Mailing-List: contact hama-dev-help@incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hama-dev@incubator.apache.org
Message-ID: <1640530364.1224670124609.JavaMail.jira@brutus>
Date: Wed, 22 Oct 2008 03:08:44 -0700 (PDT)
From: "Samuel Guo (JIRA)" <jira@apache.org>
To: hama-dev@incubator.apache.org
Subject: [jira] Commented: (HAMA-86) Implements generic load/save operations
 in hama
In-Reply-To: <1323652323.1224556724221.JavaMail.jira@brutus>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HAMA-86?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12641755#action_12641755 ] 

Samuel Guo commented on HAMA-86:
--------------------------------

I think there are different features between a matrix in HBase & a matrix in FileSystem.
- A matrix in HBase has a fixed format. But matrices in FileSystem may have different format, parsing of the file as a matrix depends on the format.
- A matrix in HBase is a full-featured matrix. we can randomly access it by its row & column. But matrices in FileSystem don't. A matrix as block-indexing format may be useful for multiplication, but may be not efficient in other computation(just my guessing).

for example:
  /**
   * Save to a table or file
   * 
   * @param path
   * @return true if saved
   * @throws IOException
   */
  public boolean save(String path) throws IOException;

if we implements "save to file" operation in matrix, we may need an OutputFormat to identify the format of the file.

so I suggest that let HamaAdmin just keep all the aliases of the matrices in HBase.
And add a method like "boolean saveAsFile(Path path, OutputFormat format);" to deal with file I/O of matrices.

Any ideas?


> Implements generic load/save operations in hama
> -----------------------------------------------
>
>                 Key: HAMA-86
>                 URL: https://issues.apache.org/jira/browse/HAMA-86
>             Project: Hama
>          Issue Type: Improvement
>          Components: implementation
>    Affects Versions: 0.1.0
>            Reporter: Samuel Guo
>             Fix For: 0.1.0
>
>
> we need a generic "pathname" class to identify the input & output.
>  "pathname" can contain two parts: one is scheme, such as HBASE or HDFS; one will be the actual name, such as table name in HBase, filename in HDFS.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.