写Pig的时候,如果需要对数据中某个字段进行 比较或者运算,尤其是使用Filter的时候。
该字段最好是 bytearray类型,而不要用 int 或者 chararray类型。否则得不到数据
--抽取话单文件中过程号为 attach的话单 统计
set job.name 'lte_kpi_user_analysis_s1mme_attatch_15min';
allData = load '/user/hdfs/data/200506/20050601/000000/0B43CED07C7438B4328F918C64DACE35-merge';
sel_fields = foreach allData generate FLOOR($10/300)*300+8*3600 as (time:chararray), $2 as (interface_type:chararray),$8 as (procedure_type:bytearray
,$7 as (msisdn:chararray),$6 as (imei:chararray),$5 as (imsi:chararray),$1 as (city:chararray),$21 as (mme_ip:chararray),$22 as (enodeb_ip:chararray)
,$25 as (tac:chararray),$26 as (cell_id:chararray),$12 as (cause:int),($10-$9) as (delay:int),$11 as (procedure_status:bytearray);
--抽取过程号等于1
s1mme_attach = filter sel_fields by procedure_type == 1;
group_by_user = group s1mme_attach by(time, interface_type, procedure_type, msisdn,imei, imsi, city,mme_ip,
enodeb_ip, tac, cell_id, cause, hour,miniute15);
tmp_result = foreach group_by_user {
failures = filter s1mme_attach by procedure_status>0;
user_num = distinct s1mme_attach.imsi;
generate FLATTEN(group), COUNT(failures) as count_fail, COUNT(user_num) as count_total, AVG(s1mme_attach.delay);
};
--测试输出
filter_tmp = limit s1mme_attach 10;
test_out = limit group_by_user 10;
test_out2 = limit tmp_result 10;
store filter_tmp into '/user/hdfs/xiatao/pigOutput/s1mme_attatch_15min_filter1';
store test_out into '/user/hdfs/xiatao/pigOutput/s1mme_attatch_15min_1';
store test_out2 into '/user/hdfs/xiatao/pigOutput/s1mme_attatch_15min_2';
如果不适用 bytearray类型,后面输出的几个文件夹中就没有数据。