kevin bailey
kevin bailey

Reputation:

How can I optimize this horrendously inefficient User Defined Function in SQL Server 08

Is there anyway to optimize this horrible inefficient UDF in SQL Server 08. I am fairly new to UDF's and especially looking at them for optimizations.

UPDATE: Should I be sending a column to a function like this if I wanted to perform it on each row and each column in a query? Is there a better way to go about this?

Thank You

** @value(float) and @fieldname(varchar(40)) are input parameters **

BEGIN
  DECLARE @UT integer, @FRM integer, @TO integer, @FACTOR float

  select @UT =  [UF_UT_ID] FROM dbo.UNIT_FIELDS where [UF_FIELD]=@fieldName
  select @FRM = [UT_UN_ID_INTERNAL_UNITS] from dbo.UNIT_TYPES where [UT_ID]=@UT
  select @TO = [UT_UN_ID_DISPLAY_UNITS] from dbo.UNIT_TYPES where [UT_ID]=@UT
  select @FACTOR = [UC_SLOPE] from dbo.UNIT_CONVERSIONS where [UC_UN_ID_UNIT_FROM]=@FRM and [UC_UN_ID_UNIT_TO]=@TO

  -- Return the result of the function dbo.
  RETURN @FACTOR*@value
END

Upvotes: 1

Views: 1353

Answers (7)

Lucero
Lucero

Reputation: 60190

A classical candidate for a inline table function...

Something like:

ALTER FUNCTION fnName(@value float, @fieldName VARCHAR(100))
RETURNS TABLE 
AS
RETURN 
(
    SELECT @value * 
        (SELECT conv.[UC_SLOPE] from dbo.UNIT_CONVERSIONS conv
        JOIN dbo.UNIT_TYPES UT_UN_ID_INTERNAL_UNITS ON where [UC_UN_ID_UNIT_FROM]=UT_UN_ID_INTERNAL_UNITS.[UT_ID]
        JOIN dbo.UNIT_TYPES UT_UN_ID_DISPLAY_UNITS ON where [UC_UN_ID_UNIT_TO]=UT_UN_ID_DISPLAY_UNITS.[UT_ID]
        JOIN dbo.UNIT_FIELDS fields ON (UT_UN_ID_INTERNAL_UNITS.[UT_ID] = fields.[UF_UT_ID]) AND (UT_UN_ID_DISPLAY_UNITS.[UT_ID] = fields.[UF_UT_ID])
        WHERE ([UF_FIELD]=@fieldName)
        )
)

Upvotes: 2

Noah
Noah

Reputation: 13959

You can try this as all one query:

--Decalre Factor var
DECLARE @FACTOR float

SELECT @FACTOR = [UC_SLOPE] 
FROM dbo.UNIT_CONVERSIONS uc
--Join the Unit Types table to Unit Conversions on the old From & To types
INNER JOIN dbo.UNIT_TYPES ut ON
    ut.[UT_UN_ID_DISPLAY_UNITS] = uc.[UC_UN_ID_UNIT_TO]
    AND ut.[UT_UN_ID_INTERNAL_UNITS] = uc.[UC_UN_ID_UNIT_FROM]
--Join the Unit Files on the Unit Types
INNER JOIN dbo.UNIT_FIELDS uf ON
    uf.[UF_UT_ID] = ut.[UT_ID]
WHERE uf.[UF_FIELD]=@fieldName

-- Return the result of the function dbo.
RETURN @FACTOR*@value

Upvotes: 0

Joe Koberg
Joe Koberg

Reputation: 26699

BEGIN
  DECLARE @FACTOR float

    select @factor = UC.UC_SLOPE
      from dbo.UNIT_CONVERSIONS UC,
           dbo.UNIT_TYPES       UT,
           dbo.UNIT_FIELDS      UF
      where UF.UF_FIELD=@fieldName
        and UT.UT_ID = UF.UF_UT_ID
        and UC.UC_UN_ID_UNIT_FROM = UT.UT_UN_ID_INTERNAL_UNITS
        and UC.UC_UN_ID_UNIT_TO = UT.UT_UN_ID_DISPLAY_UNITS


  -- Return the result of the function dbo.
  RETURN @FACTOR*@value
END

Upvotes: 0

Quassnoi
Quassnoi

Reputation: 425341

This needs to rewritten using JOIN's:

SELECT  c.UC_SLOPE * @value
FROM    unit_fields f
JOIN    unit_types t
ON      t.UT_ID = f.UF_UT_ID
JOIN    unit_conversions c
ON      c.UC_UN_ID_UNIT_FROM = t.UT_UN_ID_INTERNAL_UNITS
        AND c.UC_UN_ID_UNIT_TO = t.UT_UN_ID_DISPLAY_UNITS
WHERE   f.UF_FIELD = @field_name

Upvotes: 0

JoshBerke
JoshBerke

Reputation: 67068

One small thing you can do:

 select @FRM = [UT_UN_ID_INTERNAL_UNITS] from dbo.UNIT_TYPES where [UT_ID]=@UT  
 select @TO = [UT_UN_ID_DISPLAY_UNITS] from dbo.UNIT_TYPES where [UT_ID]=@UT 

If I read the correctly this would be the same as:

 select @TO = [UT_UN_ID_DISPLAY_UNITS],
        @FRM = [UT_UN_ID_INTERNAL_UNITS] 

 from dbo.UNIT_TYPES where [UT_ID]=@UT 

Why run a select twice on the same record?

Upvotes: 0

Jamie Ide
Jamie Ide

Reputation: 49261

If you can join the three tables based on PK/FK relationships you could get the query down to a single select. If not, the only immediately obvious optimization is to assign @FRM and @TO in a single select statement:

select @FRM = [UT_UN_ID_INTERNAL_UNITS], @TO = [UT_UN_ID_DISPLAY_UNITS] from dbo.UNIT_TYPES where [UT_ID]=@UT

Upvotes: 3

John M Gant
John M Gant

Reputation: 19308

Step 1 would be to run each select individually and see where the bottleneck is.

Upvotes: 1

Related Questions