Getting maximum value of float in SQL programmatically

Though there doesn't appear to be any inline way to get the min or max values, there's a solution somebody put together:

 CREATE TABLE datatype_extrema 
  (min_bit bit NOT NULL DEFAULT (0) CHECK (min_Bit=0) 
  ,max_bit           AS CAST(0x1 AS bit) 
  ,min_tinyint       AS CAST(0x00 AS tinyint) 
  ,max_tinyint       AS CAST(0xFF AS tinyint) 
  ,min_smallint      AS CAST(0x8000 AS smallint) 
  ,max_smallint      AS CAST(0x7FFF AS smallint) 
  ,min_int           AS CAST(0x80000000 AS int) 
  ,max_int           AS CAST(0x7FFFFFFF AS int) 
  ,min_bigint        AS CAST(0x8000000000000000 AS bigint) 
  ,max_bigint        AS CAST(0x7FFFFFFFFFFFFFFF AS bigint)
  ,min_float         AS CAST('-1.79E+308' AS float)
  ,max_float         AS CAST('1.79E+308' AS float)
  ,min_real          AS CAST('-3.40E+38' AS real)
  ,max_real          AS CAST('3.40E+38' AS real)
  ,min_smalldatetime AS CAST('19000101 00:00' AS smalldatetime) 
  ,max_smalldatetime AS CAST('20790606 23:59' AS smalldatetime) 
  ,min_datetime      AS CAST('17530101 00:00:00.000' AS datetime) 
  ,max_datetime      AS CAST('99991231 23:59:59.997' AS datetime) 
  )
  INSERT INTO datatype_extrema DEFAULT VALUES 
  GO 
  CREATE TRIGGER nochange_datatype_extrema 
  ON datatype_extrema INSTEAD OF INSERT, UPDATE, DELETE 
  AS BEGIN 
    RAISERROR ('No changes allowed for table datatype_extrema.', 16, 1) 
    ROLLBACK TRANSACTION 
  END 
  GO 

After that, you can either copy a maximum value to a local variable or (when using queries) cross join with this table.

  Declare @max_int int 
  Set @max_int=(SELECT max_int FROM datatype_extrema) 
  IF COALESCE(@FirstInt, @max_int) < COALESCE(@SecondInt, 0) 

Here are the defaults for the float and real type (that's missing in the accepted answer):

select
    CAST('-1.79E+308' AS float) as MinFloat,
    CAST('1.79E+308' AS float) as MaxFloat,
    CAST('-3.40E+38' AS real) as MinReal,
    CAST('3.40E+38' AS real) as MaxReal

Unfortunately it is not possible to convert them from a varbinary, but varchar works without any problems.