Affects Version/s: 0.2.0
Fix Version/s: 0.9.0
A = load 'studenttab10k' as (name); – note that data has more than 1 column
B = load 'votertab10k' as (name, age, reg, contrib);
D = COGROUP A by name, B by name;
E = foreach D generate flatten(A), flatten(B);
F = foreach E generate registration, contr;
The dump produces the wrong columns. This is because even though we declared only one column, we actually load all columns of A. So any place where we explicitely or implicitely use A.* as the case in flatten, we would produce the wrong results.
The long term solution is actually to push projections into the load. Shorter term the proposal is to notice if the script uses A.* and stick a project after the load. Note that we don't need to do that if types are declared because there will be already casting foreach there.