Hi Split the columns, normalize each column separately, add a sequence id to each row and store the temporary result to memory. In next subjob, read the result from memory and do a inner job based on the sequence id, and generate a output to contain all columns, the job looks like: tfileinputdelimited--main--tmap--out1--tnormalized1--tmap2-->tHashOutput1 --out2--tnormalized2--tmap3-->tHashOutput2 --out3--tnormalized3--tmap4-->tHashOutput3 -onsubjobok- thashinput1--main--tMap5-...tlogrow thashinput2-lookup- thashinput3-lookup-
out1: only has column1, out2: only has column2, out3: only has column3 tmap2: add a new column called id, and set its expression as: Numeric.sequence("s1",1,1) tmap3: add a new column called id, and set its expression as: Numeric.sequence("s2",1,1) tmap4: add a new column called id, and set its expression as: Numeric.sequence("s3",1,1)
tHashinput1: read data from thashoutput1 tHashinput2: read data from thashoutput2 tHashinput3: read data from thashoutput3
tMap5: do an inner job based on sequence id, and generate a output to contains three columns.
