> On May 9, 2014, 5:48 p.m., Ashutosh Chauhan wrote:
> > ql/src/java/org/apache/hadoop/hive/ql/exec/PTFOperator.java, line 58
> > <https://reviews.apache.org/r/21168/diff/1/?file=576144#file576144line58>
> >
> >     Can you add a comment why we need to keep track for first row processed 
> > in Map?

done


> On May 9, 2014, 5:48 p.m., Ashutosh Chauhan wrote:
> > ql/src/java/org/apache/hadoop/hive/ql/exec/PTFOperator.java, line 319
> > <https://reviews.apache.org/r/21168/diff/1/?file=576144#file576144line319>
> >
> >     Better name : outputPartRowsItr?

done


> On May 9, 2014, 5:48 p.m., Ashutosh Chauhan wrote:
> > ql/src/java/org/apache/hadoop/hive/ql/udf/ptf/TableFunctionEvaluator.java, 
> > line 96
> > <https://reviews.apache.org/r/21168/diff/1/?file=576148#file576148line96>
> >
> >     Comment made sense. Since like those fields are not present in class 
> > any more. Shall we just get rid of this?

this is needed; transient based on BeanInfo (get/set methods) in class


> On May 9, 2014, 5:48 p.m., Ashutosh Chauhan wrote:
> > ql/src/java/org/apache/hadoop/hive/ql/udf/ptf/TableFunctionEvaluator.java, 
> > line 218
> > <https://reviews.apache.org/r/21168/diff/1/?file=576148#file576148line218>
> >
> >     Better name: canAcceptInputAsStream?

done


> On May 9, 2014, 5:48 p.m., Ashutosh Chauhan wrote:
> > ql/src/java/org/apache/hadoop/hive/ql/exec/FunctionRegistry.java, line 449
> > <https://reviews.apache.org/r/21168/diff/1/?file=576143#file576143line449>
> >
> >     Instead of adding noop functions statically, we should put these 
> > functions in test/ package and than do add jar for testing. Multiple 
> > reasons:
> >     * Better to isolate test code from production code.
> >     * It will also exercise add jar functionality for PTF functions for 
> > which I am not sure we have coverage.
> >     * These functions also show up in default list of inbuilt functions. It 
> > may confuse user to wonder what good these functions are for. 
> > show_functions.q failed because of this.

Agree with your comments on Noop. This was done because for testing we need a 
PTF and Noop has some special short circuit path for Partition handling.
But can we do this as a separate Jira; removing references to Noop in the 
translation code is non trivial work.


- Harish


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/21168/#review42578
-----------------------------------------------------------


On May 14, 2014, 9:21 p.m., Harish Butani wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/21168/
> -----------------------------------------------------------
> 
> (Updated May 14, 2014, 9:21 p.m.)
> 
> 
> Review request for hive and Ashutosh Chauhan.
> 
> 
> Bugs: HIVE-6999
>     https://issues.apache.org/jira/browse/HIVE-6999
> 
> 
> Repository: hive-git
> 
> 
> Description
> -------
> 
> There are a set of use cases where the Table Function can operate on a 
> Partition row by row or on a subset(window) of rows as it is being streamed 
> to it.
> Windowing has couple of use cases of this:processing of Rank functions, 
> processing of Window Aggregations.
> But this is a generic concept: any analysis that operates on an Ordered 
> partition maybe able to operate in Streaming mode.
> This patch introduces streaming mode in PTFs and provides the mechanics to 
> handle PTF chains that contain both modes of PTFs.
> Subsequent patches will introduce Streaming mode for Windowing.
> 
> 
> Diffs
> -----
> 
>   ql/src/java/org/apache/hadoop/hive/ql/exec/FunctionRegistry.java 3bb8fa9 
>   ql/src/java/org/apache/hadoop/hive/ql/exec/PTFOperator.java 4d314b7 
>   ql/src/java/org/apache/hadoop/hive/ql/parse/PTFTranslator.java 34aebf0 
>   ql/src/java/org/apache/hadoop/hive/ql/udf/ptf/NoopStreaming.java 
> PRE-CREATION 
>   ql/src/java/org/apache/hadoop/hive/ql/udf/ptf/NoopWithMapStreaming.java 
> PRE-CREATION 
>   ql/src/java/org/apache/hadoop/hive/ql/udf/ptf/TableFunctionEvaluator.java 
> 1087bbf 
>   ql/src/test/queries/clientpositive/ptf_streaming.q PRE-CREATION 
>   ql/src/test/results/clientpositive/ptf_streaming.q.out PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/21168/diff/
> 
> 
> Testing
> -------
> 
> added new tests
> 
> 
> Thanks,
> 
> Harish Butani
> 
>

Reply via email to