Technical

book	6
book,eraser	2
book,pen	4
book,pen,pencil	3
book,pencil	5
chalk	3
chalk,eraser	2
chalk,eraser,pen	2
chalk,pen	2
eraser	6
eraser,pen	5
eraser,pen,pencil	2
eraser,pencil	3
pen	8
pen,pencil	5
pencil	7

This will be my input to Mapper 1 and Mapper 2
In Mapper 1 I will be generating rules with this input file

//OUTPUT
Rule 1: book => eraser 2
Rule 0: eraser => book 2
Rule 1: book => pen 4
Rule 0: pen => book 4

Mapper2 emits the same input file


What calculation I have to do is
P(A union B)/P(B)

ie for
book => eraser = 2/6
eraser => book 2/6

book => pen 4/6
pen => book 4/8

So my logic is
 in mapper1
context.write(book,book => eraser 2)
context.write(eraser,eraser => book 2) etc
AND
in mapper 2
context.write(book,6)
context.write(book,eraser,2)

So in Reducer we get

key -- book
values---book => eraser 2
         6

similary for
key -- eraser
values----eraser => book 2
           6

So i can clavulate as explained in my Stackoverflow question

So while doing secondary sorting I have to emit a composite key right?
so if i am emitting composite key ,I will not be able to get values in reducer as expected