数据

India 3000 1
Australia 4000 2
US 5000 1
China 3000 1
Russia 500 1

脚本

A = LOAD 'test6.txt' USING PigStorage(' ') AS (c:chararray,p:int,z:int);
B = FILTER A BY z==1;
C = FOREACH B GENERATE $0,$1;
D = GROUP C ALL;
E = FOREACH D GENERATE MAX(C.p) as P;
F = FILTER C BY p == (int)E.P;
DUMP F;

输出

请用简单的步骤解决：

country = LOAD '/home/cloudera/pig/country_population.txt' using PigStorage(',') AS (country:chararray, population:int, zone:int);

countryZone1 = filter country by zone == 1;

countryZone1Order = order countryZone1 by population DESC;

countryZone1Limit = limit countryZone1Order 1;

getMaxPopCountry = foreach countryZone1Limit generate country,population;

dump getMaxPopCountry;

输出：
（5000美元）

同样可以通过以下方式实现：

A = load 'data' using PigStorage(' ') as (c:chararray,p:int,z:int);
B = filter A by z==1;
C = foreach (group B all) {
    ordered = order B by p DESC; 
    limited = limit ordered 1; 
    generate flatten(limited)
}
dump C;

与max相比，这种方法的主要优点是您可以轻松地调整它以获得“topk”（只需替换limit语句的参数）。另外，我认为它使用更少的map reduce作业-过滤在mapper中完成，其余的都在reducer中完成。在需要两个作业之后使用max+过滤。

使用内置的top函数获取包的前n个元组

data = LOAD 'data' using PigStorage(' ') AS (c:chararray,p:int,z:int);
filtered = FILTER A by z==1;
grouped = GROUP filtered ALL;
max = FOREACH grouped {
    top = TOP(1,1,data);
    GENERATE FLATTEN(top);
}
DUMP max;