Adding an aggregate function requires some additional effort as an aggregate function on the SQL level is actually implemented by a small family of C functions and private data passed between them.

Aggregate private data is declared using the <data> tag. For each aggregate group it is initialized by the code given in <clear>. For each row of the group the <add> code is executed, receiving the declared <parameter>s just like a normal function. Finally the <result> code is executed and expected to pass the functions.

See the following example for a simple reimplementation of the standard AVG() function: 

<?xml version="1.0"?>
<udf name="x4">

  <maintainer>
    <name>Hartmut Holzgraefe</name>
    <email>hartmut@php.net</email>
  </maintainer>

  <license>LGPL</license>

  <function name="ex_avg" type="aggregate" returns="real" null="yes">
    <param name="x" type="real"/>
    <data>
      <element name="count"    type="int"  default="0"/>
      <element name="sum_x"    type="real" default="0.0"/>
    </data>
    <clear>
<![CDATA[
        data->count    = 0;
        data->sum_x    = 0.0;
]]>
    </clear>
    <add>
<![CDATA[
        if (!x_is_null) {
            data->count++;
            data->sum_x    += x;
        }
]]>
    </add>
    <result>
<![CDATA[
        if (data->count < 1)
        {
            *is_null = 1;
            return 0.0;
        }

        *is_null = 0;
        return data->sum_x / data->count;
]]>
    </result>
  </function>
</udf>