避免多关系中的前缀加入猪

时间:2013-06-22 22:19:40

标签: apache-pig

我正在尝试在猪中进行星型模式类型的连接,下面是我的代码。当我连接不同列的多个关系时,我必须每次都为前一个连接的名称添加前缀以使其正常工作。我相信应该有更好的方法,我无法通过谷歌搜索找到它。任何指针都会非常有用。

即为这样的列添加前缀“H864 :: H86 :: hs_8_d :: hs_8_desc”是我想要避免的。

hs_8 = LOAD 'hs_8_distinct' USING PigStorage('^') as (hs_8:chararray,hs_8_desc:chararray);
hs_8_d = FOREACH hs_8 GENERATE SUBSTRING(hs_8,0,2) as hs_2,SUBSTRING(hs_8,0,4) as hs_4,SUBSTRING(hs_8,0,6) as hs_6,hs_8,hs_8_desc;

hs_6_d = LOAD 'hs_6_distinct' USING PigStorage('^') as (hs_6:chararray,hs_6_desc:chararray);
hs_4_d = LOAD 'hs_4_distinct' USING PigStorage('^') as (hs_4:chararray,hs_4_desc:chararray);
hs_2_d = LOAD 'hs_2_distinct' USING PigStorage('^') as (hs_2:chararray,hs_2_desc:chararray);

H86 = JOIN hs_8_d BY hs_6, hs_6_d BY hs_6 USING 'replicated' ;
H864 = JOIN H86 BY hs_8_d::hs_4, hs_4_d BY hs_4 USING 'replicated' ;
H8642 = JOIN H864 BY H86::hs_8_d::hs_2, hs_2_d BY hs_2 USING 'replicated' ;

hs_dim = FOREACH H8642 GENERATE hs_2_d::hs_2,hs_2_d::hs_2_desc,H864::hs_4_d::hs_4,H864::hs_4_d::hs_4_desc,H864::H86::hs_6_d::hs_6,H864::H86::hs_6_d::hs_6_desc,H864::H86::hs_8_d::hs_8,H864::H86::hs_8_d::hs_8_desc;

2 个答案:

答案 0 :(得分:2)

通过向连接添加额外的foreach,您可以略微简化别名。检查统计信息,这不会向管道添加额外的MR作业。原始和此将产生4个仅限地图的作业。

E.g:

H86 = foreach (JOIN hs_8_d BY hs_6, hs_6_d BY hs_6 USING 'replicated') generate 
        hs_8_d::hs_2 as x1, 
        hs_8_d::hs_4 as x2, 
        hs_8_d::hs_6 as x3, 
        hs_8_d::hs_8 as x4,
        hs_8_d::hs_8_desc as x5, 
        hs_6_d::hs_6 as x6,
        hs_6_d::hs_6_desc as x7;

H864 = foreach (JOIN H86 BY x2, hs_4_d BY hs_4 USING 'replicated') generate 
          H86::x1 as y1,
          H86::x2 as y2, 
          H86::x3 as y3,
          H86::x4 as y4,
          H86::x5 as y5, 
          H86::x6 as y6, 
          H86::x7 as y7,
          hs_4_d::hs_4 as y8,
          hs_4_d::hs_4_desc as y9;

H8642 = foreach (JOIN H864 BY y1, hs_2_d BY hs_2 USING 'replicated') generate 
          H864::y1 as z1, 
          H864::y2 as z2,
          H864::y3 as z3, 
          H864::y4 as z4, 
          H864::y5 as z5, 
          H864::y6 as z6, 
          H864::y7 as z7,
          H864::y8 as z8, 
          H864::y9 as z9, 
          hs_2_d::hs_2 as z10, 
          hs_2_d::hs_2_desc as z11;

hs_dim = FOREACH H8642 GENERATE z10, z11, z8, z9, z6, z7, z4, z5;

如果您有一包元组,那么DatafuAliasBagFields可能会有所帮助。

答案 1 :(得分:0)

Pig将始终使用bagname::作为字段的前缀,以便在连接后消除字段歧义。不幸的是,我认为你不能避免这种情况。

相关问题