[轉]MONTHS_BETWEEN Function - Oracle to SQL Server Migration

本文轉自：http://www.sqlines.com/oracle-to-sql-server/months_between

In Oracle, MONTHS_BETWEEN(date1, date2) function returns the number of months between two dates as a decimal number. 

Note that SQL Server DATEDIFF(month, date2, date1) function does not return exactly the same result, and you have to use an user-defined function if you need to fully emulate the Oracle MONTHS_BETWEEN function (see UDF's code below).  

Oracle:

  ALTER SESSION SET NLS_DATE_FORMAT = 'YYYY-MM-DD';
 
  -- 1-day difference  
  SELECT MONTHS_BETWEEN('2013-03-01', '2013-02-28') FROM dual;
  # 0.129032258
 
  -- Still 1-day difference but the result is different
  SELECT MONTHS_BETWEEN('2013-03-02', '2013-03-01') FROM dual;
  # 0.32258065

SQL Server:

DATEDIFF always returns an integer result.

  -- 1-day difference, but 1 month returned (!)
  SELECT DATEDIFF(month, '2013-02-28', '2013-03-01');
  # 1
 
  -- Still 1-day difference but the result is different
  SELECT DATEDIFF(month, '2013-03-01', '2013-03-02');
  # 0

Also note that MONTHS_BETWEEN and DATEDIFF have different order of parameters.

Oracle MONTHS_BETWEEN in Detail

MONTHS_BETWEEN returns the number of full months between dates and a fractional part. 

An integer value is returned only if:

• Both dates specify the same day of the month (February 13 and March 13 i.e.)
• Both dates are the last days of the months (January 31 and April 30 i.e.)

Oracle:

  -- Between March 13 and February 13
  SELECT MONTHS_BETWEEN('2013-03-13', '2013-02-13') FROM dual;
  # 1
 
  -- Between April 30 and January 31
  SELECT MONTHS_BETWEEN('2013-04-30', '2013-01-31') FROM dual;
  # 3

Fractional Part

The fractional part is calculated using the following formula:

Note that when MONTHS_BETWEEN calculates the fractional part, it considers that all months have 31 days. 

Consider the following examples:

Oracle:

 -- 1-day difference  
  SELECT MONTHS_BETWEEN('2013-03-01', '2013-02-28') FROM dual;
  # 0.129032258

Although there is just 1-day difference between February 28, 2013 and March 01, 2013, MONTHS_BETWEEN considers Feb 29, Feb 30, Feb 31 and Mar 01:

Another example:

  -- Still 1-day difference but the result is different
  SELECT MONTHS_BETWEEN('2013-03-02', '2013-03-01') FROM dual;
   # 0.32258065

Now the fractional part is calculated as follows:

SQL Server User-Defined Function to Emulate Oracle MONTHS_BETWEEN

You can use the following user-defined function to emulate Oracle MONTHS_BETWEEN function:

SQL Server:

   CREATE FUNCTION MONTHS_BETWEEN (@date1 DATETIME, @date2 DATETIME) 
	RETURNS FLOAT
   AS
   /******************************************************************************
      PURPOSE: Emulate Oracle MONTHS_BETWEEN in SQL Server
 
      REVISIONS:
      Ver        Date             Author                                   Description
      ---------  ----------       ---------------                         ---------------------------
      1.1         2013-02-10  Dmitry Tolpeko (SQLines)       Created.
   ******************************************************************************/
   BEGIN
     DECLARE @months FLOAT = DATEDIFF(month, @date2, @date1);
 
     -- Both dates does not point to the same day of month
     IF DAY(@date1) <> DAY(@date2) AND
        -- Both dates does not point to the last day of month
        (MONTH(@date1) = MONTH(@date1 + 1) OR MONTH(@date2) = MONTH(@date2 + 1))
     BEGIN
        -- Correct to include full months only and calculate fraction
        IF DAY(@date1) < DAY(@date2)
          SET @months = @months + CONVERT(FLOAT, 31 - DAY(@date2) + DAY(@date1)) / 31 - 1;
        ELSE    
          SET @months = @months + CONVERT(FLOAT, DAY(@date1) - DAY(@date2)) / 31;
     END
 
     RETURN @months; 
   END;
   GO

Now you can use the UDF as follows:

SQL Server:

    -- 1-day difference  
  SELECT dbo.MONTHS_BETWEEN('2013-03-01', '2013-02-28');
  # 0.129032258
 
  -- Still 1-day difference but the result is different (as in Oracle)
  SELECT dbo.MONTHS_BETWEEN('2013-03-02', '2013-03-01');
  # 0.32258065