按两个字段分组的猪会产生奇怪的结果

Question

我在使用 apache-pig 分组数据时遇到问题。

加载数据：

client_trace_send = LOAD '/user/hduser1/adm_project/client_trace50.csv' using PigStorage(',') as (code:chararray, client_id:int, loc_ts:int, length:int, op:chararray, err_code:int, time:long, thread_id:INT);

限制和检查数据：

client_trace_send_small = LIMIT client_trace_send 10;
DUMP client_trace_send_small;

已加载数据：

(msg_snd,0,1,46,enrol_req,-1,1414250523591,9)
(res_rcv,0,1,25,enrol_resp,,1414250523655,9)
(msg_snd,1,2,48,query_queue,-1,1414250523655,9)
(res_rcv,1,2,14,err,19,1414250523661,9)
(msg_snd,1,3,59,peek_req,-1,1414250523661,9)
(res_rcv,1,3,13,err,0,1414250523662,9)
(msg_snd,1,4,59,peek_req,-1,1414250523662,9)
(res_rcv,1,4,13,err,0,1414250523663,9)
(msg_snd,1,5,59,peek_req,-1,1414250523663,9)

现在，我想在字段 "client_id" 和 "loc_ts" 上对上述数据进行分组。

GROUPED = GROUP client_trace_send_small by (client_id,loc_ts);

查看结果：

DUMP GROUPED;

奇怪的是：

((0,1),{(msg_snd,0,1,46,enrol_req,-1,1414250523591,9)})
((1,2),{(msg_snd,1,2,48,query_queue,-1,1414250523655,9)})
((1,3),{(msg_snd,1,3,59,peek_req,-1,1414250523661,9)})
((1,4),{(msg_snd,1,4,59,peek_req,-1,1414250523662,9)})
((1,5),{(msg_snd,1,5,59,peek_req,-1,1414250523663,9)})
((8,28493),{(msg_snd,8,28493,62,pop_req,-1,1414251764157,16)})
((9,25976),{(msg_snd,9,25976,66,query_sender,-1,1414251764148,17)})
((19,28250),{(msg_snd,19,28250,64,pop_req,-1,1414251764152,27)})
((31,27977),{(msg_snd,31,27977,65,peek_req,-1,1414251764152,39)})

有些值甚至没有出现在加载的数据中。对于第一组，我希望是这样的：

((0,1),{(msg_snd,0,1,46,enrol_req,-1,1414250523591,9),{(res_rcv,0,1,25,enrol_resp,,1414250523655,9)})

这里出了什么问题？

感谢您的帮助，在此先致谢。

此致

Answer 1

这是因为LIMIT取了一组随机数据：

There is no guarantee which tuples will be returned, and the tuples that are returned can change from one run to the next.

并且因为你的scipt中有两个DUMP，所以Pig会把pipeline分成两个pipeline，每个pipeline分别执行LIMIT。因此，您将为每个子管道获得两个不同的数据集。

但我们可以确定的是，转储数据显然来自输入文件，并且它们最多来自 10 行（如果文件行数较少，则为 <10）。

您可以查看您的explain plan and it's explanation。

按两个字段分组的猪会产生奇怪的结果

Pig group by two fields yields strange result

apache-pig