Kindly inform if on the mailing list , once you update it .. Also do u
suggest any external links/docs specifying the same ?

I wanted to no in detail about the way the hive optimizer works and how MR
Job plan is created .. Thanks for ur help !

Thanks!


On Tue, Oct 13, 2009 at 11:23 PM, Ashish Thusoo <[email protected]>wrote:

>  We will also add some more information on that page about the tree walker
> interface and how it is used to do the optimizations.
>
> Ashish
>
>  ------------------------------
> *From:* Zheng Shao [mailto:[email protected]]
> *Sent:* Tuesday, October 13, 2009 10:42 AM
> *To:* [email protected]
> *Subject:* Re: Current Hive Optimizer
>
> AST are built by Hive.g (Using antlr).
> AST -> OperatorTree is done by SemanticAnalyzer.java.
> Optimizations are done by Transformer.java and its sub classes.
> Hope this help you get started.
>
> Zheng
>
> On Tue, Oct 13, 2009 at 10:01 AM, bharath v <
> [email protected]> wrote:
>
>> Thanks for your reply Ning ,
>>
>> I saw that page , but we need some detailed description .. such as how
>> optimizations are carried out .. how ASTs are built etc ..
>> Do u (or any others) have any idea abt them .. Kindly reply
>>
>> Thanks !
>>
>>
>> On Tue, Oct 13, 2009 at 7:10 PM, Ning Zhang <[email protected]> wrote:
>>
>>> The design doc in Hive's wiki page provides some high level
>>> descriptions.  See http://wiki.apache.org/hadoop/Hive/Design
>>> the "compilation" and "optimization" sections.
>>>
>>> Ning
>>>
>>> On Oct 13, 2009, at 4:09 AM, bharath vissapragada wrote:
>>>
>>> > Hi all ,
>>> >
>>> > We are trying to work on the Optimizer part of hive . Can anyone
>>> > point me to a document/link containing how Hive designs it's query
>>> > plans ,
>>> > what meta data it uses , how it optimizes the plans (pushing down
>>> > selects etc) ... Iam facing difficulties is reading the source
>>> > code , so it would
>>> > be nice if any one point me to the link containing the above data or
>>> > any document containing the same.
>>> >
>>> > Any help is highly appreciated !
>>> >
>>> > Thanks !
>>>
>>>
>>
>
>
> --
> Yours,
> Zheng
>

Reply via email to