猪 - 计算成员群体

Question

说我有一个关系Students，其中包含字段grade和teacher。我希望按年级和老师分组，但保留每组中每个年级的所有学生的数量。类似的东西：

classes = GROUP Students BY (grade,teacher);
classes = FOREACH classes {
   GENERATE
      (### COUNT OF ALL STUDENTS IN GRADE ###) as grade_size,
      Students as students,
      teacher as teacher;
}

但我无法弄清楚如何从群组声明中进行过滤。某种过滤方式，但我不知道在小组内外对学生的成绩进行调整。

Answer 1

有两种方法：

1）使用Group by grade和teacher，而不是计数，而不是Flatten和Group By grade。

classes = GROUP Students BY (grade,teacher);
teachers = FOREACH classes GENEARATE FLATTEN(group) as (grade,teacher), COUNT(Students) as perTeacehr;
grade = GROUP teachers BY grade;
result = FOREACH grade GENERATE FLATTEN(teachers), SUM(teachers.perTeacher) as perGrade;
describe result;
dump result;

2）按分级分组，而不是使用来自DataFu库的BagGroup的UDF来执行内存组，但这很容易受到堆内存异常的影响，但速度更快。