PIG-619: Create one InputSplit even when the input file is zero length so that hadoop runs maps and creates output for the next job.
git-svn-id: https://svn.apache.org/repos/asf/hadoop/pig/branches/pre-multiquery-phase2@775346 13f79535-47bb-0310-9956-ffa450edef68
diff --git a/CHANGES.txt b/CHANGES.txt
index db99021..bd678f6 100644
--- a/CHANGES.txt
+++ b/CHANGES.txt
@@ -24,6 +24,10 @@
IMPROVEMENTS
+PIG-619: Create one InputSplit even when the input file is zero length
+ so that hadoop runs maps and creates output for the next
+ job (gates).
+
PIG-693: Proposed improvements to pig's optimizer (sms)
PIG-700: To automate the pig patch test process (gkesavan via sms)
diff --git a/src/org/apache/pig/backend/executionengine/PigSlicer.java b/src/org/apache/pig/backend/executionengine/PigSlicer.java
index 8cc638c..00b149e 100644
--- a/src/org/apache/pig/backend/executionengine/PigSlicer.java
+++ b/src/org/apache/pig/backend/executionengine/PigSlicer.java
@@ -100,6 +100,12 @@
// Anything that ends with a ".gz" we must process as a complete
// file
slices.add(new PigSlice(name, funcSpec, 0, size));
+ } else if (size == 0) {
+ // add one empty slice. This is a total hack to deal with the
+ // case where hadoop isn't starting maps for empty arrays of
+ // InputSplits. See PIG-619. This should be removed
+ // once we determine why this is.
+ slices.add(new PigSlice(name, funcSpec, 0, bs));
} else {
while (pos < size) {
if (pos + bs > size) {
