Same flow - initialize() is called on the GenericUDF very soon after construction, by the same methods that created the UDF.
Take a look at TypeCheckProcFactory or ExprNodeGenericFuncDesc ________________________________ From: Anirudh Paramshetti <anirudh2...@gmail.com> Sent: Tuesday, February 02, 2016 9:24 PM To: user@hive.apache.org Subject: Re: GenericUDF Thanks Jason for your inputs. I believe you are talking about the number of instances created, which explains why the constructor was called thrice. But I'm still unclear about the two calls made to the initialize method, when I use the temporary function in the query. Can you put some more light on the call flow to the initialize method.? Regards, Anirudh Paramshetti On Wed, Feb 3, 2016 at 6:08 AM, Jason Dere <jd...@hortonworks.com<mailto:jd...@hortonworks.com>> wrote: - Created once when registering the function to the FunctionRegistry. - The UDF is copied from the version in the registry during query compilation - The query plan is serialized, then deserialized by the tasks during query execution, which constructs another instance of the UDF. ________________________________ From: Anirudh Paramshetti <anirudh2...@gmail.com<mailto:anirudh2...@gmail.com>> Sent: Tuesday, February 02, 2016 6:29 AM To: user@hive.apache.org<mailto:user@hive.apache.org> Subject: GenericUDF Hi, I have written a custom UDF in Java extending the GenericUDF class. I have some print statements in the constructor and initialize method, as to understand the number of calls made to them. From what I have read about GenericUDF, I was expecting the constructor and initialize method to be called once per UDF instance. But what I found out was, the constructor was called thrice(once while creating the temporary function and twice while using it in the hive query) and the initialize method was called twice(while using it in the hive query). UDF output: hive> create temporary function replace as 'package.name.GenericNullReplacement'; Inside constructor of GenericNullReplacement hive> select replace(column_name, 0.01) from dummy_table; Inside constructor of GenericNullReplacement Inside constructor of GenericNullReplacement Inside initialize() method of GenericNullReplacement Inside initialize() method of GenericNullReplacement 1.23 4.56 4.56 0.01 4.56 9.56 It would be great if someone could explain me what is happening here? Thanks and Regards, Anirudh Paramshetti