Dare I ask why such a query would be used? AFAICT the second group operation would just stick each record in a bag and create an extra copy of group on the outside of the bag (but use up a lot more computational power than a UDF that would just do the same thing explicitly).
Cheers, Kris On Thu, Dec 09, 2010 at 03:34:58PM -0800, Lin Guo wrote: > A = load 'foo.txt' using PigStorage as (x : chararray, y : int); > > B = group A by x; > C = group B by group; > describe C; > > -- we got > -- C: {group: chararray,B: {group: chararray,A: {x: chararray,y: int}}} > > D = foreach C generate B.(group, A); -- this works > describe D; > > E = foreach C generate B.(group, A.(x)); > describe E; > --- pig returns syntax error, but should this work? Or is there a patch for > it? > > thanks, > lin -- Kris Coward http://unripe.melon.org/ GPG Fingerprint: 2BF3 957D 310A FEEC 4733 830E 21A4 05C7 1FEB 12B3