS to rename an output column to match some other table column's name.

Documentation error in SQL Server 2008

A slightly different issue with respect to aliases in the ORDER BY clause.

If column names are aliased in the SELECT list, only the alias name can be used in the ORDER BY clause.

Unless I'm insufficiently caffeinated, that's not true at all. This statement sorts by "test"."col_1" in both SQL Server 2008 and SQL Server 2012.

select col_1 as col_2
from test
order by col_1;