Hi Terry, It looks like you should FLATTEN the data relation first, so that your ids could be not nested and then join like this (or just remove GROUP statement): joined = JOIN dataFlattened by id, lookup by id USING 'replicated'; (the replacated join is recommended if your lookup relation is small) After that you can add another FOREACH to eliminate ids and another GROUP to group again. Maybe there is some other better approaches. Let me know if you are experiencing problems with this approach.
Ruslan On Tue, Sep 18, 2012 at 9:11 PM, Terry Siu <[email protected]> wrote: > Hi, > > I'm having a little difficulty figuring this out. I have the following data > in Pig, let's say aliased to 'data': > > (a1,b1,{(id1),(id2)}) > (a2,b2,{(id3)}) > (a3,b3,{(id2),(id4)}) > > Each tuple above has an associated bag of tuples that are lookup ids. Given > the following 'data', I also load the lookup id and their corresponding > values, which I'll alias to 'lookup': > > (id1,Hello) > (id2,World) > (id3,foo) > (id4,bar) > > Is it possible, from the two inputs above, to create the following: > > (a1,b1,{(Hello),(World)}) > (a2,b2,{(foo)}) > (a3,b3,{(World),(bar)}) > > Thanks, > -Terry
