...

The interface "Aggregator" is descibed described in details in the later section of the document (UDA).

...

E.g.:

select avg( col2) from /portfolio  // portfolio  Valid

select col1, avg ( col2) from /portfolio    portfolio (col1 should not be present in projection attributes) // invalid

Query Types & implementation:

...

Note: As a first cut , we will make it mandatory for the group by column to be part of projection attribute ( Will see if we can make implement in a reasonable period the code to relax this constraint. i.e group by columns not be part of projection columns)

Proposed Syntax  for using UDA:

Creation of UDA

Using API:

...

qs.createUDA(“udaAlias”, "com.gemstone.gemfire.cache.query.dunit.UDACreationDUnitTest$SumUDA");

XML based creation:

<uda-manager>

...

   <uda

...

name="uda2"

...

class="com.gemstone.gemfire.cache.query.dunit.UDACreationDUnitTest$UDACLass2"/>

...

   <uda

...

name="uda3"

...

class="com.gemstone.gemfire.cache.query.dunit.UDACreationDUnitTest$UDACLass3"/>
</uda-manager>

cache-9.0

...

xsd
<xsd:element

...

maxOccurs="1"

...

minOccurs="0"

...

name="uda-manager">
  <xsd:complexType>
     <xsd:sequence>
        <xsd:element

...

maxOccurs="unbounded"

...

minOccurs="0"

...

name="uda">
           <xsd:complexType>
             <xsd:attribute

...

name="name"

...

type="xsd:string"

...

use="required"

...

/>
                             <xsd:attribute name="class" type="xsd:string" use="required" />
                        </xsd:complexType>
                    </xsd:element>
              </xsd:sequence>
            </xsd:complexType>
        </xsd:element>  

Example of using UDA , after defining:

...

element> 

Using UDA in Query:

String queryStr = "select p.status , myUDA(p.ID) from /portfolio p where p.ID > 0 group by p.status order by p.status"; 

This will require us to maintain a registry of the aggregate functions defined & will also have to ensure that it gets created on newly joining nodes.We piggy back on the infra through which the OQL indexes get created on the peers & I assume that we can also create on newly joining nodes.

UDA Schema exchange:

The UDA schema/definitions are exchanged between cluster nodes via profile exchange mechanism.