[
https://issues.apache.org/jira/browse/HDFS-8387?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16116628#comment-16116628
]
Rakesh R commented on HDFS-8387:
--------------------------------
[~andrew.wang]
Following are different cases where it casting long to int in
{{StripedBlockUtil.java}}. Apart from first case all other cases looks mostly
OK to me and please let me know your feedback.
# I think, only below case need to be fixed and should use {{long
lastStripeDataLen}}
{code}
Line 177: final int lastStripeDataLen = (int)(dataSize % stripeSize);
{code}
# Since the Block size is long, so casting is needed here.{code}
Line 182: final int numStripes = (int) ((dataSize - 1) / stripeSize
+ 1);
Line 231: int cellIdxInBlk = (int) (offsetInBlk / cellSize);
{code}
# Uses ByteBuffer and takes integer datatype for position, limit etc, so
casting is needed here.
{code}
Line 320: (int) (rangeStartInBlockGroup % ((long) cellSize *
dataBlkNum));
Line 328: int overLapLen = (int) (overlapEnd - overlapStart +
1);
Line 331: final int pos = (int) (bufOffset + overlapStart -
cellStart);
{code}
# Uses int datatype to represent indexing and integer casting is OK.
{code}
Line 401: int firstCellIdxInBG = (int) (rangeStartInBlockGroup /
cellSize);
Line 402: int lastCellIdxInBG = (int) (rangeEndInBlockGroup /
cellSize);
{code}
# Cell size is int type, so the result will be within int value and uses int
datatype for storing the values.
{code}
Line 406: final int firstCellOffset = (int) (rangeStartInBlockGroup
% cellSize);
Line 408: (int) Math.min(cellSize - (rangeStartInBlockGroup %
cellSize), len);
Line 412: final int lastCellSize = (int) (rangeEndInBlockGroup %
cellSize) + 1;
Line 521: int overLapLen = (int) (overlapEnd - overlapStart +
1);
Line 531: (int) (done + overlapStart - cellStart),
overLapLen);
Line 919: return (int) (reportedBlock.getBlockId() &
{code}
> Erasure Coding: Revisit the long and int datatypes usage in striping logic
> --------------------------------------------------------------------------
>
> Key: HDFS-8387
> URL: https://issues.apache.org/jira/browse/HDFS-8387
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: Rakesh R
> Assignee: Rakesh R
> Labels: hdfs-ec-3.0-nice-to-have
>
> This idea of this jira is to revisit the usage of {{long}} and {{int}} data
> types in the striping logic.
> Related discussion
> [here|https://issues.apache.org/jira/browse/HDFS-8294?focusedCommentId=14540788&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14540788]
> in HDFS-8294
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
