hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jason Dere <jd...@hortonworks.com>
Subject Re: Review Request 59080: HIVE-16610
Date Tue, 09 May 2017 20:31:57 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/59080/#review174364
-----------------------------------------------------------




ql/src/java/org/apache/hadoop/hive/ql/optimizer/DynamicPartitionPruningOptimization.java
Line 216 (original), 217 (patched)
<https://reviews.apache.org/r/59080/#comment247492>

    There should be a better way to determine which table the column comes from.
    This should also be done after confirming the join key is a column reference.



ql/src/java/org/apache/hadoop/hive/ql/optimizer/DynamicPartitionPruningOptimization.java
Lines 226 (patched)
<https://reviews.apache.org/r/59080/#comment247491>

    This is automatically using the first TS in your list of candidates as the tableAlias
- which is not necessarily the table the key column comes from.



ql/src/java/org/apache/hadoop/hive/ql/optimizer/DynamicPartitionPruningOptimization.java
Line 463 (original), 490 (patched)
<https://reviews.apache.org/r/59080/#comment247489>

    Can you change this assert to Preconditions.checkNotNull()



ql/src/java/org/apache/hadoop/hive/ql/parse/SemanticAnalyzer.java
Line 9064 (original), 9058 (patched)
<https://reviews.apache.org/r/59080/#comment247487>

    This whole parsing logic seems hard to follow .. I don't see the need for a loop to parse
each hint argument. If the format is (alias, colName, number) in that order, can't you just
parse each argument in order, if the args tree has enough children?
    
    One effect of changing this is you only have to insert the hints in one place in the code.


- Jason Dere


On May 9, 2017, 7:32 a.m., Deepak Jaiswal wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/59080/
> -----------------------------------------------------------
> 
> (Updated May 9, 2017, 7:32 a.m.)
> 
> 
> Review request for hive and Jason Dere.
> 
> 
> Repository: hive-git
> 
> 
> Description
> -------
> 
> Semijoin Hint : Should able to handle more than 1 hint per alias.
> - If a user provides a hint, only that semijoin optimization is performed.
> - If there are more than 1 hints for an alias, it is upto user to make sure the (alias
+ colName) combo is unique.
> - Works with join aliases.
> - Refactored the logic to generate semijoin to keep the inner loops count down.
> 
> 
> Diffs
> -----
> 
>   ql/src/java/org/apache/hadoop/hive/ql/optimizer/DynamicPartitionPruningOptimization.java
f56b9cbb18 
>   ql/src/java/org/apache/hadoop/hive/ql/parse/ParseContext.java 6de4bcd64a 
>   ql/src/java/org/apache/hadoop/hive/ql/parse/SemanticAnalyzer.java 5115fc8090 
>   ql/src/java/org/apache/hadoop/hive/ql/parse/SemiJoinHint.java 1f24e23ff3 
>   ql/src/test/queries/clientpositive/semijoin_hint.q 5fbc273e7d 
>   ql/src/test/results/clientpositive/llap/semijoin_hint.q.out 3ffc2352a4 
> 
> 
> Diff: https://reviews.apache.org/r/59080/diff/1/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Deepak Jaiswal
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message