Thanks for a very useful function. When applied to a scalar value, the routine is fine. However, if you're applying it to thousands of rows, it's very slow because of the while loop.
Here's an alternative.
alter function dbo.UTIL_JAVA_HASHCODE(@str varchar(max))
returns int
as
-------------------------------------------------------------------------------
-- Proc: UTIL_JAVA_HASHCODE
-- Desc: Replicate Java's String.HashCode() function
-- Inputs: @STR: String
-- Outputs: Java hashcode of the string (4 byte integer)
-------------------------------------------------------------------------------
begin
declare @h bigint
set @h = 0
select @h = (@h*31 + ascii(substring(@str,X.pos,1)))%4294967296
from (select top(len(@str))
row_number() over (order by getdate()) as pos
from sys.all_objects) as X
if @h >= 2147483648 set @h = @h - 4294967296
return convert(int, @h)
end;
go
With the following logic:
set statistics time on
go
declare @table table ( hashCode int )
insert into @table
select global.dbo.UTIL_JAVA_HASHCODE( X.varchar32field )
from TableWith33000Rows as X
go
Using the original function, the insert completed in 53s (avg of 3 runs). Applying the original function to a varchar(64) field, the elapsed time grew to 100s.
After revising the function to eliminate the while loop, the varchar(32) processing completes in 11s and the varchar(64) processing in only 14s. (If you have a tally table feel free to use it instead of the derived table from sys.all_objects.)
🙂