File tree 1 file changed +11
-0
lines changed
1 file changed +11
-0
lines changed Original file line number Diff line number Diff line change
1
+ * More execution options
2
+
3
+ Parkour currently inherits Hadoop’s default job-failure behavior when a job’s
4
+ output directory already exists. It would be nice to support other behaviors,
5
+ such as skipping existing, overwriting existing, or a make-like model where
6
+ outputs are overwritten when their inputs have changed.
7
+
1
8
* Integration for writing other Hadoop classes
2
9
3
10
** Input formats
@@ -17,3 +24,7 @@ var-trampolining implementations of the various Hadoop classes which Hadoop
17
24
expects to locate via name-based reflection. A more general implementation
18
25
would use a custom ClassLoader which could dynamically generate and return new
19
26
instances of the supported types.
27
+
28
+ The ClassLoader approach may not however be possible. It’s not clear if Hadoop
29
+ allows tasks to configure an alternative ClassLoader prior to task
30
+ initialization.
You can’t perform that action at this time.
0 commit comments