[
https://issues.apache.org/jira/browse/HDFS-14585?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16874403#comment-16874403
]
stack commented on HDFS-14585:
------------------------------
nit: why the change in import order for nonnull and Preconditions? Is it about
respecting checkstyle import ordering rules?
nit: Best to not make formatting changes that are unrelated to your patch: e.g.
wrapping exception declaration on blockSeekTo (there are a few of this type of
change). Formatting bulks up your patch and distract the reviewer.
nit: ByteBuffer bb = ByteBuffer.wrap(buffer, offset, length); is offset.
Some nice cleanup and duplication removal; e.g. using BB to keep account on the
read buffer (offsets and length), purge of the HDFS-8703 unused EC version of
actualGetFromOneDataNode, pulling out long targetEnd = targetStart +
bytesToRead - 1, etc.
Is this going to be ok: 1249 tmp.limit(tmp.position() + len); ? The
EC version of actualGetFromOneDataNode had a checkReadPortions. Should there be
a check we don't go over the end of the buffer here?
Why do we drop the below in the patch?
1268 updateReadStatistics(readStatistics, nread, reader);
1269 dfsClient.updateFileSystemReadStats(
1270 reader.getNetworkDistance(), nread);
Thanks.
> Backport HDFS-8901 Use ByteBuffer in DFSInputStream#read to branch-2 and
> branch2.9
> ----------------------------------------------------------------------------------
>
> Key: HDFS-14585
> URL: https://issues.apache.org/jira/browse/HDFS-14585
> Project: Hadoop HDFS
> Issue Type: Task
> Reporter: Lisheng Sun
> Assignee: Lisheng Sun
> Priority: Major
> Attachments: HDFS-14585.branch-2.9.v1.patch,
> HDFS-14585.branch-2.v1.patch
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
