<?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>38.10. C-Language Functions</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="xfunc-internal.html" title="38.9. Internal Functions" /><link rel="next" href="xaggr.html" title="38.11. User-defined Aggregates" /></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">38.10. C-Language Functions</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="xfunc-internal.html" title="38.9. Internal Functions">Prev</a> </td><td width="10%" align="left"><a accesskey="u" href="extend.html" title="Chapter 38. Extending SQL">Up</a></td><th width="60%" align="center">Chapter 38. Extending <acronym xmlns="http://www.w3.org/1999/xhtml" class="acronym">SQL</acronym></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="xaggr.html" title="38.11. User-defined Aggregates">Next</a></td></tr></table><hr></hr></div><div class="sect1" id="XFUNC-C"><div class="titlepage"><div><div><h2 class="title" style="clear: both">38.10. C-Language Functions</h2></div></div></div><div class="toc"><dl class="toc"><dt><span class="sect2"><a href="xfunc-c.html#XFUNC-C-DYNLOAD">38.10.1. Dynamic Loading</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#XFUNC-C-BASETYPE">38.10.2. Base Types in C-Language Functions</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#id-1.8.3.13.7">38.10.3. Version 1 Calling Conventions</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#id-1.8.3.13.8">38.10.4. Writing Code</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#DFUNC">38.10.5. Compiling and Linking Dynamically-loaded Functions</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#id-1.8.3.13.10">38.10.6. Composite-type Arguments</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#id-1.8.3.13.11">38.10.7. Returning Rows (Composite Types)</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#XFUNC-C-RETURN-SET">38.10.8. Returning Sets</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#id-1.8.3.13.13">38.10.9. Polymorphic Arguments and Return Types</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#XFUNC-TRANSFORM-FUNCTIONS">38.10.10. Transform Functions</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#id-1.8.3.13.15">38.10.11. Shared Memory and LWLocks</a></span></dt><dt><span class="sect2"><a href="xfunc-c.html#EXTEND-CPP">38.10.12. Using C++ for Extensibility</a></span></dt></dl></div><a id="id-1.8.3.13.2" class="indexterm"></a><p>
User-defined functions can be written in C (or a language that can
be made compatible with C, such as C++). Such functions are
compiled into dynamically loadable objects (also called shared
libraries) and are loaded by the server on demand. The dynamic
loading feature is what distinguishes <span class="quote">“<span class="quote">C language</span>”</span> functions
from <span class="quote">“<span class="quote">internal</span>”</span> functions — the actual coding conventions
are essentially the same for both. (Hence, the standard internal
function library is a rich source of coding examples for user-defined
C functions.)
</p><p>
Currently only one calling convention is used for C functions
(<span class="quote">“<span class="quote">version 1</span>”</span>). Support for that calling convention is
indicated by writing a <code class="literal">PG_FUNCTION_INFO_V1()</code> macro
call for the function, as illustrated below.
</p><div class="sect2" id="XFUNC-C-DYNLOAD"><div class="titlepage"><div><div><h3 class="title">38.10.1. Dynamic Loading</h3></div></div></div><a id="id-1.8.3.13.5.2" class="indexterm"></a><p>
The first time a user-defined function in a particular
loadable object file is called in a session,
the dynamic loader loads that object file into memory so that the
function can be called. The <code class="command">CREATE FUNCTION</code>
for a user-defined C function must therefore specify two pieces of
information for the function: the name of the loadable
object file, and the C name (link symbol) of the specific function to call
within that object file. If the C name is not explicitly specified then
it is assumed to be the same as the SQL function name.
</p><p>
The following algorithm is used to locate the shared object file
based on the name given in the <code class="command">CREATE FUNCTION</code>
command:
</p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem"><p>
If the name is an absolute path, the given file is loaded.
</p></li><li class="listitem"><p>
If the name starts with the string <code class="literal">$libdir</code>,
that part is replaced by the <span class="productname">PostgreSQL</span> package
library directory
name, which is determined at build time.<a id="id-1.8.3.13.5.4.2.2.1.3" class="indexterm"></a>
</p></li><li class="listitem"><p>
If the name does not contain a directory part, the file is
searched for in the path specified by the configuration variable
<a class="xref" href="runtime-config-client.html#GUC-DYNAMIC-LIBRARY-PATH">dynamic_library_path</a>.<a id="id-1.8.3.13.5.4.2.3.1.2" class="indexterm"></a>
</p></li><li class="listitem"><p>
Otherwise (the file was not found in the path, or it contains a
non-absolute directory part), the dynamic loader will try to
take the name as given, which will most likely fail. (It is
unreliable to depend on the current working directory.)
</p></li></ol></div><p>
If this sequence does not work, the platform-specific shared
library file name extension (often <code class="filename">.so</code>) is
appended to the given name and this sequence is tried again. If
that fails as well, the load will fail.
</p><p>
It is recommended to locate shared libraries either relative to
<code class="literal">$libdir</code> or through the dynamic library path.
This simplifies version upgrades if the new installation is at a
different location. The actual directory that
<code class="literal">$libdir</code> stands for can be found out with the
command <code class="literal">pg_config --pkglibdir</code>.
</p><p>
The user ID the <span class="productname">PostgreSQL</span> server runs
as must be able to traverse the path to the file you intend to
load. Making the file or a higher-level directory not readable
and/or not executable by the <span class="systemitem">postgres</span>
user is a common mistake.
</p><p>
In any case, the file name that is given in the
<code class="command">CREATE FUNCTION</code> command is recorded literally
in the system catalogs, so if the file needs to be loaded again
the same procedure is applied.
</p><div class="note"><h3 class="title">Note</h3><p>
<span class="productname">PostgreSQL</span> will not compile a C function
automatically. The object file must be compiled before it is referenced
in a <code class="command">CREATE
FUNCTION</code> command. See <a class="xref" href="xfunc-c.html#DFUNC" title="38.10.5. Compiling and Linking Dynamically-loaded Functions">Section 38.10.5</a> for additional
information.
</p></div><a id="id-1.8.3.13.5.9" class="indexterm"></a><p>
To ensure that a dynamically loaded object file is not loaded into an
incompatible server, <span class="productname">PostgreSQL</span> checks that the
file contains a <span class="quote">“<span class="quote">magic block</span>”</span> with the appropriate contents.
This allows the server to detect obvious incompatibilities, such as code
compiled for a different major version of
<span class="productname">PostgreSQL</span>. To include a magic block,
write this in one (and only one) of the module source files, after having
included the header <code class="filename">fmgr.h</code>:
</p><pre class="programlisting">
PG_MODULE_MAGIC;
</pre><p>
</p><p>
After it is used for the first time, a dynamically loaded object
file is retained in memory. Future calls in the same session to
the function(s) in that file will only incur the small overhead of
a symbol table lookup. If you need to force a reload of an object
file, for example after recompiling it, begin a fresh session.
</p><a id="id-1.8.3.13.5.12" class="indexterm"></a><a id="id-1.8.3.13.5.13" class="indexterm"></a><a id="id-1.8.3.13.5.14" class="indexterm"></a><a id="id-1.8.3.13.5.15" class="indexterm"></a><p>
Optionally, a dynamically loaded file can contain initialization and
finalization functions. If the file includes a function named
<code class="function">_PG_init</code>, that function will be called immediately after
loading the file. The function receives no parameters and should
return void. If the file includes a function named
<code class="function">_PG_fini</code>, that function will be called immediately before
unloading the file. Likewise, the function receives no parameters and
should return void. Note that <code class="function">_PG_fini</code> will only be called
during an unload of the file, not during process termination.
(Presently, unloads are disabled and will never occur, but this may
change in the future.)
</p></div><div class="sect2" id="XFUNC-C-BASETYPE"><div class="titlepage"><div><div><h3 class="title">38.10.2. Base Types in C-Language Functions</h3></div></div></div><a id="id-1.8.3.13.6.2" class="indexterm"></a><p>
To know how to write C-language functions, you need to know how
<span class="productname">PostgreSQL</span> internally represents base
data types and how they can be passed to and from functions.
Internally, <span class="productname">PostgreSQL</span> regards a base
type as a <span class="quote">“<span class="quote">blob of memory</span>”</span>. The user-defined
functions that you define over a type in turn define the way that
<span class="productname">PostgreSQL</span> can operate on it. That
is, <span class="productname">PostgreSQL</span> will only store and
retrieve the data from disk and use your user-defined functions
to input, process, and output the data.
</p><p>
Base types can have one of three internal formats:
</p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
pass by value, fixed-length
</p></li><li class="listitem"><p>
pass by reference, fixed-length
</p></li><li class="listitem"><p>
pass by reference, variable-length
</p></li></ul></div><p>
</p><p>
By-value types can only be 1, 2, or 4 bytes in length
(also 8 bytes, if <code class="literal">sizeof(Datum)</code> is 8 on your machine).
You should be careful to define your types such that they will be the
same size (in bytes) on all architectures. For example, the
<code class="literal">long</code> type is dangerous because it is 4 bytes on some
machines and 8 bytes on others, whereas <code class="type">int</code> type is 4 bytes
on most Unix machines. A reasonable implementation of the
<code class="type">int4</code> type on Unix machines might be:
</p><pre class="programlisting">
/* 4-byte integer, passed by value */
typedef int int4;
</pre><p>
(The actual PostgreSQL C code calls this type <code class="type">int32</code>, because
it is a convention in C that <code class="type">int<em class="replaceable"><code>XX</code></em></code>
means <em class="replaceable"><code>XX</code></em> <span class="emphasis"><em>bits</em></span>. Note
therefore also that the C type <code class="type">int8</code> is 1 byte in size. The
SQL type <code class="type">int8</code> is called <code class="type">int64</code> in C. See also
<a class="xref" href="xfunc-c.html#XFUNC-C-TYPE-TABLE" title="Table 38.1. Equivalent C Types for Built-in SQL Types">Table 38.1</a>.)
</p><p>
On the other hand, fixed-length types of any size can
be passed by-reference. For example, here is a sample
implementation of a <span class="productname">PostgreSQL</span> type:
</p><pre class="programlisting">
/* 16-byte structure, passed by reference */
typedef struct
{
double x, y;
} Point;
</pre><p>
Only pointers to such types can be used when passing
them in and out of <span class="productname">PostgreSQL</span> functions.
To return a value of such a type, allocate the right amount of
memory with <code class="literal">palloc</code>, fill in the allocated memory,
and return a pointer to it. (Also, if you just want to return the
same value as one of your input arguments that's of the same data type,
you can skip the extra <code class="literal">palloc</code> and just return the
pointer to the input value.)
</p><p>
Finally, all variable-length types must also be passed
by reference. All variable-length types must begin
with an opaque length field of exactly 4 bytes, which will be set
by <code class="symbol">SET_VARSIZE</code>; never set this field directly! All data to
be stored within that type must be located in the memory
immediately following that length field. The
length field contains the total length of the structure,
that is, it includes the size of the length field
itself.
</p><p>
Another important point is to avoid leaving any uninitialized bits
within data type values; for example, take care to zero out any
alignment padding bytes that might be present in structs. Without
this, logically-equivalent constants of your data type might be
seen as unequal by the planner, leading to inefficient (though not
incorrect) plans.
</p><div class="warning"><h3 class="title">Warning</h3><p>
<span class="emphasis"><em>Never</em></span> modify the contents of a pass-by-reference input
value. If you do so you are likely to corrupt on-disk data, since
the pointer you are given might point directly into a disk buffer.
The sole exception to this rule is explained in
<a class="xref" href="xaggr.html" title="38.11. User-defined Aggregates">Section 38.11</a>.
</p></div><p>
As an example, we can define the type <code class="type">text</code> as
follows:
</p><pre class="programlisting">
typedef struct {
int32 length;
char data[FLEXIBLE_ARRAY_MEMBER];
} text;
</pre><p>
The <code class="literal">[FLEXIBLE_ARRAY_MEMBER]</code> notation means that the actual
length of the data part is not specified by this declaration.
</p><p>
When manipulating
variable-length types, we must be careful to allocate
the correct amount of memory and set the length field correctly.
For example, if we wanted to store 40 bytes in a <code class="structname">text</code>
structure, we might use a code fragment like this:
</p><pre class="programlisting">
#include "postgres.h"
...
char buffer[40]; /* our source data */
...
text *destination = (text *) palloc(VARHDRSZ + 40);
SET_VARSIZE(destination, VARHDRSZ + 40);
memcpy(destination->data, buffer, 40);
...
</pre><p>
<code class="literal">VARHDRSZ</code> is the same as <code class="literal">sizeof(int32)</code>, but
it's considered good style to use the macro <code class="literal">VARHDRSZ</code>
to refer to the size of the overhead for a variable-length type.
Also, the length field <span class="emphasis"><em>must</em></span> be set using the
<code class="literal">SET_VARSIZE</code> macro, not by simple assignment.
</p><p>
<a class="xref" href="xfunc-c.html#XFUNC-C-TYPE-TABLE" title="Table 38.1. Equivalent C Types for Built-in SQL Types">Table 38.1</a> specifies which C type
corresponds to which SQL type when writing a C-language function
that uses a built-in type of <span class="productname">PostgreSQL</span>.
The <span class="quote">“<span class="quote">Defined In</span>”</span> column gives the header file that
needs to be included to get the type definition. (The actual
definition might be in a different file that is included by the
listed file. It is recommended that users stick to the defined
interface.) Note that you should always include
<code class="filename">postgres.h</code> first in any source file, because
it declares a number of things that you will need anyway.
</p><div class="table" id="XFUNC-C-TYPE-TABLE"><p class="title"><strong>Table 38.1. Equivalent C Types for Built-in SQL Types</strong></p><div class="table-contents"><table class="table" summary="Equivalent C Types for Built-in SQL Types" border="1"><colgroup><col /><col /><col /></colgroup><thead><tr><th>
SQL Type
</th><th>
C Type
</th><th>
Defined In
</th></tr></thead><tbody><tr><td><code class="type">abstime</code></td><td><code class="type">AbsoluteTime</code></td><td><code class="filename">utils/nabstime.h</code></td></tr><tr><td><code class="type">bigint</code> (<code class="type">int8</code>)</td><td><code class="type">int64</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">boolean</code></td><td><code class="type">bool</code></td><td><code class="filename">postgres.h</code> (maybe compiler built-in)</td></tr><tr><td><code class="type">box</code></td><td><code class="type">BOX*</code></td><td><code class="filename">utils/geo_decls.h</code></td></tr><tr><td><code class="type">bytea</code></td><td><code class="type">bytea*</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">"char"</code></td><td><code class="type">char</code></td><td>(compiler built-in)</td></tr><tr><td><code class="type">character</code></td><td><code class="type">BpChar*</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">cid</code></td><td><code class="type">CommandId</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">date</code></td><td><code class="type">DateADT</code></td><td><code class="filename">utils/date.h</code></td></tr><tr><td><code class="type">smallint</code> (<code class="type">int2</code>)</td><td><code class="type">int16</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">int2vector</code></td><td><code class="type">int2vector*</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">integer</code> (<code class="type">int4</code>)</td><td><code class="type">int32</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">real</code> (<code class="type">float4</code>)</td><td><code class="type">float4*</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">double precision</code> (<code class="type">float8</code>)</td><td><code class="type">float8*</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">interval</code></td><td><code class="type">Interval*</code></td><td><code class="filename">datatype/timestamp.h</code></td></tr><tr><td><code class="type">lseg</code></td><td><code class="type">LSEG*</code></td><td><code class="filename">utils/geo_decls.h</code></td></tr><tr><td><code class="type">name</code></td><td><code class="type">Name</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">oid</code></td><td><code class="type">Oid</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">oidvector</code></td><td><code class="type">oidvector*</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">path</code></td><td><code class="type">PATH*</code></td><td><code class="filename">utils/geo_decls.h</code></td></tr><tr><td><code class="type">point</code></td><td><code class="type">POINT*</code></td><td><code class="filename">utils/geo_decls.h</code></td></tr><tr><td><code class="type">regproc</code></td><td><code class="type">regproc</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">reltime</code></td><td><code class="type">RelativeTime</code></td><td><code class="filename">utils/nabstime.h</code></td></tr><tr><td><code class="type">text</code></td><td><code class="type">text*</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">tid</code></td><td><code class="type">ItemPointer</code></td><td><code class="filename">storage/itemptr.h</code></td></tr><tr><td><code class="type">time</code></td><td><code class="type">TimeADT</code></td><td><code class="filename">utils/date.h</code></td></tr><tr><td><code class="type">time with time zone</code></td><td><code class="type">TimeTzADT</code></td><td><code class="filename">utils/date.h</code></td></tr><tr><td><code class="type">timestamp</code></td><td><code class="type">Timestamp*</code></td><td><code class="filename">datatype/timestamp.h</code></td></tr><tr><td><code class="type">tinterval</code></td><td><code class="type">TimeInterval</code></td><td><code class="filename">utils/nabstime.h</code></td></tr><tr><td><code class="type">varchar</code></td><td><code class="type">VarChar*</code></td><td><code class="filename">postgres.h</code></td></tr><tr><td><code class="type">xid</code></td><td><code class="type">TransactionId</code></td><td><code class="filename">postgres.h</code></td></tr></tbody></table></div></div><br class="table-break" /><p>
Now that we've gone over all of the possible structures
for base types, we can show some examples of real functions.
</p></div><div class="sect2" id="id-1.8.3.13.7"><div class="titlepage"><div><div><h3 class="title">38.10.3. Version 1 Calling Conventions</h3></div></div></div><p>
The version-1 calling convention relies on macros to suppress most
of the complexity of passing arguments and results. The C declaration
of a version-1 function is always:
</p><pre class="programlisting">
Datum funcname(PG_FUNCTION_ARGS)
</pre><p>
In addition, the macro call:
</p><pre class="programlisting">
PG_FUNCTION_INFO_V1(funcname);
</pre><p>
must appear in the same source file. (Conventionally, it's
written just before the function itself.) This macro call is not
needed for <code class="literal">internal</code>-language functions, since
<span class="productname">PostgreSQL</span> assumes that all internal functions
use the version-1 convention. It is, however, required for
dynamically-loaded functions.
</p><p>
In a version-1 function, each actual argument is fetched using a
<code class="function">PG_GETARG_<em class="replaceable"><code>xxx</code></em>()</code>
macro that corresponds to the argument's data type. In non-strict
functions there needs to be a previous check about argument null-ness
using <code class="function">PG_ARGNULL_<em class="replaceable"><code>xxx</code></em>()</code>.
The result is returned using a
<code class="function">PG_RETURN_<em class="replaceable"><code>xxx</code></em>()</code>
macro for the return type.
<code class="function">PG_GETARG_<em class="replaceable"><code>xxx</code></em>()</code>
takes as its argument the number of the function argument to
fetch, where the count starts at 0.
<code class="function">PG_RETURN_<em class="replaceable"><code>xxx</code></em>()</code>
takes as its argument the actual value to return.
</p><p>
Here are some examples using the version-1 calling convention:
</p><pre class="programlisting">
#include "postgres.h"
#include <string.h>
#include "fmgr.h"
#include "utils/geo_decls.h"
PG_MODULE_MAGIC;
/* by value */
PG_FUNCTION_INFO_V1(add_one);
Datum
add_one(PG_FUNCTION_ARGS)
{
int32 arg = PG_GETARG_INT32(0);
PG_RETURN_INT32(arg + 1);
}
/* by reference, fixed length */
PG_FUNCTION_INFO_V1(add_one_float8);
Datum
add_one_float8(PG_FUNCTION_ARGS)
{
/* The macros for FLOAT8 hide its pass-by-reference nature. */
float8 arg = PG_GETARG_FLOAT8(0);
PG_RETURN_FLOAT8(arg + 1.0);
}
PG_FUNCTION_INFO_V1(makepoint);
Datum
makepoint(PG_FUNCTION_ARGS)
{
/* Here, the pass-by-reference nature of Point is not hidden. */
Point *pointx = PG_GETARG_POINT_P(0);
Point *pointy = PG_GETARG_POINT_P(1);
Point *new_point = (Point *) palloc(sizeof(Point));
new_point->x = pointx->x;
new_point->y = pointy->y;
PG_RETURN_POINT_P(new_point);
}
/* by reference, variable length */
PG_FUNCTION_INFO_V1(copytext);
Datum
copytext(PG_FUNCTION_ARGS)
{
text *t = PG_GETARG_TEXT_PP(0);
/*
* VARSIZE_ANY_EXHDR is the size of the struct in bytes, minus the
* VARHDRSZ or VARHDRSZ_SHORT of its header. Construct the copy with a
* full-length header.
*/
text *new_t = (text *) palloc(VARSIZE_ANY_EXHDR(t) + VARHDRSZ);
SET_VARSIZE(new_t, VARSIZE_ANY_EXHDR(t) + VARHDRSZ);
/*
* VARDATA is a pointer to the data region of the new struct. The source
* could be a short datum, so retrieve its data through VARDATA_ANY.
*/
memcpy((void *) VARDATA(new_t), /* destination */
(void *) VARDATA_ANY(t), /* source */
VARSIZE_ANY_EXHDR(t)); /* how many bytes */
PG_RETURN_TEXT_P(new_t);
}
PG_FUNCTION_INFO_V1(concat_text);
Datum
concat_text(PG_FUNCTION_ARGS)
{
text *arg1 = PG_GETARG_TEXT_PP(0);
text *arg2 = PG_GETARG_TEXT_PP(1);
int32 arg1_size = VARSIZE_ANY_EXHDR(arg1);
int32 arg2_size = VARSIZE_ANY_EXHDR(arg2);
int32 new_text_size = arg1_size + arg2_size + VARHDRSZ;
text *new_text = (text *) palloc(new_text_size);
SET_VARSIZE(new_text, new_text_size);
memcpy(VARDATA(new_text), VARDATA_ANY(arg1), arg1_size);
memcpy(VARDATA(new_text) + arg1_size, VARDATA_ANY(arg2), arg2_size);
PG_RETURN_TEXT_P(new_text);
}
</pre><p>
Supposing that the above code has been prepared in file
<code class="filename">funcs.c</code> and compiled into a shared object,
we could define the functions to <span class="productname">PostgreSQL</span>
with commands like this:
</p><pre class="programlisting">
CREATE FUNCTION add_one(integer) RETURNS integer
AS '<em class="replaceable"><code>DIRECTORY</code></em>/funcs', 'add_one'
LANGUAGE C STRICT;
-- note overloading of SQL function name "add_one"
CREATE FUNCTION add_one(double precision) RETURNS double precision
AS '<em class="replaceable"><code>DIRECTORY</code></em>/funcs', 'add_one_float8'
LANGUAGE C STRICT;
CREATE FUNCTION makepoint(point, point) RETURNS point
AS '<em class="replaceable"><code>DIRECTORY</code></em>/funcs', 'makepoint'
LANGUAGE C STRICT;
CREATE FUNCTION copytext(text) RETURNS text
AS '<em class="replaceable"><code>DIRECTORY</code></em>/funcs', 'copytext'
LANGUAGE C STRICT;
CREATE FUNCTION concat_text(text, text) RETURNS text
AS '<em class="replaceable"><code>DIRECTORY</code></em>/funcs', 'concat_text'
LANGUAGE C STRICT;
</pre><p>
Here, <em class="replaceable"><code>DIRECTORY</code></em> stands for the
directory of the shared library file (for instance the
<span class="productname">PostgreSQL</span> tutorial directory, which
contains the code for the examples used in this section).
(Better style would be to use just <code class="literal">'funcs'</code> in the
<code class="literal">AS</code> clause, after having added
<em class="replaceable"><code>DIRECTORY</code></em> to the search path. In any
case, we can omit the system-specific extension for a shared
library, commonly <code class="literal">.so</code>.)
</p><p>
Notice that we have specified the functions as <span class="quote">“<span class="quote">strict</span>”</span>,
meaning that
the system should automatically assume a null result if any input
value is null. By doing this, we avoid having to check for null inputs
in the function code. Without this, we'd have to check for null values
explicitly, using <code class="function">PG_ARGISNULL()</code>.
</p><p>
At first glance, the version-1 coding conventions might appear to be just
pointless obscurantism, over using plain <code class="literal">C</code> calling
conventions. They do however allow to deal with <code class="literal">NULL</code>able
arguments/return values, and <span class="quote">“<span class="quote">toasted</span>”</span> (compressed or
out-of-line) values.
</p><p>
The macro <code class="function">PG_ARGISNULL(<em class="replaceable"><code>n</code></em>)</code>
allows a function to test whether each input is null. (Of course, doing
this is only necessary in functions not declared <span class="quote">“<span class="quote">strict</span>”</span>.)
As with the
<code class="function">PG_GETARG_<em class="replaceable"><code>xxx</code></em>()</code> macros,
the input arguments are counted beginning at zero. Note that one
should refrain from executing
<code class="function">PG_GETARG_<em class="replaceable"><code>xxx</code></em>()</code> until
one has verified that the argument isn't null.
To return a null result, execute <code class="function">PG_RETURN_NULL()</code>;
this works in both strict and nonstrict functions.
</p><p>
Other options provided by the version-1 interface are two
variants of the
<code class="function">PG_GETARG_<em class="replaceable"><code>xxx</code></em>()</code>
macros. The first of these,
<code class="function">PG_GETARG_<em class="replaceable"><code>xxx</code></em>_COPY()</code>,
guarantees to return a copy of the specified argument that is
safe for writing into. (The normal macros will sometimes return a
pointer to a value that is physically stored in a table, which
must not be written to. Using the
<code class="function">PG_GETARG_<em class="replaceable"><code>xxx</code></em>_COPY()</code>
macros guarantees a writable result.)
The second variant consists of the
<code class="function">PG_GETARG_<em class="replaceable"><code>xxx</code></em>_SLICE()</code>
macros which take three arguments. The first is the number of the
function argument (as above). The second and third are the offset and
length of the segment to be returned. Offsets are counted from
zero, and a negative length requests that the remainder of the
value be returned. These macros provide more efficient access to
parts of large values in the case where they have storage type
<span class="quote">“<span class="quote">external</span>”</span>. (The storage type of a column can be specified using
<code class="literal">ALTER TABLE <em class="replaceable"><code>tablename</code></em> ALTER
COLUMN <em class="replaceable"><code>colname</code></em> SET STORAGE
<em class="replaceable"><code>storagetype</code></em></code>. <em class="replaceable"><code>storagetype</code></em> is one of
<code class="literal">plain</code>, <code class="literal">external</code>, <code class="literal">extended</code>,
or <code class="literal">main</code>.)
</p><p>
Finally, the version-1 function call conventions make it possible
to return set results (<a class="xref" href="xfunc-c.html#XFUNC-C-RETURN-SET" title="38.10.8. Returning Sets">Section 38.10.8</a>) and
implement trigger functions (<a class="xref" href="triggers.html" title="Chapter 39. Triggers">Chapter 39</a>) and
procedural-language call handlers (<a class="xref" href="plhandler.html" title="Chapter 56. Writing A Procedural Language Handler">Chapter 56</a>). For more details
see <code class="filename">src/backend/utils/fmgr/README</code> in the
source distribution.
</p></div><div class="sect2" id="id-1.8.3.13.8"><div class="titlepage"><div><div><h3 class="title">38.10.4. Writing Code</h3></div></div></div><p>
Before we turn to the more advanced topics, we should discuss
some coding rules for <span class="productname">PostgreSQL</span>
C-language functions. While it might be possible to load functions
written in languages other than C into
<span class="productname">PostgreSQL</span>, this is usually difficult
(when it is possible at all) because other languages, such as
C++, FORTRAN, or Pascal often do not follow the same calling
convention as C. That is, other languages do not pass argument
and return values between functions in the same way. For this
reason, we will assume that your C-language functions are
actually written in C.
</p><p>
The basic rules for writing and building C functions are as follows:
</p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
Use <code class="literal">pg_config
--includedir-server</code><a id="id-1.8.3.13.8.3.1.1.1.2" class="indexterm"></a>
to find out where the <span class="productname">PostgreSQL</span> server header
files are installed on your system (or the system that your
users will be running on).
</p></li><li class="listitem"><p>
Compiling and linking your code so that it can be dynamically
loaded into <span class="productname">PostgreSQL</span> always
requires special flags. See <a class="xref" href="xfunc-c.html#DFUNC" title="38.10.5. Compiling and Linking Dynamically-loaded Functions">Section 38.10.5</a> for a
detailed explanation of how to do it for your particular
operating system.
</p></li><li class="listitem"><p>
Remember to define a <span class="quote">“<span class="quote">magic block</span>”</span> for your shared library,
as described in <a class="xref" href="xfunc-c.html#XFUNC-C-DYNLOAD" title="38.10.1. Dynamic Loading">Section 38.10.1</a>.
</p></li><li class="listitem"><p>
When allocating memory, use the
<span class="productname">PostgreSQL</span> functions
<code class="function">palloc</code><a id="id-1.8.3.13.8.3.1.4.1.3" class="indexterm"></a> and <code class="function">pfree</code><a id="id-1.8.3.13.8.3.1.4.1.5" class="indexterm"></a>
instead of the corresponding C library functions
<code class="function">malloc</code> and <code class="function">free</code>.
The memory allocated by <code class="function">palloc</code> will be
freed automatically at the end of each transaction, preventing
memory leaks.
</p></li><li class="listitem"><p>
Always zero the bytes of your structures using <code class="function">memset</code>
(or allocate them with <code class="function">palloc0</code> in the first place).
Even if you assign to each field of your structure, there might be
alignment padding (holes in the structure) that contain
garbage values. Without this, it's difficult to
support hash indexes or hash joins, as you must pick out only
the significant bits of your data structure to compute a hash.
The planner also sometimes relies on comparing constants via
bitwise equality, so you can get undesirable planning results if
logically-equivalent values aren't bitwise equal.
</p></li><li class="listitem"><p>
Most of the internal <span class="productname">PostgreSQL</span>
types are declared in <code class="filename">postgres.h</code>, while
the function manager interfaces
(<code class="symbol">PG_FUNCTION_ARGS</code>, etc.) are in
<code class="filename">fmgr.h</code>, so you will need to include at
least these two files. For portability reasons it's best to
include <code class="filename">postgres.h</code> <span class="emphasis"><em>first</em></span>,
before any other system or user header files. Including
<code class="filename">postgres.h</code> will also include
<code class="filename">elog.h</code> and <code class="filename">palloc.h</code>
for you.
</p></li><li class="listitem"><p>
Symbol names defined within object files must not conflict
with each other or with symbols defined in the
<span class="productname">PostgreSQL</span> server executable. You
will have to rename your functions or variables if you get
error messages to this effect.
</p></li></ul></div><p>
</p></div><div class="sect2" id="DFUNC"><div class="titlepage"><div><div><h3 class="title">38.10.5. Compiling and Linking Dynamically-loaded Functions</h3></div></div></div><p>
Before you are able to use your
<span class="productname">PostgreSQL</span> extension functions written in
C, they must be compiled and linked in a special way to produce a
file that can be dynamically loaded by the server. To be precise, a
<em class="firstterm">shared library</em> needs to be
created.<a id="id-1.8.3.13.9.2.3" class="indexterm"></a>
</p><p>
For information beyond what is contained in this section
you should read the documentation of your
operating system, in particular the manual pages for the C compiler,
<code class="command">cc</code>, and the link editor, <code class="command">ld</code>.
In addition, the <span class="productname">PostgreSQL</span> source code
contains several working examples in the
<code class="filename">contrib</code> directory. If you rely on these
examples you will make your modules dependent on the availability
of the <span class="productname">PostgreSQL</span> source code, however.
</p><p>
Creating shared libraries is generally analogous to linking
executables: first the source files are compiled into object files,
then the object files are linked together. The object files need to
be created as <em class="firstterm">position-independent code</em>
(<acronym class="acronym">PIC</acronym>),<a id="id-1.8.3.13.9.4.3" class="indexterm"></a> which
conceptually means that they can be placed at an arbitrary location
in memory when they are loaded by the executable. (Object files
intended for executables are usually not compiled that way.) The
command to link a shared library contains special flags to
distinguish it from linking an executable (at least in theory
— on some systems the practice is much uglier).
</p><p>
In the following examples we assume that your source code is in a
file <code class="filename">foo.c</code> and we will create a shared library
<code class="filename">foo.so</code>. The intermediate object file will be
called <code class="filename">foo.o</code> unless otherwise noted. A shared
library can contain more than one object file, but we only use one
here.
</p><div class="variablelist"><dl class="variablelist"><dt><span class="term">
<span class="systemitem">FreeBSD</span>
<a id="id-1.8.3.13.9.6.1.1.2" class="indexterm"></a>
</span></dt><dd><p>
The compiler flag to create <acronym class="acronym">PIC</acronym> is
<code class="option">-fPIC</code>. To create shared libraries the compiler
flag is <code class="option">-shared</code>.
</p><pre class="programlisting">
gcc -fPIC -c foo.c
gcc -shared -o foo.so foo.o
</pre><p>
This is applicable as of version 3.0 of
<span class="systemitem">FreeBSD</span>.
</p></dd><dt><span class="term">
<span class="systemitem">HP-UX</span>
<a id="id-1.8.3.13.9.6.2.1.2" class="indexterm"></a>
</span></dt><dd><p>
The compiler flag of the system compiler to create
<acronym class="acronym">PIC</acronym> is <code class="option">+z</code>. When using
<span class="application">GCC</span> it's <code class="option">-fPIC</code>. The
linker flag for shared libraries is <code class="option">-b</code>. So:
</p><pre class="programlisting">
cc +z -c foo.c
</pre><p>
or:
</p><pre class="programlisting">
gcc -fPIC -c foo.c
</pre><p>
and then:
</p><pre class="programlisting">
ld -b -o foo.sl foo.o
</pre><p>
<span class="systemitem">HP-UX</span> uses the extension
<code class="filename">.sl</code> for shared libraries, unlike most other
systems.
</p></dd><dt><span class="term">
<span class="systemitem">Linux</span>
<a id="id-1.8.3.13.9.6.3.1.2" class="indexterm"></a>
</span></dt><dd><p>
The compiler flag to create <acronym class="acronym">PIC</acronym> is
<code class="option">-fPIC</code>.
The compiler flag to create a shared library is
<code class="option">-shared</code>. A complete example looks like this:
</p><pre class="programlisting">
cc -fPIC -c foo.c
cc -shared -o foo.so foo.o
</pre><p>
</p></dd><dt><span class="term">
<span class="systemitem">macOS</span>
<a id="id-1.8.3.13.9.6.4.1.2" class="indexterm"></a>
</span></dt><dd><p>
Here is an example. It assumes the developer tools are installed.
</p><pre class="programlisting">
cc -c foo.c
cc -bundle -flat_namespace -undefined suppress -o foo.so foo.o
</pre><p>
</p></dd><dt><span class="term">
<span class="systemitem">NetBSD</span>
<a id="id-1.8.3.13.9.6.5.1.2" class="indexterm"></a>
</span></dt><dd><p>
The compiler flag to create <acronym class="acronym">PIC</acronym> is
<code class="option">-fPIC</code>. For <acronym class="acronym">ELF</acronym> systems, the
compiler with the flag <code class="option">-shared</code> is used to link
shared libraries. On the older non-ELF systems, <code class="literal">ld
-Bshareable</code> is used.
</p><pre class="programlisting">
gcc -fPIC -c foo.c
gcc -shared -o foo.so foo.o
</pre><p>
</p></dd><dt><span class="term">
<span class="systemitem">OpenBSD</span>
<a id="id-1.8.3.13.9.6.6.1.2" class="indexterm"></a>
</span></dt><dd><p>
The compiler flag to create <acronym class="acronym">PIC</acronym> is
<code class="option">-fPIC</code>. <code class="literal">ld -Bshareable</code> is
used to link shared libraries.
</p><pre class="programlisting">
gcc -fPIC -c foo.c
ld -Bshareable -o foo.so foo.o
</pre><p>
</p></dd><dt><span class="term">
<span class="systemitem">Solaris</span>
<a id="id-1.8.3.13.9.6.7.1.2" class="indexterm"></a>
</span></dt><dd><p>
The compiler flag to create <acronym class="acronym">PIC</acronym> is
<code class="option">-KPIC</code> with the Sun compiler and
<code class="option">-fPIC</code> with <span class="application">GCC</span>. To
link shared libraries, the compiler option is
<code class="option">-G</code> with either compiler or alternatively
<code class="option">-shared</code> with <span class="application">GCC</span>.
</p><pre class="programlisting">
cc -KPIC -c foo.c
cc -G -o foo.so foo.o
</pre><p>
or
</p><pre class="programlisting">
gcc -fPIC -c foo.c
gcc -G -o foo.so foo.o
</pre><p>
</p></dd></dl></div><div class="tip"><h3 class="title">Tip</h3><p>
If this is too complicated for you, you should consider using
<a class="ulink" href="http://www.gnu.org/software/libtool/" target="_top">
<span class="productname">GNU Libtool</span></a>,
which hides the platform differences behind a uniform interface.
</p></div><p>
The resulting shared library file can then be loaded into
<span class="productname">PostgreSQL</span>. When specifying the file name
to the <code class="command">CREATE FUNCTION</code> command, one must give it
the name of the shared library file, not the intermediate object file.
Note that the system's standard shared-library extension (usually
<code class="literal">.so</code> or <code class="literal">.sl</code>) can be omitted from
the <code class="command">CREATE FUNCTION</code> command, and normally should
be omitted for best portability.
</p><p>
Refer back to <a class="xref" href="xfunc-c.html#XFUNC-C-DYNLOAD" title="38.10.1. Dynamic Loading">Section 38.10.1</a> about where the
server expects to find the shared library files.
</p></div><div class="sect2" id="id-1.8.3.13.10"><div class="titlepage"><div><div><h3 class="title">38.10.6. Composite-type Arguments</h3></div></div></div><p>
Composite types do not have a fixed layout like C structures.
Instances of a composite type can contain null fields. In
addition, composite types that are part of an inheritance
hierarchy can have different fields than other members of the
same inheritance hierarchy. Therefore,
<span class="productname">PostgreSQL</span> provides a function
interface for accessing fields of composite types from C.
</p><p>
Suppose we want to write a function to answer the query:
</p><pre class="programlisting">
SELECT name, c_overpaid(emp, 1500) AS overpaid
FROM emp
WHERE name = 'Bill' OR name = 'Sam';
</pre><p>
Using the version-1 calling conventions, we can define
<code class="function">c_overpaid</code> as:
</p><pre class="programlisting">
#include "postgres.h"
#include "executor/executor.h" /* for GetAttributeByName() */
PG_MODULE_MAGIC;
PG_FUNCTION_INFO_V1(c_overpaid);
Datum
c_overpaid(PG_FUNCTION_ARGS)
{
HeapTupleHeader t = PG_GETARG_HEAPTUPLEHEADER(0);
int32 limit = PG_GETARG_INT32(1);
bool isnull;
Datum salary;
salary = GetAttributeByName(t, "salary", &isnull);
if (isnull)
PG_RETURN_BOOL(false);
/* Alternatively, we might prefer to do PG_RETURN_NULL() for null salary. */
PG_RETURN_BOOL(DatumGetInt32(salary) > limit);
}
</pre><p>
</p><p>
<code class="function">GetAttributeByName</code> is the
<span class="productname">PostgreSQL</span> system function that
returns attributes out of the specified row. It has
three arguments: the argument of type <code class="type">HeapTupleHeader</code> passed
into
the function, the name of the desired attribute, and a
return parameter that tells whether the attribute
is null. <code class="function">GetAttributeByName</code> returns a <code class="type">Datum</code>
value that you can convert to the proper data type by using the
appropriate <code class="function">DatumGet<em class="replaceable"><code>XXX</code></em>()</code>
macro. Note that the return value is meaningless if the null flag is
set; always check the null flag before trying to do anything with the
result.
</p><p>
There is also <code class="function">GetAttributeByNum</code>, which selects
the target attribute by column number instead of name.
</p><p>
The following command declares the function
<code class="function">c_overpaid</code> in SQL:
</p><pre class="programlisting">
CREATE FUNCTION c_overpaid(emp, integer) RETURNS boolean
AS '<em class="replaceable"><code>DIRECTORY</code></em>/funcs', 'c_overpaid'
LANGUAGE C STRICT;
</pre><p>
Notice we have used <code class="literal">STRICT</code> so that we did not have to
check whether the input arguments were NULL.
</p></div><div class="sect2" id="id-1.8.3.13.11"><div class="titlepage"><div><div><h3 class="title">38.10.7. Returning Rows (Composite Types)</h3></div></div></div><p>
To return a row or composite-type value from a C-language
function, you can use a special API that provides macros and
functions to hide most of the complexity of building composite
data types. To use this API, the source file must include:
</p><pre class="programlisting">
#include "funcapi.h"
</pre><p>
</p><p>
There are two ways you can build a composite data value (henceforth
a <span class="quote">“<span class="quote">tuple</span>”</span>): you can build it from an array of Datum values,
or from an array of C strings that can be passed to the input
conversion functions of the tuple's column data types. In either
case, you first need to obtain or construct a <code class="structname">TupleDesc</code>
descriptor for the tuple structure. When working with Datums, you
pass the <code class="structname">TupleDesc</code> to <code class="function">BlessTupleDesc</code>,
and then call <code class="function">heap_form_tuple</code> for each row. When working
with C strings, you pass the <code class="structname">TupleDesc</code> to
<code class="function">TupleDescGetAttInMetadata</code>, and then call
<code class="function">BuildTupleFromCStrings</code> for each row. In the case of a
function returning a set of tuples, the setup steps can all be done
once during the first call of the function.
</p><p>
Several helper functions are available for setting up the needed
<code class="structname">TupleDesc</code>. The recommended way to do this in most
functions returning composite values is to call:
</p><pre class="programlisting">
TypeFuncClass get_call_result_type(FunctionCallInfo fcinfo,
Oid *resultTypeId,
TupleDesc *resultTupleDesc)
</pre><p>
passing the same <code class="literal">fcinfo</code> struct passed to the calling function
itself. (This of course requires that you use the version-1
calling conventions.) <code class="varname">resultTypeId</code> can be specified
as <code class="literal">NULL</code> or as the address of a local variable to receive the
function's result type OID. <code class="varname">resultTupleDesc</code> should be the
address of a local <code class="structname">TupleDesc</code> variable. Check that the
result is <code class="literal">TYPEFUNC_COMPOSITE</code>; if so,
<code class="varname">resultTupleDesc</code> has been filled with the needed
<code class="structname">TupleDesc</code>. (If it is not, you can report an error along
the lines of <span class="quote">“<span class="quote">function returning record called in context that
cannot accept type record</span>”</span>.)
</p><div class="tip"><h3 class="title">Tip</h3><p>
<code class="function">get_call_result_type</code> can resolve the actual type of a
polymorphic function result; so it is useful in functions that return
scalar polymorphic results, not only functions that return composites.
The <code class="varname">resultTypeId</code> output is primarily useful for functions
returning polymorphic scalars.
</p></div><div class="note"><h3 class="title">Note</h3><p>
<code class="function">get_call_result_type</code> has a sibling
<code class="function">get_expr_result_type</code>, which can be used to resolve the
expected output type for a function call represented by an expression
tree. This can be used when trying to determine the result type from
outside the function itself. There is also
<code class="function">get_func_result_type</code>, which can be used when only the
function's OID is available. However these functions are not able
to deal with functions declared to return <code class="structname">record</code>, and
<code class="function">get_func_result_type</code> cannot resolve polymorphic types,
so you should preferentially use <code class="function">get_call_result_type</code>.
</p></div><p>
Older, now-deprecated functions for obtaining
<code class="structname">TupleDesc</code>s are:
</p><pre class="programlisting">
TupleDesc RelationNameGetTupleDesc(const char *relname)
</pre><p>
to get a <code class="structname">TupleDesc</code> for the row type of a named relation,
and:
</p><pre class="programlisting">
TupleDesc TypeGetTupleDesc(Oid typeoid, List *colaliases)
</pre><p>
to get a <code class="structname">TupleDesc</code> based on a type OID. This can
be used to get a <code class="structname">TupleDesc</code> for a base or
composite type. It will not work for a function that returns
<code class="structname">record</code>, however, and it cannot resolve polymorphic
types.
</p><p>
Once you have a <code class="structname">TupleDesc</code>, call:
</p><pre class="programlisting">
TupleDesc BlessTupleDesc(TupleDesc tupdesc)
</pre><p>
if you plan to work with Datums, or:
</p><pre class="programlisting">
AttInMetadata *TupleDescGetAttInMetadata(TupleDesc tupdesc)
</pre><p>
if you plan to work with C strings. If you are writing a function
returning set, you can save the results of these functions in the
<code class="structname">FuncCallContext</code> structure — use the
<code class="structfield">tuple_desc</code> or <code class="structfield">attinmeta</code> field
respectively.
</p><p>
When working with Datums, use:
</p><pre class="programlisting">
HeapTuple heap_form_tuple(TupleDesc tupdesc, Datum *values, bool *isnull)
</pre><p>
to build a <code class="structname">HeapTuple</code> given user data in Datum form.
</p><p>
When working with C strings, use:
</p><pre class="programlisting">
HeapTuple BuildTupleFromCStrings(AttInMetadata *attinmeta, char **values)
</pre><p>
to build a <code class="structname">HeapTuple</code> given user data
in C string form. <em class="parameter"><code>values</code></em> is an array of C strings,
one for each attribute of the return row. Each C string should be in
the form expected by the input function of the attribute data
type. In order to return a null value for one of the attributes,
the corresponding pointer in the <em class="parameter"><code>values</code></em> array
should be set to <code class="symbol">NULL</code>. This function will need to
be called again for each row you return.
</p><p>
Once you have built a tuple to return from your function, it
must be converted into a <code class="type">Datum</code>. Use:
</p><pre class="programlisting">
HeapTupleGetDatum(HeapTuple tuple)
</pre><p>
to convert a <code class="structname">HeapTuple</code> into a valid Datum. This
<code class="type">Datum</code> can be returned directly if you intend to return
just a single row, or it can be used as the current return value
in a set-returning function.
</p><p>
An example appears in the next section.
</p></div><div class="sect2" id="XFUNC-C-RETURN-SET"><div class="titlepage"><div><div><h3 class="title">38.10.8. Returning Sets</h3></div></div></div><p>
There is also a special API that provides support for returning
sets (multiple rows) from a C-language function. A set-returning
function must follow the version-1 calling conventions. Also,
source files must include <code class="filename">funcapi.h</code>, as
above.
</p><p>
A set-returning function (<acronym class="acronym">SRF</acronym>) is called
once for each item it returns. The <acronym class="acronym">SRF</acronym> must
therefore save enough state to remember what it was doing and
return the next item on each call.
The structure <code class="structname">FuncCallContext</code> is provided to help
control this process. Within a function, <code class="literal">fcinfo->flinfo->fn_extra</code>
is used to hold a pointer to <code class="structname">FuncCallContext</code>
across calls.
</p><pre class="programlisting">
typedef struct FuncCallContext
{
/*
* Number of times we've been called before
*
* call_cntr is initialized to 0 for you by SRF_FIRSTCALL_INIT(), and
* incremented for you every time SRF_RETURN_NEXT() is called.
*/
uint64 call_cntr;
/*
* OPTIONAL maximum number of calls
*
* max_calls is here for convenience only and setting it is optional.
* If not set, you must provide alternative means to know when the
* function is done.
*/
uint64 max_calls;
/*
* OPTIONAL pointer to result slot
*
* This is obsolete and only present for backward compatibility, viz,
* user-defined SRFs that use the deprecated TupleDescGetSlot().
*/
TupleTableSlot *slot;
/*
* OPTIONAL pointer to miscellaneous user-provided context information
*
* user_fctx is for use as a pointer to your own data to retain
* arbitrary context information between calls of your function.
*/
void *user_fctx;
/*
* OPTIONAL pointer to struct containing attribute type input metadata
*
* attinmeta is for use when returning tuples (i.e., composite data types)
* and is not used when returning base data types. It is only needed
* if you intend to use BuildTupleFromCStrings() to create the return
* tuple.
*/
AttInMetadata *attinmeta;
/*
* memory context used for structures that must live for multiple calls
*
* multi_call_memory_ctx is set by SRF_FIRSTCALL_INIT() for you, and used
* by SRF_RETURN_DONE() for cleanup. It is the most appropriate memory
* context for any memory that is to be reused across multiple calls
* of the SRF.
*/
MemoryContext multi_call_memory_ctx;
/*
* OPTIONAL pointer to struct containing tuple description
*
* tuple_desc is for use when returning tuples (i.e., composite data types)
* and is only needed if you are going to build the tuples with
* heap_form_tuple() rather than with BuildTupleFromCStrings(). Note that
* the TupleDesc pointer stored here should usually have been run through
* BlessTupleDesc() first.
*/
TupleDesc tuple_desc;
} FuncCallContext;
</pre><p>
</p><p>
An <acronym class="acronym">SRF</acronym> uses several functions and macros that
automatically manipulate the <code class="structname">FuncCallContext</code>
structure (and expect to find it via <code class="literal">fn_extra</code>). Use:
</p><pre class="programlisting">
SRF_IS_FIRSTCALL()
</pre><p>
to determine if your function is being called for the first or a
subsequent time. On the first call (only) use:
</p><pre class="programlisting">
SRF_FIRSTCALL_INIT()
</pre><p>
to initialize the <code class="structname">FuncCallContext</code>. On every function call,
including the first, use:
</p><pre class="programlisting">
SRF_PERCALL_SETUP()
</pre><p>
to properly set up for using the <code class="structname">FuncCallContext</code>
and clearing any previously returned data left over from the
previous pass.
</p><p>
If your function has data to return, use:
</p><pre class="programlisting">
SRF_RETURN_NEXT(funcctx, result)
</pre><p>
to return it to the caller. (<code class="literal">result</code> must be of type
<code class="type">Datum</code>, either a single value or a tuple prepared as
described above.) Finally, when your function is finished
returning data, use:
</p><pre class="programlisting">
SRF_RETURN_DONE(funcctx)
</pre><p>
to clean up and end the <acronym class="acronym">SRF</acronym>.
</p><p>
The memory context that is current when the <acronym class="acronym">SRF</acronym> is called is
a transient context that will be cleared between calls. This means
that you do not need to call <code class="function">pfree</code> on everything
you allocated using <code class="function">palloc</code>; it will go away anyway. However, if you want to allocate
any data structures to live across calls, you need to put them somewhere
else. The memory context referenced by
<code class="structfield">multi_call_memory_ctx</code> is a suitable location for any
data that needs to survive until the <acronym class="acronym">SRF</acronym> is finished running. In most
cases, this means that you should switch into
<code class="structfield">multi_call_memory_ctx</code> while doing the first-call setup.
</p><div class="warning"><h3 class="title">Warning</h3><p>
While the actual arguments to the function remain unchanged between
calls, if you detoast the argument values (which is normally done
transparently by the
<code class="function">PG_GETARG_<em class="replaceable"><code>xxx</code></em></code> macro)
in the transient context then the detoasted copies will be freed on
each cycle. Accordingly, if you keep references to such values in
your <code class="structfield">user_fctx</code>, you must either copy them into the
<code class="structfield">multi_call_memory_ctx</code> after detoasting, or ensure
that you detoast the values only in that context.
</p></div><p>
A complete pseudo-code example looks like the following:
</p><pre class="programlisting">
Datum
my_set_returning_function(PG_FUNCTION_ARGS)
{
FuncCallContext *funcctx;
Datum result;
<em class="replaceable"><code>further declarations as needed</code></em>
if (SRF_IS_FIRSTCALL())
{
MemoryContext oldcontext;
funcctx = SRF_FIRSTCALL_INIT();
oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
/* One-time setup code appears here: */
<em class="replaceable"><code>user code</code></em>
<em class="replaceable"><code>if returning composite</code></em>
<em class="replaceable"><code>build TupleDesc, and perhaps AttInMetadata</code></em>
<em class="replaceable"><code>endif returning composite</code></em>
<em class="replaceable"><code>user code</code></em>
MemoryContextSwitchTo(oldcontext);
}
/* Each-time setup code appears here: */
<em class="replaceable"><code>user code</code></em>
funcctx = SRF_PERCALL_SETUP();
<em class="replaceable"><code>user code</code></em>
/* this is just one way we might test whether we are done: */
if (funcctx->call_cntr < funcctx->max_calls)
{
/* Here we want to return another item: */
<em class="replaceable"><code>user code</code></em>
<em class="replaceable"><code>obtain result Datum</code></em>
SRF_RETURN_NEXT(funcctx, result);
}
else
{
/* Here we are done returning items and just need to clean up: */
<em class="replaceable"><code>user code</code></em>
SRF_RETURN_DONE(funcctx);
}
}
</pre><p>
</p><p>
A complete example of a simple <acronym class="acronym">SRF</acronym> returning a composite type
looks like:
</p><pre class="programlisting">
PG_FUNCTION_INFO_V1(retcomposite);
Datum
retcomposite(PG_FUNCTION_ARGS)
{
FuncCallContext *funcctx;
int call_cntr;
int max_calls;
TupleDesc tupdesc;
AttInMetadata *attinmeta;
/* stuff done only on the first call of the function */
if (SRF_IS_FIRSTCALL())
{
MemoryContext oldcontext;
/* create a function context for cross-call persistence */
funcctx = SRF_FIRSTCALL_INIT();
/* switch to memory context appropriate for multiple function calls */
oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
/* total number of tuples to be returned */
funcctx->max_calls = PG_GETARG_UINT32(0);
/* Build a tuple descriptor for our result type */
if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("function returning record called in context "
"that cannot accept type record")));
/*
* generate attribute metadata needed later to produce tuples from raw
* C strings
*/
attinmeta = TupleDescGetAttInMetadata(tupdesc);
funcctx->attinmeta = attinmeta;
MemoryContextSwitchTo(oldcontext);
}
/* stuff done on every call of the function */
funcctx = SRF_PERCALL_SETUP();
call_cntr = funcctx->call_cntr;
max_calls = funcctx->max_calls;
attinmeta = funcctx->attinmeta;
if (call_cntr < max_calls) /* do when there is more left to send */
{
char **values;
HeapTuple tuple;
Datum result;
/*
* Prepare a values array for building the returned tuple.
* This should be an array of C strings which will
* be processed later by the type input functions.
*/
values = (char **) palloc(3 * sizeof(char *));
values[0] = (char *) palloc(16 * sizeof(char));
values[1] = (char *) palloc(16 * sizeof(char));
values[2] = (char *) palloc(16 * sizeof(char));
snprintf(values[0], 16, "%d", 1 * PG_GETARG_INT32(1));
snprintf(values[1], 16, "%d", 2 * PG_GETARG_INT32(1));
snprintf(values[2], 16, "%d", 3 * PG_GETARG_INT32(1));
/* build a tuple */
tuple = BuildTupleFromCStrings(attinmeta, values);
/* make the tuple into a datum */
result = HeapTupleGetDatum(tuple);
/* clean up (this is not really necessary) */
pfree(values[0]);
pfree(values[1]);
pfree(values[2]);
pfree(values);
SRF_RETURN_NEXT(funcctx, result);
}
else /* do when there is no more left */
{
SRF_RETURN_DONE(funcctx);
}
}
</pre><p>
One way to declare this function in SQL is:
</p><pre class="programlisting">
CREATE TYPE __retcomposite AS (f1 integer, f2 integer, f3 integer);
CREATE OR REPLACE FUNCTION retcomposite(integer, integer)
RETURNS SETOF __retcomposite
AS '<em class="replaceable"><code>filename</code></em>', 'retcomposite'
LANGUAGE C IMMUTABLE STRICT;
</pre><p>
A different way is to use OUT parameters:
</p><pre class="programlisting">
CREATE OR REPLACE FUNCTION retcomposite(IN integer, IN integer,
OUT f1 integer, OUT f2 integer, OUT f3 integer)
RETURNS SETOF record
AS '<em class="replaceable"><code>filename</code></em>', 'retcomposite'
LANGUAGE C IMMUTABLE STRICT;
</pre><p>
Notice that in this method the output type of the function is formally
an anonymous <code class="structname">record</code> type.
</p><p>
The directory <a class="link" href="tablefunc.html" title="F.38. tablefunc"><code class="filename">contrib/tablefunc</code></a>
module in the source distribution contains more examples of
set-returning functions.
</p></div><div class="sect2" id="id-1.8.3.13.13"><div class="titlepage"><div><div><h3 class="title">38.10.9. Polymorphic Arguments and Return Types</h3></div></div></div><p>
C-language functions can be declared to accept and
return the polymorphic types
<code class="type">anyelement</code>, <code class="type">anyarray</code>, <code class="type">anynonarray</code>,
<code class="type">anyenum</code>, and <code class="type">anyrange</code>.
See <a class="xref" href="extend-type-system.html#EXTEND-TYPES-POLYMORPHIC" title="38.2.5. Polymorphic Types">Section 38.2.5</a> for a more detailed explanation
of polymorphic functions. When function arguments or return types
are defined as polymorphic types, the function author cannot know
in advance what data type it will be called with, or
need to return. There are two routines provided in <code class="filename">fmgr.h</code>
to allow a version-1 C function to discover the actual data types
of its arguments and the type it is expected to return. The routines are
called <code class="literal">get_fn_expr_rettype(FmgrInfo *flinfo)</code> and
<code class="literal">get_fn_expr_argtype(FmgrInfo *flinfo, int argnum)</code>.
They return the result or argument type OID, or <code class="symbol">InvalidOid</code> if the
information is not available.
The structure <code class="literal">flinfo</code> is normally accessed as
<code class="literal">fcinfo->flinfo</code>. The parameter <code class="literal">argnum</code>
is zero based. <code class="function">get_call_result_type</code> can also be used
as an alternative to <code class="function">get_fn_expr_rettype</code>.
There is also <code class="function">get_fn_expr_variadic</code>, which can be used to
find out whether variadic arguments have been merged into an array.
This is primarily useful for <code class="literal">VARIADIC "any"</code> functions,
since such merging will always have occurred for variadic functions
taking ordinary array types.
</p><p>
For example, suppose we want to write a function to accept a single
element of any type, and return a one-dimensional array of that type:
</p><pre class="programlisting">
PG_FUNCTION_INFO_V1(make_array);
Datum
make_array(PG_FUNCTION_ARGS)
{
ArrayType *result;
Oid element_type = get_fn_expr_argtype(fcinfo->flinfo, 0);
Datum element;
bool isnull;
int16 typlen;
bool typbyval;
char typalign;
int ndims;
int dims[MAXDIM];
int lbs[MAXDIM];
if (!OidIsValid(element_type))
elog(ERROR, "could not determine data type of input");
/* get the provided element, being careful in case it's NULL */
isnull = PG_ARGISNULL(0);
if (isnull)
element = (Datum) 0;
else
element = PG_GETARG_DATUM(0);
/* we have one dimension */
ndims = 1;
/* and one element */
dims[0] = 1;
/* and lower bound is 1 */
lbs[0] = 1;
/* get required info about the element type */
get_typlenbyvalalign(element_type, &typlen, &typbyval, &typalign);
/* now build the array */
result = construct_md_array(&element, &isnull, ndims, dims, lbs,
element_type, typlen, typbyval, typalign);
PG_RETURN_ARRAYTYPE_P(result);
}
</pre><p>
</p><p>
The following command declares the function
<code class="function">make_array</code> in SQL:
</p><pre class="programlisting">
CREATE FUNCTION make_array(anyelement) RETURNS anyarray
AS '<em class="replaceable"><code>DIRECTORY</code></em>/funcs', 'make_array'
LANGUAGE C IMMUTABLE;
</pre><p>
</p><p>
There is a variant of polymorphism that is only available to C-language
functions: they can be declared to take parameters of type
<code class="literal">"any"</code>. (Note that this type name must be double-quoted,
since it's also a SQL reserved word.) This works like
<code class="type">anyelement</code> except that it does not constrain different
<code class="literal">"any"</code> arguments to be the same type, nor do they help
determine the function's result type. A C-language function can also
declare its final parameter to be <code class="literal">VARIADIC "any"</code>. This will
match one or more actual arguments of any type (not necessarily the same
type). These arguments will <span class="emphasis"><em>not</em></span> be gathered into an array
as happens with normal variadic functions; they will just be passed to
the function separately. The <code class="function">PG_NARGS()</code> macro and the
methods described above must be used to determine the number of actual
arguments and their types when using this feature. Also, users of such
a function might wish to use the <code class="literal">VARIADIC</code> keyword in their
function call, with the expectation that the function would treat the
array elements as separate arguments. The function itself must implement
that behavior if wanted, after using <code class="function">get_fn_expr_variadic</code> to
detect that the actual argument was marked with <code class="literal">VARIADIC</code>.
</p></div><div class="sect2" id="XFUNC-TRANSFORM-FUNCTIONS"><div class="titlepage"><div><div><h3 class="title">38.10.10. Transform Functions</h3></div></div></div><p>
Some function calls can be simplified during planning based on
properties specific to the function. For example,
<code class="literal">int4mul(n, 1)</code> could be simplified to just <code class="literal">n</code>.
To define such function-specific optimizations, write a
<em class="firstterm">transform function</em> and place its OID in the
<code class="structfield">protransform</code> field of the primary function's
<code class="structname">pg_proc</code> entry. The transform function must have the SQL
signature <code class="literal">protransform(internal) RETURNS internal</code>. The
argument, actually <code class="type">FuncExpr *</code>, is a dummy node representing a
call to the primary function. If the transform function's study of the
expression tree proves that a simplified expression tree can substitute
for all possible concrete calls represented thereby, build and return
that simplified expression. Otherwise, return a <code class="literal">NULL</code>
pointer (<span class="emphasis"><em>not</em></span> a SQL null).
</p><p>
We make no guarantee that <span class="productname">PostgreSQL</span> will never call the
primary function in cases that the transform function could simplify.
Ensure rigorous equivalence between the simplified expression and an
actual call to the primary function.
</p><p>
Currently, this facility is not exposed to users at the SQL level
because of security concerns, so it is only practical to use for
optimizing built-in functions.
</p></div><div class="sect2" id="id-1.8.3.13.15"><div class="titlepage"><div><div><h3 class="title">38.10.11. Shared Memory and LWLocks</h3></div></div></div><p>
Add-ins can reserve LWLocks and an allocation of shared memory on server
startup. The add-in's shared library must be preloaded by specifying
it in
<a class="xref" href="runtime-config-client.html#GUC-SHARED-PRELOAD-LIBRARIES">shared_preload_libraries</a><a id="id-1.8.3.13.15.2.2" class="indexterm"></a>.
Shared memory is reserved by calling:
</p><pre class="programlisting">
void RequestAddinShmemSpace(int size)
</pre><p>
from your <code class="function">_PG_init</code> function.
</p><p>
LWLocks are reserved by calling:
</p><pre class="programlisting">
void RequestNamedLWLockTranche(const char *tranche_name, int num_lwlocks)
</pre><p>
from <code class="function">_PG_init</code>. This will ensure that an array of
<code class="literal">num_lwlocks</code> LWLocks is available under the name
<code class="literal">tranche_name</code>. Use <code class="function">GetNamedLWLockTranche</code>
to get a pointer to this array.
</p><p>
To avoid possible race-conditions, each backend should use the LWLock
<code class="function">AddinShmemInitLock</code> when connecting to and initializing
its allocation of shared memory, as shown here:
</p><pre class="programlisting">
static mystruct *ptr = NULL;
if (!ptr)
{
bool found;
LWLockAcquire(AddinShmemInitLock, LW_EXCLUSIVE);
ptr = ShmemInitStruct("my struct name", size, &found);
if (!found)
{
initialize contents of shmem area;
acquire any requested LWLocks using:
ptr->locks = GetNamedLWLockTranche("my tranche name");
}
LWLockRelease(AddinShmemInitLock);
}
</pre><p>
</p></div><div class="sect2" id="EXTEND-CPP"><div class="titlepage"><div><div><h3 class="title">38.10.12. Using C++ for Extensibility</h3></div></div></div><a id="id-1.8.3.13.16.2" class="indexterm"></a><p>
Although the <span class="productname">PostgreSQL</span> backend is written in
C, it is possible to write extensions in C++ if these guidelines are
followed:
</p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
All functions accessed by the backend must present a C interface
to the backend; these C functions can then call C++ functions.
For example, <code class="literal">extern C</code> linkage is required for
backend-accessed functions. This is also necessary for any
functions that are passed as pointers between the backend and
C++ code.
</p></li><li class="listitem"><p>
Free memory using the appropriate deallocation method. For example,
most backend memory is allocated using <code class="function">palloc()</code>, so use
<code class="function">pfree()</code> to free it. Using C++
<code class="function">delete</code> in such cases will fail.
</p></li><li class="listitem"><p>
Prevent exceptions from propagating into the C code (use a catch-all
block at the top level of all <code class="literal">extern C</code> functions). This
is necessary even if the C++ code does not explicitly throw any
exceptions, because events like out-of-memory can still throw
exceptions. Any exceptions must be caught and appropriate errors
passed back to the C interface. If possible, compile C++ with
<code class="option">-fno-exceptions</code> to eliminate exceptions entirely; in such
cases, you must check for failures in your C++ code, e.g. check for
NULL returned by <code class="function">new()</code>.
</p></li><li class="listitem"><p>
If calling backend functions from C++ code, be sure that the
C++ call stack contains only plain old data structures
(<acronym class="acronym">POD</acronym>). This is necessary because backend errors
generate a distant <code class="function">longjmp()</code> that does not properly
unroll a C++ call stack with non-POD objects.
</p></li></ul></div><p>
</p><p>
In summary, it is best to place C++ code behind a wall of
<code class="literal">extern C</code> functions that interface to the backend,
and avoid exception, memory, and call stack leakage.
</p></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="xfunc-internal.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="extend.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="xaggr.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">38.9. Internal Functions </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> 38.11. User-defined Aggregates</td></tr></table></div></body></html>
