[
https://issues.apache.org/jira/browse/HADOOP-4226?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yuri Pradkin updated HADOOP-4226:
---------------------------------
Status: Patch Available (was: In Progress)
Patched against current trunk.
bq. If buffer ends in \r and the following segment starts with \r, it looks
like this may not separate those lines.
I've modified tests in TestInputFormat to do additional beating on readLine.
One of the tests validates the \r\r sequence. Buffer size varies and at least
a couple of times we should have a read in between of the two \r's.
I've fixed all other nits that you commented on.
bq. It's not obvious that it should work that way at all, but backwards
compatibility is a big deal for this class.
It probably should, as it turns out \r is a newline on older macs and
commodore, I think.
> LineReader::readLine cleanup
> ----------------------------
>
> Key: HADOOP-4226
> URL: https://issues.apache.org/jira/browse/HADOOP-4226
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred
> Affects Versions: 0.19.0
> Reporter: Yuri Pradkin
> Assignee: Yuri Pradkin
> Priority: Minor
> Attachments: HADOOP-4226.patch, HADOOP-4226.patch, HADOOP-4226.patch,
> HADOOP-4226.patch, HADOOP-4226.patch
>
>
> I've been looking at HADOOP-4010 and realized that readLine is pretty
> convoluted. I changed the implementation which made it hopefully a little
> easier to read/validate/understand.
> I've had some problems testing it locally, so I'll submit it for Hudson to
> test.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
