2000-03-31 05:27:42 +02:00
|
|
|
<!--
|
|
|
|
$Header: /cvsroot/pgsql/doc/src/sgml/xfunc.sgml,v 1.13 2000/03/31 03:27:41 thomas Exp $
|
|
|
|
-->
|
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<chapter id="xfunc">
|
|
|
|
<title id="xfunc-title">Extending <acronym>SQL</acronym>: Functions</title>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
As it turns out, part of defining a new type is the
|
|
|
|
definition of functions that describe its behavior.
|
|
|
|
Consequently, while it is possible to define a new
|
|
|
|
function without defining a new type, the reverse is
|
|
|
|
not true. We therefore describe how to add new functions
|
|
|
|
to <productname>Postgres</productname> before describing
|
|
|
|
how to add new types.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
<productname>Postgres</productname> <acronym>SQL</acronym>
|
|
|
|
provides three types of functions:
|
|
|
|
|
|
|
|
<itemizedlist>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
query language functions
|
|
|
|
(functions written in <acronym>SQL</acronym>)
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
procedural language
|
|
|
|
functions (functions written in, for example, PLTCL or PLSQL)
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
programming
|
|
|
|
language functions (functions written in a compiled
|
|
|
|
programming language such as <acronym>C</acronym>)
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
</itemizedlist>
|
|
|
|
|
|
|
|
Every kind
|
|
|
|
of function can take a base type, a composite type or
|
|
|
|
some combination as arguments (parameters). In addition,
|
|
|
|
every kind of function can return a base type or
|
|
|
|
a composite type. It's easiest to define <acronym>SQL</acronym>
|
|
|
|
functions, so we'll start with those. Examples in this section
|
|
|
|
can also be found in <filename>funcs.sql</filename>
|
|
|
|
and <filename>funcs.c</filename>.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<sect1>
|
|
|
|
<title>Query Language (<acronym>SQL</acronym>) Functions</title>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
SQL functions execute an arbitrary list of SQL queries, returning
|
|
|
|
the results of the last query in the list. SQL functions in general
|
|
|
|
return sets. If their returntype is not specified as a
|
|
|
|
<literal>setof</literal>,
|
|
|
|
then an arbitrary element of the last query's result will be returned.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
The body of a SQL function following AS
|
|
|
|
should be a list of queries separated by whitespace characters and
|
|
|
|
bracketed within quotation marks. Note that quotation marks used in
|
|
|
|
the queries must be escaped, by preceding them with two
|
|
|
|
backslashes.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
Arguments to the SQL function may be referenced in the queries using
|
|
|
|
a $n syntax: $1 refers to the first argument, $2 to the second, and so
|
|
|
|
on. If an argument is complex, then a <firstterm>dot</firstterm>
|
|
|
|
notation (e.g. "$1.emp") may be
|
|
|
|
used to access attributes of the argument or
|
|
|
|
to invoke functions.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<sect2>
|
|
|
|
<title>Examples</title>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
To illustrate a simple SQL function, consider the following,
|
|
|
|
which might be used to debit a bank account:
|
|
|
|
|
|
|
|
<programlisting>
|
|
|
|
create function TP1 (int4, float8) returns int4
|
|
|
|
as 'update BANK set balance = BANK.balance - $2
|
|
|
|
where BANK.acctountno = $1
|
|
|
|
select(x = 1)'
|
|
|
|
language 'sql';
|
|
|
|
</programlisting>
|
|
|
|
|
|
|
|
A user could execute this function to debit account 17 by $100.00 as
|
|
|
|
follows:
|
|
|
|
|
|
|
|
<programlisting>
|
|
|
|
select (x = TP1( 17,100.0));
|
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
The following more interesting example takes a single argument of type
|
|
|
|
EMP, and retrieves multiple results:
|
|
|
|
|
|
|
|
<programlisting>
|
|
|
|
select function hobbies (EMP) returns set of HOBBIES
|
|
|
|
as 'select (HOBBIES.all) from HOBBIES
|
|
|
|
where $1.name = HOBBIES.person'
|
|
|
|
language 'sql';
|
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
</sect2>
|
|
|
|
|
|
|
|
<sect2>
|
|
|
|
<title><acronym>SQL</acronym> Functions on Base Types</title>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
The simplest possible <acronym>SQL</acronym> function has no arguments and
|
1999-10-04 17:18:54 +02:00
|
|
|
simply returns a base type, such as <literal>int4</literal>:
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
CREATE FUNCTION one() RETURNS int4
|
|
|
|
AS 'SELECT 1 as RESULT' LANGUAGE 'sql';
|
|
|
|
|
|
|
|
SELECT one() AS answer;
|
|
|
|
|
|
|
|
+-------+
|
|
|
|
|answer |
|
|
|
|
+-------+
|
|
|
|
|1 |
|
|
|
|
+-------+
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
Notice that we defined a target list for the function
|
|
|
|
(with the name RESULT), but the target list of the
|
|
|
|
query that invoked the function overrode the function's
|
|
|
|
target list. Hence, the result is labelled answer
|
|
|
|
instead of one.
|
1999-07-22 17:11:05 +02:00
|
|
|
</para>
|
|
|
|
<para>
|
|
|
|
It's almost as easy to define <acronym>SQL</acronym> functions
|
1999-03-14 16:24:15 +01:00
|
|
|
that take base types as arguments. In the example below, notice
|
1998-03-01 09:16:16 +01:00
|
|
|
how we refer to the arguments within the function as $1
|
1999-07-22 17:11:05 +02:00
|
|
|
and $2:
|
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
CREATE FUNCTION add_em(int4, int4) RETURNS int4
|
|
|
|
AS 'SELECT $1 + $2;' LANGUAGE 'sql';
|
|
|
|
|
|
|
|
SELECT add_em(1, 2) AS answer;
|
|
|
|
|
|
|
|
+-------+
|
|
|
|
|answer |
|
|
|
|
+-------+
|
|
|
|
|3 |
|
|
|
|
+-------+
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
</sect2>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<sect2>
|
|
|
|
<title><acronym>SQL</acronym> Functions on Composite Types</title>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
When specifying functions with arguments of composite
|
|
|
|
types (such as EMP), we must not only specify which
|
|
|
|
argument we want (as we did above with $1 and $2) but
|
|
|
|
also the attributes of that argument. For example,
|
|
|
|
take the function double_salary that computes what your
|
1999-07-22 17:11:05 +02:00
|
|
|
salary would be if it were doubled:
|
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
CREATE FUNCTION double_salary(EMP) RETURNS int4
|
|
|
|
AS 'SELECT $1.salary * 2 AS salary;' LANGUAGE 'sql';
|
|
|
|
|
|
|
|
SELECT name, double_salary(EMP) AS dream
|
|
|
|
FROM EMP
|
1999-03-14 16:24:15 +01:00
|
|
|
WHERE EMP.cubicle ~= '(2,1)'::point;
|
|
|
|
|
1998-03-01 09:16:16 +01:00
|
|
|
|
|
|
|
+-----+-------+
|
|
|
|
|name | dream |
|
|
|
|
+-----+-------+
|
|
|
|
|Sam | 2400 |
|
|
|
|
+-----+-------+
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
Notice the use of the syntax $1.salary.
|
|
|
|
Before launching into the subject of functions that
|
|
|
|
return composite types, we must first introduce the
|
|
|
|
function notation for projecting attributes. The simple way
|
|
|
|
to explain this is that we can usually use the
|
1999-07-22 17:11:05 +02:00
|
|
|
notation attribute(class) and class.attribute interchangably:
|
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
--
|
|
|
|
-- this is the same as:
|
|
|
|
-- SELECT EMP.name AS youngster FROM EMP WHERE EMP.age < 30
|
|
|
|
--
|
|
|
|
SELECT name(EMP) AS youngster
|
|
|
|
FROM EMP
|
|
|
|
WHERE age(EMP) < 30;
|
|
|
|
|
|
|
|
+----------+
|
|
|
|
|youngster |
|
|
|
|
+----------+
|
|
|
|
|Sam |
|
|
|
|
+----------+
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
As we shall see, however, this is not always the case.
|
|
|
|
This function notation is important when we want to use
|
|
|
|
a function that returns a single instance. We do this
|
|
|
|
by assembling the entire instance within the function,
|
|
|
|
attribute by attribute. This is an example of a function
|
|
|
|
that returns a single EMP instance:
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
CREATE FUNCTION new_emp() RETURNS EMP
|
|
|
|
AS 'SELECT \'None\'::text AS name,
|
|
|
|
1000 AS salary,
|
|
|
|
25 AS age,
|
1999-03-14 16:24:15 +01:00
|
|
|
\'(2,2)\'::point AS cubicle'
|
1998-03-01 09:16:16 +01:00
|
|
|
LANGUAGE 'sql';
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
In this case we have specified each of the attributes
|
|
|
|
with a constant value, but any computation or expression
|
|
|
|
could have been substituted for these constants.
|
|
|
|
Defining a function like this can be tricky. Some of
|
|
|
|
the more important caveats are as follows:
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<itemizedlist>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
The target list order must be exactly the same as
|
|
|
|
that in which the attributes appear in the CREATE
|
|
|
|
TABLE statement (or when you execute a .* query).
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
You must typecast the expressions (using ::) very carefully
|
|
|
|
or you will see the following error:
|
|
|
|
|
|
|
|
<programlisting>
|
|
|
|
<computeroutput>
|
|
|
|
WARN::function declared to return type EMP does not retrieve (EMP.*)
|
|
|
|
</computeroutput>
|
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
When calling a function that returns an instance, we
|
1998-03-01 09:16:16 +01:00
|
|
|
cannot retrieve the entire instance. We must either
|
|
|
|
project an attribute out of the instance or pass the
|
|
|
|
entire instance into another function.
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
SELECT name(new_emp()) AS nobody;
|
|
|
|
|
|
|
|
+-------+
|
|
|
|
|nobody |
|
|
|
|
+-------+
|
|
|
|
|None |
|
|
|
|
+-------+
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
The reason why, in general, we must use the function
|
1998-03-01 09:16:16 +01:00
|
|
|
syntax for projecting attributes of function return
|
|
|
|
values is that the parser just doesn't understand
|
|
|
|
the other (dot) syntax for projection when combined
|
|
|
|
with function calls.
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
SELECT new_emp().name AS nobody;
|
|
|
|
WARN:parser: syntax error at or near "."
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
</itemizedlist>
|
|
|
|
</para>
|
|
|
|
<para>
|
|
|
|
Any collection of commands in the <acronym>SQL</acronym> query
|
1999-03-14 16:24:15 +01:00
|
|
|
language can be packaged together and defined as a function.
|
1999-10-04 17:18:54 +02:00
|
|
|
The commands can include updates (i.e.,
|
|
|
|
<command>INSERT</command>, <command>UPDATE</command>, and
|
|
|
|
<command>DELETE</command>) as well
|
|
|
|
as <command>SELECT</command> queries. However, the final command
|
|
|
|
must be a <command>SELECT</command> that returns whatever is
|
1998-03-01 09:16:16 +01:00
|
|
|
specified as the function's returntype.
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
CREATE FUNCTION clean_EMP () RETURNS int4
|
|
|
|
AS 'DELETE FROM EMP WHERE EMP.salary <= 0;
|
|
|
|
SELECT 1 AS ignore_this'
|
|
|
|
LANGUAGE 'sql';
|
|
|
|
|
|
|
|
SELECT clean_EMP();
|
|
|
|
|
|
|
|
+--+
|
|
|
|
|x |
|
|
|
|
+--+
|
|
|
|
|1 |
|
|
|
|
+--+
|
1999-03-14 16:24:15 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
</sect2>
|
|
|
|
</sect1>
|
|
|
|
|
|
|
|
<sect1>
|
|
|
|
<title>Procedural Language Functions</title>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
Procedural languages aren't built into Postgres. They are offered
|
|
|
|
by loadable modules. Please refer to the documentation for the
|
|
|
|
PL in question for details about the syntax and how the AS
|
|
|
|
clause is interpreted by the PL handler.
|
|
|
|
</para>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
|
|
|
There are two procedural languages available with the standard
|
|
|
|
<productname>Postgres</productname> distribution (PLTCL and PLSQL), and other
|
|
|
|
languages can be defined.
|
|
|
|
Refer to <xref linkend="xplang-title" endterm="xplang-title"> for
|
|
|
|
more information.
|
|
|
|
</para>
|
|
|
|
</sect1>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<sect1>
|
|
|
|
<title>Internal Functions</title>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
|
|
|
Internal functions are functions written in C which have been statically
|
|
|
|
linked into the <productname>Postgres</productname> backend
|
|
|
|
process. The AS
|
|
|
|
clause gives the C-language name of the function, which need not be the
|
|
|
|
same as the name being declared for SQL use.
|
|
|
|
(For reasons of backwards compatibility, an empty AS
|
|
|
|
string is accepted as meaning that the C-language function name is the
|
|
|
|
same as the SQL name.) Normally, all internal functions present in the
|
|
|
|
backend are declared as SQL functions during database initialization,
|
|
|
|
but a user could use <command>CREATE FUNCTION</command>
|
|
|
|
to create additional alias names for an internal function.
|
|
|
|
</para>
|
|
|
|
</sect1>
|
|
|
|
|
|
|
|
<sect1>
|
|
|
|
<title>Compiled (C) Language Functions</title>
|
|
|
|
|
|
|
|
<para>
|
I have been working with user defined types and user defined c
functions. One problem that I have encountered with the function
manager is that it does not allow the user to define type conversion
functions that convert between user types. For instance if mytype1,
mytype2, and mytype3 are three Postgresql user types, and if I wish to
define Postgresql conversion functions like
I run into problems, because the Postgresql dynamic loader would look
for a single link symbol, mytype3, for both pieces of object code. If
I just change the name of one of the Postgresql functions (to make the
symbols distinct), the automatic type conversion that Postgresql uses,
for example, when matching operators to arguments no longer finds the
type conversion function.
The solution that I propose, and have implemented in the attatched
patch extends the CREATE FUNCTION syntax as follows. In the first case
above I use the link symbol mytype2_to_mytype3 for the link object
that implements the first conversion function, and define the
Postgresql operator with the following syntax
The patch includes changes to the parser to include the altered
syntax, changes to the ProcedureStmt node in nodes/parsenodes.h,
changes to commands/define.c to handle the extra information in the AS
clause, and changes to utils/fmgr/dfmgr.c that alter the way that the
dynamic loader figures out what link symbol to use. I store the
string for the link symbol in the prosrc text attribute of the pg_proc
table which is currently unused in rows that reference dynamically
loaded
functions.
Bernie Frankpitt
1999-09-28 06:34:56 +02:00
|
|
|
Functions written in C can be compiled into dynamically loadable
|
|
|
|
objects, and used to implement user-defined SQL functions. The
|
|
|
|
first time the user defined function is called inside the backend,
|
|
|
|
the dynamic loader loads the function's object code into memory,
|
|
|
|
and links the function with the running
|
|
|
|
<productname>Postgres</productname> executable. The SQL syntax
|
1999-10-01 17:20:06 +02:00
|
|
|
for <command>CREATE FUNCTION</command>
|
|
|
|
links the SQL function
|
I have been working with user defined types and user defined c
functions. One problem that I have encountered with the function
manager is that it does not allow the user to define type conversion
functions that convert between user types. For instance if mytype1,
mytype2, and mytype3 are three Postgresql user types, and if I wish to
define Postgresql conversion functions like
I run into problems, because the Postgresql dynamic loader would look
for a single link symbol, mytype3, for both pieces of object code. If
I just change the name of one of the Postgresql functions (to make the
symbols distinct), the automatic type conversion that Postgresql uses,
for example, when matching operators to arguments no longer finds the
type conversion function.
The solution that I propose, and have implemented in the attatched
patch extends the CREATE FUNCTION syntax as follows. In the first case
above I use the link symbol mytype2_to_mytype3 for the link object
that implements the first conversion function, and define the
Postgresql operator with the following syntax
The patch includes changes to the parser to include the altered
syntax, changes to the ProcedureStmt node in nodes/parsenodes.h,
changes to commands/define.c to handle the extra information in the AS
clause, and changes to utils/fmgr/dfmgr.c that alter the way that the
dynamic loader figures out what link symbol to use. I store the
string for the link symbol in the prosrc text attribute of the pg_proc
table which is currently unused in rows that reference dynamically
loaded
functions.
Bernie Frankpitt
1999-09-28 06:34:56 +02:00
|
|
|
to the C source function in one of two ways. If the SQL function
|
|
|
|
has the same name as the C source function the first form of the
|
|
|
|
statement is used. The string argument in the AS clause is the
|
|
|
|
full pathname of the file that contains the dynamically loadable
|
1999-10-02 23:27:49 +02:00
|
|
|
compiled object. If the name of the C function is different from the
|
|
|
|
desired name of the SQL function, then the second form is used. In this
|
I have been working with user defined types and user defined c
functions. One problem that I have encountered with the function
manager is that it does not allow the user to define type conversion
functions that convert between user types. For instance if mytype1,
mytype2, and mytype3 are three Postgresql user types, and if I wish to
define Postgresql conversion functions like
I run into problems, because the Postgresql dynamic loader would look
for a single link symbol, mytype3, for both pieces of object code. If
I just change the name of one of the Postgresql functions (to make the
symbols distinct), the automatic type conversion that Postgresql uses,
for example, when matching operators to arguments no longer finds the
type conversion function.
The solution that I propose, and have implemented in the attatched
patch extends the CREATE FUNCTION syntax as follows. In the first case
above I use the link symbol mytype2_to_mytype3 for the link object
that implements the first conversion function, and define the
Postgresql operator with the following syntax
The patch includes changes to the parser to include the altered
syntax, changes to the ProcedureStmt node in nodes/parsenodes.h,
changes to commands/define.c to handle the extra information in the AS
clause, and changes to utils/fmgr/dfmgr.c that alter the way that the
dynamic loader figures out what link symbol to use. I store the
string for the link symbol in the prosrc text attribute of the pg_proc
table which is currently unused in rows that reference dynamically
loaded
functions.
Bernie Frankpitt
1999-09-28 06:34:56 +02:00
|
|
|
form the AS clause takes two string arguments, the first is the
|
|
|
|
full pathname of the dynamically loadable object file, and the
|
|
|
|
second is the link symbol that the dynamic loader should search
|
|
|
|
for. This link symbol is just the function name in the C source
|
|
|
|
code.
|
|
|
|
|
1999-10-01 17:20:06 +02:00
|
|
|
<note>
|
|
|
|
<para>
|
|
|
|
After it is used for the first time, a dynamically loaded, user
|
|
|
|
function is retained in memory, and future calls to the function
|
|
|
|
only incur the small overhead of a symbol table lookup.
|
|
|
|
</para>
|
|
|
|
</note>
|
1999-07-22 17:11:05 +02:00
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
I have been working with user defined types and user defined c
functions. One problem that I have encountered with the function
manager is that it does not allow the user to define type conversion
functions that convert between user types. For instance if mytype1,
mytype2, and mytype3 are three Postgresql user types, and if I wish to
define Postgresql conversion functions like
I run into problems, because the Postgresql dynamic loader would look
for a single link symbol, mytype3, for both pieces of object code. If
I just change the name of one of the Postgresql functions (to make the
symbols distinct), the automatic type conversion that Postgresql uses,
for example, when matching operators to arguments no longer finds the
type conversion function.
The solution that I propose, and have implemented in the attatched
patch extends the CREATE FUNCTION syntax as follows. In the first case
above I use the link symbol mytype2_to_mytype3 for the link object
that implements the first conversion function, and define the
Postgresql operator with the following syntax
The patch includes changes to the parser to include the altered
syntax, changes to the ProcedureStmt node in nodes/parsenodes.h,
changes to commands/define.c to handle the extra information in the AS
clause, and changes to utils/fmgr/dfmgr.c that alter the way that the
dynamic loader figures out what link symbol to use. I store the
string for the link symbol in the prosrc text attribute of the pg_proc
table which is currently unused in rows that reference dynamically
loaded
functions.
Bernie Frankpitt
1999-09-28 06:34:56 +02:00
|
|
|
The string which specifies the object file (the string in the AS
|
|
|
|
clause) should be the <emphasis>full path</emphasis> of the object
|
|
|
|
code file for the function, bracketed by quotation marks. If a
|
|
|
|
link symbol is used in the AS clause, the link symbol should also be
|
|
|
|
bracketed by single quotation marks, and should be exactly the
|
1999-10-04 17:18:54 +02:00
|
|
|
same as the name of the function in the C source code. On Unix systems
|
I have been working with user defined types and user defined c
functions. One problem that I have encountered with the function
manager is that it does not allow the user to define type conversion
functions that convert between user types. For instance if mytype1,
mytype2, and mytype3 are three Postgresql user types, and if I wish to
define Postgresql conversion functions like
I run into problems, because the Postgresql dynamic loader would look
for a single link symbol, mytype3, for both pieces of object code. If
I just change the name of one of the Postgresql functions (to make the
symbols distinct), the automatic type conversion that Postgresql uses,
for example, when matching operators to arguments no longer finds the
type conversion function.
The solution that I propose, and have implemented in the attatched
patch extends the CREATE FUNCTION syntax as follows. In the first case
above I use the link symbol mytype2_to_mytype3 for the link object
that implements the first conversion function, and define the
Postgresql operator with the following syntax
The patch includes changes to the parser to include the altered
syntax, changes to the ProcedureStmt node in nodes/parsenodes.h,
changes to commands/define.c to handle the extra information in the AS
clause, and changes to utils/fmgr/dfmgr.c that alter the way that the
dynamic loader figures out what link symbol to use. I store the
string for the link symbol in the prosrc text attribute of the pg_proc
table which is currently unused in rows that reference dynamically
loaded
functions.
Bernie Frankpitt
1999-09-28 06:34:56 +02:00
|
|
|
the command <command>nm</command> will print all of the link
|
|
|
|
symbols in a dynamically loadable object.
|
|
|
|
(<productname>Postgres</productname> will not compile a function
|
|
|
|
automatically; it must be compiled before it is used in a CREATE
|
|
|
|
FUNCTION command. See below for additional information.)
|
1999-07-22 17:11:05 +02:00
|
|
|
</para>
|
|
|
|
|
|
|
|
<sect2>
|
|
|
|
<title>C Language Functions on Base Types</title>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
The following table gives the C type required for parameters in the C
|
|
|
|
functions that will be loaded into Postgres. The "Defined In"
|
|
|
|
column gives the actual header file (in the
|
|
|
|
<filename>.../src/backend/</filename>
|
|
|
|
directory) that the equivalent C type is defined. However, if you
|
|
|
|
include <filename>utils/builtins.h</filename>,
|
|
|
|
these files will automatically be
|
|
|
|
included.
|
|
|
|
|
|
|
|
<table tocentry="1">
|
|
|
|
<title>Equivalent C Types
|
|
|
|
for Built-In <productname>Postgres</productname> Types</title>
|
|
|
|
<titleabbrev>Equivalent C Types</titleabbrev>
|
|
|
|
<tgroup cols="3">
|
|
|
|
<thead>
|
|
|
|
<row>
|
|
|
|
<entry>
|
|
|
|
Built-In Type
|
|
|
|
</entry>
|
|
|
|
<entry>
|
|
|
|
C Type
|
|
|
|
</entry>
|
|
|
|
<entry>
|
|
|
|
Defined In
|
|
|
|
</entry>
|
|
|
|
</row>
|
|
|
|
</thead>
|
|
|
|
<tbody>
|
|
|
|
<row>
|
|
|
|
<entry>abstime</entry>
|
|
|
|
<entry>AbsoluteTime</entry>
|
|
|
|
<entry>utils/nabstime.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>bool</entry>
|
|
|
|
<entry>bool</entry>
|
|
|
|
<entry>include/c.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>box</entry>
|
|
|
|
<entry>(BOX *)</entry>
|
|
|
|
<entry>utils/geo-decls.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>bytea</entry>
|
|
|
|
<entry>(bytea *)</entry>
|
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>char</entry>
|
|
|
|
<entry>char</entry>
|
|
|
|
<entry>N/A</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>cid</entry>
|
|
|
|
<entry>CID</entry>
|
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>datetime</entry>
|
|
|
|
<entry>(DateTime *)</entry>
|
|
|
|
<entry>include/c.h or include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>int2</entry>
|
|
|
|
<entry>int2</entry>
|
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
2000-01-11 02:40:04 +01:00
|
|
|
<entry>int2vector</entry>
|
|
|
|
<entry>(int2vector *)</entry>
|
1999-07-22 17:11:05 +02:00
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>int4</entry>
|
|
|
|
<entry>int4</entry>
|
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>float4</entry>
|
|
|
|
<entry>float32 or (float4 *)</entry>
|
|
|
|
<entry>include/c.h or include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>float8</entry>
|
|
|
|
<entry>float64 or (float8 *)</entry>
|
|
|
|
<entry>include/c.h or include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>lseg</entry>
|
|
|
|
<entry>(LSEG *)</entry>
|
|
|
|
<entry>include/geo-decls.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>name</entry>
|
|
|
|
<entry>(Name)</entry>
|
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>oid</entry>
|
|
|
|
<entry>oid</entry>
|
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
2000-01-11 02:40:04 +01:00
|
|
|
<entry>oidvector</entry>
|
|
|
|
<entry>(oidvector *)</entry>
|
1999-07-22 17:11:05 +02:00
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>path</entry>
|
|
|
|
<entry>(PATH *)</entry>
|
|
|
|
<entry>utils/geo-decls.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>point</entry>
|
|
|
|
<entry>(POINT *)</entry>
|
|
|
|
<entry>utils/geo-decls.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>regproc</entry>
|
|
|
|
<entry>regproc or REGPROC</entry>
|
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>reltime</entry>
|
|
|
|
<entry>RelativeTime</entry>
|
|
|
|
<entry>utils/nabstime.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>text</entry>
|
|
|
|
<entry>(text *)</entry>
|
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>tid</entry>
|
|
|
|
<entry>ItemPointer</entry>
|
|
|
|
<entry>storage/itemptr.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>timespan</entry>
|
|
|
|
<entry>(TimeSpan *)</entry>
|
|
|
|
<entry>include/c.h or include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>tinterval</entry>
|
|
|
|
<entry>TimeInterval</entry>
|
|
|
|
<entry>utils/nabstime.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>uint2</entry>
|
|
|
|
<entry>uint16</entry>
|
|
|
|
<entry>include/c.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>uint4</entry>
|
|
|
|
<entry>uint32</entry>
|
|
|
|
<entry>include/c.h</entry>
|
|
|
|
</row>
|
|
|
|
<row>
|
|
|
|
<entry>xid</entry>
|
|
|
|
<entry>(XID *)</entry>
|
|
|
|
<entry>include/postgres.h</entry>
|
|
|
|
</row>
|
|
|
|
</tbody>
|
|
|
|
</tgroup>
|
|
|
|
</table>
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
Internally, <productname>Postgres</productname> regards a
|
1999-03-14 16:24:15 +01:00
|
|
|
base type as a "blob of memory." The user-defined
|
|
|
|
functions that you define over a type in turn define the
|
1999-07-22 17:11:05 +02:00
|
|
|
way that <productname>Postgres</productname> can operate
|
|
|
|
on it. That is, <productname>Postgres</productname> will
|
1999-03-14 16:24:15 +01:00
|
|
|
only store and retrieve the data from disk and use your
|
|
|
|
user-defined functions to input, process, and output the data.
|
1998-03-01 09:16:16 +01:00
|
|
|
Base types can have one of three internal formats:
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<itemizedlist>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
pass by value, fixed-length
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
pass by reference, fixed-length
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
pass by reference, variable-length
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
</itemizedlist>
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
By-value types can only be 1, 2 or 4 bytes in length
|
|
|
|
(even if your computer supports by-value types of other
|
1999-07-22 17:11:05 +02:00
|
|
|
sizes). <productname>Postgres</productname> itself
|
1999-03-14 16:24:15 +01:00
|
|
|
only passes integer types by value. You should be careful
|
|
|
|
to define your types such that they will be the same
|
|
|
|
size (in bytes) on all architectures. For example, the
|
1999-10-04 17:18:54 +02:00
|
|
|
<literal>long</literal> type is dangerous because it
|
1999-03-14 16:24:15 +01:00
|
|
|
is 4 bytes on some machines and 8 bytes on others, whereas
|
1999-10-04 17:18:54 +02:00
|
|
|
<literal>int</literal> type is 4 bytes on most
|
|
|
|
Unix machines (though not on most
|
1999-03-14 16:24:15 +01:00
|
|
|
personal computers). A reasonable implementation of
|
1999-10-04 17:18:54 +02:00
|
|
|
the <literal>int4</literal> type on Unix
|
1998-03-01 09:16:16 +01:00
|
|
|
machines might be:
|
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<programlisting>
|
|
|
|
/* 4-byte integer, passed by value */
|
|
|
|
typedef int int4;
|
|
|
|
</programlisting>
|
|
|
|
</para>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
On the other hand, fixed-length types of any size may
|
|
|
|
be passed by-reference. For example, here is a sample
|
1999-07-22 17:11:05 +02:00
|
|
|
implementation of a <productname>Postgres</productname> type:
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<programlisting>
|
|
|
|
/* 16-byte structure, passed by reference */
|
|
|
|
typedef struct
|
|
|
|
{
|
|
|
|
double x, y;
|
|
|
|
} Point;
|
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
Only pointers to such types can be used when passing
|
1999-07-22 17:11:05 +02:00
|
|
|
them in and out of <productname>Postgres</productname> functions.
|
1998-03-01 09:16:16 +01:00
|
|
|
Finally, all variable-length types must also be passed
|
|
|
|
by reference. All variable-length types must begin
|
|
|
|
with a length field of exactly 4 bytes, and all data to
|
|
|
|
be stored within that type must be located in the memory
|
|
|
|
immediately following that length field. The
|
|
|
|
length field is the total length of the structure
|
|
|
|
(i.e., it includes the size of the length field
|
|
|
|
itself). We can define the text type as follows:
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<programlisting>
|
|
|
|
typedef struct {
|
|
|
|
int4 length;
|
|
|
|
char data[1];
|
|
|
|
} text;
|
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
Obviously, the data field is not long enough to hold
|
1999-07-22 17:11:05 +02:00
|
|
|
all possible strings; it's impossible to declare such
|
|
|
|
a structure in <acronym>C</acronym>. When manipulating
|
1999-03-14 16:24:15 +01:00
|
|
|
variable-length types, we must be careful to allocate
|
|
|
|
the correct amount of memory and initialize the length field.
|
|
|
|
For example, if we wanted to store 40 bytes in a text
|
1998-03-01 09:16:16 +01:00
|
|
|
structure, we might use a code fragment like this:
|
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<programlisting>
|
|
|
|
#include "postgres.h"
|
|
|
|
...
|
|
|
|
char buffer[40]; /* our source data */
|
|
|
|
...
|
|
|
|
text *destination = (text *) palloc(VARHDRSZ + 40);
|
|
|
|
destination->length = VARHDRSZ + 40;
|
|
|
|
memmove(destination->data, buffer, 40);
|
|
|
|
...
|
|
|
|
</programlisting>
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
Now that we've gone over all of the possible structures
|
|
|
|
for base types, we can show some examples of real functions.
|
1999-07-22 17:11:05 +02:00
|
|
|
Suppose <filename>funcs.c</filename> look like:
|
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
#include <string.h>
|
1998-04-26 06:18:06 +02:00
|
|
|
#include "postgres.h"
|
1999-03-14 16:24:15 +01:00
|
|
|
|
|
|
|
/* By Value */
|
|
|
|
|
1998-03-01 09:16:16 +01:00
|
|
|
int
|
|
|
|
add_one(int arg)
|
|
|
|
{
|
|
|
|
return(arg + 1);
|
|
|
|
}
|
1999-03-14 16:24:15 +01:00
|
|
|
|
|
|
|
/* By Reference, Fixed Length */
|
|
|
|
|
|
|
|
Point *
|
|
|
|
makepoint(Point *pointx, Point *pointy )
|
1998-03-01 09:16:16 +01:00
|
|
|
{
|
1999-03-14 16:24:15 +01:00
|
|
|
Point *new_point = (Point *) palloc(sizeof(Point));
|
|
|
|
|
|
|
|
new_point->x = pointx->x;
|
|
|
|
new_point->y = pointy->y;
|
|
|
|
|
|
|
|
return new_point;
|
1998-03-01 09:16:16 +01:00
|
|
|
}
|
1999-03-14 16:24:15 +01:00
|
|
|
|
|
|
|
/* By Reference, Variable Length */
|
|
|
|
|
1998-03-01 09:16:16 +01:00
|
|
|
text *
|
|
|
|
copytext(text *t)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* VARSIZE is the total size of the struct in bytes.
|
|
|
|
*/
|
|
|
|
text *new_t = (text *) palloc(VARSIZE(t));
|
|
|
|
memset(new_t, 0, VARSIZE(t));
|
|
|
|
VARSIZE(new_t) = VARSIZE(t);
|
|
|
|
/*
|
|
|
|
* VARDATA is a pointer to the data region of the struct.
|
|
|
|
*/
|
|
|
|
memcpy((void *) VARDATA(new_t), /* destination */
|
|
|
|
(void *) VARDATA(t), /* source */
|
|
|
|
VARSIZE(t)-VARHDRSZ); /* how many bytes */
|
|
|
|
return(new_t);
|
|
|
|
}
|
1999-03-14 16:24:15 +01:00
|
|
|
|
|
|
|
text *
|
|
|
|
concat_text(text *arg1, text *arg2)
|
|
|
|
{
|
|
|
|
int32 new_text_size = VARSIZE(arg1) + VARSIZE(arg2) - VARHDRSZ;
|
|
|
|
text *new_text = (text *) palloc(new_text_size);
|
|
|
|
|
|
|
|
memset((void *) new_text, 0, new_text_size);
|
|
|
|
VARSIZE(new_text) = new_text_size;
|
|
|
|
strncpy(VARDATA(new_text), VARDATA(arg1), VARSIZE(arg1)-VARHDRSZ);
|
|
|
|
strncat(VARDATA(new_text), VARDATA(arg2), VARSIZE(arg2)-VARHDRSZ);
|
|
|
|
return (new_text);
|
|
|
|
}
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
|
|
|
On <acronym>OSF/1</acronym> we would type:
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
CREATE FUNCTION add_one(int4) RETURNS int4
|
1999-07-22 17:11:05 +02:00
|
|
|
AS '<replaceable>PGROOT</replaceable>/tutorial/funcs.so' LANGUAGE 'c';
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-03-14 16:24:15 +01:00
|
|
|
CREATE FUNCTION makepoint(point, point) RETURNS point
|
1999-07-22 17:11:05 +02:00
|
|
|
AS '<replaceable>PGROOT</replaceable>/tutorial/funcs.so' LANGUAGE 'c';
|
1999-03-14 16:24:15 +01:00
|
|
|
|
1998-04-26 06:18:06 +02:00
|
|
|
CREATE FUNCTION concat_text(text, text) RETURNS text
|
1999-07-22 17:11:05 +02:00
|
|
|
AS '<replaceable>PGROOT</replaceable>/tutorial/funcs.so' LANGUAGE 'c';
|
1999-03-14 16:24:15 +01:00
|
|
|
|
1998-03-01 09:16:16 +01:00
|
|
|
CREATE FUNCTION copytext(text) RETURNS text
|
1999-07-22 17:11:05 +02:00
|
|
|
AS '<replaceable>PGROOT</replaceable>/tutorial/funcs.so' LANGUAGE 'c';
|
|
|
|
</programlisting>
|
|
|
|
</para>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
On other systems, we might have to make the filename
|
|
|
|
end in .sl (to indicate that it's a shared library).
|
1999-07-22 17:11:05 +02:00
|
|
|
</para>
|
|
|
|
</sect2>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<sect2>
|
|
|
|
<title>C Language Functions on Composite Types</title>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
Composite types do not have a fixed layout like C
|
|
|
|
structures. Instances of a composite type may contain
|
|
|
|
null fields. In addition, composite types that are
|
|
|
|
part of an inheritance hierarchy may have different
|
|
|
|
fields than other members of the same inheritance hierarchy.
|
1999-07-22 17:11:05 +02:00
|
|
|
Therefore, <productname>Postgres</productname> provides
|
1999-03-14 16:24:15 +01:00
|
|
|
a procedural interface for accessing fields of composite types
|
1999-07-22 17:11:05 +02:00
|
|
|
from C. As <productname>Postgres</productname> processes
|
1999-03-14 16:24:15 +01:00
|
|
|
a set of instances, each instance will be passed into your
|
1999-10-04 17:18:54 +02:00
|
|
|
function as an opaque structure of type <literal>TUPLE</literal>.
|
1998-03-01 09:16:16 +01:00
|
|
|
Suppose we want to write a function to answer the query
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
* SELECT name, c_overpaid(EMP, 1500) AS overpaid
|
|
|
|
FROM EMP
|
|
|
|
WHERE name = 'Bill' or name = 'Sam';
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
|
1998-03-01 09:16:16 +01:00
|
|
|
In the query above, we can define c_overpaid as:
|
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<programlisting>
|
1998-04-26 06:18:06 +02:00
|
|
|
#include "postgres.h"
|
1999-03-14 16:24:15 +01:00
|
|
|
#include "executor/executor.h" /* for GetAttributeByName() */
|
|
|
|
|
1998-03-01 09:16:16 +01:00
|
|
|
bool
|
1999-03-14 16:24:15 +01:00
|
|
|
c_overpaid(TupleTableSlot *t, /* the current instance of EMP */
|
1998-03-01 09:16:16 +01:00
|
|
|
int4 limit)
|
|
|
|
{
|
|
|
|
bool isnull = false;
|
|
|
|
int4 salary;
|
|
|
|
salary = (int4) GetAttributeByName(t, "salary", &isnull);
|
|
|
|
if (isnull)
|
|
|
|
return (false);
|
|
|
|
return(salary > limit);
|
|
|
|
}
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
1999-10-04 17:18:54 +02:00
|
|
|
<function>GetAttributeByName</function> is the
|
1999-07-22 17:11:05 +02:00
|
|
|
<productname>Postgres</productname> system function that
|
1998-03-01 09:16:16 +01:00
|
|
|
returns attributes out of the current instance. It has
|
|
|
|
three arguments: the argument of type TUPLE passed into
|
|
|
|
the function, the name of the desired attribute, and a
|
|
|
|
return parameter that describes whether the attribute
|
1999-10-04 17:18:54 +02:00
|
|
|
is null. <function>GetAttributeByName</function> will
|
1999-03-14 16:24:15 +01:00
|
|
|
align data properly so you can cast its return value to
|
|
|
|
the desired type. For example, if you have an attribute
|
1999-10-04 17:18:54 +02:00
|
|
|
name which is of the type name, the <function>GetAttributeByName</function>
|
1999-03-14 16:24:15 +01:00
|
|
|
call would look like:
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
char *str;
|
|
|
|
...
|
|
|
|
str = (char *) GetAttributeByName(t, "name", &isnull)
|
1999-07-22 17:11:05 +02:00
|
|
|
</programlisting>
|
|
|
|
</para>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
|
|
|
The following query lets <productname>Postgres</productname>
|
1999-03-14 16:24:15 +01:00
|
|
|
know about the c_overpaid function:
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<programlisting>
|
1998-03-01 09:16:16 +01:00
|
|
|
* CREATE FUNCTION c_overpaid(EMP, int4) RETURNS bool
|
1999-07-22 17:11:05 +02:00
|
|
|
AS '<replaceable>PGROOT</replaceable>/tutorial/obj/funcs.so' LANGUAGE 'c';
|
|
|
|
</programlisting>
|
|
|
|
</para>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
While there are ways to construct new instances or modify
|
|
|
|
existing instances from within a C function, these
|
|
|
|
are far too complex to discuss in this manual.
|
1999-07-22 17:11:05 +02:00
|
|
|
</para>
|
|
|
|
</sect2>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<sect2>
|
|
|
|
<title>Writing Code</title>
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<para>
|
1998-03-01 09:16:16 +01:00
|
|
|
We now turn to the more difficult task of writing
|
|
|
|
programming language functions. Be warned: this section
|
|
|
|
of the manual will not make you a programmer. You must
|
1999-07-22 17:11:05 +02:00
|
|
|
have a good understanding of <acronym>C</acronym>
|
1999-03-14 16:24:15 +01:00
|
|
|
(including the use of pointers and the malloc memory manager)
|
1999-07-22 17:11:05 +02:00
|
|
|
before trying to write <acronym>C</acronym> functions for
|
|
|
|
use with <productname>Postgres</productname>. While it may
|
1999-03-14 16:24:15 +01:00
|
|
|
be possible to load functions written in languages other
|
1999-07-22 17:11:05 +02:00
|
|
|
than <acronym>C</acronym> into <productname>Postgres</productname>,
|
1999-03-14 16:24:15 +01:00
|
|
|
this is often difficult (when it is possible at all)
|
1999-07-22 17:11:05 +02:00
|
|
|
because other languages, such as <acronym>FORTRAN</acronym>
|
|
|
|
and <acronym>Pascal</acronym> often do not follow the same
|
|
|
|
<firstterm>calling convention</firstterm>
|
|
|
|
as <acronym>C</acronym>. That is, other
|
1998-03-01 09:16:16 +01:00
|
|
|
languages do not pass argument and return values
|
|
|
|
between functions in the same way. For this reason, we
|
|
|
|
will assume that your programming language functions
|
1999-07-22 17:11:05 +02:00
|
|
|
are written in <acronym>C</acronym>.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
C functions with base type arguments can be written in a
|
|
|
|
straightforward fashion. The C equivalents of built-in Postgres types
|
|
|
|
are accessible in a C file if
|
|
|
|
<filename><replaceable>PGROOT</replaceable>/src/backend/utils/builtins.h</filename>
|
|
|
|
is included as a header file. This can be achieved by having
|
|
|
|
|
|
|
|
<programlisting>
|
|
|
|
#include <utils/builtins.h>
|
|
|
|
</programlisting>
|
|
|
|
|
|
|
|
at the top of the C source file.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
The basic rules for building <acronym>C</acronym> functions
|
1999-03-14 16:24:15 +01:00
|
|
|
are as follows:
|
1998-03-01 09:16:16 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<itemizedlist>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
Most of the header (include) files for
|
|
|
|
<productname>Postgres</productname>
|
|
|
|
should already be installed in
|
|
|
|
<filename><replaceable>PGROOT</replaceable>/include</filename> (see Figure 2).
|
|
|
|
You should always include
|
|
|
|
|
|
|
|
<programlisting>
|
|
|
|
-I$PGROOT/include
|
|
|
|
</programlisting>
|
|
|
|
|
|
|
|
on your cc command lines. Sometimes, you may
|
|
|
|
find that you require header files that are in
|
|
|
|
the server source itself (i.e., you need a file
|
|
|
|
we neglected to install in include). In those
|
|
|
|
cases you may need to add one or more of
|
|
|
|
|
|
|
|
<programlisting>
|
|
|
|
-I$PGROOT/src/backend
|
|
|
|
-I$PGROOT/src/backend/include
|
|
|
|
-I$PGROOT/src/backend/port/<PORTNAME>
|
|
|
|
-I$PGROOT/src/backend/obj
|
|
|
|
</programlisting>
|
|
|
|
|
|
|
|
(where <PORTNAME> is the name of the port, e.g.,
|
|
|
|
alpha or sparc).
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
When allocating memory, use the
|
|
|
|
<productname>Postgres</productname>
|
|
|
|
routines palloc and pfree instead of the
|
|
|
|
corresponding <acronym>C</acronym> library routines
|
|
|
|
malloc and free.
|
|
|
|
The memory allocated by palloc will be freed
|
|
|
|
automatically at the end of each transaction,
|
|
|
|
preventing memory leaks.
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
Always zero the bytes of your structures using
|
|
|
|
memset or bzero. Several routines (such as the
|
|
|
|
hash access method, hash join and the sort algorithm)
|
|
|
|
compute functions of the raw bits contained in
|
|
|
|
your structure. Even if you initialize all fields
|
|
|
|
of your structure, there may be
|
|
|
|
several bytes of alignment padding (holes in the
|
|
|
|
structure) that may contain garbage values.
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
Most of the internal <productname>Postgres</productname>
|
|
|
|
types are declared in <filename>postgres.h</filename>,
|
|
|
|
so it's a good
|
|
|
|
idea to always include that file as well. Including
|
|
|
|
postgres.h will also include elog.h and palloc.h for you.
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
<listitem>
|
|
|
|
<para>
|
|
|
|
Compiling and loading your object code so that
|
|
|
|
it can be dynamically loaded into
|
|
|
|
<productname>Postgres</productname>
|
|
|
|
always requires special flags.
|
|
|
|
See <xref linkend="dfunc-title" endterm="dfunc-title">
|
|
|
|
for a detailed explanation of how to do it for
|
|
|
|
your particular operating system.
|
|
|
|
</para>
|
|
|
|
</listitem>
|
|
|
|
</itemizedlist>
|
|
|
|
</para>
|
|
|
|
</sect2>
|
|
|
|
</sect1>
|
|
|
|
|
|
|
|
<sect1>
|
|
|
|
<title>Function Overloading</title>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
More than one function may be defined with the same name, as long as
|
|
|
|
the arguments they take are different. In other words, function names
|
|
|
|
can be <firstterm>overloaded</firstterm>.
|
|
|
|
A function may also have the same name as an attribute. In the case
|
|
|
|
that there is an ambiguity between a function on a complex type and
|
|
|
|
an attribute of the complex type, the attribute will always be used.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<sect2>
|
|
|
|
<title>Name Space Conflicts</title>
|
|
|
|
|
|
|
|
<para>
|
2000-03-30 04:51:12 +02:00
|
|
|
As of <productname>Postgres</productname> v7.0, the alternative
|
1999-10-01 17:20:06 +02:00
|
|
|
form of the AS clause for the SQL
|
|
|
|
<command>CREATE FUNCTION</command> command
|
|
|
|
decouples the SQL function name from the function name in the C
|
|
|
|
source code. This is now the preferred technique to accomplish
|
|
|
|
function overloading.
|
1999-07-22 17:11:05 +02:00
|
|
|
</para>
|
|
|
|
|
|
|
|
<sect3>
|
2000-03-30 04:51:12 +02:00
|
|
|
<title>Pre-v7.0</title>
|
1999-07-22 17:11:05 +02:00
|
|
|
|
|
|
|
<para>
|
|
|
|
For functions written in C, the SQL name declared in
|
|
|
|
<command>CREATE FUNCTION</command>
|
|
|
|
must be exactly the same as the actual name of the function in the
|
|
|
|
C code (hence it must be a legal C function name).
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
There is a subtle implication of this restriction: while the
|
|
|
|
dynamic loading routines in most operating systems are more than
|
|
|
|
happy to allow you to load any number of shared libraries that
|
|
|
|
contain conflicting (identically-named) function names, they may
|
|
|
|
in fact botch the load in interesting ways. For example, if you
|
|
|
|
define a dynamically-loaded function that happens to have the
|
|
|
|
same name as a function built into Postgres, the DEC OSF/1 dynamic
|
|
|
|
loader causes Postgres to call the function within itself rather than
|
|
|
|
allowing Postgres to call your function. Hence, if you want your
|
|
|
|
function to be used on different architectures, we recommend that
|
|
|
|
you do not overload C function names.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
There is a clever trick to get around the problem just described.
|
|
|
|
Since there is no problem overloading SQL functions, you can
|
|
|
|
define a set of C functions with different names and then define
|
|
|
|
a set of identically-named SQL function wrappers that take the
|
|
|
|
appropriate argument types and call the matching C function.
|
|
|
|
</para>
|
|
|
|
|
|
|
|
<para>
|
|
|
|
Another solution is not to use dynamic loading, but to link your
|
|
|
|
functions into the backend statically and declare them as INTERNAL
|
|
|
|
functions. Then, the functions must all have distinct C names but
|
|
|
|
they can be declared with the same SQL names (as long as their
|
|
|
|
argument types differ, of course). This way avoids the overhead of
|
|
|
|
an SQL wrapper function, at the cost of more effort to prepare a
|
1999-10-02 23:27:49 +02:00
|
|
|
custom backend executable. (This option is only available in version
|
|
|
|
6.5 and later, since prior versions required internal functions to
|
|
|
|
have the same name in SQL as in the C code.)
|
1999-07-22 17:11:05 +02:00
|
|
|
</para>
|
|
|
|
</sect3>
|
|
|
|
</sect2>
|
|
|
|
</sect1>
|
|
|
|
</chapter>
|
1998-12-29 03:24:47 +01:00
|
|
|
|
1999-07-22 17:11:05 +02:00
|
|
|
<!-- Keep this comment at the end of the file
|
|
|
|
Local variables:
|
2000-03-31 05:27:42 +02:00
|
|
|
mode:sgml
|
1999-07-22 17:11:05 +02:00
|
|
|
sgml-omittag:nil
|
|
|
|
sgml-shorttag:t
|
|
|
|
sgml-minimize-attributes:nil
|
|
|
|
sgml-always-quote-attributes:t
|
|
|
|
sgml-indent-step:1
|
|
|
|
sgml-indent-data:t
|
|
|
|
sgml-parent-document:nil
|
|
|
|
sgml-default-dtd-file:"./reference.ced"
|
|
|
|
sgml-exposed-tags:nil
|
2000-03-31 05:27:42 +02:00
|
|
|
sgml-local-catalogs:("/usr/lib/sgml/catalog")
|
1999-07-22 17:11:05 +02:00
|
|
|
sgml-local-ecat-files:nil
|
|
|
|
End:
|
|
|
|
-->
|