[ 
https://issues.apache.org/jira/browse/PHOENIX-939?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13982056#comment-13982056
 ] 

James Taylor commented on PHOENIX-939:
--------------------------------------

Thanks for following up on this, [~maghamravikiran]. Perhaps we can just leave 
it up to the user? How is this Pig schema that you generated used down-the-road 
by the user? Are they creating a Pig table through querying Phoenix? What do 
you think,  [~prkommireddi]?

> Generalize SELECT expressions for Pig Loader
> --------------------------------------------
>
>                 Key: PHOENIX-939
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-939
>             Project: Phoenix
>          Issue Type: Improvement
>    Affects Versions: 5.0.0, 3.1, 4.1
>            Reporter: James Taylor
>            Assignee: maghamravikiran
>             Fix For: 5.0.0, 3.1, 4.1
>
>
> The current Pig Loader requires that the query contain only column references 
> in the SELECT expressions. Instead, we should allow any expression as that 
> will provide more general utility. For example, built-in functions, sequence 
> references, etc. could be used then.
> Validation can be done by simply compiling the query. This does all the 
> validation required. It's ok if it's compiled twice if need be too.
> Pig doesn't know and likely wouldn't care if the expressions in the SELECT 
> correspond to columns in Phoenix or general expressions. You can use the 
> ColumnProjection.getName() method to get back an alias of the SELECT 
> expression. If no alias is provided, then the String of the expression is 
> returned. [~prkommireddi] - can you weigh in here? You can use the Phoenix 
> RowProjector to iterate through each ColumnProjector you get back after the 
> compile to get the alias of the select expression (i.e. you'd give this to 
> Pig as the "column" name) plus the data type. Note that if this is 
> problemattic, then you could likely generate an alias name if one is not 
> present.
> For example:
> {code}
>     RowProjector rowProj = queryPlan.getProjector();
>     for (ColumnProjector colProj : rowProj.getColumnProjectors()) {
>         String columnName = colProj.getName();
>         PDataType dataType = colProj.getExpression().getDataType();
>     }
> {code}
> If the SELECT expressions are simple column references, this would be exactly 
> the same as is being done now. If the SELECT expressions are more complex 
> expressions, this would work as well.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to