1) Map method : 1) Reads the input 2) Converts the input as key-value pair. In this case key=JoincolumnValue Value=Table object( this object will be encapsulted by Writable). 2) Reduce method : 1) The reducer gets a list of values of each key ex : key=2 Value={[R, 2, Don, Larson, Newark, 555-3221],[S, 2, 18000, 2000, part1]} 2) Put all the values of a key in a list 3) If the size of the list is more than one then Iterate through the list using two pointers: outer and inner. if the value pointed by outer is not equal to inner then it means that they are two different tables and hence output the values to the output file 3) Driver class : equijoin.java 1) Initialise the mapper and reducer class 2) Gets the input file 3) Call the map method 4) Writes the output to the output location. Reason behind using Value= Table instead of Text: In the reduce phase , we need to find the table to which a row belongs If the Value was a Text, then we cannot compare two objects(If you try adding then it just replaces the existing value) as Map Reduce uses same hashValue for the all values of a key. Hence in order to avoid this, I created a class "Table" which has the following attributes : TableName and Column. I have overriden equals method in order to compare two object based on tablename. Comparison between two objects is done by overriding equals and hashcode()
