Skip to content

Commit 0d3f440

Browse files
tglsfdctglsfdc
committed
Allow aggregate functions to be VARIADIC.
There's no inherent reason why an aggregate function can't be variadic (even VARIADIC ANY) if its transition function can handle the case. Indeed, this patch to add the feature touches none of the planner or executor, and little of the parser; the main missing stuff was DDL and pg_dump support. It is true that variadic aggregates can create the same sort of ambiguity about parameters versus ORDER BY keys that was complained of when we (briefly) had both one- and two-argument forms of string_agg(). However, the policy formed in response to that discussion only said that we'd not create any built-in aggregates with varying numbers of arguments, not that we shouldn't allow users to do it. So the logical extension of that is we can allow users to make variadic aggregates as long as we're wary about shipping any such in core. In passing, this patch allows aggregate function arguments to be named, to the extent of remembering the names in pg_proc and dumping them in pg_dump. You can't yet call an aggregate using named-parameter notation. That seems like a likely future extension, but it'll take some work, and it's not what this patch is really about. Likewise, there's still some work needed to make window functions handle VARIADIC fully, but I left that for another day. initdb forced because of new aggvariadic field in Aggref parse nodes.
1 parent 8b290f3 commit 0d3f440

35 files changed

+448
-116
lines changed

‎doc/src/sgml/ref/alter_aggregate.sgml

Lines changed: 32 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -21,9 +21,12 @@ PostgreSQL documentation
2121

2222
<refsynopsisdiv>
2323
<synopsis>
24-
ALTER AGGREGATE <replaceable>name</replaceable> ( <replaceable>argtype</replaceable> [ , ... ] ) RENAME TO <replaceable>new_name</replaceable>
25-
ALTER AGGREGATE <replaceable>name</replaceable> ( <replaceable>argtype</replaceable> [ , ... ] ) OWNER TO <replaceable>new_owner</replaceable>
26-
ALTER AGGREGATE <replaceable>name</replaceable> ( <replaceable>argtype</replaceable> [ , ... ] ) SET SCHEMA <replaceable>new_schema</replaceable>
24+
ALTER AGGREGATE <replaceable>name</replaceable> ( [ <replaceable>argmode</replaceable> ] [ <replaceable>arg_name</replaceable> ] <replaceable>arg_data_type</replaceable> [ , ... ] )
25+
RENAME TO <replaceable>new_name</replaceable>
26+
ALTER AGGREGATE <replaceable>name</replaceable> ( [ <replaceable>argmode</replaceable> ] [ <replaceable>arg_name</replaceable> ] <replaceable>arg_data_type</replaceable> [ , ... ] )
27+
OWNER TO <replaceable>new_owner</replaceable>
28+
ALTER AGGREGATE <replaceable>name</replaceable> ( [ <replaceable>argmode</replaceable> ] [ <replaceable>arg_name</replaceable> ] <replaceable>arg_data_type</replaceable> [ , ... ] )
29+
SET SCHEMA <replaceable>new_schema</replaceable>
2730
</synopsis>
2831
</refsynopsisdiv>
2932

@@ -62,12 +65,36 @@ ALTER AGGREGATE <replaceable>name</replaceable> ( <replaceable>argtype</replacea
6265
</varlistentry>
6366

6467
<varlistentry>
65-
<term><replaceable class="parameter">argtype</replaceable></term>
68+
<term><replaceable class="parameter">argmode</replaceable></term>
69+
70+
<listitem>
71+
<para>
72+
The mode of an argument: <literal>IN</> or <literal>VARIADIC</>.
73+
If omitted, the default is <literal>IN</>.
74+
</para>
75+
</listitem>
76+
</varlistentry>
77+
78+
<varlistentry>
79+
<term><replaceable class="parameter">arg_name</replaceable></term>
80+
81+
<listitem>
82+
<para>
83+
The name of an argument.
84+
Note that <command>ALTER AGGREGATE</command> does not actually pay
85+
any attention to argument names, since only the argument data
86+
types are needed to determine the aggregate function's identity.
87+
</para>
88+
</listitem>
89+
</varlistentry>
90+
91+
<varlistentry>
92+
<term><replaceable class="parameter">arg_data_type</replaceable></term>
6693
<listitem>
6794
<para>
6895
An input data type on which the aggregate function operates.
6996
To reference a zero-argument aggregate function, write <literal>*</>
70-
in place of the list of input data types.
97+
in place of the list of argument specifications.
7198
</para>
7299
</listitem>
73100
</varlistentry>

‎doc/src/sgml/ref/alter_extension.sgml

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -30,7 +30,7 @@ ALTER EXTENSION <replaceable class="PARAMETER">name</replaceable> DROP <replacea
3030

3131
<phrase>where <replaceable class="PARAMETER">member_object</replaceable> is:</phrase>
3232

33-
AGGREGATE <replaceable class="PARAMETER">agg_name</replaceable> (<replaceable class="PARAMETER">agg_type</replaceable> [, ...] ) |
33+
AGGREGATE <replaceable class="PARAMETER">agg_name</replaceable> ( [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">argname</replaceable> ] <replaceable class="parameter">agg_type</replaceable> [, ...] ) |
3434
CAST (<replaceable>source_type</replaceable> AS <replaceable>target_type</replaceable>) |
3535
COLLATION <replaceable class="PARAMETER">object_name</replaceable> |
3636
CONVERSION <replaceable class="PARAMETER">object_name</replaceable> |
@@ -179,7 +179,7 @@ ALTER EXTENSION <replaceable class="PARAMETER">name</replaceable> DROP <replacea
179179
<para>
180180
An input data type on which the aggregate function operates.
181181
To reference a zero-argument aggregate function, write <literal>*</>
182-
in place of the list of input data types.
182+
in place of the list of argument specifications.
183183
</para>
184184
</listitem>
185185
</varlistentry>

‎doc/src/sgml/ref/comment.sgml

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -23,7 +23,7 @@ PostgreSQL documentation
2323
<synopsis>
2424
COMMENT ON
2525
{
26-
AGGREGATE <replaceable class="PARAMETER">agg_name</replaceable> (<replaceable class="PARAMETER">agg_type</replaceable> [, ...] ) |
26+
AGGREGATE <replaceable class="PARAMETER">agg_name</replaceable> ( [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">argname</replaceable> ] <replaceable class="parameter">agg_type</replaceable> [, ...] ) |
2727
CAST (<replaceable>source_type</replaceable> AS <replaceable>target_type</replaceable>) |
2828
COLLATION <replaceable class="PARAMETER">object_name</replaceable> |
2929
COLUMN <replaceable class="PARAMETER">relation_name</replaceable>.<replaceable class="PARAMETER">column_name</replaceable> |
@@ -126,7 +126,7 @@ COMMENT ON
126126
<para>
127127
An input data type on which the aggregate function operates.
128128
To reference a zero-argument aggregate function, write <literal>*</>
129-
in place of the list of input data types.
129+
in place of the list of argument specifications.
130130
</para>
131131
</listitem>
132132
</varlistentry>
@@ -156,7 +156,7 @@ COMMENT ON
156156
The mode of a function argument: <literal>IN</>, <literal>OUT</>,
157157
<literal>INOUT</>, or <literal>VARIADIC</>.
158158
If omitted, the default is <literal>IN</>.
159-
Note that <command>COMMENT ON FUNCTION</command> does not actually pay
159+
Note that <command>COMMENT</command> does not actually pay
160160
any attention to <literal>OUT</> arguments, since only the input
161161
arguments are needed to determine the function's identity.
162162
So it is sufficient to list the <literal>IN</>, <literal>INOUT</>,
@@ -170,7 +170,7 @@ COMMENT ON
170170
<listitem>
171171
<para>
172172
The name of a function argument.
173-
Note that <command>COMMENT ON FUNCTION</command> does not actually pay
173+
Note that <command>COMMENT</command> does not actually pay
174174
any attention to argument names, since only the argument data
175175
types are needed to determine the function's identity.
176176
</para>

‎doc/src/sgml/ref/create_aggregate.sgml

Lines changed: 30 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -21,7 +21,7 @@ PostgreSQL documentation
2121

2222
<refsynopsisdiv>
2323
<synopsis>
24-
CREATE AGGREGATE <replaceable class="PARAMETER">name</replaceable> ( <replaceable class="PARAMETER">input_data_type</replaceable> [ , ... ] ) (
24+
CREATE AGGREGATE <replaceable class="parameter">name</replaceable> ( [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">arg_name</replaceable> ] <replaceable class="parameter">arg_data_type</replaceable> [ , ... ] ) (
2525
SFUNC = <replaceable class="PARAMETER">sfunc</replaceable>,
2626
STYPE = <replaceable class="PARAMETER">state_data_type</replaceable>
2727
[ , FINALFUNC = <replaceable class="PARAMETER">ffunc</replaceable> ]
@@ -118,7 +118,7 @@ CREATE AGGREGATE <replaceable class="PARAMETER">name</replaceable> (
118118
Note that this behavior is only available when
119119
<replaceable class="PARAMETER">state_data_type</replaceable>
120120
is the same as the first
121-
<replaceable class="PARAMETER">input_data_type</replaceable>.
121+
<replaceable class="PARAMETER">arg_data_type</replaceable>.
122122
When these types are different, you must supply a nonnull initial
123123
condition or use a nonstrict transition function.
124124
</para>
@@ -187,12 +187,36 @@ SELECT col FROM tab ORDER BY col USING sortop LIMIT 1;
187187
</varlistentry>
188188

189189
<varlistentry>
190-
<term><replaceable class="PARAMETER">input_data_type</replaceable></term>
190+
<term><replaceable class="parameter">argmode</replaceable></term>
191+
192+
<listitem>
193+
<para>
194+
The mode of an argument: <literal>IN</> or <literal>VARIADIC</>.
195+
(Aggregate functions do not support <literal>OUT</> arguments.)
196+
If omitted, the default is <literal>IN</>. Only the last argument
197+
can be marked <literal>VARIADIC</>.
198+
</para>
199+
</listitem>
200+
</varlistentry>
201+
202+
<varlistentry>
203+
<term><replaceable class="parameter">arg_name</replaceable></term>
204+
205+
<listitem>
206+
<para>
207+
The name of an argument. This is currently only useful for
208+
documentation purposes. If omitted, the argument has no name.
209+
</para>
210+
</listitem>
211+
</varlistentry>
212+
213+
<varlistentry>
214+
<term><replaceable class="PARAMETER">arg_data_type</replaceable></term>
191215
<listitem>
192216
<para>
193217
An input data type on which this aggregate function operates.
194218
To create a zero-argument aggregate function, write <literal>*</>
195-
in place of the list of input data types. (An example of such an
219+
in place of the list of argument specifications. (An example of such an
196220
aggregate is <function>count(*)</function>.)
197221
</para>
198222
</listitem>
@@ -205,8 +229,8 @@ SELECT col FROM tab ORDER BY col USING sortop LIMIT 1;
205229
In the old syntax for <command>CREATE AGGREGATE</>, the input data type
206230
is specified by a <literal>basetype</> parameter rather than being
207231
written next to the aggregate name. Note that this syntax allows
208-
only one input parameter. To define a zero-argument aggregate function,
209-
specify the <literal>basetype</> as
232+
only one input parameter. To define a zero-argument aggregate function
233+
with this syntax, specify the <literal>basetype</> as
210234
<literal>"ANY"</> (not <literal>*</>).
211235
</para>
212236
</listitem>

‎doc/src/sgml/ref/drop_aggregate.sgml

Lines changed: 29 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -21,7 +21,9 @@ PostgreSQL documentation
2121

2222
<refsynopsisdiv>
2323
<synopsis>
24-
DROP AGGREGATE [ IF EXISTS ] <replaceable class="PARAMETER">name</replaceable> ( <replaceable class="PARAMETER">argtype</replaceable> [ , ... ] ) [ CASCADE | RESTRICT ]
24+
DROP AGGREGATE [ IF EXISTS ]
25+
<replaceable class="parameter">name</replaceable> ( [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">arg_name</replaceable> ] <replaceable class="parameter">arg_data_type</replaceable> [ , ... ] )
26+
[ CASCADE | RESTRICT ]
2527
</synopsis>
2628
</refsynopsisdiv>
2729

@@ -60,12 +62,36 @@ DROP AGGREGATE [ IF EXISTS ] <replaceable class="PARAMETER">name</replaceable> (
6062
</varlistentry>
6163

6264
<varlistentry>
63-
<term><replaceable class="parameter">argtype</replaceable></term>
65+
<term><replaceable class="parameter">argmode</replaceable></term>
66+
67+
<listitem>
68+
<para>
69+
The mode of an argument: <literal>IN</> or <literal>VARIADIC</>.
70+
If omitted, the default is <literal>IN</>.
71+
</para>
72+
</listitem>
73+
</varlistentry>
74+
75+
<varlistentry>
76+
<term><replaceable class="parameter">arg_name</replaceable></term>
77+
78+
<listitem>
79+
<para>
80+
The name of an argument.
81+
Note that <command>DROP AGGREGATE</command> does not actually pay
82+
any attention to argument names, since only the argument data
83+
types are needed to determine the aggregate function's identity.
84+
</para>
85+
</listitem>
86+
</varlistentry>
87+
88+
<varlistentry>
89+
<term><replaceable class="parameter">arg_data_type</replaceable></term>
6490
<listitem>
6591
<para>
6692
An input data type on which the aggregate function operates.
6793
To reference a zero-argument aggregate function, write <literal>*</>
68-
in place of the list of input data types.
94+
in place of the list of argument specifications.
6995
</para>
7096
</listitem>
7197
</varlistentry>

‎doc/src/sgml/ref/security_label.sgml

Lines changed: 5 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -25,7 +25,7 @@ SECURITY LABEL [ FOR <replaceable class="PARAMETER">provider</replaceable> ] ON
2525
{
2626
TABLE <replaceable class="PARAMETER">object_name</replaceable> |
2727
COLUMN <replaceable class="PARAMETER">table_name</replaceable>.<replaceable class="PARAMETER">column_name</replaceable> |
28-
AGGREGATE <replaceable class="PARAMETER">agg_name</replaceable> (<replaceable class="PARAMETER">agg_type</replaceable> [, ...] ) |
28+
AGGREGATE <replaceable class="PARAMETER">agg_name</replaceable> ( [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">argname</replaceable> ] <replaceable class="parameter">agg_type</replaceable> [, ...] ) |
2929
DATABASE <replaceable class="PARAMETER">object_name</replaceable> |
3030
DOMAIN <replaceable class="PARAMETER">object_name</replaceable> |
3131
EVENT TRIGGER <replaceable class="PARAMETER">object_name</replaceable> |
@@ -107,12 +107,12 @@ SECURITY LABEL [ FOR <replaceable class="PARAMETER">provider</replaceable> ] ON
107107
</varlistentry>
108108

109109
<varlistentry>
110-
<term><replaceable class="parameter">arg_type</replaceable></term>
110+
<term><replaceable class="parameter">agg_type</replaceable></term>
111111
<listitem>
112112
<para>
113113
An input data type on which the aggregate function operates.
114114
To reference a zero-argument aggregate function, write <literal>*</>
115-
in place of the list of input data types.
115+
in place of the list of argument specifications.
116116
</para>
117117
</listitem>
118118
</varlistentry>
@@ -125,7 +125,7 @@ SECURITY LABEL [ FOR <replaceable class="PARAMETER">provider</replaceable> ] ON
125125
The mode of a function argument: <literal>IN</>, <literal>OUT</>,
126126
<literal>INOUT</>, or <literal>VARIADIC</>.
127127
If omitted, the default is <literal>IN</>.
128-
Note that <command>SECURITY LABEL ON FUNCTION</command> does not actually
128+
Note that <command>SECURITY LABEL</command> does not actually
129129
pay any attention to <literal>OUT</> arguments, since only the input
130130
arguments are needed to determine the function's identity.
131131
So it is sufficient to list the <literal>IN</>, <literal>INOUT</>,
@@ -140,7 +140,7 @@ SECURITY LABEL [ FOR <replaceable class="PARAMETER">provider</replaceable> ] ON
140140
<listitem>
141141
<para>
142142
The name of a function argument.
143-
Note that <command>SECURITY LABEL ON FUNCTION</command> does not actually
143+
Note that <command>SECURITY LABEL</command> does not actually
144144
pay any attention to argument names, since only the argument data
145145
types are needed to determine the function's identity.
146146
</para>

‎doc/src/sgml/syntax.sgml

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -2524,6 +2524,13 @@ SELECT concat_lower_or_upper('Hello', 'World', uppercase := true);
25242524
having numerous parameters that have default values, named or mixed
25252525
notation can save a great deal of writing and reduce chances for error.
25262526
</para>
2527+
2528+
<note>
2529+
<para>
2530+
Named and mixed call notations can currently be used only with regular
2531+
functions, not with aggregate functions or window functions.
2532+
</para>
2533+
</note>
25272534
</sect2>
25282535
</sect1>
25292536

‎doc/src/sgml/xaggr.sgml

Lines changed: 36 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -169,6 +169,42 @@ SELECT attrelid::regclass, array_accum(atttypid::regtype)
169169
</programlisting>
170170
</para>
171171

172+
<para>
173+
An aggregate function can be made to accept a varying number of arguments
174+
by declaring its last argument as a <literal>VARIADIC</> array, in much
175+
the same fashion as for regular functions; see
176+
<xref linkend="xfunc-sql-variadic-functions">. The aggregate's transition
177+
function must have the same array type as its last argument. The
178+
transition function typically would also be marked <literal>VARIADIC</>,
179+
but this is not strictly required.
180+
</para>
181+
182+
<note>
183+
<para>
184+
Variadic aggregates are easily misused in connection with
185+
the <literal>ORDER BY</> option (see <xref linkend="syntax-aggregates">),
186+
since the parser cannot tell whether the wrong number of actual arguments
187+
have been given in such a combination. Keep in mind that everything to
188+
the right of <literal>ORDER BY</> is a sort key, not an argument to the
189+
aggregate. For example, in
190+
<programlisting>
191+
SELECT myaggregate(a ORDER BY a, b, c) FROM ...
192+
</programlisting>
193+
the parser will see this as a single aggregate function argument and
194+
three sort keys. However, the user might have intended
195+
<programlisting>
196+
SELECT myaggregate(a, b, c ORDER BY a) FROM ...
197+
</programlisting>
198+
If <literal>myaggregate</> is variadic, both these calls could be
199+
perfectly valid.
200+
</para>
201+
202+
<para>
203+
For the same reason, it's wise to think twice before creating aggregate
204+
functions with the same names and different numbers of regular arguments.
205+
</para>
206+
</note>
207+
172208
<para>
173209
A function written in C can detect that it is being called as an
174210
aggregate transition or final function by calling

‎src/backend/catalog/pg_aggregate.c

Lines changed: 11 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -45,8 +45,12 @@ static Oid lookup_agg_function(List *fnName, int nargs, Oid *input_types,
4545
Oid
4646
AggregateCreate(const char *aggName,
4747
Oid aggNamespace,
48-
Oid *aggArgTypes,
4948
int numArgs,
49+
oidvector *parameterTypes,
50+
Datum allParameterTypes,
51+
Datum parameterModes,
52+
Datum parameterNames,
53+
List *parameterDefaults,
5054
List *aggtransfnName,
5155
List *aggfinalfnName,
5256
List *aggsortopName,
@@ -61,6 +65,7 @@ AggregateCreate(const char *aggName,
6165
Oid transfn;
6266
Oid finalfn = InvalidOid; /* can be omitted */
6367
Oid sortop = InvalidOid; /* can be omitted */
68+
Oid *aggArgTypes = parameterTypes->values;
6469
bool hasPolyArg;
6570
bool hasInternalArg;
6671
Oid rettype;
@@ -244,12 +249,11 @@ AggregateCreate(const char *aggName,
244249
false, /* isStrict (not needed for agg) */
245250
PROVOLATILE_IMMUTABLE, /* volatility (not
246251
* needed for agg) */
247-
buildoidvector(aggArgTypes,
248-
numArgs), /* paramTypes */
249-
PointerGetDatum(NULL), /* allParamTypes */
250-
PointerGetDatum(NULL), /* parameterModes */
251-
PointerGetDatum(NULL), /* parameterNames */
252-
NIL, /* parameterDefaults */
252+
parameterTypes, /* paramTypes */
253+
allParameterTypes, /* allParamTypes */
254+
parameterModes, /* parameterModes */
255+
parameterNames, /* parameterNames */
256+
parameterDefaults, /* parameterDefaults */
253257
PointerGetDatum(NULL), /* proconfig */
254258
1, /* procost */
255259
0); /* prorows */

0 commit comments

Comments
 (0)