[ 
https://issues.apache.org/jira/browse/FLINK-8903?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16401160#comment-16401160
 ] 

ASF GitHub Bot commented on FLINK-8903:
---------------------------------------

GitHub user fhueske opened a pull request:

    https://github.com/apache/flink/pull/5706

    [FLINK-8903] [table] Fix VAR_SAMP, VAR_POP, STDEV_SAMP, STDEV_POP functions 
on GROUP BY windows.

    ## What is the purpose of the change
    
    * Fixes the computation of `VAR_SAMP`, `VAR_POP`, `STDDEV_SAMP`, 
`STDDEV_POP` aggregations in the context of `GROUP BY` windows (`TUMBLE`, 
`HOP`, `SESSION`). Right now, these methods are computed as `AVG`.
    
    ## Brief change log
    
    * copy Calcite's `AggregateReduceFunctionsRule` to Flink and improve its 
extensibility
    * add a `WindowAggregateReduceFunctionsRule` based on the copied 
`AggregateReduceFunctionsRule` to decompose the faulty aggregation functions 
into `COUNT` and `SUM` functions.
    * add restriction to `FlinkLogicalWindowAggregateConverter` to prevent 
translation of group window aggregates with failing aggregation functions
    * prevent translation of `VAR_SAMP`, `VAR_POP`, `STDDEV_SAMP`, `STDDEV_POP` 
in `AggregateUtil`
    * add unit tests (plan validation) for batch (SQL, Table API) and stream 
(SQL, Table API)
    
    ## Verifying this change
    
    * run the added plan tests
    
    ## Does this pull request potentially affect one of the following parts:
    
      - Dependencies (does it add or upgrade a dependency): **no**
      - The public API, i.e., is any changed class annotated with 
`@Public(Evolving)`: **no**
      - The serializers: **no**
      - The runtime per-record code paths (performance sensitive): **no**
      - Anything that affects deployment or recovery: JobManager (and its 
components), Checkpointing, Yarn/Mesos, ZooKeeper: **no**
      - The S3 file system connector: **no**
    
    ## Documentation
    
      - Does this pull request introduce a new feature? **no**
      - If yes, how is the feature documented? **n/a**


You can merge this pull request into a Git repository by running:

    $ git pull https://github.com/fhueske/flink tableVarStddevAggFix

Alternatively you can review and apply these changes as the patch at:

    https://github.com/apache/flink/pull/5706.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

    This closes #5706
    
----
commit 517567348b0ec0c23ef0c1dcc05c54a91d5c5671
Author: Fabian Hueske <fhueske@...>
Date:   2018-03-15T20:04:00Z

    [FLINK-8903] [table] Fix VAR_SAMP, VAR_POP, STDEV_SAMP, STDEV_POP functions 
on GROUP BY windows.

----


> Built-in agg functions VAR_POP, VAR_SAMP, STDEV_POP, STDEV_SAMP are broken in 
> Group Windows
> -------------------------------------------------------------------------------------------
>
>                 Key: FLINK-8903
>                 URL: https://issues.apache.org/jira/browse/FLINK-8903
>             Project: Flink
>          Issue Type: Bug
>          Components: Table API &amp; SQL
>    Affects Versions: 1.3.2, 1.5.0, 1.4.2
>            Reporter: lilizhao
>            Assignee: Fabian Hueske
>            Priority: Critical
>             Fix For: 1.5.0
>
>         Attachments: QQ图片20180312180143.jpg, TableAndSQLTest.java
>
>
> The built-in aggregation functions VAR_POP, VAR_SAMP, STDEV_POP, STDEV_SAMP 
> are translated into regular AVG functions if they are applied in the context 
> of a Group Window aggregation (\{{GROUP BY TUMBLE/HOP/SESSION}}).
> The reason is that these functions are internally represented as 
> {{SqlAvgAggFunction}} but with different {{SqlKind}}. When translating 
> Calcite aggregation functions to Flink Table agg functions, we only look at 
> the type of the class, not at the value of the {{kind}} field. We did not 
> notice that before, because in all other cases (regular {{GROUP BY}} without 
> windows or {{OVER}} windows, we have a translation rule 
> {{AggregateReduceFunctionsRule}} that decomposes the more complex functions 
> into expressions of {{COUNT}} and {{SUM}} functions such that we never 
> execute an {{AVG}} Flink function. That rule can only be applied on 
> {{LogicalAggregate}}, however, we represent group windows as 
> {{LogicalWindowAggregate}}, so the rule does not match.
> We should fix this by:
> 1. restrict the translation to Flink avg functions in {{AggregateUtil}} to 
> {{SqlKind.AVG}}. 
> 2. implement a rule (hopefully based on {{AggregateReduceFunctionsRule}}) 
> that decomposes the complex agg functions into the {{SUM}} and {{COUNT}}.
> Step 1. is easy and a quick fix but we would get an exception "Unsupported 
> Function" if {{VAR_POP}} is used in a {{GROUP BY}} window.
> Step 2. might be more involved, depending on how difficult it is to port the 
> rule.
>  



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to