1 <!-- $Header: /cvsroot/pgsql/doc/src/sgml/plpython.sgml,v 1.20 2003/09/12 22:17:23 tgl Exp $ -->
3 <chapter id="plpython">
4 <title>PL/Python - Python Procedural Language</title>
6 <indexterm zone="plpython"><primary>PL/Python</></>
7 <indexterm zone="plpython"><primary>Python</></>
10 The <application>PL/Python</application> procedural language allows
11 <productname>PostgreSQL</productname> functions to be written in the
12 <ulink url="http://www.python.org">Python</ulink> language.
16 To install PL/Python in a particular database, use
17 <literal>createlang plpythonu <replaceable>dbname</></literal>.
22 As of <productname>PostgreSQL</productname> 7.4,
23 PL/Python is only available as an <quote>untrusted</> language
24 (meaning it does not offer any way of restricting what users
25 can do in it). It has therefore been renamed to <literal>plpythonu</>.
26 The trusted variant <literal>plpython</> may become available again in
27 future, if a new secure execution mechanism is developed by the Python
34 If a language is installed into <literal>template1</>, all subsequently
35 created databases will have the language installed automatically.
41 Users of source packages must specially enable the build of
42 PL/Python during the installation process. (Refer to the
43 installation instructions for more information.) Users of binary
44 packages might find PL/Python in a separate subpackage.
48 <sect1 id="plpython-funcs">
49 <title>PL/Python Functions</title>
52 The Python code you write gets transformed into a Python function. E.g.,
54 CREATE FUNCTION myfunc(text) RETURNS text
62 def __plpython_procedure_myfunc_23456():
66 assuming that 23456 is the OID of the function.
70 If you do not provide a return value, Python returns the default
71 <symbol>None</symbol>. The language module translates Python's
72 <symbol>None</symbol> into the SQL null
73 value.<indexterm><primary>null value</><secondary
74 sortas="PL/Python">in PL/Python</></indexterm>
78 The <productname>PostgreSQL</> function parameters are available in
79 the global <varname>args</varname> list. In the
80 <function>myfunc</function> example, <varname>args[0]</> contains
81 whatever was passed in as the text argument. For
82 <literal>myfunc2(text, integer)</literal>, <varname>args[0]</>
83 would contain the <type>text</type> argument and
84 <varname>args[1]</varname> the <type>integer</type> argument.
88 The global dictionary <varname>SD</varname> is available to store
89 data between function calls. This variable is private static data.
90 The global dictionary <varname>GD</varname> is public data,
91 available to all Python functions within a session. Use with
92 care.<indexterm><primary>global data</><secondary>in
93 PL/Python</></indexterm>
97 Each function gets its own execution environment in the
98 Python interpreter, so that global data and function arguments from
99 <function>myfunc</function> are not available to
100 <function>myfunc2</function>. The exception is the data in the
101 <varname>GD</varname> dictionary, as mentioned above.
105 <sect1 id="plpython-trigger">
106 <title>Trigger Functions</title>
108 <indexterm zone="plpython-trigger">
109 <primary>trigger</primary>
110 <secondary>in PL/Python</secondary>
114 When a function is used in a trigger, the dictionary
115 <literal>TD</literal> contains trigger-related values. The trigger
116 rows are in <literal>TD["new"]</> and/or <literal>TD["old"]</>
117 depending on the trigger event. <literal>TD["event"]</> contains
118 the event as a string (<literal>INSERT</>, <literal>UPDATE</>,
119 <literal>DELETE</>, or <literal>UNKNOWN</>).
120 <literal>TD["when"]</> contains one of <literal>BEFORE</>,
121 <literal>AFTER</>, and <literal>UNKNOWN</>.
122 <literal>TD["level"]</> contains one of <literal>ROW</>,
123 <literal>STATEMENT</>, and <literal>UNKNOWN</>.
124 <literal>TD["name"]</> contains the trigger name, and
125 <literal>TD["relid"]</> contains the OID of the table on
126 which the trigger occurred. If the trigger was called with
127 arguments they are available in <literal>TD["args"][0]</> to
128 <literal>TD["args"][(n-1)]</>.
132 If <literal>TD["when"]</literal> is <literal>BEFORE</>, you may
133 return <literal>None</literal> or <literal>"OK"</literal> from the
134 Python function to indicate the row is unmodified,
135 <literal>"SKIP"</> to abort the event, or <literal>"MODIFY"</> to
136 indicate you've modified the row.
140 <sect1 id="plpython-database">
141 <title>Database Access</title>
144 The PL/Python language module automatically imports a Python module
145 called <literal>plpy</literal>. The functions and constants in
146 this module are available to you in the Python code as
147 <literal>plpy.<replaceable>foo</replaceable></literal>. At present
148 <literal>plpy</literal> implements the functions
149 <literal>plpy.debug("msg")</literal>,
150 <literal>plpy.log("msg")</literal>,
151 <literal>plpy.info("msg")</literal>,
152 <literal>plpy.notice("msg")</literal>,
153 <literal>plpy.warning("msg")</literal>,
154 <literal>plpy.error("msg")</literal>, and
155 <literal>plpy.fatal("msg")</literal>. They are mostly equivalent
156 to calling <literal>elog(<replaceable>LEVEL</>, "msg")</literal>
157 from C code.<indexterm><primary>elog</><secondary>in
158 PL/Python</></indexterm> <function>plpy.error</function> and
159 <function>plpy.fatal</function> actually raise a Python exception
160 which, if uncaught, causes the PL/Python module to call
161 <literal>elog(ERROR, msg)</literal> when the function handler
162 returns from the Python interpreter. Long-jumping out of the
163 Python interpreter is probably not good. <literal>raise
164 plpy.ERROR("msg")</literal> and <literal>raise
165 plpy.FATAL("msg")</literal> are equivalent to calling
166 <function>plpy.error</function> and
167 <function>plpy.fatal</function>, respectively.
171 Additionally, the <literal>plpy</literal> module provides two
172 functions called <function>execute</function> and
173 <function>prepare</function>. Calling
174 <function>plpy.execute</function> with a query string and an
175 optional limit argument causes that query to be run and the result
176 to be returned in a result object. The result object emulates a
177 list or dictionary object. The result object can be accessed by
178 row number and column name. It has these additional methods:
179 <function>nrows</function> which returns the number of rows
180 returned by the query, and <function>status</function> which is the
181 <function>SPI_exec()</function> return value. The result object
188 rv = plpy.execute("SELECT * FROM my_table", 5)
190 returns up to 5 rows from <literal>my_table</literal>. If
191 <literal>my_table</literal> has a column
192 <literal>my_column</literal>, it would be accessed as
194 foo = rv[i]["my_column"]
199 <indexterm><primary>preparing a query</><secondary>in PL/Python</></indexterm>
200 The second function, <function>plpy.prepare</function>, prepares
201 the execution plan for a query. It is called with a query string
202 and a list of parameter types, if you have parameter references in
203 the query. For example:
205 plan = plpy.prepare("SELECT last_name FROM my_users WHERE first_name = $1", [ "text" ])
207 <literal>text</literal> is the type of the variable you will be
208 passing for <literal>$1</literal>. After preparing a statement, you
209 use the function <function>plpy.execute</function> to run it:
211 rv = plpy.execute(plan, [ "name" ], 5)
213 The third argument is the limit and is optional.
217 In the current version, any database error encountered while
218 running a <application>PL/Python</application> function will result
219 in the immediate termination of that function by the server; it is
220 not possible to trap error conditions using Python <literal>try
221 ... catch</literal> constructs. For example, a syntax error in an
222 SQL statement passed to the <literal>plpy.execute</literal> call
223 will terminate the function. This behavior may be changed in a
228 When you prepare a plan using the PL/Python module it is
229 automatically saved. Read the SPI documentation (<xref
230 linkend="spi">) for a description of what this means.
231 In order to make effective use of this across function calls
232 one needs to use one of the persistent storage dictionaries
233 <literal>SD</literal> or <literal>GD</literal> (see
234 <xref linkend="plpython-funcs">). For example:
236 CREATE FUNCTION usesavedplan() RETURNS trigger AS '
237 if SD.has_key("plan"):
240 plan = plpy.prepare("SELECT 1")
243 ' LANGUAGE plpythonu;
249 <!-- NOT CURRENTLY SUPPORTED -->
251 <sect1 id="plpython-trusted">
252 <title>Restricted Environment</title>
255 The current version of <application>PL/Python</application>
256 functions as a trusted language only; access to the file system and
257 other local resources is disabled. Specifically,
258 <application>PL/Python</application> uses the Python restricted
259 execution environment, further restricts it to prevent the use of
260 the file <function>open</> call, and allows only modules from a
261 specific list to be imported. Presently, that list includes:
262 <literal>array</>, <literal>bisect</>, <literal>binascii</>,
263 <literal>calendar</>, <literal>cmath</>, <literal>codecs</>,
264 <literal>errno</>, <literal>marshal</>, <literal>math</>, <literal>md5</>,
265 <literal>mpz</>, <literal>operator</>, <literal>pcre</>,
266 <literal>pickle</>, <literal>random</>, <literal>re</>, <literal>regex</>,
267 <literal>sre</>, <literal>sha</>, <literal>string</>, <literal>StringIO</>,
268 <literal>struct</>, <literal>time</>, <literal>whrandom</>, and