[
https://issues.apache.org/jira/browse/TAJO-1436?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14386339#comment-14386339
]
ASF GitHub Bot commented on TAJO-1436:
--------------------------------------
Github user navis commented on a diff in the pull request:
https://github.com/apache/tajo/pull/456#discussion_r27371654
--- Diff:
tajo-core/src/main/java/org/apache/tajo/engine/planner/physical/SeqScanExec.java
---
@@ -69,9 +70,8 @@ public SeqScanExec(TaskAttemptContext context, ScanNode
plan,
this.qual = plan.getQual();
this.fragments = fragments;
- if (fragments != null
- && plan.getTableDesc().hasPartition()
--- End diff --
Even fragments is null, partition columns in expr should be replaced with
null. Without this, some tests fails with exception("No Such Column Reference").
> Add Bind method to EvalNode
> ---------------------------
>
> Key: TAJO-1436
> URL: https://issues.apache.org/jira/browse/TAJO-1436
> Project: Tajo
> Issue Type: Improvement
> Components: operator/expression
> Reporter: Hyunsik Choi
> Assignee: Navis
> Fix For: 0.11.0
>
>
> EvalNode just has eval() method to evaluate some tuples. It does not have any
> method to prepare actual execution.
> For example, FieldEval should know an actual field index of an input tuple.
> This process is also performed in eval() method.
> It has some problems:
> * eval() method should involve two works, and it does different behavior
> with some flag or null check.
> * 'if condition' in eval() method for different behaviors involves
> unnecessary branches, causing performance degradation.
> So, we should add {{bind(Schema)}} method to EvalNode and then move some
> preparation code from eval() to bind(Schame) method.
> Also, we should refactor the signature {{eval(Schema, Tuple)}} to
> {{eval(Tuple)}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
