phoenix-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Rajeshbabu Chintaguntla (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (PHOENIX-1060) Replace ReferencingColumn with ColumnReference
Date Fri, 06 Mar 2015 19:24:38 GMT

     [ https://issues.apache.org/jira/browse/PHOENIX-1060?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Rajeshbabu Chintaguntla updated PHOENIX-1060:
---------------------------------------------
    Attachment: PHOENIX-1060_v2.patch

Thanks for the review [~jamestaylor].
Here is the Patch addressing the comments.
bq. Might be cleanest to just add a ReadOnlyImmutableBytesPtr as this might come up again.
Added ReadOnlyImmutableBytesPtr.
bq. An alternative might be to get rid of those methods, but there are probably a lot of callers
(though it's possible they could all be converted to be ok with an ImmutableBytesWritable).
It's not good because most of the places we want family or qualifier preparing filters,scan
objects which will not accept ImmutableBytesWritable mostly. So just creating family,qualifier
byte arrays on demand. 

> Replace ReferencingColumn with ColumnReference
> ----------------------------------------------
>
>                 Key: PHOENIX-1060
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-1060
>             Project: Phoenix
>          Issue Type: Sub-task
>            Reporter: rajeshbabu
>            Assignee: Rajeshbabu Chintaguntla
>            Priority: Minor
>             Fix For: 5.0.0, 4.3.1
>
>         Attachments: PHOENIX-1060.patch, PHOENIX-1060_v2.patch
>
>
> We currently have two separate classes that represent a column family/column qualifier
combo: ColumnReference and ReferencingColumn. I believe the only difference is that ReferencingColumn
allows a backing byte array to be used with an offset and length. Since ColumnReference already
has an ImmutableBytesPtr we can add a constructor that takes two offsets and lengths for the
family and qualifier and do a bit of minor refactoring (removing the family and qualifier
member variables and just use the familyPtr and qualifierPtr ones exclusivly) so that we don't
have two classes for the same thing:
> {code}
> public class ColumnReference implements Comparable<ColumnReference> {
>     
>   public ColumnReference(byte[] family, int familyOffset, int familyLength, byte[] qualifier,
int qualifierOffset, int qualiferLength) {
>     this.familyPtr = new ImmutableBytesPtr(family, familyOffset, familyLength);
>     this.qualifierPtr = new ImmutableBytesPtr(qualifier, qualifierOffset, qualifierLength);
>     this.hashCode = calcHashCode(familyPtr.get(), qualifierPtr.get());
>   }
> {code}
> Another useful addition would be to implement Writable so we don't duplicate the serialization
logic.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message