Do not push down filter to ORC for union type schema (#123)

linkedin · Aug 23, 2022 · f2e5bd2 · f2e5bd2
1 parent 21c3a80
commit f2e5bd2
Showing 1 changed file with 5 additions and 1 deletion.
diff --git a/orc/src/main/java/org/apache/iceberg/orc/OrcIterable.java b/orc/src/main/java/org/apache/iceberg/orc/OrcIterable.java
@@ -102,8 +102,12 @@ public CloseableIterator<T> iterator() {
     // to push down filters to ORC's SearchArgument, since we are not reading anything from files at all
     boolean isEmptyStruct = readOrcSchema.getChildren().size() == 0;
 
+    // If the projected ORC schema has union type, we can't reliably build a pushdown filter for the
+    // underlying ORC files, since the Iceberg schema has a different representation than the ORC schema.
+    boolean containsUnion = readOrcSchema.toString().contains("uniontype");
+
     SearchArgument sarg = null;
-    if (filter != null && !isEmptyStruct) {
+    if (filter != null && !isEmptyStruct && !containsUnion) {
       Expression boundFilter = Binder.bind(schema.asStruct(), filter, caseSensitive);
       sarg = ExpressionToSearchArgument.convert(boundFilter, readOrcSchema);
     }