sql - performing calculations on a column for every 2 rows and displaying the result for the group -

i have table (current) extracted sql query. each benchmark has 2 data points. need calculate difference of 2 data points (i.e. more recent data point - less recent data point) , have result displayed alongside rows.

i guess has group by, weakest area in sql , not sure how return result display alongside grouped rows. i've maanged calculate difference doing "group benchmark" , have max(value) - min(value), realised dumb have no way of knowing signage. if can so, not know how display calculated value. assitance pls?

thanks. (oracle pls)

current				  	datef	    benchmark	value	  	31-jul-14	a	      100	  	31-aug-14	a	      101	  	31-jul-14	b	      101	  	31-aug-14	b	      99	  	31-jul-14	c	      100	  	31-aug-14	c	      101	  	31-jul-14	d	      100	  	31-aug-14	d	      100	  	31-jul-14	e	      101	  	31-aug-14	e	      102	  				    either this:				  				  	datef	    benchmark  	value	difference  	31-jul-14	a	        100	    1  	31-aug-14	a	        101	    1  	31-jul-14	b	        101     -2  	31-aug-14	b          	99	    -2  	31-jul-14	c	        100	    1  	31-aug-14	c          	101	    1    	31-jul-14	d	        100	    0  	31-aug-14	d        	100	    0  	31-jul-14	e	        101	    1  	31-aug-14	e	        102	    1        or this:				  				  	datef	    benchmark  	value	difference  	31-jul-14	a	        100	      	31-aug-14	a	        101	    1  	31-jul-14	b	        101       	31-aug-14	b          	99	    -2  	31-jul-14	c	        100	      	31-aug-14	c          	101	    1    	31-jul-14	d	        100	      	31-aug-14	d        	100	    0  	31-jul-14	e	        101	      	31-aug-14	e	        102	    1

oracle offers first_value() , last_value() functions seem need. problem 1 of analytic/window functions, not aggregation.

select c.*,        (last_value(value) on (partition benchmark order datef range between unbounded preceding , unbounded following) -         first_value(value) on (partition benchmark order datef range between unbounded preceding , unbounded following)        ) difference current c; 

edit:

or, think can simplify above to:

select c.*,        (first_value(value) on (partition benchmark order datef desc) -         first_value(value) on (partition benchmark order datef)        ) difference current c;