<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"><html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>CREATE AGGREGATE</title><link rel="stylesheet" type="text/css" href="stylesheet.css" /><link rev="made" href="pgsql-docs@lists.postgresql.org" /><meta name="generator" content="DocBook XSL Stylesheets Vsnapshot" /><link rel="prev" href="sql-create-access-method.html" title="CREATE ACCESS METHOD" /><link rel="next" href="sql-createcast.html" title="CREATE CAST" /></head><body><div xmlns="http://www.w3.org/TR/xhtml1/transitional" class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="5" align="center">CREATE AGGREGATE</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="sql-create-access-method.html" title="CREATE ACCESS METHOD">Prev</a> </td><td width="10%" align="left"><a accesskey="u" href="sql-commands.html" title="SQL Commands">Up</a></td><th width="60%" align="center">SQL Commands</th><td width="10%" align="right"><a accesskey="h" href="index.html" title="PostgreSQL 11.5 Documentation">Home</a></td><td width="10%" align="right"> <a accesskey="n" href="sql-createcast.html" title="CREATE CAST">Next</a></td></tr></table><hr></hr></div><div class="refentry" id="SQL-CREATEAGGREGATE"><div class="titlepage"></div><a id="id-1.9.3.57.1" class="indexterm"></a><div class="refnamediv"><h2><span class="refentrytitle">CREATE AGGREGATE</span></h2><p>CREATE AGGREGATE — define a new aggregate function</p></div><div class="refsynopsisdiv"><h2>Synopsis</h2><pre class="synopsis">
CREATE AGGREGATE <em class="replaceable"><code>name</code></em> ( [ <em class="replaceable"><code>argmode</code></em> ] [ <em class="replaceable"><code>argname</code></em> ] <em class="replaceable"><code>arg_data_type</code></em> [ , ... ] ) (
SFUNC = <em class="replaceable"><code>sfunc</code></em>,
STYPE = <em class="replaceable"><code>state_data_type</code></em>
[ , SSPACE = <em class="replaceable"><code>state_data_size</code></em> ]
[ , FINALFUNC = <em class="replaceable"><code>ffunc</code></em> ]
[ , FINALFUNC_EXTRA ]
[ , FINALFUNC_MODIFY = { READ_ONLY | SHAREABLE | READ_WRITE } ]
[ , COMBINEFUNC = <em class="replaceable"><code>combinefunc</code></em> ]
[ , SERIALFUNC = <em class="replaceable"><code>serialfunc</code></em> ]
[ , DESERIALFUNC = <em class="replaceable"><code>deserialfunc</code></em> ]
[ , INITCOND = <em class="replaceable"><code>initial_condition</code></em> ]
[ , MSFUNC = <em class="replaceable"><code>msfunc</code></em> ]
[ , MINVFUNC = <em class="replaceable"><code>minvfunc</code></em> ]
[ , MSTYPE = <em class="replaceable"><code>mstate_data_type</code></em> ]
[ , MSSPACE = <em class="replaceable"><code>mstate_data_size</code></em> ]
[ , MFINALFUNC = <em class="replaceable"><code>mffunc</code></em> ]
[ , MFINALFUNC_EXTRA ]
[ , MFINALFUNC_MODIFY = { READ_ONLY | SHAREABLE | READ_WRITE } ]
[ , MINITCOND = <em class="replaceable"><code>minitial_condition</code></em> ]
[ , SORTOP = <em class="replaceable"><code>sort_operator</code></em> ]
[ , PARALLEL = { SAFE | RESTRICTED | UNSAFE } ]
)
CREATE AGGREGATE <em class="replaceable"><code>name</code></em> ( [ [ <em class="replaceable"><code>argmode</code></em> ] [ <em class="replaceable"><code>argname</code></em> ] <em class="replaceable"><code>arg_data_type</code></em> [ , ... ] ]
ORDER BY [ <em class="replaceable"><code>argmode</code></em> ] [ <em class="replaceable"><code>argname</code></em> ] <em class="replaceable"><code>arg_data_type</code></em> [ , ... ] ) (
SFUNC = <em class="replaceable"><code>sfunc</code></em>,
STYPE = <em class="replaceable"><code>state_data_type</code></em>
[ , SSPACE = <em class="replaceable"><code>state_data_size</code></em> ]
[ , FINALFUNC = <em class="replaceable"><code>ffunc</code></em> ]
[ , FINALFUNC_EXTRA ]
[ , FINALFUNC_MODIFY = { READ_ONLY | SHAREABLE | READ_WRITE } ]
[ , INITCOND = <em class="replaceable"><code>initial_condition</code></em> ]
[ , PARALLEL = { SAFE | RESTRICTED | UNSAFE } ]
[ , HYPOTHETICAL ]
)
<span class="phrase">or the old syntax</span>
CREATE AGGREGATE <em class="replaceable"><code>name</code></em> (
BASETYPE = <em class="replaceable"><code>base_type</code></em>,
SFUNC = <em class="replaceable"><code>sfunc</code></em>,
STYPE = <em class="replaceable"><code>state_data_type</code></em>
[ , SSPACE = <em class="replaceable"><code>state_data_size</code></em> ]
[ , FINALFUNC = <em class="replaceable"><code>ffunc</code></em> ]
[ , FINALFUNC_EXTRA ]
[ , FINALFUNC_MODIFY = { READ_ONLY | SHAREABLE | READ_WRITE } ]
[ , COMBINEFUNC = <em class="replaceable"><code>combinefunc</code></em> ]
[ , SERIALFUNC = <em class="replaceable"><code>serialfunc</code></em> ]
[ , DESERIALFUNC = <em class="replaceable"><code>deserialfunc</code></em> ]
[ , INITCOND = <em class="replaceable"><code>initial_condition</code></em> ]
[ , MSFUNC = <em class="replaceable"><code>msfunc</code></em> ]
[ , MINVFUNC = <em class="replaceable"><code>minvfunc</code></em> ]
[ , MSTYPE = <em class="replaceable"><code>mstate_data_type</code></em> ]
[ , MSSPACE = <em class="replaceable"><code>mstate_data_size</code></em> ]
[ , MFINALFUNC = <em class="replaceable"><code>mffunc</code></em> ]
[ , MFINALFUNC_EXTRA ]
[ , MFINALFUNC_MODIFY = { READ_ONLY | SHAREABLE | READ_WRITE } ]
[ , MINITCOND = <em class="replaceable"><code>minitial_condition</code></em> ]
[ , SORTOP = <em class="replaceable"><code>sort_operator</code></em> ]
)
</pre></div><div class="refsect1" id="id-1.9.3.57.5"><h2>Description</h2><p>
<code class="command">CREATE AGGREGATE</code> defines a new aggregate
function. Some basic and commonly-used aggregate functions are
included with the distribution; they are documented in <a class="xref" href="functions-aggregate.html" title="9.20. Aggregate Functions">Section 9.20</a>. If one defines new types or needs
an aggregate function not already provided, then <code class="command">CREATE
AGGREGATE</code> can be used to provide the desired features.
</p><p>
If a schema name is given (for example, <code class="literal">CREATE AGGREGATE
myschema.myagg ...</code>) then the aggregate function is created in the
specified schema. Otherwise it is created in the current schema.
</p><p>
An aggregate function is identified by its name and input data type(s).
Two aggregates in the same schema can have the same name if they operate on
different input types. The
name and input data type(s) of an aggregate must also be distinct from
the name and input data type(s) of every ordinary function in the same
schema.
This behavior is identical to overloading of ordinary function names
(see <a class="xref" href="sql-createfunction.html" title="CREATE FUNCTION"><span class="refentrytitle">CREATE FUNCTION</span></a>).
</p><p>
A simple aggregate function is made from one or two ordinary
functions:
a state transition function
<em class="replaceable"><code>sfunc</code></em>,
and an optional final calculation function
<em class="replaceable"><code>ffunc</code></em>.
These are used as follows:
</p><pre class="programlisting">
<em class="replaceable"><code>sfunc</code></em>( internal-state, next-data-values ) ---> next-internal-state
<em class="replaceable"><code>ffunc</code></em>( internal-state ) ---> aggregate-value
</pre><p>
</p><p>
<span class="productname">PostgreSQL</span> creates a temporary variable
of data type <em class="replaceable"><code>stype</code></em>
to hold the current internal state of the aggregate. At each input row,
the aggregate argument value(s) are calculated and
the state transition function is invoked with the current state value
and the new argument value(s) to calculate a new
internal state value. After all the rows have been processed,
the final function is invoked once to calculate the aggregate's return
value. If there is no final function then the ending state value
is returned as-is.
</p><p>
An aggregate function can provide an initial condition,
that is, an initial value for the internal state value.
This is specified and stored in the database as a value of type
<code class="type">text</code>, but it must be a valid external representation
of a constant of the state value data type. If it is not supplied
then the state value starts out null.
</p><p>
If the state transition function is declared <span class="quote">“<span class="quote">strict</span>”</span>,
then it cannot be called with null inputs. With such a transition
function, aggregate execution behaves as follows. Rows with any null input
values are ignored (the function is not called and the previous state value
is retained). If the initial state value is null, then at the first row
with all-nonnull input values, the first argument value replaces the state
value, and the transition function is invoked at each subsequent row with
all-nonnull input values.
This is handy for implementing aggregates like <code class="function">max</code>.
Note that this behavior is only available when
<em class="replaceable"><code>state_data_type</code></em>
is the same as the first
<em class="replaceable"><code>arg_data_type</code></em>.
When these types are different, you must supply a nonnull initial
condition or use a nonstrict transition function.
</p><p>
If the state transition function is not strict, then it will be called
unconditionally at each input row, and must deal with null inputs
and null state values for itself. This allows the aggregate
author to have full control over the aggregate's handling of null values.
</p><p>
If the final function is declared <span class="quote">“<span class="quote">strict</span>”</span>, then it will not
be called when the ending state value is null; instead a null result
will be returned automatically. (Of course this is just the normal
behavior of strict functions.) In any case the final function has
the option of returning a null value. For example, the final function for
<code class="function">avg</code> returns null when it sees there were zero
input rows.
</p><p>
Sometimes it is useful to declare the final function as taking not just
the state value, but extra parameters corresponding to the aggregate's
input values. The main reason for doing this is if the final function
is polymorphic and the state value's data type would be inadequate to
pin down the result type. These extra parameters are always passed as
NULL (and so the final function must not be strict when
the <code class="literal">FINALFUNC_EXTRA</code> option is used), but nonetheless they
are valid parameters. The final function could for example make use
of <code class="function">get_fn_expr_argtype</code> to identify the actual argument type
in the current call.
</p><p>
An aggregate can optionally support <em class="firstterm">moving-aggregate mode</em>,
as described in <a class="xref" href="xaggr.html#XAGGR-MOVING-AGGREGATES" title="38.11.1. Moving-Aggregate Mode">Section 38.11.1</a>. This requires
specifying the <code class="literal">MSFUNC</code>, <code class="literal">MINVFUNC</code>,
and <code class="literal">MSTYPE</code> parameters, and optionally
the <code class="literal">MSPACE</code>, <code class="literal">MFINALFUNC</code>,
<code class="literal">MFINALFUNC_EXTRA</code>, <code class="literal">MFINALFUNC_MODIFY</code>,
and <code class="literal">MINITCOND</code> parameters. Except for <code class="literal">MINVFUNC</code>,
these parameters work like the corresponding simple-aggregate parameters
without <code class="literal">M</code>; they define a separate implementation of the
aggregate that includes an inverse transition function.
</p><p>
The syntax with <code class="literal">ORDER BY</code> in the parameter list creates
a special type of aggregate called an <em class="firstterm">ordered-set
aggregate</em>; or if <code class="literal">HYPOTHETICAL</code> is specified, then
a <em class="firstterm">hypothetical-set aggregate</em> is created. These
aggregates operate over groups of sorted values in order-dependent ways,
so that specification of an input sort order is an essential part of a
call. Also, they can have <em class="firstterm">direct</em> arguments, which are
arguments that are evaluated only once per aggregation rather than once
per input row. Hypothetical-set aggregates are a subclass of ordered-set
aggregates in which some of the direct arguments are required to match,
in number and data types, the aggregated argument columns. This allows
the values of those direct arguments to be added to the collection of
aggregate-input rows as an additional <span class="quote">“<span class="quote">hypothetical</span>”</span> row.
</p><p>
An aggregate can optionally support <em class="firstterm">partial aggregation</em>,
as described in <a class="xref" href="xaggr.html#XAGGR-PARTIAL-AGGREGATES" title="38.11.4. Partial Aggregation">Section 38.11.4</a>.
This requires specifying the <code class="literal">COMBINEFUNC</code> parameter.
If the <em class="replaceable"><code>state_data_type</code></em>
is <code class="type">internal</code>, it's usually also appropriate to provide the
<code class="literal">SERIALFUNC</code> and <code class="literal">DESERIALFUNC</code> parameters so that
parallel aggregation is possible. Note that the aggregate must also be
marked <code class="literal">PARALLEL SAFE</code> to enable parallel aggregation.
</p><p>
Aggregates that behave like <code class="function">MIN</code> or <code class="function">MAX</code> can
sometimes be optimized by looking into an index instead of scanning every
input row. If this aggregate can be so optimized, indicate it by
specifying a <em class="firstterm">sort operator</em>. The basic requirement is that
the aggregate must yield the first element in the sort ordering induced by
the operator; in other words:
</p><pre class="programlisting">
SELECT agg(col) FROM tab;
</pre><p>
must be equivalent to:
</p><pre class="programlisting">
SELECT col FROM tab ORDER BY col USING sortop LIMIT 1;
</pre><p>
Further assumptions are that the aggregate ignores null inputs, and that
it delivers a null result if and only if there were no non-null inputs.
Ordinarily, a data type's <code class="literal"><</code> operator is the proper sort
operator for <code class="function">MIN</code>, and <code class="literal">></code> is the proper sort
operator for <code class="function">MAX</code>. Note that the optimization will never
actually take effect unless the specified operator is the <span class="quote">“<span class="quote">less
than</span>”</span> or <span class="quote">“<span class="quote">greater than</span>”</span> strategy member of a B-tree
index operator class.
</p><p>
To be able to create an aggregate function, you must
have <code class="literal">USAGE</code> privilege on the argument types, the state
type(s), and the return type, as well as <code class="literal">EXECUTE</code>
privilege on the supporting functions.
</p></div><div class="refsect1" id="id-1.9.3.57.6"><h2>Parameters</h2><div class="variablelist"><dl class="variablelist"><dt><span class="term"><em class="replaceable"><code>name</code></em></span></dt><dd><p>
The name (optionally schema-qualified) of the aggregate function
to create.
</p></dd><dt><span class="term"><em class="replaceable"><code>argmode</code></em></span></dt><dd><p>
The mode of an argument: <code class="literal">IN</code> or <code class="literal">VARIADIC</code>.
(Aggregate functions do not support <code class="literal">OUT</code> arguments.)
If omitted, the default is <code class="literal">IN</code>. Only the last argument
can be marked <code class="literal">VARIADIC</code>.
</p></dd><dt><span class="term"><em class="replaceable"><code>argname</code></em></span></dt><dd><p>
The name of an argument. This is currently only useful for
documentation purposes. If omitted, the argument has no name.
</p></dd><dt><span class="term"><em class="replaceable"><code>arg_data_type</code></em></span></dt><dd><p>
An input data type on which this aggregate function operates.
To create a zero-argument aggregate function, write <code class="literal">*</code>
in place of the list of argument specifications. (An example of such an
aggregate is <code class="function">count(*)</code>.)
</p></dd><dt><span class="term"><em class="replaceable"><code>base_type</code></em></span></dt><dd><p>
In the old syntax for <code class="command">CREATE AGGREGATE</code>, the input data type
is specified by a <code class="literal">basetype</code> parameter rather than being
written next to the aggregate name. Note that this syntax allows
only one input parameter. To define a zero-argument aggregate function
with this syntax, specify the <code class="literal">basetype</code> as
<code class="literal">"ANY"</code> (not <code class="literal">*</code>).
Ordered-set aggregates cannot be defined with the old syntax.
</p></dd><dt><span class="term"><em class="replaceable"><code>sfunc</code></em></span></dt><dd><p>
The name of the state transition function to be called for each
input row. For a normal <em class="replaceable"><code>N</code></em>-argument
aggregate function, the <em class="replaceable"><code>sfunc</code></em>
must take <em class="replaceable"><code>N</code></em>+1 arguments,
the first being of type <em class="replaceable"><code>state_data_type</code></em> and the rest
matching the declared input data type(s) of the aggregate.
The function must return a value of type <em class="replaceable"><code>state_data_type</code></em>. This function
takes the current state value and the current input data value(s),
and returns the next state value.
</p><p>
For ordered-set (including hypothetical-set) aggregates, the state
transition function receives only the current state value and the
aggregated arguments, not the direct arguments. Otherwise it is the
same.
</p></dd><dt><span class="term"><em class="replaceable"><code>state_data_type</code></em></span></dt><dd><p>
The data type for the aggregate's state value.
</p></dd><dt><span class="term"><em class="replaceable"><code>state_data_size</code></em></span></dt><dd><p>
The approximate average size (in bytes) of the aggregate's state value.
If this parameter is omitted or is zero, a default estimate is used
based on the <em class="replaceable"><code>state_data_type</code></em>.
The planner uses this value to estimate the memory required for a
grouped aggregate query. The planner will consider using hash
aggregation for such a query only if the hash table is estimated to fit
in <a class="xref" href="runtime-config-resource.html#GUC-WORK-MEM">work_mem</a>; therefore, large values of this
parameter discourage use of hash aggregation.
</p></dd><dt><span class="term"><em class="replaceable"><code>ffunc</code></em></span></dt><dd><p>
The name of the final function called to compute the aggregate's
result after all input rows have been traversed.
For a normal aggregate, this function
must take a single argument of type <em class="replaceable"><code>state_data_type</code></em>. The return
data type of the aggregate is defined as the return type of this
function. If <em class="replaceable"><code>ffunc</code></em>
is not specified, then the ending state value is used as the
aggregate's result, and the return type is <em class="replaceable"><code>state_data_type</code></em>.
</p><p>
For ordered-set (including hypothetical-set) aggregates, the
final function receives not only the final state value,
but also the values of all the direct arguments.
</p><p>
If <code class="literal">FINALFUNC_EXTRA</code> is specified, then in addition to the
final state value and any direct arguments, the final function
receives extra NULL values corresponding to the aggregate's regular
(aggregated) arguments. This is mainly useful to allow correct
resolution of the aggregate result type when a polymorphic aggregate
is being defined.
</p></dd><dt><span class="term"><code class="literal">FINALFUNC_MODIFY</code> = { <code class="literal">READ_ONLY</code> | <code class="literal">SHAREABLE</code> | <code class="literal">READ_WRITE</code> }</span></dt><dd><p>
This option specifies whether the final function is a pure function
that does not modify its arguments. <code class="literal">READ_ONLY</code> indicates
it does not; the other two values indicate that it may change the
transition state value. See <a class="xref" href="sql-createaggregate.html#SQL-CREATEAGGREGATE-NOTES" title="Notes">Notes</a> below for more detail. The
default is <code class="literal">READ_ONLY</code>, except for ordered-set aggregates,
for which the default is <code class="literal">READ_WRITE</code>.
</p></dd><dt><span class="term"><em class="replaceable"><code>combinefunc</code></em></span></dt><dd><p>
The <em class="replaceable"><code>combinefunc</code></em> function
may optionally be specified to allow the aggregate function to support
partial aggregation. If provided,
the <em class="replaceable"><code>combinefunc</code></em> must
combine two <em class="replaceable"><code>state_data_type</code></em>
values, each containing the result of aggregation over some subset of
the input values, to produce a
new <em class="replaceable"><code>state_data_type</code></em> that
represents the result of aggregating over both sets of inputs. This
function can be thought of as
an <em class="replaceable"><code>sfunc</code></em>, where instead of
acting upon an individual input row and adding it to the running
aggregate state, it adds another aggregate state to the running state.
</p><p>
The <em class="replaceable"><code>combinefunc</code></em> must be
declared as taking two arguments of
the <em class="replaceable"><code>state_data_type</code></em> and
returning a value of
the <em class="replaceable"><code>state_data_type</code></em>.
Optionally this function may be <span class="quote">“<span class="quote">strict</span>”</span>. In this case the
function will not be called when either of the input states are null;
the other state will be taken as the correct result.
</p><p>
For aggregate functions
whose <em class="replaceable"><code>state_data_type</code></em>
is <code class="type">internal</code>,
the <em class="replaceable"><code>combinefunc</code></em> must not
be strict. In this case
the <em class="replaceable"><code>combinefunc</code></em> must
ensure that null states are handled correctly and that the state being
returned is properly stored in the aggregate memory context.
</p></dd><dt><span class="term"><em class="replaceable"><code>serialfunc</code></em></span></dt><dd><p>
An aggregate function
whose <em class="replaceable"><code>state_data_type</code></em>
is <code class="type">internal</code> can participate in parallel aggregation only if it
has a <em class="replaceable"><code>serialfunc</code></em> function,
which must serialize the aggregate state into a <code class="type">bytea</code> value for
transmission to another process. This function must take a single
argument of type <code class="type">internal</code> and return type <code class="type">bytea</code>. A
corresponding <em class="replaceable"><code>deserialfunc</code></em>
is also required.
</p></dd><dt><span class="term"><em class="replaceable"><code>deserialfunc</code></em></span></dt><dd><p>
Deserialize a previously serialized aggregate state back into
<em class="replaceable"><code>state_data_type</code></em>. This
function must take two arguments of types <code class="type">bytea</code>
and <code class="type">internal</code>, and produce a result of type <code class="type">internal</code>.
(Note: the second, <code class="type">internal</code> argument is unused, but is required
for type safety reasons.)
</p></dd><dt><span class="term"><em class="replaceable"><code>initial_condition</code></em></span></dt><dd><p>
The initial setting for the state value. This must be a string
constant in the form accepted for the data type <em class="replaceable"><code>state_data_type</code></em>. If not
specified, the state value starts out null.
</p></dd><dt><span class="term"><em class="replaceable"><code>msfunc</code></em></span></dt><dd><p>
The name of the forward state transition function to be called for each
input row in moving-aggregate mode. This is exactly like the regular
transition function, except that its first argument and result are of
type <em class="replaceable"><code>mstate_data_type</code></em>, which might be different
from <em class="replaceable"><code>state_data_type</code></em>.
</p></dd><dt><span class="term"><em class="replaceable"><code>minvfunc</code></em></span></dt><dd><p>
The name of the inverse state transition function to be used in
moving-aggregate mode. This function has the same argument and
result types as <em class="replaceable"><code>msfunc</code></em>, but it is used to remove
a value from the current aggregate state, rather than add a value to
it. The inverse transition function must have the same strictness
attribute as the forward state transition function.
</p></dd><dt><span class="term"><em class="replaceable"><code>mstate_data_type</code></em></span></dt><dd><p>
The data type for the aggregate's state value, when using
moving-aggregate mode.
</p></dd><dt><span class="term"><em class="replaceable"><code>mstate_data_size</code></em></span></dt><dd><p>
The approximate average size (in bytes) of the aggregate's state
value, when using moving-aggregate mode. This works the same as
<em class="replaceable"><code>state_data_size</code></em>.
</p></dd><dt><span class="term"><em class="replaceable"><code>mffunc</code></em></span></dt><dd><p>
The name of the final function called to compute the aggregate's
result after all input rows have been traversed, when using
moving-aggregate mode. This works the same as <em class="replaceable"><code>ffunc</code></em>,
except that its first argument's type
is <em class="replaceable"><code>mstate_data_type</code></em> and extra dummy arguments are
specified by writing <code class="literal">MFINALFUNC_EXTRA</code>.
The aggregate result type determined by <em class="replaceable"><code>mffunc</code></em>
or <em class="replaceable"><code>mstate_data_type</code></em> must match that determined by the
aggregate's regular implementation.
</p></dd><dt><span class="term"><code class="literal">MFINALFUNC_MODIFY</code> = { <code class="literal">READ_ONLY</code> | <code class="literal">SHAREABLE</code> | <code class="literal">READ_WRITE</code> }</span></dt><dd><p>
This option is like <code class="literal">FINALFUNC_MODIFY</code>, but it describes
the behavior of the moving-aggregate final function.
</p></dd><dt><span class="term"><em class="replaceable"><code>minitial_condition</code></em></span></dt><dd><p>
The initial setting for the state value, when using moving-aggregate
mode. This works the same as <em class="replaceable"><code>initial_condition</code></em>.
</p></dd><dt><span class="term"><em class="replaceable"><code>sort_operator</code></em></span></dt><dd><p>
The associated sort operator for a <code class="function">MIN</code>- or
<code class="function">MAX</code>-like aggregate.
This is just an operator name (possibly schema-qualified).
The operator is assumed to have the same input data types as
the aggregate (which must be a single-argument normal aggregate).
</p></dd><dt><span class="term"><code class="literal">PARALLEL =</code> { <code class="literal">SAFE</code> | <code class="literal">RESTRICTED</code> | <code class="literal">UNSAFE</code> }</span></dt><dd><p>
The meanings of <code class="literal">PARALLEL SAFE</code>, <code class="literal">PARALLEL
RESTRICTED</code>, and <code class="literal">PARALLEL UNSAFE</code> are the same as
in <a class="xref" href="sql-createfunction.html" title="CREATE FUNCTION"><span class="refentrytitle">CREATE FUNCTION</span></a>. An aggregate will not be
considered for parallelization if it is marked <code class="literal">PARALLEL
UNSAFE</code> (which is the default!) or <code class="literal">PARALLEL RESTRICTED</code>.
Note that the parallel-safety markings of the aggregate's support
functions are not consulted by the planner, only the marking of the
aggregate itself.
</p></dd><dt><span class="term"><code class="literal">HYPOTHETICAL</code></span></dt><dd><p>
For ordered-set aggregates only, this flag specifies that the aggregate
arguments are to be processed according to the requirements for
hypothetical-set aggregates: that is, the last few direct arguments must
match the data types of the aggregated (<code class="literal">WITHIN GROUP</code>)
arguments. The <code class="literal">HYPOTHETICAL</code> flag has no effect on
run-time behavior, only on parse-time resolution of the data types and
collations of the aggregate's arguments.
</p></dd></dl></div><p>
The parameters of <code class="command">CREATE AGGREGATE</code> can be
written in any order, not just the order illustrated above.
</p></div><div class="refsect1" id="SQL-CREATEAGGREGATE-NOTES"><h2>Notes</h2><p>
In parameters that specify support function names, you can write
a schema name if needed, for example <code class="literal">SFUNC = public.sum</code>.
Do not write argument types there, however — the argument types
of the support functions are determined from other parameters.
</p><p>
Ordinarily, PostgreSQL functions are expected to be true functions that
do not modify their input values. However, an aggregate transition
function, <span class="emphasis"><em>when used in the context of an aggregate</em></span>,
is allowed to cheat and modify its transition-state argument in place.
This can provide substantial performance benefits compared to making
a fresh copy of the transition state each time.
</p><p>
Likewise, while an aggregate final function is normally expected not to
modify its input values, sometimes it is impractical to avoid modifying
the transition-state argument. Such behavior must be declared using
the <code class="literal">FINALFUNC_MODIFY</code> parameter.
The <code class="literal">READ_WRITE</code>
value indicates that the final function modifies the transition state in
unspecified ways. This value prevents use of the aggregate as a window
function, and it also prevents merging of transition states for aggregate
calls that share the same input values and transition functions.
The <code class="literal">SHAREABLE</code> value indicates that the transition function
cannot be applied after the final function, but multiple final-function
calls can be performed on the ending transition state value. This value
prevents use of the aggregate as a window function, but it allows merging
of transition states. (That is, the optimization of interest here is not
applying the same final function repeatedly, but applying different final
functions to the same ending transition state value. This is allowed as
long as none of the final functions are marked <code class="literal">READ_WRITE</code>.)
</p><p>
If an aggregate supports moving-aggregate mode, it will improve
calculation efficiency when the aggregate is used as a window function
for a window with moving frame start (that is, a frame start mode other
than <code class="literal">UNBOUNDED PRECEDING</code>). Conceptually, the forward
transition function adds input values to the aggregate's state when
they enter the window frame from the bottom, and the inverse transition
function removes them again when they leave the frame at the top. So,
when values are removed, they are always removed in the same order they
were added. Whenever the inverse transition function is invoked, it will
thus receive the earliest added but not yet removed argument value(s).
The inverse transition function can assume that at least one row will
remain in the current state after it removes the oldest row. (When this
would not be the case, the window function mechanism simply starts a
fresh aggregation, rather than using the inverse transition function.)
</p><p>
The forward transition function for moving-aggregate mode is not
allowed to return NULL as the new state value. If the inverse
transition function returns NULL, this is taken as an indication that
the inverse function cannot reverse the state calculation for this
particular input, and so the aggregate calculation will be redone from
scratch for the current frame starting position. This convention
allows moving-aggregate mode to be used in situations where there are
some infrequent cases that are impractical to reverse out of the
running state value.
</p><p>
If no moving-aggregate implementation is supplied,
the aggregate can still be used with moving frames,
but <span class="productname">PostgreSQL</span> will recompute the whole
aggregation whenever the start of the frame moves.
Note that whether or not the aggregate supports moving-aggregate
mode, <span class="productname">PostgreSQL</span> can handle a moving frame
end without recalculation; this is done by continuing to add new values
to the aggregate's state. This is why use of an aggregate as a window
function requires that the final function be read-only: it must
not damage the aggregate's state value, so that the aggregation can be
continued even after an aggregate result value has been obtained for
one set of frame boundaries.
</p><p>
The syntax for ordered-set aggregates allows <code class="literal">VARIADIC</code>
to be specified for both the last direct parameter and the last
aggregated (<code class="literal">WITHIN GROUP</code>) parameter. However, the
current implementation restricts use of <code class="literal">VARIADIC</code>
in two ways. First, ordered-set aggregates can only use
<code class="literal">VARIADIC "any"</code>, not other variadic array types.
Second, if the last direct parameter is <code class="literal">VARIADIC "any"</code>,
then there can be only one aggregated parameter and it must also
be <code class="literal">VARIADIC "any"</code>. (In the representation used in the
system catalogs, these two parameters are merged into a single
<code class="literal">VARIADIC "any"</code> item, since <code class="structname">pg_proc</code> cannot
represent functions with more than one <code class="literal">VARIADIC</code> parameter.)
If the aggregate is a hypothetical-set aggregate, the direct arguments
that match the <code class="literal">VARIADIC "any"</code> parameter are the hypothetical
ones; any preceding parameters represent additional direct arguments
that are not constrained to match the aggregated arguments.
</p><p>
Currently, ordered-set aggregates do not need to support
moving-aggregate mode, since they cannot be used as window functions.
</p><p>
Partial (including parallel) aggregation is currently not supported for
ordered-set aggregates. Also, it will never be used for aggregate calls
that include <code class="literal">DISTINCT</code> or <code class="literal">ORDER BY</code> clauses, since
those semantics cannot be supported during partial aggregation.
</p></div><div class="refsect1" id="id-1.9.3.57.8"><h2>Examples</h2><p>
See <a class="xref" href="xaggr.html" title="38.11. User-defined Aggregates">Section 38.11</a>.
</p></div><div class="refsect1" id="id-1.9.3.57.9"><h2>Compatibility</h2><p>
<code class="command">CREATE AGGREGATE</code> is a
<span class="productname">PostgreSQL</span> language extension. The SQL
standard does not provide for user-defined aggregate functions.
</p></div><div class="refsect1" id="id-1.9.3.57.10"><h2>See Also</h2><span class="simplelist"><a class="xref" href="sql-alteraggregate.html" title="ALTER AGGREGATE"><span class="refentrytitle">ALTER AGGREGATE</span></a>, <a class="xref" href="sql-dropaggregate.html" title="DROP AGGREGATE"><span class="refentrytitle">DROP AGGREGATE</span></a></span></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="sql-create-access-method.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="sql-commands.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="sql-createcast.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">CREATE ACCESS METHOD </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> CREATE CAST</td></tr></table></div></body></html>
