Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Won't Fix
-
1.4.4, 1.5.0, 1.6.0
-
None
-
None
Description
Right now, the classpath parsing for hte AccumuloClassLoader supports
- non-file URLs
- paths to a jar file
- paths to a directory
- paths with regex matching at the file level
This behavior is undocumented and incompatible with the output of other ecosystem tools, most importantly the "hadoop classpath" command. that command may use shell globbing (depending on the hadoop distro).
Proposed solution:
- abstract the matching behavior
- add matching by shell globbing
- add a flag that determines if regex, globbing, or both (regex falling back to globbing) is used
the flag should default to regex to maintain current out-of-the-box behavior.