[
https://issues.apache.org/jira/browse/TAJO-1686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15186590#comment-15186590
]
ASF GitHub Bot commented on TAJO-1686:
--------------------------------------
Github user jihoonson commented on a diff in the pull request:
https://github.com/apache/tajo/pull/929#discussion_r55475795
--- Diff: tajo-core/src/main/java/org/apache/tajo/master/TajoMaster.java ---
@@ -215,12 +216,48 @@ public void serviceInit(Configuration conf) throws
Exception {
LOG.info("Tajo Master is initialized.");
}
+ private Collection<FunctionDesc> loadFunctions() throws IOException,
AmbiguousFunctionException {
+ List<FunctionDesc> functionList = new
ArrayList<>(FunctionLoader.loadBuiltinFunctions().values());
+
+ HashMap<Integer, FunctionDesc> funcSet = new HashMap<>();
+
+ for (FunctionDesc desc: functionList) {
+ funcSet.put(desc.hashCodeWithoutType(), desc);
+ }
+
+
checkUDFduplicateAndMerge(FunctionLoader.loadUserDefinedFunctions(systemConf),
funcSet, functionList);
+ checkUDFduplicateAndMerge(HiveFunctionLoader.loadHiveUDFs(systemConf),
funcSet, functionList);
+
+ return functionList;
+ }
+
+ /**
+ * Checks duplicates between pre-loaded functions and UDFs. And they are
meged to funcList.
+ *
+ * @param udfs UDF list
+ * @param funcSet set for pre-loaded functions to match signature
+ * @param funcList list to be merged
+ * @throws AmbiguousFunctionException
+ */
+ private void checkUDFduplicateAndMerge(Optional<List<FunctionDesc>>
udfs, HashMap<Integer, FunctionDesc> funcSet, List<FunctionDesc> funcList)
--- End diff --
It would be better if you add a test for this method.
> Allow Tajo to use Hive UDF
> --------------------------
>
> Key: TAJO-1686
> URL: https://issues.apache.org/jira/browse/TAJO-1686
> Project: Tajo
> Issue Type: New Feature
> Components: Function/UDF
> Reporter: Jaehwa Jung
> Assignee: Jongyoung Park
>
> Hive has been widely used in this area. Many users have maintained lots of
> big tables through Hive metastore using HiveQL and UDFs. Currently, Tajo
> provides own UDF and Hive users can implement their UDFs in Tajo. But if we
> can wrap Hive UDF in Tajo, it seems that they would be able to use Tajo
> easily for their analysis infrastructure.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
