[
http://issues.apache.org/jira/browse/HADOOP-473?page=comments#action_12429789 ]
James P. White commented on HADOOP-473:
---------------------------------------
The proposed patch does read-ahead/seek-back even when the initial newline
character is '\n'.
That is needlessly wasteful. The could should be simplified to:
if (c == '\n')
break;
if (c == '\r') {
long curPos = in.getPos();
char nextC = (char)in.read();
// EOF -1 is also != '\n' so will work with Mac and EOF
if (nextC != '\n')) {
// seek back 1 char because the line ending is only 1 char
in.seek(curPos);
}
break;
}
> TextInputFormat does not correctly handle all line endings
> ----------------------------------------------------------
>
> Key: HADOOP-473
> URL: http://issues.apache.org/jira/browse/HADOOP-473
> Project: Hadoop
> Issue Type: Bug
> Components: mapred
> Affects Versions: 0.5.0, 0.6.0
> Environment: All environments
> Reporter: Dennis Kubes
> Attachments: text-input-format.patch
>
>
> The current TextInputFormat readLine method calls break on either a single
> '\r' or '\n' character. This causes windows formatted text files '\r' '\n'
> to leave a trailing '\n' character and the next time the readLine method is
> called on the same input stream it returns a blank string. The patch
> attached corrects this issue by looking for either single or double character
> line endings and positions the input stream to the next line. It correctly
> handles windows, mac, and unix line endings.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
