[MAPREDUCE-6596] MultipleInputs does not escape Path characters

• Type: Bug
• Status: Patch Available
• Priority: Major
• Resolution: Unresolved
• Component/s: mrv2
• Labels: newbie
• Release Note: Fixed path escaping in MultipleInputs for paths containing commas or semicolons.
• Flags: Patch
• Assignee: Unassigned
• Reporter: Zac Hopkinson
• Created: 30/Dec/15 19:12
• Updated: 07/Jan/16 14:40

Description

Filenames containing commas or semicolons cause MultipleInputs to break since these characters are used for joining and storing the path names.

MultipleInputs stores mapreduce.input.multipleinputs.dir.formats as:

path;inputFormatClass,path2;inputFormatClass2[, ...]

If a filename contains one of the characters used for joining the data then getInputFormatMap and getMapperTypeMap will fail.

Looking at FileInputFormat.addInputPath() it uses escapeString and unescapeString from StringUtils. I took the same approach for escaping in MultipleInputs.