[ https://issues.apache.org/jira/browse/MAPREDUCE-5061?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13601707#comment-13601707 ]
Thierry Granger commented on MAPREDUCE-5061: -------------------------------------------- Maybe = is not a good idea for separator as that would break the backward compatibility, so a smart regexp seems to be more appropriate. > MultipleInputs can not manage path with glob characters like {a,b} > ------------------------------------------------------------------ > > Key: MAPREDUCE-5061 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5061 > Project: Hadoop Map/Reduce > Issue Type: Bug > Reporter: Thierry Granger > Priority: Minor > > Hello, > The org.apache.hadoop.mapreduce.lib.input.MultipleInputs expects a > configuration with : > * the parameter mapreduce.input.multipleinputs.dir.mappers in the format > <path 1>,<mapper class 1>;<path 2>,<mapper class 2> > * the parameter mapreduce.input.multipleinputs.dir.formats in the format > <path 1>,<input format class 1>;<path 2>,<input format class 2> > The usage of the comma as separator unable the use of the glob characters > {a,b} in the path (like : hdfs://localhost/user/foo/data201{2,3} ). > I suggest to change the separator to another non-problematic one (maybe =). > I patched the MultipleInputs code locally and found an other related issue : > when the path are created, the string representing the path is not unescape > (like it's done in the > org.apache.hadoop.mapreduce.lib.input.FileInputFormat.getInputPaths method) > which makes an invalid path in case of glob characters {a\,b} > Thank you -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira