[
https://issues.apache.org/jira/browse/PIG-158?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12586582#action_12586582
]
Alan Gates commented on PIG-158:
--------------------------------
In answer to Shravan's question, consider a script like the following:
{code}
a = load 'myfile';
b = group a by $0;
c = foreach b { cc = distinct $1; generate myudf($1); }
store c into 'output';
{code}
And consider a visitor whose job it is to find all of the UDFs used in a script
(for whatever reason). If we go with the nested plan model, but we don't allow
visitors to switch plans, then when the outer most visitor (V) comes to the
foreach, it has to spawn a child visitor (VC) to visit the nested plan. When
VC is done, V has to then ask VC what it learned visiting the nested plan.
(That is it has to incorporate VCs state into its state.) On the other hand,
if V can instead change walkers when it hits the foreach, it can still visit
the nested plan. The state change that resulted from visiting the nested plan
(in this case, learning that the UDF myudf is called) need not be transmitted
between visitors.
> Rework logical plan
> -------------------
>
> Key: PIG-158
> URL: https://issues.apache.org/jira/browse/PIG-158
> Project: Pig
> Issue Type: Sub-task
> Components: impl
> Reporter: Alan Gates
> Assignee: Alan Gates
> Attachments: logical_operators.patch, logical_operators_rev_1.patch,
> logical_operators_rev_2.patch, logical_operators_rev_3.patch,
> visitorWalker.patch
>
>
> Rework the logical plan in line with
> http://wiki.apache.org/pig/PigExecutionModel
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
