tajo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hyunsik Choi (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (TAJO-338) Add Query Optimization Part for Column-Partitioned Tables
Date Wed, 18 Dec 2013 13:06:07 GMT

     [ https://issues.apache.org/jira/browse/TAJO-338?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Hyunsik Choi updated TAJO-338:

    Attachment: TAJO-338.patch

This patch does as follows:
 * Add PartitionedTableScanNode
 * Add a rewrite rule named PartitionedTableRewriter
   ** First of all, PartitionedTableRewriter finds partition predicates from search condition.
     *** A partition predicate must be a simple predicate, meaning one column reference and
one constant with one of =, <, <=, >, >=, IN, and LIKE predicate operators.
  ** PartitionedTableRewriter also finds necessary partition paths via PathFilters generated
from partition predicates.
  ** Then, it rewrites ScanNode to PartitionedTableScanNode with filtered partition paths.
 * Some methods in EvalTreeUtil and AlgebraUtil are refactored and renamed to more comprehensive
 * Rename Partitions to PartitionDesc
 * Add some schema rewrite methods to ColumnPartitionedTableStoreExec and SeqScan
  ** Column Partitioned Table aims at hive-compatible table partition.
  ** As a result, scan does not read field values corresponding to partition key columns from
files, and store does not store field values to files.
  ** Instead of, field values corresponding to partition key columns are just retrieved from
a partition path.
  ** For that, the rewriter usually removes partition key columns from actual input schema
of scan plan and output schema of store plan.

> Add Query Optimization Part for Column-Partitioned Tables
> ---------------------------------------------------------
>                 Key: TAJO-338
>                 URL: https://issues.apache.org/jira/browse/TAJO-338
>             Project: Tajo
>          Issue Type: Sub-task
>          Components: storage
>    Affects Versions: 0.8-incubating
>            Reporter: JaeHwa Jung
>            Assignee: Hyunsik Choi
>             Fix For: 0.8-incubating
>         Attachments: TAJO-338.patch
> If user write partitioned column at where clause, Tajo must make a plan to scan a partition
directory path. So, Planner needs to improve for partitioned table.

This message was sent by Atlassian JIRA

View raw message