[ 
https://issues.apache.org/jira/browse/CASSANDRA-11580?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15292830#comment-15292830
 ] 

Stefania commented on CASSANDRA-11580:
--------------------------------------

Really great stuff Yuki, I could only find minor things and I had to look 
really hard to find them :) The items marked as optional are left to your 
decision since they are more down to individual stylistic preferences:

*SegmentedFile:*

* Class comments are no longer up-to-date
* protected accessors can be private
* Optional: class can be final
* Comments of onDiskLength mention SegmentIterator that no longer exists
* Trivial: alignment error in the constructor parameters at line 66
* Cleanup.Tidy(): the chunk cache is invalidated only when the metadata is not 
null, this was the existing behavior but is it correct? Index files use a chunk 
cache even if they are not compressed?
* Cleanup.Tidy(): can either metdata.close or the cache invalidation throw? 
* Optional: should we rename metadata to compressionMetadata in Cleanup?
* Class comments of {{Builder}} are also no longer up-to-date
* Suppress resource warnings for rebufferer in Builder.complete() since it is 
owned by the SegmentedFile

*Other files:*

* SStableReader ln 436: {{// special implementation of load to use non-pooled 
SegmentedFile builders}} can be removed
* The EMPTY BufferHolder in Rebuffered.java at line 62 can still be static final
* SegmentedFileTest should be renamed to DiskOptimizationStrategyTest
* Optional: There is a bit of code duplications in BigTableWriter openFinal and 
openForBatch as well as SSTableReader openForBatch and load. I wonder if we 
could introduce helper methods in SSTable to create the index and data builders 
and to create the index and data segmented files.

> remove DatabaseDescriptor dependency from SegmentedFile
> -------------------------------------------------------
>
>                 Key: CASSANDRA-11580
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11580
>             Project: Cassandra
>          Issue Type: Sub-task
>            Reporter: Yuki Morishita
>            Assignee: Yuki Morishita
>             Fix For: 3.x
>
>
> Several configurable parameters are pulled from {{DatabaseDescriptor}} from 
> {{SegmentedFile}} and its subclasses.



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

Reply via email to