[
https://issues.apache.org/jira/browse/TAJO-2027?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15287860#comment-15287860
]
ASF GitHub Bot commented on TAJO-2027:
--------------------------------------
Github user hyunsik commented on a diff in the pull request:
https://github.com/apache/tajo/pull/1026#discussion_r63620568
--- Diff: tajo-docs/src/main/sphinx/functions/hivefunc.rst ---
@@ -0,0 +1,81 @@
+##############
+Hive Functions
+##############
+
+Tajo provides a feature to use Hive functions directly without
re-compilation or additional code.
+
+*************
+Configuration
+*************
+
+Only thing to do is registering path to a directory for jar files
containing your hive functions.
+You can do this by set ``tajo.function.hive.code-dir`` in
``tajo-site.xml`` like the following.
+
+.. code-block:: xml
+
+ <property>
+ <name>tajo.function.hive.code-dir</name>
+ <value>/path/to/hive/function/jar</value>
+ </property>
+
+.. note::
+ The path should be one in local filesystem. HDFS directory is not
supported because of JAVA URI compatability problem.
+
+.. warning::
+
+ The path must point to a directory, not a file. And multiple directory
entries are not allowed.
+ However, it is possible to load multiple jar files.
+
+***************
+Using in detail
+***************
+
+=============
+Function Name
+=============
+
+Tajo reads hive functions override ``org.apache.hadoop.hive.ql.exec.UDF``
class. Function name is used as specified in
+``@Description`` annotation. If it doesn't exist, Tajo uses full qualified
class name as function name. For example,
+it can be like this : ``select com_example_hive_udf_myupper('abcd')``, so
it is recommended to use Description annotation.
+
+And if some function conflict occurs, it may throw
``AmbiguousFunctionException``. This conflict means about function signature,
+not only about function name.
+
+============================
+Parameter type / Return type
+============================
+
+Hive uses *Writable* type of Hadoop in functions, but Tajo uses its
internal *Datum* type.
+Because Tajo doesn't support a kind of pluggable type system yet, only
some Writable types are supported currently by internal converting.
--- End diff --
I think the statement 'Because Tajo doesn't support a kind of pluggable
type system yet' is unnecessary. IMO, it may cause other misunderstanding.
> Writing Hive UDF integration document
> -------------------------------------
>
> Key: TAJO-2027
> URL: https://issues.apache.org/jira/browse/TAJO-2027
> Project: Tajo
> Issue Type: Task
> Components: Documentation
> Reporter: Jongyoung Park
> Assignee: Jongyoung Park
> Priority: Minor
> Fix For: 0.11.0
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
