由于pig处理关系的方式，您无法真正实现所需的输出（以那种精确的格式）。你的模式是什么 File 3 是吗？所有的行没有相同的字段，这是pig所期望的。

-- F1 is File 1
-- F2 is File 2

-- If order does matter for File 2, then you are going to need to RANK the data
-- right of the bat because most Pig operations do not preserve order.
-- This means you will have to carry the first field of F2_R in later operations,
-- and use F2_R in place of F2.

-- F2_R = RANK F2 ;

G = GROUP F2 BY EmpID ;
H = FOREACH G GENERATE group AS EmpID, F2.(Agreeno, POLNO) AS agree_details ;

J = JOIN F1 BY EmpID LEFT, H BY EmpID ;
-- Now, the output of J should look like:
-- (122292,ABCDEF,1985,60000,122292,{(UNI456,AWS12345),(UNI457,AWS12346),(UNI458,AWS12347),(UNI459,AWS12348)})
-- etc.
``` `J` 是你可能想要表达的意思 `File 3` 在Pig身上。这样可以存储 `File 2` 作为一个袋子。如果真的需要输出完全平坦，那么就必须使用 `FLATTEN` 在 `agree_details` 之后 `J` .