[Pig-dev] [jira] Created: (PIG-429) Self join wth implicit split has the join output in wrong order

Self join wth implicit split has the join output in wrong order
---------------------------------------------------------------

Key: PIG-429
URL: https://issues.apache.org/jira/browse/PIG-429
Project: Pig
Issue Type: Bug
Affects Versions: types_branch
Reporter: Pradeep Kamath
Fix For: types_branch

Attachments: PIG-429.patch


Query:
{code}
A = load 'st10k' split by 'file';
B = filter A by $1 > 25;
D = join A by $0, B by $0;
dump D;
{code}
In the output the columns from B are projected out first and from A next. On closer examination of the code, the ImplicitSplitInserter class adds in the split and two splitoutput operators into the plan and tries the connect the successors of LOad to these. However it does this by iterating over its successors and disconnecting from them and connecting up the split-splitoutput to the successors. However the order in which it gets its successors is NOT the same as the order in which cogroup (join) expects its inputs. Hence the discrepancy.

Self join wth implicit split has the join output in wrong order
---------------------------------------------------------------

Key: PIG-429
URL: https://issues.apache.org/jira/browse/PIG-429
Project: Pig
Issue Type: Bug
Affects Versions: types_branch
Reporter: Pradeep Kamath
Fix For: types_branch

Attachments: PIG-429.patch


Query:
{code}
A = load 'st10k' split by 'file';
B = filter A by $1 > 25;
D = join A by $0, B by $0;
dump D;
{code}
In the output the columns from B are projected out first and from A next. On closer examination of the code, the ImplicitSplitInserter class adds in the split and two splitoutput operators into the plan and tries the connect the successors of LOad to these. However it does this by iterating over its successors and disconnecting from them and connecting up the split-splitoutput to the successors. However the order in which it gets its successors is NOT the same as the order in which cogroup (join) expects its inputs. Hence the discrepancy.

Self join wth implicit split has the join output in wrong order
---------------------------------------------------------------

Key: PIG-429
URL: https://issues.apache.org/jira/browse/PIG-429
Project: Pig
Issue Type: Bug
Affects Versions: types_branch
Reporter: Pradeep Kamath
Fix For: types_branch

Attachments: PIG-429.patch


Query:
{code}
A = load 'st10k' split by 'file';
B = filter A by $1 > 25;
D = join A by $0, B by $0;
dump D;
{code}
In the output the columns from B are projected out first and from A next. On closer examination of the code, the ImplicitSplitInserter class adds in the split and two splitoutput operators into the plan and tries the connect the successors of LOad to these. However it does this by iterating over its successors and disconnecting from them and connecting up the split-splitoutput to the successors. However the order in which it gets its successors is NOT the same as the order in which cogroup (join) expects its inputs. Hence the discrepancy.