postgresql/doc/src/sgml/xaggr.sgml

<!-- doc/src/sgml/xaggr.sgml -->

 <sect1 id="xaggr">
  <title>User-defined Aggregates</title>

  <indexterm zone="xaggr">
   <primary>aggregate function</primary>
   <secondary>user-defined</secondary>
  </indexterm>

  <para>
   Aggregate functions  in <productname>PostgreSQL</productname>
   are expressed in terms of <firstterm>state values</firstterm>
   and <firstterm>state transition functions</firstterm>.
   That is, an aggregate operates using a state value that is updated
   as each successive input row is processed.
   To define a new aggregate
   function, one selects a data type for the state value,
   an initial value for the state, and a state transition
   function.  The state transition function is just an
   ordinary function that could also be used outside the
   context of the aggregate.  A <firstterm>final function</firstterm>
   can also be specified, in case the desired result of the aggregate
   is different from the data that needs to be kept in the running
   state value.
  </para>

  <para>
   Thus, in addition to the argument and result data types seen by a user
   of the aggregate, there is an internal state-value data type that
   might be different from both the argument and result types.
  </para>

  <para>
   If we define an aggregate that does not use a final function,
   we have an aggregate that computes a running function of
   the column values from each row.  <function>sum</>  is  an
   example  of  this  kind  of aggregate.  <function>sum</> starts at
   zero and always adds the current  row's  value  to
   its  running  total.  For example, if we want to make a <function>sum</>
   aggregate to work on a data type for complex numbers,
   we only need the addition function for that data type.
   The aggregate definition would be:

<screen>
CREATE AGGREGATE sum (complex)
(
    sfunc = complex_add,
    stype = complex,
    initcond = '(0,0)'
);

SELECT sum(a) FROM test_complex;

   sum
-----------
 (34,53.9)
</screen>

   (Notice that we are relying on function overloading: there is more than
    one aggregate named <function>sum</>, but
   <productname>PostgreSQL</productname> can figure out which kind
   of sum applies to a column of type <type>complex</type>.)
  </para>

  <para>
   The above definition of <function>sum</function> will return zero (the initial
   state condition) if there are no nonnull input values.
   Perhaps we want to return null in that case instead &mdash; the SQL standard
   expects <function>sum</function> to behave that way.  We can do this simply by
   omitting the <literal>initcond</literal> phrase, so that the initial state
   condition is null.  Ordinarily this would mean that the <literal>sfunc</literal>
   would need to check for a null state-condition input.  But for
   <function>sum</function> and some other simple aggregates like
   <function>max</> and <function>min</>,
   it is sufficient to insert the first nonnull input value into
   the state variable and then start applying the transition function
   at the second nonnull input value.  <productname>PostgreSQL</productname>
   will do that automatically if the initial condition is null and
   the transition function is marked <quote>strict</> (i.e., not to be called
   for null inputs).
  </para>

  <para>
   Another bit of default behavior for a <quote>strict</> transition function
   is that the previous state value is retained unchanged whenever a
   null input value is encountered.  Thus, null values are ignored.  If you
   need some other behavior for null inputs, do not declare your
   transition function as strict; instead code it to test for null inputs and
   do whatever is needed.
  </para>

  <para>
   <function>avg</> (average) is a more complex example of an aggregate.
   It requires
   two pieces of running state: the sum of the inputs and the count
   of the number of inputs.  The final result is obtained by dividing
   these quantities.  Average is typically implemented by using an
   array as the state value.  For example,
   the built-in implementation of <function>avg(float8)</function>
   looks like:

<programlisting>
CREATE AGGREGATE avg (float8)
(
    sfunc = float8_accum,
    stype = float8[],
    finalfunc = float8_avg,
    initcond = '{0,0,0}'
);
</programlisting>

   (<function>float8_accum</> requires a three-element array, not just
   two elements, because it accumulates the sum of squares as well as
   the sum and count of the inputs.  This is so that it can be used for
   some other aggregates besides <function>avg</>.)
  </para>

  <para>
   Aggregate functions can use polymorphic
   state transition functions or final functions, so that the same functions
   can be used to implement multiple aggregates.
   See <xref linkend="extend-types-polymorphic">
   for an explanation of polymorphic functions.
   Going a step further, the aggregate function itself can be specified
   with polymorphic input type(s) and state type, allowing a single
   aggregate definition to serve for multiple input data types.
   Here is an example of a polymorphic aggregate:

<programlisting>
CREATE AGGREGATE array_accum (anyelement)
(
    sfunc = array_append,
    stype = anyarray,
    initcond = '{}'
);
</programlisting>

   Here, the actual state type for any aggregate call is the array type
   having the actual input type as elements.  The behavior of the aggregate
   is to concatenate all the inputs into an array of that type.
   (Note: the built-in aggregate <function>array_agg</> provides similar
   functionality, with better performance than this definition would have.)
  </para>

  <para>
   Here's the output using two different actual data types as arguments:

<programlisting>
SELECT attrelid::regclass, array_accum(attname)
    FROM pg_attribute
    WHERE attnum &gt; 0 AND attrelid = 'pg_tablespace'::regclass
    GROUP BY attrelid;

   attrelid    |              array_accum              
---------------+---------------------------------------
 pg_tablespace | {spcname,spcowner,spcacl,spcoptions}
(1 row)

SELECT attrelid::regclass, array_accum(atttypid::regtype)
    FROM pg_attribute
    WHERE attnum &gt; 0 AND attrelid = 'pg_tablespace'::regclass
    GROUP BY attrelid;

   attrelid    |        array_accum        
---------------+---------------------------
 pg_tablespace | {name,oid,aclitem[],text[]}
(1 row)
</programlisting>
  </para>

  <para>
   An aggregate function can be made to accept a varying number of arguments
   by declaring its last argument as a <literal>VARIADIC</> array, in much
   the same fashion as for regular functions; see
   <xref linkend="xfunc-sql-variadic-functions">.  The aggregate's transition
   function must have the same array type as its last argument.  The
   transition function typically would also be marked <literal>VARIADIC</>,
   but this is not strictly required.
  </para>

  <note>
   <para>
    Variadic aggregates are easily misused in connection with
    the <literal>ORDER BY</> option (see <xref linkend="syntax-aggregates">),
    since the parser cannot tell whether the wrong number of actual arguments
    have been given in such a combination.  Keep in mind that everything to
    the right of <literal>ORDER BY</> is a sort key, not an argument to the
    aggregate.  For example, in
<programlisting>
SELECT myaggregate(a ORDER BY a, b, c) FROM ...
</programlisting>
    the parser will see this as a single aggregate function argument and
    three sort keys.  However, the user might have intended
<programlisting>
SELECT myaggregate(a, b, c ORDER BY a) FROM ...
</programlisting>
    If <literal>myaggregate</> is variadic, both these calls could be
    perfectly valid.
   </para>

   <para>
    For the same reason, it's wise to think twice before creating aggregate
    functions with the same names and different numbers of regular arguments.
   </para>
  </note>

  <para>
   A function written in C can detect that it is being called as an
   aggregate transition or final function by calling
   <function>AggCheckCallContext</>, for example:
<programlisting>
if (AggCheckCallContext(fcinfo, NULL))
</programlisting>
   One reason for checking this is that when it is true for a transition
   function, the first input
   must be a temporary transition value and can therefore safely be modified
   in-place rather than allocating a new copy.
   See <literal>int8inc()</> for an example.
   (This is the <emphasis>only</>
   case where it is safe for a function to modify a pass-by-reference input.
   In particular, aggregate final functions should not modify their inputs in
   any case, because in some cases they will be re-executed on the same
   final transition value.)
  </para>

  <para>
   For further details see the
   <xref linkend="sql-createaggregate">
   command.
  </para>
 </sect1>