hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "stack (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-631) HTable.getRow() for only a column family
Date Mon, 07 Jul 2008 22:52:31 GMT

    [ https://issues.apache.org/jira/browse/HBASE-631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12611386#action_12611386
] 

stack commented on HBASE-631:
-----------------------------

Patch looks good (especially the unit test additions).

Couple of comments/questions:

+ So, if a user adds a cell at row/columnfamily name -- i.e. if row is x and column family
is y:, at x/y: -- they can use get to extract that explicit cell but if they use x/y: with
getFull/getRow, the behavior will be different; instead they will get the explicit x/y: cell
AND all other cells that might on the 'y:' column family such y:1, y:2, etc.  I think this
is OK.
+ I was wondering if the code below:

{code}
+      if(columns != null) {
+        for (byte[] bs : columns) {
+          if(Bytes.toString(bs).split(":").length == 1) {
+            HStore store = stores.get(Bytes.mapKey(HStoreKey.getFamily(bs)));
+            store.getFull(key, null, result);
+          }
+        }
{code}
 
...should be integrated with the code above it?  If only a column family specified, pass null
to the store.getFull?



> HTable.getRow() for only a column family
> ----------------------------------------
>
>                 Key: HBASE-631
>                 URL: https://issues.apache.org/jira/browse/HBASE-631
>             Project: Hadoop HBase
>          Issue Type: New Feature
>    Affects Versions: 0.1.2
>            Reporter: Ryan Lynch
>            Assignee: Jean-Daniel Cryans
>         Attachments: hbase-631-v1.patch
>
>
> I have a use case in which we want to access all the columns for a supplied column family
while also avoiding the overhead of reading any other column families.  There seems to be
no method for doing this but doing a quick dive into the code it "seems" like this should
be relatively easy to allow. This seems desirable since the architecture describes column
families being stored together so reading only a single column family should be a lighter
operation.

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


Mime
View raw message