Hi Swapna,
All our aggregate functions allow expressions as arguments and it wouldn't
make sense to have these new ones be different. A reference to a column is
also an expression. It doesn't change the HBase data model being sparse.

I think the next step should be for you to submit a patch that the
community can take a look at, as it's too difficult to discuss this without
that.

Thanks,
James

On Tuesday, May 3, 2016, Swapna Swapna <[email protected]> wrote:

> Hi James,
>
> Thanks for your swift response.
>
> I wouldn't be able to use the expression in the below query rather I would
> have to provide the columns (as arguments) which I'm interested in to
> perform the aggregation on respective provided columns.
>
> myaggFunc(col1,col2)
>
> the reason being, the hbase data is sparsed and I would not know the column
> values. Data fetch is based on a row key.
>
> expression example:
>
> ID=1 OR NAME='Hi'
>
> Regards
>
> Swapna
>
>
>
> On Tue, May 3, 2016 at 7:17 PM, James Taylor <[email protected]
> <javascript:;>> wrote:
>
> > Hi Swapna,
> > The return type is typically derived from looking at the return types of
> > each of the input arguments and choosing what'll work without losing
> > precision. For example, take a look at this loop in ExpressionCompiler
> that
> > determines this for expressions that are added together:
> >
> >         new ArithmeticExpressionFactory() {
> >             @Override
> >             public Expression create(ArithmeticParseNode node,
> > List<Expression> children) throws SQLException {
> >                 boolean foundDate = false;
> >                 Determinism determinism = Determinism.ALWAYS;
> >                 PDataType theType = null;
> >                 for(int i = 0; i < children.size(); i++) {
> >
> > Your probably already doing this, but make sure you don't assume the
> > arguments are column references, but allow them to be any expression.
> >
> > Also, it'd be great to see what you've got so far without handling
> multiple
> > arguments to your function (in the form of a pull request) so folks can
> get
> > you feedback on your work so far.
> >
> > Thanks, and we appreciate the contributions!
> >
> > James
> >
> > On Tue, May 3, 2016 at 12:59 PM, Swapna Swapna <[email protected]
> <javascript:;>>
> > wrote:
> >
> > > Sure,
> > >
> > > Hbase data that I have is:
> > >
> > > rowkey                us         uk
> > > 20161001           3            4
> > > 20161002           1            2
> > >
> > >
> > > select myaggFunc(us) from table :    // this is returning output as :
> > > 4
> > > select myaggFunc(uk) from table :    // this is returning output as :
> > > 6
> > >
> > > In similar to that, i'm visualizing the query like: select
> > > myaggFunc1(us,uk)
> > > from table;  //with multiple columns
> > >
> > > to return output:   (based on the aggregation logic, below results are
> > for
> > > sum aggregation)
> > > us   4
> > > uk   6
> > >
> > >
> > >
> > > On Tue, May 3, 2016 at 11:33 AM, James Taylor <[email protected]
> <javascript:;>>
> > > wrote:
> > >
> > > > Removing user list (please don't cross post)
> > > >
> > > > Can you give us a full example of the query you have in mind?
> > > >
> > > > Thanks,
> > > > James
> > > >
> > > > On Tue, May 3, 2016 at 11:14 AM, Swapna Swapna <
> [email protected] <javascript:;>
> > >
> > > > wrote:
> > > >
> > > > > Hi,
> > > > >
> > > > > I'm trying to implement aggregate function on multiple columns (as
> an
> > > > > arguments) like:
> > > > >
> > > > > myaggFunc(col1,col2)
> > > > >
> > > > > And I would want to return the results by each column after
> applying
> > > > > aggregate operation.
> > > > >
> > > > > The output would be something like:
> > > > >
> > > > > col1, count ( aggregate of all records for col1)
> > > > > col2, count
> > > > >
> > > > > Inorder to return the results in the above format, what is the
> return
> > > > data
> > > > > type (of the method) should I have to choose?
> > > > >
> > > > > Thanks
> > > > >
> > > >
> > >
> >
>

Reply via email to