This feature/syntax seems be causing confusion in many cases , so I have proposed deprecating this syntax in the next release . See - https://issues.apache.org/jira/browse/PIG-1967 .
-Thejas On 4/6/11 12:30 PM, "Thejas M Nair" <[email protected]> wrote: In the relation TCRaw, there is no column called SrcFuid. As a result, you end up using this feature - http://pig.apache.org/docs/r0.8.0/piglatin_ref2.html#Casting+Relations+to+Sc alars . Change your statement to - TCGroupedByFuid = group TCRaw by (citingdocid, col, seq); Thanks, Thejas On 4/6/11 11:09 AM, "[email protected]" <[email protected]> wrote: > > >> Do you need the group-key to be concatenated ? If not, you can just group on >> all the three columns - > >> TCGroupedByFuid = group TCRaw by (SrcFuid.citingdocid, > SrcFuid.col, > SrcFuid.seq); > > Hi Thejas, > > I had tried that originally before introducing CONCAT(), but I got this error > message: > > ERROR 0: Scalar has more than one row in the output. > 1st : (14159274,BCI,6), 2nd :(45937168,BCI,17) > > I don't understand that, since TCRaw is > > (14159274,BCI,6,14159274,14159163,BCI,5,1999,BCI.BCI) > (14159274,BCI,6,14159274,14159163,WOS,11,1999,WOS.SCI) > (14159274,WOS,16,14159274,14159163,BCI,5,1999,BCI.BCI) > (14159274,WOS,16,14159274,14159163,WOS,11,1999,WOS.SCI) > > and the 2nd tuple is not a (projection of any) member of TCRaw (though it is a > member of SrcFuid). So I think my understanding of GROUP is incorrect. > > Thanks for your help! > > Will > >
