You might be able to reduce a few instructions by using BETWEEN instead of x >= y AND x <= z.  So,   ...WHEN x BETWEEN y AND z.

There seems to be an oportunity to simply write a BASE 10 to BASE 36 conversion routine.  I don't know if there is a more straightforward conversion than what you've preposed here but it might be worth a thought.

Enjoy

The email header for your article states: "User defined functions were added in SQL Server 7 and enhanced in SQL Server 2000...".

I was not aware of any capabiltiy for user defined functions in SQL Server 7. Did I miss something?

You may be right. I will try it. But the article was not written to show the best code, but to show the idea of how to use user defined function for the case.

As I pointed, it will be more than 999 number of records in a table.

Column data must be assigned automatically by the database. No one wants to rebuid the applications. Anyway any mechanism must create a value and place it as default for the column.

User defined function is set the value by default if NULL is inserted.

This reminds me of a function I had to write recently, to convert alphanumeric keys to numeric keys.  The company had char(8) keys as '3AJPUL68' and desired to map the alphanumeric space to a numeric space, in order to start using numeric keys.  I've written a function that takes a string and makes up a numeric hash for it.

The resolution parameter @p controls how many bits of information are kept, at each position of the string.  Entering 0 will suppress the position, 1 will do (ascii & 2^1 => giving either 0 or 1), entering 2 will do (ascii & 2^2 => resulting in either 0, 1, 2, or 3), etc, up to 6, being the highest resolution, which would make the ascii code correspond to a number between 0 and 63.  As the function iterates through the string positions, from right to left, it shifts the intermediate left results before adding the next hash value.

By setting the resolution parameter @p to '00016666' you'd always get integers in return (because the sum of all bits is less than 32), that's what we did, because the leftmost characters were almost always for us non-significant.  You may need to change the return value to bigint if you need a higher resolution for the left-most positions.

CREATE  function fn_char8hash ( @s char(8) , @p char(8) )
returns int
as
begin
if @s is null return @s

set @s = replicate(' ', 8 - len(@s)) + @s
declare @c char(1)
declare @a smallint

declare @e tinyint
declare @l tinyint
set @l = 0

declare @r int
set @r = 0

declare @i tinyint
set @i = 8
while @i > 0
begin
 set @e = cast(substring(@p, @i, 1) as tinyint)
 set @c = substring(@s, @i, 1)

 set @c = replace(upper(@c), '_', ' ')
 set @a = (power(2, @e) - 1) & (ascii(@c) - 
  case when @c < 'A' then 32 else 
  case when @e < 6 then 64 else 32 
  end end)
-- print @a
 if @a > 0
 begin
  set @r = @a * power(2, @l) + @r
  set @l = @l + @e
 end
-- print - @r
-- print @l

 set @i = @i - 1
end

return @r

end