doc/src/sgml/plpgsql.sgml - master - githubmirror/postgres

doc/src/sgml/plpgsql.sgml
6 174 строки227 KB
David Rowley
Doc: include MERGE in variable substitution command list
17 ноя 2025, 00:51
17 ноя 2025, 00:512b54a1a
<!-- doc/src/sgml/plpgsql.sgml -->

<chapter id="plpgsql">
  <title><application>PL/pgSQL</application> &mdash; <acronym>SQL</acronym> Procedural Language</title>

 <indexterm zone="plpgsql">
  <primary>PL/pgSQL</primary>
 </indexterm>

 <sect1 id="plpgsql-overview">
  <title>Overview</title>

 <para>
  <application>PL/pgSQL</application> is a loadable procedural
  language for the <productname>PostgreSQL</productname> database
  system.  The design goals of <application>PL/pgSQL</application> were to create
  a loadable procedural language that

    <itemizedlist>
     <listitem>
      <para>
       can be used to create functions, procedures, and triggers,
      </para>
     </listitem>
     <listitem>
      <para>
       adds control structures to the <acronym>SQL</acronym> language,
      </para>
     </listitem>
     <listitem>
      <para>
       can perform complex computations,
      </para>
     </listitem>
     <listitem>
      <para>
       inherits all user-defined types, functions, procedures, and operators,
      </para>
     </listitem>
     <listitem>
      <para>
       can be defined to be trusted by the server,
      </para>
     </listitem>
     <listitem>
      <para>
       is easy to use.
      </para>
     </listitem>
    </itemizedlist>
   </para>

   <para>
    Functions created with <application>PL/pgSQL</application> can be
    used anywhere that built-in functions could be used.
    For example, it is possible to
    create complex conditional computation functions and later use
    them to define operators or use them in index expressions.
   </para>

   <para>
    In <productname>PostgreSQL</productname> 9.0 and later,
    <application>PL/pgSQL</application> is installed by default.
    However it is still a loadable module, so especially security-conscious
    administrators could choose to remove it.
   </para>

  <sect2 id="plpgsql-advantages">
   <title>Advantages of Using <application>PL/pgSQL</application></title>

    <para>
     <acronym>SQL</acronym> is the language <productname>PostgreSQL</productname>
     and most other relational databases use as query language. It's
     portable and easy to learn. But every <acronym>SQL</acronym>
     statement must be executed individually by the database server.
    </para>

    <para>
     That means that your client application must send each query to
     the database server, wait for it to be processed, receive and
     process the results, do some computation, then send further
     queries to the server.  All this incurs interprocess
     communication and will also incur network overhead if your client
     is on a different machine than the database server.
    </para>

    <para>
     With <application>PL/pgSQL</application> you can group a block of
     computation and a series of queries <emphasis>inside</emphasis>
     the database server, thus having the power of a procedural
     language and the ease of use of SQL, but with considerable
     savings of client/server communication overhead.
    </para>
    <itemizedlist>

     <listitem><para> Extra round trips between
     client and server are eliminated </para></listitem>

     <listitem><para> Intermediate results that the client does not
     need do not have to be marshaled or transferred between server
     and client </para></listitem>

     <listitem><para> Multiple rounds of query
     parsing can be avoided </para></listitem>

    </itemizedlist>
    <para> This can result in a considerable performance increase as
    compared to an application that does not use stored functions.
    </para>

    <para>
     Also, with <application>PL/pgSQL</application> you can use all
     the data types, operators and functions of SQL.
    </para>
  </sect2>

  <sect2 id="plpgsql-args-results">
   <title>Supported Argument and Result Data Types</title>

    <para>
     Functions written in <application>PL/pgSQL</application> can accept
     as arguments any scalar or array data type supported by the server,
     and they can return a result of any of these types.  They can also
     accept or return any composite type (row type) specified by name.
     It is also possible to declare a <application>PL/pgSQL</application>
     function as accepting <type>record</type>, which means that any
     composite type will do as input, or
     as returning <type>record</type>, which means that the result
     is a row type whose columns are determined by specification in the
     calling query, as discussed in <xref linkend="queries-tablefunctions"/>.
    </para>

    <para>
     <application>PL/pgSQL</application> functions can be declared to accept a variable
     number of arguments by using the <literal>VARIADIC</literal> marker.  This
     works exactly the same way as for SQL functions, as discussed in
     <xref linkend="xfunc-sql-variadic-functions"/>.
    </para>

    <para>
     <application>PL/pgSQL</application> functions can also be declared to
     accept and return the polymorphic types described in
     <xref linkend="extend-types-polymorphic"/>, thus allowing the actual data
     types handled by the function to vary from call to call.
     Examples appear in <xref linkend="plpgsql-declaration-parameters"/>.
    </para>

    <para>
     <application>PL/pgSQL</application> functions can also be declared to return
     a <quote>set</quote> (or table) of any data type that can be returned as
     a single instance.  Such a function generates its output by executing
     <command>RETURN NEXT</command> for each desired element of the result
     set, or by using <command>RETURN QUERY</command> to output the result of
     evaluating a query.
    </para>

    <para>
     Finally, a <application>PL/pgSQL</application> function can be declared to return
     <type>void</type> if it has no useful return value.  (Alternatively, it
     could be written as a procedure in that case.)
    </para>

    <para>
     <application>PL/pgSQL</application> functions can also be declared with output
     parameters in place of an explicit specification of the return type.
     This does not add any fundamental capability to the language, but
     it is often convenient, especially for returning multiple values.
     The <literal>RETURNS TABLE</literal> notation can also be used in place
     of <literal>RETURNS SETOF</literal>.
    </para>

    <para>
     Specific examples appear in
     <xref linkend="plpgsql-declaration-parameters"/> and
     <xref linkend="plpgsql-statements-returning"/>.
    </para>
  </sect2>
 </sect1>

 <sect1 id="plpgsql-structure">
  <title>Structure of <application>PL/pgSQL</application></title>

  <para>
   Functions written in <application>PL/pgSQL</application> are defined
   to the server by executing <xref linkend="sql-createfunction"/> commands.
   Such a command would normally look like, say,
<programlisting>
CREATE FUNCTION somefunc(integer, text) RETURNS integer
AS '<replaceable>function body text</replaceable>'
LANGUAGE plpgsql;
</programlisting>
   The function body is simply a string literal so far as <command>CREATE
   FUNCTION</command> is concerned.  It is often helpful to use dollar quoting
   (see <xref linkend="sql-syntax-dollar-quoting"/>) to write the function
   body, rather than the normal single quote syntax.  Without dollar quoting,
   any single quotes or backslashes in the function body must be escaped by
   doubling them.  Almost all the examples in this chapter use dollar-quoted
   literals for their function bodies.
  </para>

  <para>
   <application>PL/pgSQL</application> is a block-structured language.
   The complete text of a function body must be a
   <firstterm>block</firstterm>. A block is defined as:

<synopsis>
<optional> &lt;&lt;<replaceable>label</replaceable>&gt;&gt; </optional>
<optional> DECLARE
    <replaceable>declarations</replaceable> </optional>
BEGIN
    <replaceable>statements</replaceable>
END <optional> <replaceable>label</replaceable> </optional>;
</synopsis>
    </para>

    <para>
     Each declaration and each statement within a block is terminated
     by a semicolon.  A block that appears within another block must
     have a semicolon after <literal>END</literal>, as shown above;
     however the final <literal>END</literal> that
     concludes a function body does not require a semicolon.
    </para>

    <tip>
     <para>
      A common mistake is to write a semicolon immediately after
      <literal>BEGIN</literal>.  This is incorrect and will result in a syntax error.
     </para>
    </tip>

    <para>
     A <replaceable>label</replaceable> is only needed if you want to
     identify the block for use
     in an <literal>EXIT</literal> statement, or to qualify the names of the
     variables declared in the block.  If a label is given after
     <literal>END</literal>, it must match the label at the block's beginning.
    </para>

    <para>
     All key words are case-insensitive.
     Identifiers are implicitly converted to lower case
     unless double-quoted, just as they are in ordinary SQL commands.
    </para>

    <para>
     Comments work the same way in <application>PL/pgSQL</application> code as in
     ordinary SQL.  A double dash (<literal>--</literal>) starts a comment
     that extends to the end of the line. A <literal>/*</literal> starts a
     block comment that extends to the matching occurrence of
     <literal>*/</literal>.  Block comments nest.
    </para>

    <para>
     Any statement in the statement section of a block
     can be a <firstterm>subblock</firstterm>.  Subblocks can be used for
     logical grouping or to localize variables to a small group
     of statements.  Variables declared in a subblock mask any
     similarly-named variables of outer blocks for the duration
     of the subblock; but you can access the outer variables anyway
     if you qualify their names with their block's label. For example:
<programlisting>
CREATE FUNCTION somefunc() RETURNS integer AS $$
&lt;&lt; outerblock &gt;&gt;
DECLARE
    quantity integer := 30;
BEGIN
    RAISE NOTICE 'Quantity here is %', quantity;  -- Prints 30
    quantity := 50;
    --
    -- Create a subblock
    --
    DECLARE
        quantity integer := 80;
    BEGIN
        RAISE NOTICE 'Quantity here is %', quantity;  -- Prints 80
        RAISE NOTICE 'Outer quantity here is %', outerblock.quantity;  -- Prints 50
    END;

    RAISE NOTICE 'Quantity here is %', quantity;  -- Prints 50

    RETURN quantity;
END;
$$ LANGUAGE plpgsql;
</programlisting>
    </para>

    <note>
     <para>
      There is actually a hidden <quote>outer block</quote> surrounding the body
      of any <application>PL/pgSQL</application> function.  This block provides the
      declarations of the function's parameters (if any), as well as some
      special variables such as <literal>FOUND</literal> (see
      <xref linkend="plpgsql-statements-diagnostics"/>).  The outer block is
      labeled with the function's name, meaning that parameters and special
      variables can be qualified with the function's name.
     </para>
    </note>

    <para>
     It is important not to confuse the use of
     <command>BEGIN</command>/<command>END</command> for grouping statements in
     <application>PL/pgSQL</application> with the similarly-named SQL commands
     for transaction
     control.  <application>PL/pgSQL</application>'s <command>BEGIN</command>/<command>END</command>
     are only for grouping; they do not start or end a transaction.
     See <xref linkend="plpgsql-transactions"/> for information on managing
     transactions in <application>PL/pgSQL</application>.
     Also, a block containing an <literal>EXCEPTION</literal> clause effectively
     forms a subtransaction that can be rolled back without affecting the
     outer transaction.  For more about that see <xref
     linkend="plpgsql-error-trapping"/>.
    </para>
  </sect1>

  <sect1 id="plpgsql-declarations">
    <title>Declarations</title>

    <para>
     All variables used in a block must be declared in the
     declarations section of the block.
     (The only exceptions are that the loop variable of a <literal>FOR</literal> loop
     iterating over a range of integer values is automatically declared as an
     integer variable, and likewise the loop variable of a <literal>FOR</literal> loop
     iterating over a cursor's result is automatically declared as a
     record variable.)
    </para>

    <para>
     <application>PL/pgSQL</application> variables can have any SQL data type, such as
     <type>integer</type>, <type>varchar</type>, and
     <type>char</type>.
    </para>

    <para>
     Here are some examples of variable declarations:
<programlisting>
user_id integer;
quantity numeric(5);
url varchar;
myrow tablename%ROWTYPE;
myfield tablename.columnname%TYPE;
arow RECORD;
</programlisting>
    </para>

    <para>
     The general syntax of a variable declaration is:
<synopsis>
<replaceable>name</replaceable> <optional> CONSTANT </optional> <replaceable>type</replaceable> <optional> COLLATE <replaceable>collation_name</replaceable> </optional> <optional> NOT NULL </optional> <optional> { DEFAULT | := | = } <replaceable>expression</replaceable> </optional>;
</synopsis>
      The <literal>DEFAULT</literal> clause, if given, specifies the initial value assigned
      to the variable when the block is entered.  If the <literal>DEFAULT</literal> clause
      is not given then the variable is initialized to the
      <acronym>SQL</acronym> null value.
      The <literal>CONSTANT</literal> option prevents the variable from being
      assigned to after initialization, so that its value will remain constant
      for the duration of the block.
      The <literal>COLLATE</literal> option specifies a collation to use for the
      variable (see <xref linkend="plpgsql-declaration-collation"/>).
      If <literal>NOT NULL</literal>
      is specified, an assignment of a null value results in a run-time
      error. All variables declared as <literal>NOT NULL</literal>
      must have a nonnull default value specified.
      Equal (<literal>=</literal>) can be used instead of PL/SQL-compliant
      <literal>:=</literal>.
     </para>

     <para>
      A variable's default value is evaluated and assigned to the variable
      each time the block is entered (not just once per function call).
      So, for example, assigning <literal>now()</literal> to a variable of type
      <type>timestamp</type> causes the variable to have the
      time of the current function call, not the time when the function was
      precompiled.
     </para>

     <para>
      Examples:
<programlisting>
quantity integer DEFAULT 32;
url varchar := 'http://mysite.com';
transaction_time CONSTANT timestamp with time zone := now();
</programlisting>
     </para>

     <para>
      Once declared, a variable's value can be used in later initialization
      expressions in the same block, for example:
<programlisting>
DECLARE
  x integer := 1;
  y integer := x + 1;
</programlisting>
     </para>

    <sect2 id="plpgsql-declaration-parameters">
     <title>Declaring Function Parameters</title>

     <para>
      Parameters passed to functions are named with the identifiers
      <literal>$1</literal>, <literal>$2</literal>,
      etc.  Optionally, aliases can be declared for
      <literal>$<replaceable>n</replaceable></literal>
      parameter names for increased readability.  Either the alias or the
      numeric identifier can then be used to refer to the parameter value.
     </para>

     <para>
      There are two ways to create an alias.  The preferred way is to give a
      name to the parameter in the <command>CREATE FUNCTION</command> command,
      for example:
<programlisting>
CREATE FUNCTION sales_tax(subtotal real) RETURNS real AS $$
BEGIN
    RETURN subtotal * 0.06;
END;
$$ LANGUAGE plpgsql;
</programlisting>
      The other way is to explicitly declare an alias, using the
      declaration syntax

<synopsis>
<replaceable>name</replaceable> ALIAS FOR $<replaceable>n</replaceable>;
</synopsis>

      The same example in this style looks like:
<programlisting>
CREATE FUNCTION sales_tax(real) RETURNS real AS $$
DECLARE
    subtotal ALIAS FOR $1;
BEGIN
    RETURN subtotal * 0.06;
END;
$$ LANGUAGE plpgsql;
</programlisting>
     </para>

    <note>
     <para>
      These two examples are not perfectly equivalent.  In the first case,
      <literal>subtotal</literal> could be referenced as
      <literal>sales_tax.subtotal</literal>, but in the second case it could not.
      (Had we attached a label to the inner block, <literal>subtotal</literal> could
      be qualified with that label, instead.)
     </para>
    </note>

     <para>
      Some more examples:
<programlisting>
CREATE FUNCTION instr(varchar, integer) RETURNS integer AS $$
DECLARE
    v_string ALIAS FOR $1;
    index ALIAS FOR $2;
BEGIN
    -- some computations using v_string and index here
END;
$$ LANGUAGE plpgsql;


CREATE FUNCTION concat_selected_fields(in_t sometablename) RETURNS text AS $$
BEGIN
    RETURN in_t.f1 || in_t.f3 || in_t.f5 || in_t.f7;
END;
$$ LANGUAGE plpgsql;
</programlisting>
     </para>

     <para>
      When a <application>PL/pgSQL</application> function is declared
      with output parameters, the output parameters are given
      <literal>$<replaceable>n</replaceable></literal> names and optional
      aliases in just the same way as the normal input parameters.  An
      output parameter is effectively a variable that starts out NULL;
      it should be assigned to during the execution of the function.
      The final value of the parameter is what is returned.  For instance,
      the sales-tax example could also be done this way:

<programlisting>
CREATE FUNCTION sales_tax(subtotal real, OUT tax real) AS $$
BEGIN
    tax := subtotal * 0.06;
END;
$$ LANGUAGE plpgsql;
</programlisting>

      Notice that we omitted <literal>RETURNS real</literal> &mdash; we could have
      included it, but it would be redundant.
     </para>

     <para>
      To call a function with <literal>OUT</literal> parameters, omit the
      output parameter(s) in the function call:
<programlisting>
SELECT sales_tax(100.00);
</programlisting>
     </para>

     <para>
      Output parameters are most useful when returning multiple values.
      A trivial example is:

<programlisting>
CREATE FUNCTION sum_n_product(x int, y int, OUT sum int, OUT prod int) AS $$
BEGIN
    sum := x + y;
    prod := x * y;
END;
$$ LANGUAGE plpgsql;

SELECT * FROM sum_n_product(2, 4);
 sum | prod
-----+------
   6 |    8
</programlisting>

      As discussed in <xref linkend="xfunc-output-parameters"/>, this
      effectively creates an anonymous record type for the function's
      results.  If a <literal>RETURNS</literal> clause is given, it must say
      <literal>RETURNS record</literal>.
     </para>

     <para>
      This also works with procedures, for example:

<programlisting>
CREATE PROCEDURE sum_n_product(x int, y int, OUT sum int, OUT prod int) AS $$
BEGIN
    sum := x + y;
    prod := x * y;
END;
$$ LANGUAGE plpgsql;
</programlisting>

      In a call to a procedure, all the parameters must be specified.  For
      output parameters, <literal>NULL</literal> may be specified when
      calling the procedure from plain SQL:
<programlisting>
CALL sum_n_product(2, 4, NULL, NULL);
 sum | prod
-----+------
   6 |    8
</programlisting>

      However, when calling a procedure
      from <application>PL/pgSQL</application>, you should instead write a
      variable for any output parameter; the variable will receive the result
      of the call.  See <xref linkend="plpgsql-statements-calling-procedure"/>
      for details.
     </para>

     <para>
      Another way to declare a <application>PL/pgSQL</application> function
      is with <literal>RETURNS TABLE</literal>, for example:

<programlisting>
CREATE FUNCTION extended_sales(p_itemno int)
RETURNS TABLE(quantity int, total numeric) AS $$
BEGIN
    RETURN QUERY SELECT s.quantity, s.quantity * s.price FROM sales AS s
                 WHERE s.itemno = p_itemno;
END;
$$ LANGUAGE plpgsql;
</programlisting>

      This is exactly equivalent to declaring one or more <literal>OUT</literal>
      parameters and specifying <literal>RETURNS SETOF
      <replaceable>sometype</replaceable></literal>.
     </para>

     <para>
      When the return type of a <application>PL/pgSQL</application> function
      is declared as a polymorphic type (see
      <xref linkend="extend-types-polymorphic"/>), a special
      parameter <literal>$0</literal> is created.  Its data type is the actual
      return type of the function, as deduced from the actual input types.
      This allows the function to access its actual return type
      as shown in <xref linkend="plpgsql-declaration-type"/>.
      <literal>$0</literal> is initialized to null and can be modified by
      the function, so it can be used to hold the return value if desired,
      though that is not required.  <literal>$0</literal> can also be
      given an alias.  For example, this function works on any data type
      that has a <literal>+</literal> operator:

<programlisting>
CREATE FUNCTION add_three_values(v1 anyelement, v2 anyelement, v3 anyelement)
RETURNS anyelement AS $$
DECLARE
    result ALIAS FOR $0;
BEGIN
    result := v1 + v2 + v3;
    RETURN result;
END;
$$ LANGUAGE plpgsql;
</programlisting>
     </para>

     <para>
      The same effect can be obtained by declaring one or more output parameters as
      polymorphic types.  In this case the
      special <literal>$0</literal> parameter is not used; the output
      parameters themselves serve the same purpose.  For example:

<programlisting>
CREATE FUNCTION add_three_values(v1 anyelement, v2 anyelement, v3 anyelement,
                                 OUT sum anyelement)
AS $$
BEGIN
    sum := v1 + v2 + v3;
END;
$$ LANGUAGE plpgsql;
</programlisting>
     </para>

     <para>
      In practice it might be more useful to declare a polymorphic function
      using the <type>anycompatible</type> family of types, so that automatic
      promotion of the input arguments to a common type will occur.
      For example:

<programlisting>
CREATE FUNCTION add_three_values(v1 anycompatible, v2 anycompatible, v3 anycompatible)
RETURNS anycompatible AS $$
BEGIN
    RETURN v1 + v2 + v3;
END;
$$ LANGUAGE plpgsql;
</programlisting>

      With this example, a call such as

<programlisting>
SELECT add_three_values(1, 2, 4.7);
</programlisting>

      will work, automatically promoting the integer inputs to numeric.
      The function using <type>anyelement</type> would require you to
      cast the three inputs to the same type manually.
     </para>
    </sect2>

  <sect2 id="plpgsql-declaration-alias">
   <title><literal>ALIAS</literal></title>

<synopsis>
<replaceable>newname</replaceable> ALIAS FOR <replaceable>oldname</replaceable>;
</synopsis>

   <para>
    The <literal>ALIAS</literal> syntax is more general than is suggested in the
    previous section: you can declare an alias for any variable, not just
    function parameters.  The main practical use for this is to assign
    a different name for variables with predetermined names, such as
    <varname>NEW</varname> or <varname>OLD</varname> within
    a trigger function.
   </para>

   <para>
    Examples:
<programlisting>
DECLARE
  prior ALIAS FOR old;
  updated ALIAS FOR new;
</programlisting>
   </para>

   <para>
    Since <literal>ALIAS</literal> creates two different ways to name the same
    object, unrestricted use can be confusing.  It's best to use it only
    for the purpose of overriding predetermined names.
   </para>
   </sect2>

  <sect2 id="plpgsql-declaration-type">
   <title>Copying Types</title>

<synopsis>
<replaceable>name</replaceable> <replaceable>table</replaceable>.<replaceable>column</replaceable>%TYPE
<replaceable>name</replaceable> <replaceable>variable</replaceable>%TYPE
</synopsis>

   <para>
    <literal>%TYPE</literal> provides the data type of a table column
    or a previously-declared <application>PL/pgSQL</application>
    variable. You can use this to declare variables that will hold
    database values. For example, let's say you have a column named
    <literal>user_id</literal> in your <literal>users</literal>
    table. To declare a variable with the same data type as
    <literal>users.user_id</literal> you write:
<programlisting>
user_id users.user_id%TYPE;
</programlisting>
   </para>

   <para>
    It is also possible to write array decoration
    after <literal>%TYPE</literal>, thereby creating a variable that holds
    an array of the referenced type:
<programlisting>
user_ids users.user_id%TYPE[];
user_ids users.user_id%TYPE ARRAY[4];  -- equivalent to the above
</programlisting>
    Just as when declaring table columns that are arrays, it doesn't
    matter whether you write multiple bracket pairs or specific array
    dimensions: <productname>PostgreSQL</productname> treats all arrays of
    a given element type as the same type, regardless of dimensionality.
    (See <xref linkend="arrays-declaration"/>.)
   </para>

   <para>
    By using <literal>%TYPE</literal> you don't need to know the data
    type of the structure you are referencing, and most importantly,
    if the data type of the referenced item changes in the future (for
    instance: you change the type of <literal>user_id</literal>
    from <type>integer</type> to <type>real</type>), you might not need
    to change your function definition.
   </para>

   <para>
    <literal>%TYPE</literal> is particularly valuable in polymorphic
    functions, since the data types needed for internal variables can
    change from one call to the next.  Appropriate variables can be
    created by applying <literal>%TYPE</literal> to the function's
    arguments or result placeholders.
   </para>

  </sect2>

    <sect2 id="plpgsql-declaration-rowtypes">
     <title>Row Types</title>

<synopsis>
<replaceable>name</replaceable> <replaceable>table_name</replaceable><literal>%ROWTYPE</literal>;
<replaceable>name</replaceable> <replaceable>composite_type_name</replaceable>;
</synopsis>

   <para>
    A variable of a composite type is called a <firstterm>row</firstterm>
    variable (or <firstterm>row-type</firstterm> variable).  Such a variable
    can hold a whole row of a <command>SELECT</command> or <command>FOR</command>
    query result, so long as that query's column set matches the
    declared type of the variable.
    The individual fields of the row value
    are accessed using the usual dot notation, for example
    <literal>rowvar.field</literal>.
   </para>

   <para>
    A row variable can be declared to have the same type as the rows of
    an existing table or view, by using the
    <replaceable>table_name</replaceable><literal>%ROWTYPE</literal>
    notation; or it can be declared by giving a composite type's name.
    (Since every table has an associated composite type of the same name,
    it actually does not matter in <productname>PostgreSQL</productname> whether you
    write <literal>%ROWTYPE</literal> or not.  But the form with
    <literal>%ROWTYPE</literal> is more portable.)
   </para>

   <para>
    As with <literal>%TYPE</literal>, <literal>%ROWTYPE</literal> can be
    followed by array decoration to declare a variable that holds an array
    of the referenced composite type.
   </para>

   <para>
    Parameters to a function can be
    composite types (complete table rows). In that case, the
    corresponding identifier <literal>$<replaceable>n</replaceable></literal> will be a row variable, and fields can
    be selected from it, for example <literal>$1.user_id</literal>.
   </para>

   <para>
    Here is an example of using composite types.  <structname>table1</structname>
    and <structname>table2</structname> are existing tables having at least the
    mentioned fields:

<programlisting>
CREATE FUNCTION merge_fields(t_row table1) RETURNS text AS $$
DECLARE
    t2_row table2%ROWTYPE;
BEGIN
    SELECT * INTO t2_row FROM table2 WHERE ... ;
    RETURN t_row.f1 || t2_row.f3 || t_row.f5 || t2_row.f7;
END;
$$ LANGUAGE plpgsql;

SELECT merge_fields(t.*) FROM table1 t WHERE ... ;
</programlisting>
   </para>
  </sect2>

  <sect2 id="plpgsql-declaration-records">
   <title>Record Types</title>

<synopsis>
<replaceable>name</replaceable> RECORD;
</synopsis>

   <para>
    Record variables are similar to row-type variables, but they have no
    predefined structure.  They take on the actual row structure of the
    row they are assigned during a <command>SELECT</command> or <command>FOR</command> command.  The substructure
    of a record variable can change each time it is assigned to.
    A consequence of this is that until a record variable is first assigned
    to, it has no substructure, and any attempt to access a
    field in it will draw a run-time error.
   </para>

   <para>
    Note that <literal>RECORD</literal> is not a true data type, only a placeholder.
    One should also realize that when a <application>PL/pgSQL</application>
    function is declared to return type <type>record</type>, this is not quite the
    same concept as a record variable, even though such a function might
    use a record variable to hold its result.  In both cases the actual row
    structure is unknown when the function is written, but for a function
    returning <type>record</type> the actual structure is determined when the
    calling query is parsed, whereas a record variable can change its row
    structure on-the-fly.
   </para>
  </sect2>

  <sect2 id="plpgsql-declaration-collation">
   <title>Collation of <application>PL/pgSQL</application> Variables</title>

   <indexterm>
    <primary>collation</primary>
    <secondary>in PL/pgSQL</secondary>
   </indexterm>

   <para>
    When a <application>PL/pgSQL</application> function has one or more
    parameters of collatable data types, a collation is identified for each
    function call depending on the collations assigned to the actual
    arguments, as described in <xref linkend="collation"/>.  If a collation is
    successfully identified (i.e., there are no conflicts of implicit
    collations among the arguments) then all the collatable parameters are
    treated as having that collation implicitly.  This will affect the
    behavior of collation-sensitive operations within the function.
    For example, consider

<programlisting>
CREATE FUNCTION less_than(a text, b text) RETURNS boolean AS $$
BEGIN
    RETURN a &lt; b;
END;
$$ LANGUAGE plpgsql;

SELECT less_than(text_field_1, text_field_2) FROM table1;
SELECT less_than(text_field_1, text_field_2 COLLATE "C") FROM table1;
</programlisting>

    The first use of <function>less_than</function> will use the common collation
    of <structfield>text_field_1</structfield> and <structfield>text_field_2</structfield> for
    the comparison, while the second use will use <literal>C</literal> collation.
   </para>

   <para>
    Furthermore, the identified collation is also assumed as the collation of
    any local variables that are of collatable types.  Thus this function
    would not work any differently if it were written as

<programlisting>
CREATE FUNCTION less_than(a text, b text) RETURNS boolean AS $$
DECLARE
    local_a text := a;
    local_b text := b;
BEGIN
    RETURN local_a &lt; local_b;
END;
$$ LANGUAGE plpgsql;
</programlisting>
   </para>

   <para>
    If there are no parameters of collatable data types, or no common
    collation can be identified for them, then parameters and local variables
    use the default collation of their data type (which is usually the
    database's default collation, but could be different for variables of
    domain types).
   </para>

   <para>
    A local variable of a collatable data type can have a different collation
    associated with it by including the <literal>COLLATE</literal> option in its
    declaration, for example

<programlisting>
DECLARE
    local_a text COLLATE "en_US";
</programlisting>

    This option overrides the collation that would otherwise be
    given to the variable according to the rules above.
   </para>

   <para>
    Also, of course explicit <literal>COLLATE</literal> clauses can be written inside
    a function if it is desired to force a particular collation to be used in
    a particular operation.  For example,

<programlisting>
CREATE FUNCTION less_than_c(a text, b text) RETURNS boolean AS $$
BEGIN
    RETURN a &lt; b COLLATE "C";
END;
$$ LANGUAGE plpgsql;
</programlisting>

    This overrides the collations associated with the table columns,
    parameters, or local variables used in the expression, just as would
    happen in a plain SQL command.
   </para>
  </sect2>
  </sect1>

  <sect1 id="plpgsql-expressions">
  <title>Expressions</title>

    <para>
     All expressions used in <application>PL/pgSQL</application>
     statements are processed using the server's main
     <acronym>SQL</acronym> executor.  For example, when you write
     a <application>PL/pgSQL</application> statement like
<synopsis>
IF <replaceable>expression</replaceable> THEN ...
</synopsis>
     <application>PL/pgSQL</application> will evaluate the expression by
     feeding a query like
<synopsis>
SELECT <replaceable>expression</replaceable>
</synopsis>
     to the main SQL engine.  While forming the <command>SELECT</command> command,
     any occurrences of <application>PL/pgSQL</application> variable names
     are replaced by query parameters, as discussed in detail in
     <xref linkend="plpgsql-var-subst"/>.
     This allows the query plan for the <command>SELECT</command> to
     be prepared just once and then reused for subsequent
     evaluations with different values of the variables.  Thus, what
     really happens on first use of an expression is essentially a
     <command>PREPARE</command> command.  For example, if we have declared
     two integer variables <literal>x</literal> and <literal>y</literal>, and we write
<programlisting>
IF x &lt; y THEN ...
</programlisting>
     what happens behind the scenes is equivalent to
<programlisting>
PREPARE <replaceable>statement_name</replaceable>(integer, integer) AS SELECT $1 &lt; $2;
</programlisting>
     and then this prepared statement is <command>EXECUTE</command>d for each
     execution of the <command>IF</command> statement, with the current values
     of the <application>PL/pgSQL</application> variables supplied as
     parameter values.  Normally these details are
     not important to a <application>PL/pgSQL</application> user, but
     they are useful to know when trying to diagnose a problem.
     More information appears in <xref linkend="plpgsql-plan-caching"/>.
    </para>

    <para>
     Since an <replaceable>expression</replaceable> is converted to a
     <literal>SELECT</literal> command, it can contain the same clauses
     that an ordinary <literal>SELECT</literal> would, except that it
     cannot include a top-level <literal>UNION</literal>,
     <literal>INTERSECT</literal>, or <literal>EXCEPT</literal> clause.
     Thus for example one could test whether a table is non-empty with
<programlisting>
IF count(*) &gt; 0 FROM my_table THEN ...
</programlisting>
     since the <replaceable>expression</replaceable>
     between <literal>IF</literal> and <literal>THEN</literal> is parsed as
     though it were <literal>SELECT count(*) &gt; 0 FROM my_table</literal>.
     The <literal>SELECT</literal> must produce a single column, and not
     more than one row.  (If it produces no rows, the result is taken as
     NULL.)
    </para>
  </sect1>

  <sect1 id="plpgsql-statements">
  <title>Basic Statements</title>

   <para>
    In this section and the following ones, we describe all the statement
    types that are explicitly understood by
    <application>PL/pgSQL</application>.
    Anything not recognized as one of these statement types is presumed
    to be an SQL command and is sent to the main database engine to execute,
    as described in <xref linkend="plpgsql-statements-general-sql"/>.
   </para>

   <sect2 id="plpgsql-statements-assignment">
    <title>Assignment</title>

    <para>
     An assignment of a value to a <application>PL/pgSQL</application>
     variable is written as:
<synopsis>
<replaceable>variable</replaceable> { := | = } <replaceable>expression</replaceable>;
</synopsis>
     As explained previously, the expression in such a statement is evaluated
     by means of an SQL <command>SELECT</command> command sent to the main
     database engine.  The expression must yield a single value (possibly
     a row value, if the variable is a row or record variable).  The target
     variable can be a simple variable (optionally qualified with a block
     name), a field of a row or record target, or an element or slice of
     an array target.  Equal (<literal>=</literal>) can be
     used instead of PL/SQL-compliant <literal>:=</literal>.
    </para>

    <para>
     If the expression's result data type doesn't match the variable's
     data type, the value will be coerced as though by an assignment cast
     (see <xref linkend="typeconv-query"/>).  If no assignment cast is known
     for the pair of data types involved, the <application>PL/pgSQL</application>
     interpreter will attempt to convert the result value textually, that is
     by applying the result type's output function followed by the variable
     type's input function.  Note that this could result in run-time errors
     generated by the input function, if the string form of the result value
     is not acceptable to the input function.
    </para>

    <para>
     Examples:
<programlisting>
tax := subtotal * 0.06;
my_record.user_id := 20;
my_array[j] := 20;
my_array[1:3] := ARRAY[1, 2, 3];
complex_array[n].realpart = 12.3;
</programlisting>
    </para>
   </sect2>

   <sect2 id="plpgsql-statements-general-sql">
    <title>Executing SQL Commands</title>

    <para>
     In general, any SQL command that does not return rows can be executed
     within a <application>PL/pgSQL</application> function just by writing
     the command.  For example, you could create and fill a table by writing
<programlisting>
CREATE TABLE mytable (id int PRIMARY KEY, data text);
INSERT INTO mytable VALUES (1,'one'), (2,'two');
</programlisting>
    </para>

    <para>
     If the command does return rows (for example <command>SELECT</command>, or
     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
     with <literal>RETURNING</literal>), there are two ways to proceed.
     When the command will return at most one row, or you only care about
     the first row of output, write the command as usual but add
     an <literal>INTO</literal> clause to capture the output, as described
     in <xref linkend="plpgsql-statements-sql-onerow"/>.
     To process all of the output rows, write the command as the data
     source for a <command>FOR</command> loop, as described in
     <xref linkend="plpgsql-records-iterating"/>.
    </para>

    <para>
     Usually it is not sufficient just to execute statically-defined SQL
     commands.  Typically you'll want a command to use varying data values,
     or even to vary in more fundamental ways such as by using different
     table names at different times.  Again, there are two ways to proceed
     depending on the situation.
    </para>

    <para>
     <application>PL/pgSQL</application> variable values can be
     automatically inserted into optimizable SQL commands, which
     are <command>SELECT</command>, <command>INSERT</command>,
     <command>UPDATE</command>, <command>DELETE</command>,
     <command>MERGE</command>, and certain
     utility commands that incorporate one of these, such
     as <command>EXPLAIN</command> and <command>CREATE TABLE ... AS
     SELECT</command>.  In these commands,
     any <application>PL/pgSQL</application> variable name appearing
     in the command text is replaced by a query parameter, and then the
     current value of the variable is provided as the parameter value
     at run time.  This is exactly like the processing described earlier
     for expressions; for details see <xref linkend="plpgsql-var-subst"/>.
    </para>

    <para>
     When executing an optimizable SQL command in this way,
     <application>PL/pgSQL</application> may cache and re-use the execution
     plan for the command, as discussed in
     <xref linkend="plpgsql-plan-caching"/>.
    </para>

    <para>
     Non-optimizable SQL commands (also called utility commands) are not
     capable of accepting query parameters.  So automatic substitution
     of <application>PL/pgSQL</application> variables does not work in such
     commands.  To include non-constant text in a utility command executed
     from <application>PL/pgSQL</application>, you must build the utility
     command as a string and then <command>EXECUTE</command> it, as
     discussed in <xref linkend="plpgsql-statements-executing-dyn"/>.
    </para>

    <para>
     <command>EXECUTE</command> must also be used if you want to modify
     the command in some other way than supplying a data value, for example
     by changing a table name.
    </para>

    <para>
     Sometimes it is useful to evaluate an expression or <command>SELECT</command>
     query but discard the result, for example when calling a function
     that has side-effects but no useful result value.  To do
     this in <application>PL/pgSQL</application>, use the
     <command>PERFORM</command> statement:

<synopsis>
PERFORM <replaceable>query</replaceable>;
</synopsis>

     This executes <replaceable>query</replaceable> and discards the
     result.  Write the <replaceable>query</replaceable> the same
     way you would write an SQL <command>SELECT</command> command, but replace the
     initial keyword <command>SELECT</command> with <command>PERFORM</command>.
     For <command>WITH</command> queries, use <command>PERFORM</command> and then
     place the query in parentheses.  (In this case, the query can only
     return one row.)
     <application>PL/pgSQL</application> variables will be
     substituted into the query just as described above,
     and the plan is cached in the same way.  Also, the special variable
     <literal>FOUND</literal> is set to true if the query produced at
     least one row, or false if it produced no rows (see
     <xref linkend="plpgsql-statements-diagnostics"/>).
    </para>

    <note>
     <para>
      One might expect that writing <command>SELECT</command> directly
      would accomplish this result, but at
      present the only accepted way to do it is
      <command>PERFORM</command>.  An SQL command that can return rows,
      such as <command>SELECT</command>, will be rejected as an error
      unless it has an <literal>INTO</literal> clause as discussed in the
      next section.
     </para>
    </note>

    <para>
     An example:
<programlisting>
PERFORM create_mv('cs_session_page_requests_mv', my_query);
</programlisting>
    </para>
   </sect2>

   <sect2 id="plpgsql-statements-sql-onerow">
    <title>Executing a Command with a Single-Row Result</title>

    <indexterm zone="plpgsql-statements-sql-onerow">
     <primary>SELECT INTO</primary>
     <secondary>in PL/pgSQL</secondary>
    </indexterm>

    <indexterm zone="plpgsql-statements-sql-onerow">
     <primary>RETURNING INTO</primary>
     <secondary>in PL/pgSQL</secondary>
    </indexterm>

    <para>
     The result of an SQL command yielding a single row (possibly of multiple
     columns) can be assigned to a record variable, row-type variable, or list
     of scalar variables.  This is done by writing the base SQL command and
     adding an <literal>INTO</literal> clause.  For example,

<synopsis>
SELECT <replaceable>select_expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable> FROM ...;
INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
</synopsis>

     where <replaceable>target</replaceable> can be a record variable, a row
     variable, or a comma-separated list of simple variables and
     record/row fields.
     <application>PL/pgSQL</application> variables will be
     substituted into the rest of the command (that is, everything but the
     <literal>INTO</literal> clause) just as described above,
     and the plan is cached in the same way.
     This works for <command>SELECT</command>,
     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
     with <literal>RETURNING</literal>, and certain utility commands
     that return row sets, such as <command>EXPLAIN</command>.
     Except for the <literal>INTO</literal> clause, the SQL command is the same
     as it would be written outside <application>PL/pgSQL</application>.
    </para>

   <tip>
    <para>
     Note that this interpretation of <command>SELECT</command> with <literal>INTO</literal>
     is quite different from <productname>PostgreSQL</productname>'s regular
     <command>SELECT INTO</command> command, wherein the <literal>INTO</literal>
     target is a newly created table.  If you want to create a table from a
     <command>SELECT</command> result inside a
     <application>PL/pgSQL</application> function, use the syntax
     <command>CREATE TABLE ... AS SELECT</command>.
    </para>
   </tip>

    <para>
     If a row variable or a variable list is used as target,
     the command's result columns
     must exactly match the structure of the target as to number and data
     types, or else a run-time error
     occurs.  When a record variable is the target, it automatically
     configures itself to the row type of the command's result columns.
    </para>

    <para>
     The <literal>INTO</literal> clause can appear almost anywhere in the SQL
     command.  Customarily it is written either just before or just after
     the list of <replaceable>select_expressions</replaceable> in a
     <command>SELECT</command> command, or at the end of the command for other
     command types.  It is recommended that you follow this convention
     in case the <application>PL/pgSQL</application> parser becomes
     stricter in future versions.
    </para>

    <para>
     If <literal>STRICT</literal> is not specified in the <literal>INTO</literal>
     clause, then <replaceable>target</replaceable> will be set to the first
     row returned by the command, or to nulls if the command returned no rows.
     (Note that <quote>the first row</quote> is not
     well-defined unless you've used <literal>ORDER BY</literal>.)  Any result rows
     after the first row are discarded.
     You can check the special <literal>FOUND</literal> variable (see
     <xref linkend="plpgsql-statements-diagnostics"/>) to
     determine whether a row was returned:

<programlisting>
SELECT * INTO myrec FROM emp WHERE empname = myname;
IF NOT FOUND THEN
    RAISE EXCEPTION 'employee % not found', myname;
END IF;
</programlisting>

     If the <literal>STRICT</literal> option is specified, the command must
     return exactly one row or a run-time error will be reported, either
     <literal>NO_DATA_FOUND</literal> (no rows) or <literal>TOO_MANY_ROWS</literal>
     (more than one row). You can use an exception block if you wish
     to catch the error, for example:

<programlisting>
BEGIN
    SELECT * INTO STRICT myrec FROM emp WHERE empname = myname;
    EXCEPTION
        WHEN NO_DATA_FOUND THEN
            RAISE EXCEPTION 'employee % not found', myname;
        WHEN TOO_MANY_ROWS THEN
            RAISE EXCEPTION 'employee % not unique', myname;
END;
</programlisting>
     Successful execution of a command with <literal>STRICT</literal>
     always sets <literal>FOUND</literal> to true.
    </para>

    <para>
     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
     <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
     an error for more than one returned row, even when
     <literal>STRICT</literal> is not specified.  This is because there
     is no option such as <literal>ORDER BY</literal> with which to determine
     which affected row should be returned.
    </para>

    <para>
     If <literal>print_strict_params</literal> is enabled for the function,
     then when an error is thrown because the requirements
     of <literal>STRICT</literal> are not met, the <literal>DETAIL</literal> part of
     the error message will include information about the parameters
     passed to the command.
     You can change the <literal>print_strict_params</literal>
     setting for all functions by setting
     <varname>plpgsql.print_strict_params</varname>, though only subsequent
     function compilations will be affected.  You can also enable it
     on a per-function basis by using a compiler option, for example:
<programlisting>
CREATE FUNCTION get_userid(username text) RETURNS int
AS $$
#print_strict_params on
DECLARE
userid int;
BEGIN
    SELECT users.userid INTO STRICT userid
        FROM users WHERE users.username = get_userid.username;
    RETURN userid;
END;
$$ LANGUAGE plpgsql;
</programlisting>
     On failure, this function might produce an error message such as
<programlisting>
ERROR:  query returned no rows
DETAIL:  parameters: username = 'nosuchuser'
CONTEXT:  PL/pgSQL function get_userid(text) line 6 at SQL statement
</programlisting>
    </para>

    <note>
     <para>
      The <literal>STRICT</literal> option matches the behavior of
      Oracle PL/SQL's <command>SELECT INTO</command> and related statements.
     </para>
    </note>

   </sect2>

   <sect2 id="plpgsql-statements-executing-dyn">
    <title>Executing Dynamic Commands</title>

    <para>
     Oftentimes you will want to generate dynamic commands inside your
     <application>PL/pgSQL</application> functions, that is, commands
     that will involve different tables or different data types each
     time they are executed.  <application>PL/pgSQL</application>'s
     normal attempts to cache plans for commands (as discussed in
     <xref linkend="plpgsql-plan-caching"/>) will not work in such
     scenarios.  To handle this sort of problem, the
     <command>EXECUTE</command> statement is provided:

<synopsis>
EXECUTE <replaceable class="command">command-string</replaceable> <optional> INTO <optional>STRICT</optional> <replaceable>target</replaceable> </optional> <optional> USING <replaceable>expression</replaceable> <optional>, ... </optional> </optional>;
</synopsis>

     where <replaceable>command-string</replaceable> is an expression
     yielding a string (of type <type>text</type>) containing the
     command to be executed.  The optional <replaceable>target</replaceable>
     is a record variable, a row variable, or a comma-separated list of
     simple variables and record/row fields, into which the results of
     the command will be stored.  The optional <literal>USING</literal> expressions
     supply values to be inserted into the command.
    </para>

    <para>
     No substitution of <application>PL/pgSQL</application> variables is done on the
     computed command string.  Any required variable values must be inserted
     in the command string as it is constructed; or you can use parameters
     as described below.
    </para>

    <para>
     Also, there is no plan caching for commands executed via
     <command>EXECUTE</command>.  Instead, the command is always planned
     each time the statement is run. Thus the command
     string can be dynamically created within the function to perform
     actions on different tables and columns.
    </para>

    <para>
     The <literal>INTO</literal> clause specifies where the results of
     an SQL command returning rows should be assigned. If a row variable
     or variable list is provided, it must exactly match the structure
     of the command's results; if a
     record variable is provided, it will configure itself to match the
     result structure automatically. If multiple rows are returned,
     only the first will be assigned to the <literal>INTO</literal>
     variable(s). If no rows are returned, NULL is assigned to the
     <literal>INTO</literal> variable(s). If no <literal>INTO</literal>
     clause is specified, the command results are discarded.
    </para>

    <para>
     If the <literal>STRICT</literal> option is given, an error is reported
     unless the command produces exactly one row.
    </para>

    <para>
     The command string can use parameter values, which are referenced
     in the command as <literal>$1</literal>, <literal>$2</literal>, etc.
     These symbols refer to values supplied in the <literal>USING</literal>
     clause.  This method is often preferable to inserting data values
     into the command string as text: it avoids run-time overhead of
     converting the values to text and back, and it is much less prone
     to SQL-injection attacks since there is no need for quoting or escaping.
     An example is:
<programlisting>
EXECUTE 'SELECT count(*) FROM mytable WHERE inserted_by = $1 AND inserted &lt;= $2'
   INTO c
   USING checked_user, checked_date;
</programlisting>
    </para>

    <para>
     Note that parameter symbols can only be used for data values
     &mdash; if you want to use dynamically determined table or column
     names, you must insert them into the command string textually.
     For example, if the preceding query needed to be done against a
     dynamically selected table, you could do this:
<programlisting>
EXECUTE 'SELECT count(*) FROM '
    || quote_ident(tabname)
    || ' WHERE inserted_by = $1 AND inserted &lt;= $2'
   INTO c
   USING checked_user, checked_date;
</programlisting>
     A cleaner approach is to use <function>format()</function>'s <literal>%I</literal>
     specification to insert table or column names with automatic quoting:
<programlisting>
EXECUTE format('SELECT count(*) FROM %I '
   'WHERE inserted_by = $1 AND inserted &lt;= $2', tabname)
   INTO c
   USING checked_user, checked_date;
</programlisting>
     (This example relies on the SQL rule that string literals separated by a
     newline are implicitly concatenated.)
    </para>

    <para>
     Another restriction on parameter symbols is that they only work in
     optimizable SQL commands
     (<command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
     <command>DELETE</command>, <command>MERGE</command>, and certain commands containing one of these).
     In other statement
     types (generically called utility statements), you must insert
     values textually even if they are just data values.
    </para>

    <para>
     An <command>EXECUTE</command> with a simple constant command string and some
     <literal>USING</literal> parameters, as in the first example above, is
     functionally equivalent to just writing the command directly in
     <application>PL/pgSQL</application> and allowing replacement of
     <application>PL/pgSQL</application> variables to happen automatically.
     The important difference is that <command>EXECUTE</command> will re-plan
     the command on each execution, generating a plan that is specific
     to the current parameter values; whereas
     <application>PL/pgSQL</application> may otherwise create a generic plan
     and cache it for re-use.  In situations where the best plan depends
     strongly on the parameter values, it can be helpful to use
     <command>EXECUTE</command> to positively ensure that a generic plan is not
     selected.
    </para>

    <para>
     <command>SELECT INTO</command> is not currently supported within
     <command>EXECUTE</command>; instead, execute a plain <command>SELECT</command>
     command and specify <literal>INTO</literal> as part of the <command>EXECUTE</command>
     itself.
    </para>

   <note>
    <para>
     The <application>PL/pgSQL</application>
     <command>EXECUTE</command> statement is not related to the
     <link linkend="sql-execute"><command>EXECUTE</command></link> SQL
     statement supported by the
     <productname>PostgreSQL</productname> server. The server's
     <command>EXECUTE</command> statement cannot be used directly within
     <application>PL/pgSQL</application> functions (and is not needed).
    </para>
   </note>

   <example id="plpgsql-quote-literal-example">
   <title>Quoting Values in Dynamic Queries</title>

    <indexterm>
     <primary>quote_ident</primary>
     <secondary>use in PL/pgSQL</secondary>
    </indexterm>

    <indexterm>
     <primary>quote_literal</primary>
     <secondary>use in PL/pgSQL</secondary>
    </indexterm>

    <indexterm>
     <primary>quote_nullable</primary>
     <secondary>use in PL/pgSQL</secondary>
    </indexterm>

    <indexterm>
     <primary>format</primary>
     <secondary>use in PL/pgSQL</secondary>
    </indexterm>

    <para>
     When working with dynamic commands you will often have to handle escaping
     of single quotes.  The recommended method for quoting fixed text in your
     function body is dollar quoting.  (If you have legacy code that does
     not use dollar quoting, please refer to the
     overview in <xref linkend="plpgsql-quote-tips"/>, which can save you
     some effort when translating said code to a more reasonable scheme.)
    </para>

    <para>
     Dynamic values require careful handling since they might contain
     quote characters.
     An example using <function>format()</function> (this assumes that you are
     dollar quoting the function body so quote marks need not be doubled):
<programlisting>
EXECUTE format('UPDATE tbl SET %I = $1 '
   'WHERE key = $2', colname) USING newvalue, keyvalue;
</programlisting>
     It is also possible to call the quoting functions directly:
<programlisting>
EXECUTE 'UPDATE tbl SET '
        || quote_ident(colname)
        || ' = '
        || quote_literal(newvalue)
        || ' WHERE key = '
        || quote_literal(keyvalue);
</programlisting>
    </para>

    <para>
     This example demonstrates the use of the
     <function>quote_ident</function> and
     <function>quote_literal</function> functions (see <xref
     linkend="functions-string"/>).  For safety, expressions containing column
     or table identifiers should be passed through
     <function>quote_ident</function> before insertion in a dynamic query.
     Expressions containing values that should be literal strings in the
     constructed command should be passed through <function>quote_literal</function>.
     These functions take the appropriate steps to return the input text
     enclosed in double or single quotes respectively, with any embedded
     special characters properly escaped.
    </para>

    <para>
     Because <function>quote_literal</function> is labeled
     <literal>STRICT</literal>, it will always return null when called with a
     null argument.  In the above example, if <literal>newvalue</literal> or
     <literal>keyvalue</literal> were null, the entire dynamic query string would
     become null, leading to an error from <command>EXECUTE</command>.
     You can avoid this problem by using the <function>quote_nullable</function>
     function, which works the same as <function>quote_literal</function> except that
     when called with a null argument it returns the string <literal>NULL</literal>.
     For example,
<programlisting>
EXECUTE 'UPDATE tbl SET '
        || quote_ident(colname)
        || ' = '
        || quote_nullable(newvalue)
        || ' WHERE key = '
        || quote_nullable(keyvalue);
</programlisting>
     If you are dealing with values that might be null, you should usually
     use <function>quote_nullable</function> in place of <function>quote_literal</function>.
    </para>

    <para>
     As always, care must be taken to ensure that null values in a query do
     not deliver unintended results.  For example the <literal>WHERE</literal> clause
<programlisting>
'WHERE key = ' || quote_nullable(keyvalue)
</programlisting>
     will never succeed if <literal>keyvalue</literal> is null, because the
     result of using the equality operator <literal>=</literal> with a null operand
     is always null.  If you wish null to work like an ordinary key value,
     you would need to rewrite the above as
<programlisting>
'WHERE key IS NOT DISTINCT FROM ' || quote_nullable(keyvalue)
</programlisting>
     (At present, <literal>IS NOT DISTINCT FROM</literal> is handled much less
     efficiently than <literal>=</literal>, so don't do this unless you must.
     See <xref linkend="functions-comparison"/> for
     more information on nulls and <literal>IS DISTINCT</literal>.)
    </para>

    <para>
     Note that dollar quoting is only useful for quoting fixed text.
     It would be a very bad idea to try to write this example as:
<programlisting>
EXECUTE 'UPDATE tbl SET '
        || quote_ident(colname)
        || ' = $$'
        || newvalue
        || '$$ WHERE key = '
        || quote_literal(keyvalue);
</programlisting>
     because it would break if the contents of <literal>newvalue</literal>
     happened to contain <literal>$$</literal>.  The same objection would
     apply to any other dollar-quoting delimiter you might pick.
     So, to safely quote text that is not known in advance, you
     <emphasis>must</emphasis> use <function>quote_literal</function>,
     <function>quote_nullable</function>, or <function>quote_ident</function>, as appropriate.
    </para>

    <para>
     Dynamic SQL statements can also be safely constructed using the
     <function>format</function> function (see <xref
     linkend="functions-string-format"/>). For example:
<programlisting>
EXECUTE format('UPDATE tbl SET %I = %L '
   'WHERE key = %L', colname, newvalue, keyvalue);
</programlisting>
     <literal>%I</literal> is equivalent to <function>quote_ident</function>, and
     <literal>%L</literal> is equivalent to <function>quote_nullable</function>.
     The <function>format</function> function can be used in conjunction with
     the <literal>USING</literal> clause:
<programlisting>
EXECUTE format('UPDATE tbl SET %I = $1 WHERE key = $2', colname)
   USING newvalue, keyvalue;
</programlisting>
     This form is better because the variables are handled in their native
     data type format, rather than unconditionally converting them to
     text and quoting them via <literal>%L</literal>.  It is also more efficient.
    </para>
   </example>

    <para>
     A much larger example of a dynamic command and
     <command>EXECUTE</command> can be seen in <xref
     linkend="plpgsql-porting-ex2"/>, which builds and executes a
     <command>CREATE FUNCTION</command> command to define a new function.
    </para>
   </sect2>

   <sect2 id="plpgsql-statements-diagnostics">
    <title>Obtaining the Result Status</title>

    <para>
     There are several ways to determine the effect of a command. The
     first method is to use the <command>GET DIAGNOSTICS</command>
     command, which has the form:

<synopsis>
GET <optional> CURRENT </optional> DIAGNOSTICS <replaceable>variable</replaceable> { = | := } <replaceable>item</replaceable> <optional> , ... </optional>;
</synopsis>

     This command allows retrieval of system status indicators.
     <literal>CURRENT</literal> is a noise word (but see also <command>GET STACKED
     DIAGNOSTICS</command> in <xref linkend="plpgsql-exception-diagnostics"/>).
     Each <replaceable>item</replaceable> is a key word identifying a status
     value to be assigned to the specified <replaceable>variable</replaceable>
     (which should be of the right data type to receive it).  The currently
     available status items are shown
     in <xref linkend="plpgsql-current-diagnostics-values"/>.  Colon-equal
     (<literal>:=</literal>) can be used instead of the SQL-standard <literal>=</literal>
     token.  An example:
<programlisting>
GET DIAGNOSTICS integer_var = ROW_COUNT;
</programlisting>
    </para>

     <table id="plpgsql-current-diagnostics-values">
      <title>Available Diagnostics Items</title>
      <tgroup cols="3">
       <colspec colname="col1" colwidth="1*"/>
       <colspec colname="col2" colwidth="1*"/>
       <colspec colname="col3" colwidth="2*"/>
       <thead>
        <row>
         <entry>Name</entry>
         <entry>Type</entry>
         <entry>Description</entry>
        </row>
       </thead>
       <tbody>
        <row>
         <entry><varname>ROW_COUNT</varname></entry>
         <entry><type>bigint</type></entry>
         <entry>the number of rows processed by the most
          recent <acronym>SQL</acronym> command</entry>
        </row>
        <row>
         <entry><literal>PG_CONTEXT</literal></entry>
         <entry><type>text</type></entry>
         <entry>line(s) of text describing the current call stack
          (see <xref linkend="plpgsql-call-stack"/>)</entry>
        </row>
        <row>
         <entry><literal>PG_ROUTINE_OID</literal></entry>
         <entry><type>oid</type></entry>
         <entry>OID of the current function</entry>
        </row>
       </tbody>
      </tgroup>
     </table>

    <para>
     The second method to determine the effects of a command is to check the
     special variable named <literal>FOUND</literal>, which is of
     type <type>boolean</type>.  <literal>FOUND</literal> starts out
     false within each <application>PL/pgSQL</application> function call.
     It is set by each of the following types of statements:

         <itemizedlist>
          <listitem>
           <para>
            A <command>SELECT INTO</command> statement sets
            <literal>FOUND</literal> true if a row is assigned, false if no
            row is returned.
           </para>
          </listitem>
          <listitem>
           <para>
            A <command>PERFORM</command> statement sets <literal>FOUND</literal>
            true if it produces (and discards) one or more rows, false if
            no row is produced.
           </para>
          </listitem>
          <listitem>
           <para>
            <command>UPDATE</command>, <command>INSERT</command>, <command>DELETE</command>,
            and <command>MERGE</command>
            statements set <literal>FOUND</literal> true if at least one
            row is affected, false if no row is affected.
           </para>
          </listitem>
          <listitem>
           <para>
            A <command>FETCH</command> statement sets <literal>FOUND</literal>
            true if it returns a row, false if no row is returned.
           </para>
          </listitem>
          <listitem>
           <para>
            A <command>MOVE</command> statement sets <literal>FOUND</literal>
            true if it successfully repositions the cursor, false otherwise.
           </para>
          </listitem>
          <listitem>
           <para>
            A <command>FOR</command> or <command>FOREACH</command> statement sets
            <literal>FOUND</literal> true
            if it iterates one or more times, else false.
            <literal>FOUND</literal> is set this way when the
            loop exits; inside the execution of the loop,
            <literal>FOUND</literal> is not modified by the
            loop statement, although it might be changed by the
            execution of other statements within the loop body.
           </para>
          </listitem>
          <listitem>
           <para>
            <command>RETURN QUERY</command> and <command>RETURN QUERY
            EXECUTE</command> statements set <literal>FOUND</literal>
            true if the query returns at least one row, false if no row
            is returned.
           </para>
          </listitem>
         </itemizedlist>

     Other <application>PL/pgSQL</application> statements do not change
     the state of <literal>FOUND</literal>.
     Note in particular that <command>EXECUTE</command>
     changes the output of <command>GET DIAGNOSTICS</command>, but
     does not change <literal>FOUND</literal>.
    </para>

    <para>
     <literal>FOUND</literal> is a local variable within each
     <application>PL/pgSQL</application> function; any changes to it
     affect only the current function.
    </para>

   </sect2>

   <sect2 id="plpgsql-statements-null">
    <title>Doing Nothing At All</title>

    <para>
     Sometimes a placeholder statement that does nothing is useful.
     For example, it can indicate that one arm of an if/then/else
     chain is deliberately empty.  For this purpose, use the
     <command>NULL</command> statement:

<synopsis>
NULL;
</synopsis>
    </para>

    <para>
     For example, the following two fragments of code are equivalent:
<programlisting>
BEGIN
    y := x / 0;
EXCEPTION
    WHEN division_by_zero THEN
        NULL;  -- ignore the error
END;
</programlisting>

<programlisting>
BEGIN
    y := x / 0;
EXCEPTION
    WHEN division_by_zero THEN  -- ignore the error
END;
</programlisting>
     Which is preferable is a matter of taste.
    </para>

    <note>
     <para>
      In Oracle's PL/SQL, empty statement lists are not allowed, and so
      <command>NULL</command> statements are <emphasis>required</emphasis> for situations
      such as this.  <application>PL/pgSQL</application> allows you to
      just write nothing, instead.
     </para>
    </note>

   </sect2>
  </sect1>

  <sect1 id="plpgsql-control-structures">
   <title>Control Structures</title>

   <para>
    Control structures are probably the most useful (and
    important) part of <application>PL/pgSQL</application>. With
    <application>PL/pgSQL</application>'s control structures,
    you can manipulate <productname>PostgreSQL</productname> data in a very
    flexible and powerful way.
   </para>

   <sect2 id="plpgsql-statements-returning">
    <title>Returning from a Function</title>

    <para>
     There are two commands available that allow you to return data
     from a function: <command>RETURN</command> and <command>RETURN
     NEXT</command>.
    </para>

    <sect3 id="plpgsql-statements-returning-return">
     <title><command>RETURN</command></title>

<synopsis>
RETURN <replaceable>expression</replaceable>;
</synopsis>

     <para>
      <command>RETURN</command> with an expression terminates the
      function and returns the value of
      <replaceable>expression</replaceable> to the caller.  This form
      is used for <application>PL/pgSQL</application> functions that do
      not return a set.
     </para>

     <para>
      In a function that returns a scalar type, the expression's result will
      automatically be cast into the function's return type as described for
      assignments.  But to return a composite (row) value, you must write an
      expression delivering exactly the requested column set.  This may
      require use of explicit casting.
     </para>

     <para>
      If you declared the function with output parameters, write just
      <command>RETURN</command> with no expression.  The current values
      of the output parameter variables will be returned.
     </para>

     <para>
      If you declared the function to return <type>void</type>, a
      <command>RETURN</command> statement can be used to exit the function
      early; but do not write an expression following
      <command>RETURN</command>.
     </para>

     <para>
      The return value of a function cannot be left undefined. If
      control reaches the end of the top-level block of the function
      without hitting a <command>RETURN</command> statement, a run-time
      error will occur.  This restriction does not apply to functions
      with output parameters and functions returning <type>void</type>,
      however.  In those cases a <command>RETURN</command> statement is
      automatically executed if the top-level block finishes.
     </para>

     <para>
      Some examples:

<programlisting>
-- functions returning a scalar type
RETURN 1 + 2;
RETURN scalar_var;

-- functions returning a composite type
RETURN composite_type_var;
RETURN (1, 2, 'three'::text);  -- must cast columns to correct types
</programlisting>
     </para>
    </sect3>

    <sect3 id="plpgsql-statements-returning-return-next">
     <title><command>RETURN NEXT</command> and <command>RETURN QUERY</command></title>
    <indexterm>
     <primary>RETURN NEXT</primary>
     <secondary>in PL/pgSQL</secondary>
    </indexterm>
    <indexterm>
     <primary>RETURN QUERY</primary>
     <secondary>in PL/pgSQL</secondary>
    </indexterm>

<synopsis>
RETURN NEXT <replaceable>expression</replaceable>;
RETURN QUERY <replaceable>query</replaceable>;
RETURN QUERY EXECUTE <replaceable class="command">command-string</replaceable> <optional> USING <replaceable>expression</replaceable> <optional>, ... </optional> </optional>;
</synopsis>

     <para>
      When a <application>PL/pgSQL</application> function is declared to return
      <literal>SETOF <replaceable>sometype</replaceable></literal>, the procedure
      to follow is slightly different.  In that case, the individual
      items to return are specified by a sequence of <command>RETURN
      NEXT</command> or <command>RETURN QUERY</command> commands, and
      then a final <command>RETURN</command> command with no argument
      is used to indicate that the function has finished executing.
      <command>RETURN NEXT</command> can be used with both scalar and
      composite data types; with a composite result type, an entire
      <quote>table</quote> of results will be returned.
      <command>RETURN QUERY</command> appends the results of executing
      a query to the function's result set. <command>RETURN
      NEXT</command> and <command>RETURN QUERY</command> can be freely
      intermixed in a single set-returning function, in which case
      their results will be concatenated.
     </para>

     <para>
      <command>RETURN NEXT</command> and <command>RETURN
      QUERY</command> do not actually return from the function &mdash;
      they simply append zero or more rows to the function's result
      set.  Execution then continues with the next statement in the
      <application>PL/pgSQL</application> function.  As successive
      <command>RETURN NEXT</command> or <command>RETURN
      QUERY</command> commands are executed, the result set is built
      up.  A final <command>RETURN</command>, which should have no
      argument, causes control to exit the function (or you can just
      let control reach the end of the function).
     </para>

     <para>
      <command>RETURN QUERY</command> has a variant
      <command>RETURN QUERY EXECUTE</command>, which specifies the
      query to be executed dynamically.  Parameter expressions can
      be inserted into the computed query string via <literal>USING</literal>,
      in just the same way as in the <command>EXECUTE</command> command.
     </para>

     <para>
      If you declared the function with output parameters, write just
      <command>RETURN NEXT</command> with no expression.  On each
      execution, the current values of the output parameter
      variable(s) will be saved for eventual return as a row of the
      result.  Note that you must declare the function as returning
      <literal>SETOF record</literal> when there are multiple output
      parameters, or <literal>SETOF <replaceable>sometype</replaceable></literal>
      when there is just one output parameter of type
      <replaceable>sometype</replaceable>, in order to create a set-returning
      function with output parameters.
     </para>

     <para>
      Here is an example of a function using <command>RETURN
      NEXT</command>:

<programlisting>
CREATE TABLE foo (fooid INT, foosubid INT, fooname TEXT);
INSERT INTO foo VALUES (1, 2, 'three');
INSERT INTO foo VALUES (4, 5, 'six');

CREATE OR REPLACE FUNCTION get_all_foo() RETURNS SETOF foo AS
$BODY$
DECLARE
    r foo%rowtype;
BEGIN
    FOR r IN
        SELECT * FROM foo WHERE fooid &gt; 0
    LOOP
        -- can do some processing here
        RETURN NEXT r; -- return current row of SELECT
    END LOOP;
    RETURN;
END;
$BODY$
LANGUAGE plpgsql;

SELECT * FROM get_all_foo();
</programlisting>
     </para>

     <para>
      Here is an example of a function using <command>RETURN
      QUERY</command>:

<programlisting>
CREATE FUNCTION get_available_flightid(date) RETURNS SETOF integer AS
$BODY$
BEGIN
    RETURN QUERY SELECT flightid
                   FROM flight
                  WHERE flightdate &gt;= $1
                    AND flightdate &lt; ($1 + 1);

    -- Since execution is not finished, we can check whether rows were returned
    -- and raise exception if not.
    IF NOT FOUND THEN
        RAISE EXCEPTION 'No flight at %.', $1;
    END IF;

    RETURN;
 END;
$BODY$
LANGUAGE plpgsql;

-- Returns available flights or raises exception if there are no
-- available flights.
SELECT * FROM get_available_flightid(CURRENT_DATE);
</programlisting>
     </para>

     <note>
      <para>
       The current implementation of <command>RETURN NEXT</command>
       and <command>RETURN QUERY</command> stores the entire result set
       before returning from the function, as discussed above.  That
       means that if a <application>PL/pgSQL</application> function produces a
       very large result set, performance might be poor: data will be
       written to disk to avoid memory exhaustion, but the function
       itself will not return until the entire result set has been
       generated.  A future version of <application>PL/pgSQL</application> might
       allow users to define set-returning functions
       that do not have this limitation.  Currently, the point at
       which data begins being written to disk is controlled by the
       <xref linkend="guc-work-mem"/>
       configuration variable.  Administrators who have sufficient
       memory to store larger result sets in memory should consider
       increasing this parameter.
      </para>
     </note>
    </sect3>
   </sect2>

   <sect2 id="plpgsql-statements-returning-procedure">
    <title>Returning from a Procedure</title>

    <para>
     A procedure does not have a return value.  A procedure can therefore end
     without a <command>RETURN</command> statement.  If you wish to use
     a <command>RETURN</command> statement to exit the code early, write
     just <command>RETURN</command> with no expression.
    </para>

    <para>
     If the procedure has output parameters, the final values of the output
     parameter variables will be returned to the caller.
    </para>
   </sect2>

   <sect2 id="plpgsql-statements-calling-procedure">
    <title>Calling a Procedure</title>

    <para>
     A <application>PL/pgSQL</application> function, procedure,
     or <command>DO</command> block can call a procedure
     using <command>CALL</command>.  Output parameters are handled
     differently from the way that <command>CALL</command> works in plain
     SQL.  Each <literal>OUT</literal> or <literal>INOUT</literal>
     parameter of the procedure must
     correspond to a variable in the <command>CALL</command> statement, and
     whatever the procedure returns is assigned back to that variable after
     it returns.  For example:
<programlisting>
CREATE PROCEDURE triple(INOUT x int)
LANGUAGE plpgsql
AS $$
BEGIN
    x := x * 3;
END;
$$;

DO $$
DECLARE myvar int := 5;
BEGIN
  CALL triple(myvar);
  RAISE NOTICE 'myvar = %', myvar;  -- prints 15
END;
$$;
</programlisting>
     The variable corresponding to an output parameter can be a simple
     variable or a field of a composite-type variable.  Currently,
     it cannot be an element of an array.
    </para>
   </sect2>

   <sect2 id="plpgsql-conditionals">
    <title>Conditionals</title>

    <para>
     <command>IF</command> and <command>CASE</command> statements let you execute
     alternative commands based on certain conditions.
     <application>PL/pgSQL</application> has three forms of <command>IF</command>:
    <itemizedlist>
     <listitem>
      <para><literal>IF ... THEN ... END IF</literal></para>
     </listitem>
     <listitem>
      <para><literal>IF ... THEN ... ELSE ... END IF</literal></para>
     </listitem>
     <listitem>
      <para><literal>IF ... THEN ... ELSIF ... THEN ... ELSE ... END IF</literal></para>
     </listitem>
    </itemizedlist>

    and two forms of <command>CASE</command>:
    <itemizedlist>
     <listitem>
      <para><literal>CASE ... WHEN ... THEN ... ELSE ... END CASE</literal></para>
     </listitem>
     <listitem>
      <para><literal>CASE WHEN ... THEN ... ELSE ... END CASE</literal></para>
     </listitem>
    </itemizedlist>
    </para>

    <sect3 id="plpgsql-conditionals-if-then">
     <title><literal>IF-THEN</literal></title>

<synopsis>
IF <replaceable>boolean-expression</replaceable> THEN
    <replaceable>statements</replaceable>
END IF;
</synopsis>

       <para>
        <literal>IF-THEN</literal> statements are the simplest form of
        <literal>IF</literal>. The statements between
        <literal>THEN</literal> and <literal>END IF</literal> will be
        executed if the condition is true. Otherwise, they are
        skipped.
       </para>

       <para>
        Example:
<programlisting>
IF v_user_id &lt;&gt; 0 THEN
    UPDATE users SET email = v_email WHERE user_id = v_user_id;
END IF;
</programlisting>
       </para>
     </sect3>

     <sect3 id="plpgsql-conditionals-if-then-else">
      <title><literal>IF-THEN-ELSE</literal></title>

<synopsis>
IF <replaceable>boolean-expression</replaceable> THEN
    <replaceable>statements</replaceable>
ELSE
    <replaceable>statements</replaceable>
END IF;
</synopsis>

       <para>
        <literal>IF-THEN-ELSE</literal> statements add to
        <literal>IF-THEN</literal> by letting you specify an
        alternative set of statements that should be executed if the
        condition is not true.  (Note this includes the case where the
        condition evaluates to NULL.)
       </para>

       <para>
        Examples:
<programlisting>
IF parentid IS NULL OR parentid = ''
THEN
    RETURN fullname;
ELSE
    RETURN hp_true_filename(parentid) || '/' || fullname;
END IF;
</programlisting>

<programlisting>
IF v_count &gt; 0 THEN
    INSERT INTO users_count (count) VALUES (v_count);
    RETURN 't';
ELSE
    RETURN 'f';
END IF;
</programlisting>
     </para>
    </sect3>

     <sect3 id="plpgsql-conditionals-if-then-elsif">
      <title><literal>IF-THEN-ELSIF</literal></title>

<synopsis>
IF <replaceable>boolean-expression</replaceable> THEN
    <replaceable>statements</replaceable>
<optional> ELSIF <replaceable>boolean-expression</replaceable> THEN
    <replaceable>statements</replaceable>
<optional> ELSIF <replaceable>boolean-expression</replaceable> THEN
    <replaceable>statements</replaceable>
    ...
</optional>
</optional>
<optional> ELSE
    <replaceable>statements</replaceable> </optional>
END IF;
</synopsis>

       <para>
        Sometimes there are more than just two alternatives.
        <literal>IF-THEN-ELSIF</literal> provides a convenient
        method of checking several alternatives in turn.
        The <literal>IF</literal> conditions are tested successively
        until the first one that is true is found.  Then the
        associated statement(s) are executed, after which control
        passes to the next statement after <literal>END IF</literal>.
        (Any subsequent <literal>IF</literal> conditions are <emphasis>not</emphasis>
        tested.)  If none of the <literal>IF</literal> conditions is true,
        then the <literal>ELSE</literal> block (if any) is executed.
       </para>

       <para>
        Here is an example:

<programlisting>
IF number = 0 THEN
    result := 'zero';
ELSIF number &gt; 0 THEN
    result := 'positive';
ELSIF number &lt; 0 THEN
    result := 'negative';
ELSE
    -- hmm, the only other possibility is that number is null
    result := 'NULL';
END IF;
</programlisting>
       </para>

       <para>
        The key word <literal>ELSIF</literal> can also be spelled
        <literal>ELSEIF</literal>.
       </para>

       <para>
        An alternative way of accomplishing the same task is to nest
        <literal>IF-THEN-ELSE</literal> statements, as in the
        following example:

<programlisting>
IF demo_row.sex = 'm' THEN
    pretty_sex := 'man';
ELSE
    IF demo_row.sex = 'f' THEN
        pretty_sex := 'woman';
    END IF;
END IF;
</programlisting>
       </para>

       <para>
        However, this method requires writing a matching <literal>END IF</literal>
        for each <literal>IF</literal>, so it is much more cumbersome than
        using <literal>ELSIF</literal> when there are many alternatives.
       </para>
     </sect3>

     <sect3 id="plpgsql-conditionals-simple-case">
      <title>Simple <literal>CASE</literal></title>

<synopsis>
CASE <replaceable>search-expression</replaceable>
    WHEN <replaceable>expression</replaceable> <optional>, <replaceable>expression</replaceable> <optional> ... </optional></optional> THEN
      <replaceable>statements</replaceable>
  <optional> WHEN <replaceable>expression</replaceable> <optional>, <replaceable>expression</replaceable> <optional> ... </optional></optional> THEN
      <replaceable>statements</replaceable>
    ... </optional>
  <optional> ELSE
      <replaceable>statements</replaceable> </optional>
END CASE;
</synopsis>

      <para>
       The simple form of <command>CASE</command> provides conditional execution
       based on equality of operands.  The <replaceable>search-expression</replaceable>
       is evaluated (once) and successively compared to each
       <replaceable>expression</replaceable> in the <literal>WHEN</literal> clauses.
       If a match is found, then the corresponding
       <replaceable>statements</replaceable> are executed, and then control
       passes to the next statement after <literal>END CASE</literal>.  (Subsequent
       <literal>WHEN</literal> expressions are not evaluated.)  If no match is
       found, the <literal>ELSE</literal> <replaceable>statements</replaceable> are
       executed; but if <literal>ELSE</literal> is not present, then a
       <literal>CASE_NOT_FOUND</literal> exception is raised.
      </para>

      <para>
       Here is a simple example:

<programlisting>
CASE x
    WHEN 1, 2 THEN
        msg := 'one or two';
    ELSE
        msg := 'other value than one or two';
END CASE;
</programlisting>
      </para>
     </sect3>

     <sect3 id="plpgsql-conditionals-searched-case">
      <title>Searched <literal>CASE</literal></title>

<synopsis>
CASE
    WHEN <replaceable>boolean-expression</replaceable> THEN
      <replaceable>statements</replaceable>
  <optional> WHEN <replaceable>boolean-expression</replaceable> THEN
      <replaceable>statements</replaceable>
    ... </optional>
  <optional> ELSE
      <replaceable>statements</replaceable> </optional>
END CASE;
</synopsis>

      <para>
       The searched form of <command>CASE</command> provides conditional execution
       based on truth of Boolean expressions.  Each <literal>WHEN</literal> clause's
       <replaceable>boolean-expression</replaceable> is evaluated in turn,
       until one is found that yields <literal>true</literal>.  Then the
       corresponding <replaceable>statements</replaceable> are executed, and
       then control passes to the next statement after <literal>END CASE</literal>.
       (Subsequent <literal>WHEN</literal> expressions are not evaluated.)
       If no true result is found, the <literal>ELSE</literal>
       <replaceable>statements</replaceable> are executed;
       but if <literal>ELSE</literal> is not present, then a
       <literal>CASE_NOT_FOUND</literal> exception is raised.
      </para>

      <para>
       Here is an example:

<programlisting>
CASE
    WHEN x BETWEEN 0 AND 10 THEN
        msg := 'value is between zero and ten';
    WHEN x BETWEEN 11 AND 20 THEN
        msg := 'value is between eleven and twenty';
END CASE;
</programlisting>
      </para>

      <para>
       This form of <command>CASE</command> is entirely equivalent to
       <literal>IF-THEN-ELSIF</literal>, except for the rule that reaching
       an omitted <literal>ELSE</literal> clause results in an error rather
       than doing nothing.
      </para>

     </sect3>
   </sect2>

   <sect2 id="plpgsql-control-structures-loops">
    <title>Simple Loops</title>

    <indexterm zone="plpgsql-control-structures-loops">
     <primary>loop</primary>
     <secondary>in PL/pgSQL</secondary>
    </indexterm>

    <para>
     With the <literal>LOOP</literal>, <literal>EXIT</literal>,
     <literal>CONTINUE</literal>, <literal>WHILE</literal>, <literal>FOR</literal>,
     and <literal>FOREACH</literal> statements, you can arrange for your
     <application>PL/pgSQL</application> function to repeat a series of commands.
    </para>

    <sect3 id="plpgsql-control-structures-loops-loop">
     <title><literal>LOOP</literal></title>

<synopsis>
<optional> &lt;&lt;<replaceable>label</replaceable>&gt;&gt; </optional>
LOOP
    <replaceable>statements</replaceable>
END LOOP <optional> <replaceable>label</replaceable> </optional>;
</synopsis>

     <para>
      <literal>LOOP</literal> defines an unconditional loop that is repeated
      indefinitely until terminated by an <literal>EXIT</literal> or
      <command>RETURN</command> statement.  The optional
      <replaceable>label</replaceable> can be used by <literal>EXIT</literal>
      and <literal>CONTINUE</literal> statements within nested loops to
      specify which loop those statements refer to.
     </para>
    </sect3>

     <sect3 id="plpgsql-control-structures-loops-exit">
      <title><literal>EXIT</literal></title>

     <indexterm>
      <primary>EXIT</primary>
      <secondary>in PL/pgSQL</secondary>
     </indexterm>

<synopsis>
EXIT <optional> <replaceable>label</replaceable> </optional> <optional> WHEN <replaceable>boolean-expression</replaceable> </optional>;
</synopsis>

       <para>
        If no <replaceable>label</replaceable> is given, the innermost
        loop is terminated and the statement following <literal>END
        LOOP</literal> is executed next.  If <replaceable>label</replaceable>
        is given, it must be the label of the current or some outer
        level of nested loop or block. Then the named loop or block is
        terminated and control continues with the statement after the
        loop's/block's corresponding <literal>END</literal>.
       </para>

       <para>
        If <literal>WHEN</literal> is specified, the loop exit occurs only if
        <replaceable>boolean-expression</replaceable> is true. Otherwise, control passes
        to the statement after <literal>EXIT</literal>.
       </para>

       <para>
        <literal>EXIT</literal> can be used with all types of loops; it is
        not limited to use with unconditional loops.
       </para>

       <para>
        When used with a
        <literal>BEGIN</literal> block, <literal>EXIT</literal> passes
        control to the next statement after the end of the block.
        Note that a label must be used for this purpose; an unlabeled
        <literal>EXIT</literal> is never considered to match a
        <literal>BEGIN</literal> block.  (This is a change from
        pre-8.4 releases of <productname>PostgreSQL</productname>, which
        would allow an unlabeled <literal>EXIT</literal> to match
        a <literal>BEGIN</literal> block.)
       </para>

       <para>
        Examples:
<programlisting>
LOOP
    -- some computations
    IF count &gt; 0 THEN
        EXIT;  -- exit loop
    END IF;
END LOOP;

LOOP
    -- some computations
    EXIT WHEN count &gt; 0;  -- same result as previous example
END LOOP;

&lt;&lt;ablock&gt;&gt;
BEGIN
    -- some computations
    IF stocks &gt; 100000 THEN
        EXIT ablock;  -- causes exit from the BEGIN block
    END IF;
    -- computations here will be skipped when stocks &gt; 100000
END;
</programlisting>
       </para>
     </sect3>

     <sect3 id="plpgsql-control-structures-loops-continue">
      <title><literal>CONTINUE</literal></title>

     <indexterm>
      <primary>CONTINUE</primary>
      <secondary>in PL/pgSQL</secondary>
     </indexterm>

<synopsis>
CONTINUE <optional> <replaceable>label</replaceable> </optional> <optional> WHEN <replaceable>boolean-expression</replaceable> </optional>;
</synopsis>

       <para>
        If no <replaceable>label</replaceable> is given, the next iteration of
        the innermost loop is begun. That is, all statements remaining
        in the loop body are skipped, and control returns
        to the loop control expression (if any) to determine whether
        another loop iteration is needed.
        If <replaceable>label</replaceable> is present, it
        specifies the label of the loop whose execution will be
        continued.
       </para>

       <para>
        If <literal>WHEN</literal> is specified, the next iteration of the
        loop is begun only if <replaceable>boolean-expression</replaceable> is
        true. Otherwise, control passes to the statement after
        <literal>CONTINUE</literal>.
       </para>

       <para>
        <literal>CONTINUE</literal> can be used with all types of loops; it
        is not limited to use with unconditional loops.
       </para>

       <para>
        Examples:
<programlisting>
LOOP
    -- some computations
    EXIT WHEN count &gt; 100;
    CONTINUE WHEN count &lt; 50;
    -- some computations for count IN [50 .. 100]
END LOOP;
</programlisting>
       </para>
     </sect3>


     <sect3 id="plpgsql-control-structures-loops-while">
      <title><literal>WHILE</literal></title>

     <indexterm>
      <primary>WHILE</primary>
      <secondary>in PL/pgSQL</secondary>
     </indexterm>

<synopsis>
<optional> &lt;&lt;<replaceable>label</replaceable>&gt;&gt; </optional>
WHILE <replaceable>boolean-expression</replaceable> LOOP
    <replaceable>statements</replaceable>
END LOOP <optional> <replaceable>label</replaceable> </optional>;
</synopsis>

       <para>
        The <literal>WHILE</literal> statement repeats a
        sequence of statements so long as the
        <replaceable>boolean-expression</replaceable>
        evaluates to true.  The expression is checked just before
        each entry to the loop body.
       </para>

       <para>
        For example:
<programlisting>
WHILE amount_owed &gt; 0 AND gift_certificate_balance &gt; 0 LOOP
    -- some computations here
END LOOP;

WHILE NOT done LOOP
    -- some computations here
END LOOP;
</programlisting>
       </para>
     </sect3>

     <sect3 id="plpgsql-integer-for">
      <title><literal>FOR</literal> (Integer Variant)</title>

<synopsis>
<optional> &lt;&lt;<replaceable>label</replaceable>&gt;&gt; </optional>
FOR <replaceable>name</replaceable> IN <optional> REVERSE </optional> <replaceable>expression</replaceable> .. <replaceable>expression</replaceable> <optional> BY <replaceable>expression</replaceable> </optional> LOOP
    <replaceable>statements</replaceable>
END LOOP <optional> <replaceable>label</replaceable> </optional>;
</synopsis>

       <para>
        This form of <literal>FOR</literal> creates a loop that iterates over a range
        of integer values. The variable
        <replaceable>name</replaceable> is automatically defined as type
        <type>integer</type> and exists only inside the loop (any existing
        definition of the variable name is ignored within the loop).
        The two expressions giving
        the lower and upper bound of the range are evaluated once when entering
        the loop. If the <literal>BY</literal> clause isn't specified the iteration
        step is 1, otherwise it's the value specified in the <literal>BY</literal>
        clause, which again is evaluated once on loop entry.
        If <literal>REVERSE</literal> is specified then the step value is
        subtracted, rather than added, after each iteration.
       </para>

       <para>
        Some examples of integer <literal>FOR</literal> loops:
<programlisting>
FOR i IN 1..10 LOOP
    -- i will take on the values 1,2,3,4,5,6,7,8,9,10 within the loop
END LOOP;

FOR i IN REVERSE 10..1 LOOP
    -- i will take on the values 10,9,8,7,6,5,4,3,2,1 within the loop
END LOOP;

FOR i IN REVERSE 10..1 BY 2 LOOP
    -- i will take on the values 10,8,6,4,2 within the loop
END LOOP;
</programlisting>
       </para>

       <para>
        If the lower bound is greater than the upper bound (or less than,
        in the <literal>REVERSE</literal> case), the loop body is not
        executed at all.  No error is raised.
       </para>

       <para>
        If a <replaceable>label</replaceable> is attached to the
        <literal>FOR</literal> loop then the integer loop variable can be
        referenced with a qualified name, using that
        <replaceable>label</replaceable>.
       </para>
     </sect3>
   </sect2>

   <sect2 id="plpgsql-records-iterating">
    <title>Looping through Query Results</title>

    <para>
     Using a different type of <literal>FOR</literal> loop, you can iterate through
     the results of a query and manipulate that data
     accordingly. The syntax is:
<synopsis>
<optional> &lt;&lt;<replaceable>label</replaceable>&gt;&gt; </optional>
FOR <replaceable>target</replaceable> IN <replaceable>query</replaceable> LOOP
    <replaceable>statements</replaceable>
END LOOP <optional> <replaceable>label</replaceable> </optional>;
</synopsis>
     The <replaceable>target</replaceable> is a record variable, row variable,
     or comma-separated list of scalar variables.
     The <replaceable>target</replaceable> is successively assigned each row
     resulting from the <replaceable>query</replaceable> and the loop body is
     executed for each row. Here is an example:
<programlisting>
CREATE FUNCTION refresh_mviews() RETURNS integer AS $$
DECLARE
    mviews RECORD;
BEGIN
    RAISE NOTICE 'Refreshing all materialized views...';

    FOR mviews IN
       SELECT n.nspname AS mv_schema,
              c.relname AS mv_name,
              pg_catalog.pg_get_userbyid(c.relowner) AS owner
         FROM pg_catalog.pg_class c
    LEFT JOIN pg_catalog.pg_namespace n ON (n.oid = c.relnamespace)
        WHERE c.relkind = 'm'
     ORDER BY 1
    LOOP

        -- Now "mviews" has one record with information about the materialized view

        RAISE NOTICE 'Refreshing materialized view %.% (owner: %)...',
                     quote_ident(mviews.mv_schema),
                     quote_ident(mviews.mv_name),
                     quote_ident(mviews.owner);
        EXECUTE format('REFRESH MATERIALIZED VIEW %I.%I', mviews.mv_schema, mviews.mv_name);
    END LOOP;

    RAISE NOTICE 'Done refreshing materialized views.';
    RETURN 1;
END;
$$ LANGUAGE plpgsql;
</programlisting>

     If the loop is terminated by an <literal>EXIT</literal> statement, the last
     assigned row value is still accessible after the loop.
    </para>

    <para>
     The <replaceable>query</replaceable> used in this type of <literal>FOR</literal>
     statement can be any SQL command that returns rows to the caller:
     <command>SELECT</command> is the most common case,
     but you can also use <command>INSERT</command>, <command>UPDATE</command>,
     <command>DELETE</command>, or <command>MERGE</command> with a
     <literal>RETURNING</literal> clause.  Some utility
     commands such as <command>EXPLAIN</command> will work too.
    </para>

    <para>
     <application>PL/pgSQL</application> variables are replaced by query parameters,
     and the query plan is cached for possible re-use, as discussed in
     detail in <xref linkend="plpgsql-var-subst"/> and
     <xref linkend="plpgsql-plan-caching"/>.
    </para>

    <para>
     The <literal>FOR-IN-EXECUTE</literal> statement is another way to iterate over
     rows:
<synopsis>
<optional> &lt;&lt;<replaceable>label</replaceable>&gt;&gt; </optional>
FOR <replaceable>target</replaceable> IN EXECUTE <replaceable>text_expression</replaceable> <optional> USING <replaceable>expression</replaceable> <optional>, ... </optional> </optional> LOOP
    <replaceable>statements</replaceable>
END LOOP <optional> <replaceable>label</replaceable> </optional>;
</synopsis>
     This is like the previous form, except that the source query
     is specified as a string expression, which is evaluated and replanned
     on each entry to the <literal>FOR</literal> loop.  This allows the programmer to
     choose the speed of a preplanned query or the flexibility of a dynamic
     query, just as with a plain <command>EXECUTE</command> statement.
     As with <command>EXECUTE</command>, parameter values can be inserted
     into the dynamic command via <literal>USING</literal>.
    </para>

    <para>
     Another way to specify the query whose results should be iterated
     through is to declare it as a cursor.  This is described in
     <xref linkend="plpgsql-cursor-for-loop"/>.
    </para>
   </sect2>

   <sect2 id="plpgsql-foreach-array">
    <title>Looping through Arrays</title>

    <para>
     The <literal>FOREACH</literal> loop is much like a <literal>FOR</literal> loop,
     but instead of iterating through the rows returned by an SQL query,
     it iterates through the elements of an array value.
     (In general, <literal>FOREACH</literal> is meant for looping through
     components of a composite-valued expression; variants for looping
     through composites besides arrays may be added in future.)
     The <literal>FOREACH</literal> statement to loop over an array is:

<synopsis>
<optional> &lt;&lt;<replaceable>label</replaceable>&gt;&gt; </optional>
FOREACH <replaceable>target</replaceable> <optional> SLICE <replaceable>number</replaceable> </optional> IN ARRAY <replaceable>expression</replaceable> LOOP
    <replaceable>statements</replaceable>
END LOOP <optional> <replaceable>label</replaceable> </optional>;
</synopsis>
    </para>

    <para>
     Without <literal>SLICE</literal>, or if <literal>SLICE 0</literal> is specified,
     the loop iterates through individual elements of the array produced
     by evaluating the <replaceable>expression</replaceable>.
     The <replaceable>target</replaceable> variable is assigned each
     element value in sequence, and the loop body is executed for each element.
     Here is an example of looping through the elements of an integer
     array:

<programlisting>
CREATE FUNCTION sum(int[]) RETURNS int8 AS $$
DECLARE
  s int8 := 0;
  x int;
BEGIN
  FOREACH x IN ARRAY $1
  LOOP
    s := s + x;
  END LOOP;
  RETURN s;
END;
$$ LANGUAGE plpgsql;
</programlisting>

     The elements are visited in storage order, regardless of the number of
     array dimensions.  Although the <replaceable>target</replaceable> is
     usually just a single variable, it can be a list of variables when
     looping through an array of composite values (records).  In that case,
     for each array element, the variables are assigned from successive
     columns of the composite value.
    </para>

    <para>
     With a positive <literal>SLICE</literal> value, <literal>FOREACH</literal>
     iterates through slices of the array rather than single elements.
     The <literal>SLICE</literal> value must be an integer constant not larger
     than the number of dimensions of the array.  The
     <replaceable>target</replaceable> variable must be an array,
     and it receives successive slices of the array value, where each slice
     is of the number of dimensions specified by <literal>SLICE</literal>.
     Here is an example of iterating through one-dimensional slices:

<programlisting>
CREATE FUNCTION scan_rows(int[]) RETURNS void AS $$
DECLARE
  x int[];
BEGIN
  FOREACH x SLICE 1 IN ARRAY $1
  LOOP
    RAISE NOTICE 'row = %', x;
  END LOOP;
END;
$$ LANGUAGE plpgsql;

SELECT scan_rows(ARRAY[[1,2,3],[4,5,6],[7,8,9],[10,11,12]]);

NOTICE:  row = {1,2,3}
NOTICE:  row = {4,5,6}
NOTICE:  row = {7,8,9}
NOTICE:  row = {10,11,12}
</programlisting>
    </para>
   </sect2>

   <sect2 id="plpgsql-error-trapping">
    <title>Trapping Errors</title>

    <indexterm>
     <primary>exceptions</primary>
     <secondary>in PL/pgSQL</secondary>
    </indexterm>

    <para>
     By default, any error occurring in a <application>PL/pgSQL</application>
     function aborts execution of the function and the
     surrounding transaction.  You can trap errors and recover
     from them by using a <command>BEGIN</command> block with an
     <literal>EXCEPTION</literal> clause.  The syntax is an extension of the
     normal syntax for a <command>BEGIN</command> block:

<synopsis>
<optional> &lt;&lt;<replaceable>label</replaceable>&gt;&gt; </optional>
<optional> DECLARE
    <replaceable>declarations</replaceable> </optional>
BEGIN
    <replaceable>statements</replaceable>
EXCEPTION
    WHEN <replaceable>condition</replaceable> <optional> OR <replaceable>condition</replaceable> ... </optional> THEN
        <replaceable>handler_statements</replaceable>
    <optional> WHEN <replaceable>condition</replaceable> <optional> OR <replaceable>condition</replaceable> ... </optional> THEN
          <replaceable>handler_statements</replaceable>
      ... </optional>
END;
</synopsis>
    </para>

    <para>
     If no error occurs, this form of block simply executes all the
     <replaceable>statements</replaceable>, and then control passes
     to the next statement after <literal>END</literal>.  But if an error
     occurs within the <replaceable>statements</replaceable>, further
     processing of the <replaceable>statements</replaceable> is
     abandoned, and control passes to the <literal>EXCEPTION</literal> list.
     The list is searched for the first <replaceable>condition</replaceable>
     matching the error that occurred.  If a match is found, the
     corresponding <replaceable>handler_statements</replaceable> are
     executed, and then control passes to the next statement after
     <literal>END</literal>.  If no match is found, the error propagates out
     as though the <literal>EXCEPTION</literal> clause were not there at all:
     the error can be caught by an enclosing block with
     <literal>EXCEPTION</literal>, or if there is none it aborts processing
     of the function.
    </para>

    <para>
     The <replaceable>condition</replaceable> names can be any of
     those shown in <xref linkend="errcodes-appendix"/>.  A category
     name matches any error within its category.  The special
     condition name <literal>OTHERS</literal> matches every error type except
     <literal>QUERY_CANCELED</literal> and <literal>ASSERT_FAILURE</literal>.
     (It is possible, but often unwise, to trap those two error types
     by name.)  Condition names are
     not case-sensitive.  Also, an error condition can be specified
     by <literal>SQLSTATE</literal> code; for example these are equivalent:
<programlisting>
WHEN division_by_zero THEN ...
WHEN SQLSTATE '22012' THEN ...
</programlisting>
    </para>

    <para>
     If a new error occurs within the selected
     <replaceable>handler_statements</replaceable>, it cannot be caught
     by this <literal>EXCEPTION</literal> clause, but is propagated out.
     A surrounding <literal>EXCEPTION</literal> clause could catch it.
    </para>

    <para>
     When an error is caught by an <literal>EXCEPTION</literal> clause,
     the local variables of the <application>PL/pgSQL</application> function
     remain as they were when the error occurred, but all changes
     to persistent database state within the block are rolled back.
     As an example, consider this fragment:

<programlisting>
INSERT INTO mytab(firstname, lastname) VALUES('Tom', 'Jones');
BEGIN
    UPDATE mytab SET firstname = 'Joe' WHERE lastname = 'Jones';
    x := x + 1;
    y := x / 0;
EXCEPTION
    WHEN division_by_zero THEN
        RAISE NOTICE 'caught division_by_zero';
        RETURN x;
END;
</programlisting>

     When control reaches the assignment to <literal>y</literal>, it will
     fail with a <literal>division_by_zero</literal> error.  This will be caught by
     the <literal>EXCEPTION</literal> clause.  The value returned in the
     <command>RETURN</command> statement will be the incremented value of
     <literal>x</literal>, but the effects of the <command>UPDATE</command> command will
     have been rolled back.  The <command>INSERT</command> command preceding the
     block is not rolled back, however, so the end result is that the database
     contains <literal>Tom Jones</literal> not <literal>Joe Jones</literal>.
    </para>

    <tip>
     <para>
      A block containing an <literal>EXCEPTION</literal> clause is significantly
      more expensive to enter and exit than a block without one.  Therefore,
      don't use <literal>EXCEPTION</literal> without need.
     </para>
    </tip>

    <example id="plpgsql-upsert-example">
    <title>Exceptions with <command>UPDATE</command>/<command>INSERT</command></title>
    <para>

    This example uses exception handling to perform either
    <command>UPDATE</command> or <command>INSERT</command>, as appropriate.  It is
    recommended that applications use <command>INSERT</command> with
    <literal>ON CONFLICT DO UPDATE</literal> rather than actually using
    this pattern.  This example serves primarily to illustrate use of
    <application>PL/pgSQL</application> control flow structures:

<programlisting>
CREATE TABLE db (a INT PRIMARY KEY, b TEXT);

CREATE FUNCTION merge_db(key INT, data TEXT) RETURNS VOID AS
$$
BEGIN
    LOOP
        -- first try to update the key
        UPDATE db SET b = data WHERE a = key;
        IF found THEN
            RETURN;
        END IF;
        -- not there, so try to insert the key
        -- if someone else inserts the same key concurrently,
        -- we could get a unique-key failure
        BEGIN
            INSERT INTO db(a,b) VALUES (key, data);
            RETURN;
        EXCEPTION WHEN unique_violation THEN
            -- Do nothing, and loop to try the UPDATE again.
        END;
    END LOOP;
END;
$$
LANGUAGE plpgsql;

SELECT merge_db(1, 'david');
SELECT merge_db(1, 'dennis');
</programlisting>

     This coding assumes the <literal>unique_violation</literal> error is caused by
     the <command>INSERT</command>, and not by, say, an <command>INSERT</command> in a
     trigger function on the table.  It might also misbehave if there is
     more than one unique index on the table, since it will retry the
     operation regardless of which index caused the error.
     More safety could be had by using the
     features discussed next to check that the trapped error was the one
     expected.
    </para>
    </example>

   <sect3 id="plpgsql-exception-diagnostics">
    <title>Obtaining Information about an Error</title>

    <para>
     Exception handlers frequently need to identify the specific error that
     occurred.  There are two ways to get information about the current
     exception in <application>PL/pgSQL</application>: special variables and the
     <command>GET STACKED DIAGNOSTICS</command> command.
    </para>

    <para>
     Within an exception handler, the special variable
     <varname>SQLSTATE</varname> contains the error code that corresponds to
     the exception that was raised (refer to <xref linkend="errcodes-table"/>
     for a list of possible error codes). The special variable
     <varname>SQLERRM</varname> contains the error message associated with the
     exception. These variables are undefined outside exception handlers.
    </para>

    <para>
     Within an exception handler, one may also retrieve
     information about the current exception by using the
     <command>GET STACKED DIAGNOSTICS</command> command, which has the form:

<synopsis>
GET STACKED DIAGNOSTICS <replaceable>variable</replaceable> { = | := } <replaceable>item</replaceable> <optional> , ... </optional>;
</synopsis>

     Each <replaceable>item</replaceable> is a key word identifying a status
     value to be assigned to the specified <replaceable>variable</replaceable>
     (which should be of the right data type to receive it).  The currently
     available status items are shown
     in <xref linkend="plpgsql-exception-diagnostics-values"/>.
    </para>

     <table id="plpgsql-exception-diagnostics-values">
      <title>Error Diagnostics Items</title>
      <tgroup cols="3">
       <colspec colname="col1" colwidth="2*"/>
       <colspec colname="col2" colwidth="1*"/>
       <colspec colname="col3" colwidth="2*"/>
       <thead>
        <row>
         <entry>Name</entry>
         <entry>Type</entry>
         <entry>Description</entry>
        </row>
       </thead>
       <tbody>
        <row>
         <entry><literal>RETURNED_SQLSTATE</literal></entry>
         <entry><type>text</type></entry>
         <entry>the SQLSTATE error code of the exception</entry>
        </row>
        <row>
         <entry><literal>COLUMN_NAME</literal></entry>
         <entry><type>text</type></entry>
         <entry>the name of the column related to exception</entry>
        </row>
        <row>
         <entry><literal>CONSTRAINT_NAME</literal></entry>
         <entry><type>text</type></entry>
         <entry>the name of the constraint related to exception</entry>
        </row>
        <row>
         <entry><literal>PG_DATATYPE_NAME</literal></entry>
         <entry><type>text</type></entry>
         <entry>the name of the data type related to exception</entry>
        </row>
        <row>
         <entry><literal>MESSAGE_TEXT</literal></entry>
         <entry><type>text</type></entry>
         <entry>the text of the exception's primary message</entry>
        </row>
        <row>
         <entry><literal>TABLE_NAME</literal></entry>
         <entry><type>text</type></entry>
         <entry>the name of the table related to exception</entry>
        </row>
        <row>
         <entry><literal>SCHEMA_NAME</literal></entry>
         <entry><type>text</type></entry>
         <entry>the name of the schema related to exception</entry>
        </row>
        <row>
         <entry><literal>PG_EXCEPTION_DETAIL</literal></entry>
         <entry><type>text</type></entry>
         <entry>the text of the exception's detail message, if any</entry>
        </row>
        <row>
         <entry><literal>PG_EXCEPTION_HINT</literal></entry>
         <entry><type>text</type></entry>
         <entry>the text of the exception's hint message, if any</entry>
        </row>
        <row>
         <entry><literal>PG_EXCEPTION_CONTEXT</literal></entry>
         <entry><type>text</type></entry>
         <entry>line(s) of text describing the call stack at the time of the
          exception (see <xref linkend="plpgsql-call-stack"/>)</entry>
        </row>
       </tbody>
      </tgroup>
     </table>

    <para>
     If the exception did not set a value for an item, an empty string
     will be returned.
    </para>

    <para>
     Here is an example:
<programlisting>
DECLARE
  text_var1 text;
  text_var2 text;
  text_var3 text;
BEGIN
  -- some processing which might cause an exception
  ...
EXCEPTION WHEN OTHERS THEN
  GET STACKED DIAGNOSTICS text_var1 = MESSAGE_TEXT,
                          text_var2 = PG_EXCEPTION_DETAIL,
                          text_var3 = PG_EXCEPTION_HINT;
END;
</programlisting>
    </para>
   </sect3>
  </sect2>

  <sect2 id="plpgsql-call-stack">
   <title>Obtaining Execution Location Information</title>

   <para>
    The <command>GET DIAGNOSTICS</command> command, previously described
    in <xref linkend="plpgsql-statements-diagnostics"/>, retrieves information
    about current execution state (whereas the <command>GET STACKED
    DIAGNOSTICS</command> command discussed above reports information about
    the execution state as of a previous error).  Its <literal>PG_CONTEXT</literal>
    status item is useful for identifying the current execution
    location.  <literal>PG_CONTEXT</literal> returns a text string with line(s)
    of text describing the call stack.  The first line refers to the current
    function and currently executing <command>GET DIAGNOSTICS</command>
    command.  The second and any subsequent lines refer to calling functions
    further up the call stack.  For example:

<programlisting>
CREATE OR REPLACE FUNCTION outer_func() RETURNS integer AS $$
BEGIN
  RETURN inner_func();
END;
$$ LANGUAGE plpgsql;

CREATE OR REPLACE FUNCTION inner_func() RETURNS integer AS $$
DECLARE
  stack text;
BEGIN
  GET DIAGNOSTICS stack = PG_CONTEXT;
  RAISE NOTICE E'--- Call Stack ---\n%', stack;
  RETURN 1;
END;
$$ LANGUAGE plpgsql;

SELECT outer_func();

NOTICE:  --- Call Stack ---
PL/pgSQL function inner_func() line 5 at GET DIAGNOSTICS
PL/pgSQL function outer_func() line 3 at RETURN
CONTEXT:  PL/pgSQL function outer_func() line 3 at RETURN
 outer_func
 ------------
           1
(1 row)
</programlisting>

   </para>

   <para>
    <literal>GET STACKED DIAGNOSTICS ... PG_EXCEPTION_CONTEXT</literal>
    returns the same sort of stack trace, but describing the location
    at which an error was detected, rather than the current location.
   </para>
  </sect2>
  </sect1>

  <sect1 id="plpgsql-cursors">
   <title>Cursors</title>

   <indexterm zone="plpgsql-cursors">
    <primary>cursor</primary>
    <secondary>in PL/pgSQL</secondary>
   </indexterm>

   <para>
    Rather than executing a whole query at once, it is possible to set
    up a <firstterm>cursor</firstterm> that encapsulates the query, and then read
    the query result a few rows at a time. One reason for doing this is
    to avoid memory overrun when the result contains a large number of
    rows. (However, <application>PL/pgSQL</application> users do not normally need
    to worry about that, since <literal>FOR</literal> loops automatically use a cursor
    internally to avoid memory problems.) A more interesting usage is to
    return a reference to a cursor that a function has created, allowing the
    caller to read the rows. This provides an efficient way to return
    large row sets from functions.
   </para>

   <sect2 id="plpgsql-cursor-declarations">
    <title>Declaring Cursor Variables</title>

    <para>
     All access to cursors in <application>PL/pgSQL</application> goes through
     cursor variables, which are always of the special data type
     <type>refcursor</type>.  One way to create a cursor variable
     is just to declare it as a variable of type <type>refcursor</type>.
     Another way is to use the cursor declaration syntax,
     which in general is:
<synopsis>
<replaceable>name</replaceable> <optional> <optional> NO </optional> SCROLL </optional> CURSOR <optional> ( <replaceable>arguments</replaceable> ) </optional> FOR <replaceable>query</replaceable>;
</synopsis>
     (<literal>FOR</literal> can be replaced by <literal>IS</literal> for
     <productname>Oracle</productname> compatibility.)
     If <literal>SCROLL</literal> is specified, the cursor will be capable of
     scrolling backward; if <literal>NO SCROLL</literal> is specified, backward
     fetches will be rejected; if neither specification appears, it is
     query-dependent whether backward fetches will be allowed.
     <replaceable>arguments</replaceable>, if specified, is a
     comma-separated list of pairs <literal><replaceable>name</replaceable>
     <replaceable>datatype</replaceable></literal> that define names to be
     replaced by parameter values in the given query.  The actual
     values to substitute for these names will be specified later,
     when the cursor is opened.
    </para>
    <para>
     Some examples:
<programlisting>
DECLARE
    curs1 refcursor;
    curs2 CURSOR FOR SELECT * FROM tenk1;
    curs3 CURSOR (key integer) FOR SELECT * FROM tenk1 WHERE unique1 = key;
</programlisting>
     All three of these variables have the data type <type>refcursor</type>,
     but the first can be used with any query, while the second has
     a fully specified query already <firstterm>bound</firstterm> to it, and the last
     has a parameterized query bound to it.  (<literal>key</literal> will be
     replaced by an integer parameter value when the cursor is opened.)
     The variable <literal>curs1</literal>
     is said to be <firstterm>unbound</firstterm> since it is not bound to
     any particular query.
    </para>

    <para>
     The <literal>SCROLL</literal> option cannot be used when the cursor's
     query uses <literal>FOR UPDATE/SHARE</literal>.  Also, it is
     best to use <literal>NO SCROLL</literal> with a query that involves
     volatile functions.  The implementation of <literal>SCROLL</literal>
     assumes that re-reading the query's output will give consistent
     results, which a volatile function might not do.
    </para>
   </sect2>

   <sect2 id="plpgsql-cursor-opening">
    <title>Opening Cursors</title>

    <para>
     Before a cursor can be used to retrieve rows, it must be
     <firstterm>opened</firstterm>. (This is the equivalent action to the SQL
     command <link linkend="sql-declare"><command>DECLARE
     CURSOR</command></link>.)
     <application>PL/pgSQL</application> has
     three forms of the <command>OPEN</command> statement, two of which use unbound
     cursor variables while the third uses a bound cursor variable.
    </para>

    <note>
     <para>
      Bound cursor variables can also be used without explicitly opening the cursor,
      via the <command>FOR</command> statement described in
      <xref linkend="plpgsql-cursor-for-loop"/>.
      A <command>FOR</command> loop will open the cursor and then
      close it again when the loop completes.
     </para>
    </note>

    <indexterm>
     <primary>portal</primary>
     <secondary>in PL/pgSQL</secondary>
    </indexterm>

    <para>
     Opening a cursor involves creating a server-internal data structure
     called a <firstterm>portal</firstterm>, which holds the execution
     state for the cursor's query.  A portal has a name, which must be
     unique within the session for the duration of the portal's existence.
     By default, <application>PL/pgSQL</application> will assign a unique
     name to each portal it creates.  However, if you assign a non-null
     string value to a cursor variable, that string will be used as its
     portal name.  This feature can be used as described in
     <xref linkend="plpgsql-cursor-returning"/>.
    </para>

    <sect3 id="plpgsql-cursor-opening-open-for-query">
     <title><command>OPEN FOR</command> <replaceable>query</replaceable></title>

<synopsis>
OPEN <replaceable>unbound_cursorvar</replaceable> <optional> <optional> NO </optional> SCROLL </optional> FOR <replaceable>query</replaceable>;
</synopsis>

       <para>
        The cursor variable is opened and given the specified query to
        execute.  The cursor cannot be open already, and it must have been
        declared as an unbound cursor variable (that is, as a simple
        <type>refcursor</type> variable).  The query must be a
        <command>SELECT</command>, or something else that returns rows
        (such as <command>EXPLAIN</command>).  The query
        is treated in the same way as other SQL commands in
        <application>PL/pgSQL</application>: <application>PL/pgSQL</application>
        variable names are substituted, and the query plan is cached for
        possible reuse.  When a <application>PL/pgSQL</application>
        variable is substituted into the cursor query, the value that is
        substituted is the one it has at the time of the <command>OPEN</command>;
        subsequent changes to the variable will not affect the cursor's
        behavior.
        The <literal>SCROLL</literal> and <literal>NO SCROLL</literal>
        options have the same meanings as for a bound cursor.
       </para>

       <para>
        An example:
<programlisting>
OPEN curs1 FOR SELECT * FROM foo WHERE key = mykey;
</programlisting>
       </para>
     </sect3>

    <sect3 id="plpgsql-cursor-opening-open-for-execute">
     <title><command>OPEN FOR EXECUTE</command></title>

<synopsis>
OPEN <replaceable>unbound_cursorvar</replaceable> <optional> <optional> NO </optional> SCROLL </optional> FOR EXECUTE <replaceable class="command">query_string</replaceable>
                                     <optional> USING <replaceable>expression</replaceable> <optional>, ... </optional> </optional>;
</synopsis>

         <para>
          The cursor variable is opened and given the specified query to
          execute.  The cursor cannot be open already, and it must have been
          declared as an unbound cursor variable (that is, as a simple
          <type>refcursor</type> variable).  The query is specified as a string
          expression, in the same way as in the <command>EXECUTE</command>
          command.  As usual, this gives flexibility so the query plan can vary
          from one run to the next (see <xref linkend="plpgsql-plan-caching"/>),
          and it also means that variable substitution is not done on the
          command string. As with <command>EXECUTE</command>, parameter values
          can be inserted into the dynamic command via
          <literal>format()</literal> and <literal>USING</literal>.
          The <literal>SCROLL</literal> and
          <literal>NO SCROLL</literal> options have the same meanings as for a bound
          cursor.
         </para>

       <para>
        An example:
<programlisting>
OPEN curs1 FOR EXECUTE format('SELECT * FROM %I WHERE col1 = $1',tabname) USING keyvalue;
</programlisting>
        In this example, the table name is inserted into the query via
        <function>format()</function>.  The comparison value for <literal>col1</literal>
        is inserted via a <literal>USING</literal> parameter, so it needs
        no quoting.
       </para>
     </sect3>

    <sect3 id="plpgsql-open-bound-cursor">
     <title>Opening a Bound Cursor</title>

<synopsis>
OPEN <replaceable>bound_cursorvar</replaceable> <optional> ( <optional> <replaceable>argument_name</replaceable> { := | =&gt; } </optional> <replaceable>argument_value</replaceable> <optional>, ...</optional> ) </optional>;
</synopsis>

         <para>
          This form of <command>OPEN</command> is used to open a cursor
          variable whose query was bound to it when it was declared.  The
          cursor cannot be open already.  A list of actual argument value
          expressions must appear if and only if the cursor was declared to
          take arguments.  These values will be substituted in the query.
         </para>

         <para>
          The query plan for a bound cursor is always considered cacheable;
          there is no equivalent of <command>EXECUTE</command> in this case.
          Notice that <literal>SCROLL</literal> and <literal>NO SCROLL</literal> cannot be
          specified in <command>OPEN</command>, as the cursor's scrolling
          behavior was already determined.
         </para>

         <para>
          Argument values can be passed using either <firstterm>positional</firstterm>
          or <firstterm>named</firstterm> notation.  In positional
          notation, all arguments are specified in order.  In named notation,
          each argument's name is specified using <literal>:=</literal>
          or <literal>=&gt;</literal> to
          separate it from the argument expression. Similar to calling
          functions, described in <xref linkend="sql-syntax-calling-funcs"/>, it
          is also allowed to mix positional and named notation.
         </para>

         <para>
          Examples (these use the cursor declaration examples above):
<programlisting>
OPEN curs2;
OPEN curs3(42);
OPEN curs3(key := 42);
OPEN curs3(key =&gt; 42);
</programlisting>
         </para>

         <para>
          Because variable substitution is done on a bound cursor's query,
          there are really two ways to pass values into the cursor: either
          with an explicit argument to <command>OPEN</command>, or implicitly by
          referencing a <application>PL/pgSQL</application> variable in the query.
          However, only variables declared before the bound cursor was
          declared will be substituted into it.  In either case the value to
          be passed is determined at the time of the <command>OPEN</command>.
          For example, another way to get the same effect as the
          <literal>curs3</literal> example above is
<programlisting>
DECLARE
    key integer;
    curs4 CURSOR FOR SELECT * FROM tenk1 WHERE unique1 = key;
BEGIN
    key := 42;
    OPEN curs4;
</programlisting>
         </para>
     </sect3>
   </sect2>

   <sect2 id="plpgsql-cursor-using">
    <title>Using Cursors</title>

    <para>
     Once a cursor has been opened, it can be manipulated with the
     statements described here.
    </para>

    <para>
     These manipulations need not occur in the same function that
     opened the cursor to begin with.  You can return a <type>refcursor</type>
     value out of a function and let the caller operate on the cursor.
     (Internally, a <type>refcursor</type> value is simply the string name
     of the portal containing the active query for the cursor.  This name
     can be passed around, assigned to other <type>refcursor</type> variables,
     and so on, without disturbing the portal.)
    </para>

    <para>
     All portals are implicitly closed at transaction end.  Therefore
     a <type>refcursor</type> value is usable to reference an open cursor
     only until the end of the transaction.
    </para>

    <sect3 id="plpgsql-cursor-using-fetch">
     <title><literal>FETCH</literal></title>

<synopsis>
FETCH <optional> <replaceable>direction</replaceable> { FROM | IN } </optional> <replaceable>cursor</replaceable> INTO <replaceable>target</replaceable>;
</synopsis>

    <para>
     <command>FETCH</command> retrieves the next row (in the indicated
     direction) from the
     cursor into a target, which might be a row variable, a record
     variable, or a comma-separated list of simple variables, just like
     <command>SELECT INTO</command>.  If there is no suitable row, the
     target is set to NULL(s).  As with <command>SELECT
     INTO</command>, the special variable <literal>FOUND</literal> can
     be checked to see whether a row was obtained or not.  If no row is
     obtained, the cursor is positioned after the last row or before the
     first row, depending on the movement direction.
    </para>

    <para>
     The <replaceable>direction</replaceable> clause can be any of the
     variants allowed in the SQL <xref linkend="sql-fetch"/>
     command except the ones that can fetch
     more than one row; namely, it can be
     <literal>NEXT</literal>,
     <literal>PRIOR</literal>,
     <literal>FIRST</literal>,
     <literal>LAST</literal>,
     <literal>ABSOLUTE</literal> <replaceable>count</replaceable>,
     <literal>RELATIVE</literal> <replaceable>count</replaceable>,
     <literal>FORWARD</literal>, or
     <literal>BACKWARD</literal>.
     Omitting <replaceable>direction</replaceable> is the same
     as specifying <literal>NEXT</literal>.
     In the forms using a <replaceable>count</replaceable>,
     the <replaceable>count</replaceable> can be any integer-valued
     expression (unlike the SQL <command>FETCH</command> command,
     which only allows an integer constant).
     <replaceable>direction</replaceable> values that require moving
     backward are likely to fail unless the cursor was declared or opened
     with the <literal>SCROLL</literal> option.
    </para>

    <para>
     <replaceable>cursor</replaceable> must be the name of a <type>refcursor</type>
     variable that references an open cursor portal.
    </para>

    <para>
     Examples:
<programlisting>
FETCH curs1 INTO rowvar;
FETCH curs2 INTO foo, bar, baz;
FETCH LAST FROM curs3 INTO x, y;
FETCH RELATIVE -2 FROM curs4 INTO x;
</programlisting>
       </para>
     </sect3>

    <sect3 id="plpgsql-cursor-using-move">
     <title><literal>MOVE</literal></title>

<synopsis>
MOVE <optional> <replaceable>direction</replaceable> { FROM | IN } </optional> <replaceable>cursor</replaceable>;
</synopsis>

    <para>
     <command>MOVE</command> repositions a cursor without retrieving
     any data. <command>MOVE</command> works like the
     <command>FETCH</command> command, except it only repositions the
     cursor and does not return the row moved to.
     The <replaceable>direction</replaceable> clause can be any of the
     variants allowed in the SQL <xref linkend="sql-fetch"/>
     command, including those that can fetch more than one row;
     the cursor is positioned to the last such row.
     (However, the case in which the <replaceable>direction</replaceable>
     clause is simply a <replaceable>count</replaceable> expression with
     no key word is deprecated in <application>PL/pgSQL</application>.
     That syntax is ambiguous with the case where
     the <replaceable>direction</replaceable> clause is omitted
     altogether, and hence it may fail if
     the <replaceable>count</replaceable> is not a constant.)
     As with <command>SELECT
     INTO</command>, the special variable <literal>FOUND</literal> can
     be checked to see whether there was a row to move to.  If there is no
     such row, the cursor is positioned after the last row or before the
     first row, depending on the movement direction.
    </para>

    <para>
     Examples:
<programlisting>
MOVE curs1;
MOVE LAST FROM curs3;
MOVE RELATIVE -2 FROM curs4;
MOVE FORWARD 2 FROM curs4;
</programlisting>
       </para>
     </sect3>

    <sect3 id="plpgsql-cursor-using-update-delete">
     <title><literal>UPDATE/DELETE WHERE CURRENT OF</literal></title>

<synopsis>
UPDATE <replaceable>table</replaceable> SET ... WHERE CURRENT OF <replaceable>cursor</replaceable>;
DELETE FROM <replaceable>table</replaceable> WHERE CURRENT OF <replaceable>cursor</replaceable>;
</synopsis>

       <para>
        When a cursor is positioned on a table row, that row can be updated
        or deleted using the cursor to identify the row.  There are
        restrictions on what the cursor's query can be (in particular,
        no grouping) and it's best to use <literal>FOR UPDATE</literal> in the
        cursor.  For more information see the
        <xref linkend="sql-declare"/>
        reference page.
       </para>

       <para>
        An example:
<programlisting>
UPDATE foo SET dataval = myval WHERE CURRENT OF curs1;
</programlisting>
       </para>
     </sect3>

    <sect3 id="plpgsql-cursor-using-close">
     <title><literal>CLOSE</literal></title>

<synopsis>
CLOSE <replaceable>cursor</replaceable>;
</synopsis>

       <para>
        <command>CLOSE</command> closes the portal underlying an open
        cursor.  This can be used to release resources earlier than end of
        transaction, or to free up the cursor variable to be opened again.
       </para>

       <para>
        An example:
<programlisting>
CLOSE curs1;
</programlisting>
       </para>
     </sect3>

    <sect3 id="plpgsql-cursor-returning">
     <title>Returning Cursors</title>

       <para>
        <application>PL/pgSQL</application> functions can return cursors to the
        caller. This is useful to return multiple rows or columns,
        especially with very large result sets.  To do this, the function
        opens the cursor and returns the cursor name to the caller (or simply
        opens the cursor using a portal name specified by or otherwise known
        to the caller).  The caller can then fetch rows from the cursor. The
        cursor can be closed by the caller, or it will be closed automatically
        when the transaction closes.
       </para>

       <para>
        The portal name used for a cursor can be specified by the
        programmer or automatically generated.  To specify a portal name,
        simply assign a string to the <type>refcursor</type> variable before
        opening it.  The string value of the <type>refcursor</type> variable
        will be used by <command>OPEN</command> as the name of the underlying portal.
        However, if the <type>refcursor</type> variable's value is null
        (as it will be by default), then
        <command>OPEN</command> automatically generates a name that does not
        conflict with any existing portal, and assigns it to the
        <type>refcursor</type> variable.
       </para>

       <note>
        <para>
         Prior to <productname>PostgreSQL</productname> 16, bound cursor
         variables were initialized to contain their own names, rather
         than being left as null, so that the underlying portal name would
         be the same as the cursor variable's name by default.  This was
         changed because it created too much risk of conflicts between
         similarly-named cursors in different functions.
        </para>
       </note>

       <para>
        The following example shows one way a cursor name can be supplied by
        the caller:

<programlisting>
CREATE TABLE test (col text);
INSERT INTO test VALUES ('123');

CREATE FUNCTION reffunc(refcursor) RETURNS refcursor AS '
BEGIN
    OPEN $1 FOR SELECT col FROM test;
    RETURN $1;
END;
' LANGUAGE plpgsql;

BEGIN;
SELECT reffunc('funccursor');
FETCH ALL IN funccursor;
COMMIT;
</programlisting>
       </para>

       <para>
        The following example uses automatic cursor name generation:

<programlisting>
CREATE FUNCTION reffunc2() RETURNS refcursor AS '
DECLARE
    ref refcursor;
BEGIN
    OPEN ref FOR SELECT col FROM test;
    RETURN ref;
END;
' LANGUAGE plpgsql;

-- need to be in a transaction to use cursors.
BEGIN;
SELECT reffunc2();

      reffunc2
--------------------
 &lt;unnamed cursor 1&gt;
(1 row)

FETCH ALL IN "&lt;unnamed cursor 1&gt;";
COMMIT;
</programlisting>
       </para>

       <para>
        The following example shows one way to return multiple cursors
        from a single function:

<programlisting>
CREATE FUNCTION myfunc(refcursor, refcursor) RETURNS SETOF refcursor AS $$
BEGIN
    OPEN $1 FOR SELECT * FROM table_1;
    RETURN NEXT $1;
    OPEN $2 FOR SELECT * FROM table_2;
    RETURN NEXT $2;
END;
$$ LANGUAGE plpgsql;

-- need to be in a transaction to use cursors.
BEGIN;

SELECT * FROM myfunc('a', 'b');

FETCH ALL FROM a;
FETCH ALL FROM b;
COMMIT;
</programlisting>
       </para>
     </sect3>
   </sect2>

   <sect2 id="plpgsql-cursor-for-loop">
    <title>Looping through a Cursor's Result</title>

    <para>
     There is a variant of the <command>FOR</command> statement that allows
     iterating through the rows returned by a cursor.  The syntax is:

<synopsis>
<optional> &lt;&lt;<replaceable>label</replaceable>&gt;&gt; </optional>
FOR <replaceable>recordvar</replaceable> IN <replaceable>bound_cursorvar</replaceable> <optional> ( <optional> <replaceable>argument_name</replaceable> { := | =&gt; } </optional> <replaceable>argument_value</replaceable> <optional>, ...</optional> ) </optional> LOOP
    <replaceable>statements</replaceable>
END LOOP <optional> <replaceable>label</replaceable> </optional>;
</synopsis>

     The cursor variable must have been bound to some query when it was
     declared, and it <emphasis>cannot</emphasis> be open already.  The
     <command>FOR</command> statement automatically opens the cursor, and it closes
     the cursor again when the loop exits.  A list of actual argument value
     expressions must appear if and only if the cursor was declared to take
     arguments.  These values will be substituted in the query, in just
     the same way as during an <command>OPEN</command> (see <xref
     linkend="plpgsql-open-bound-cursor"/>).
   </para>

   <para>
     The variable <replaceable>recordvar</replaceable> is automatically
     defined as type <type>record</type> and exists only inside the loop (any
     existing definition of the variable name is ignored within the loop).
     Each row returned by the cursor is successively assigned to this
     record variable and the loop body is executed.
    </para>
   </sect2>

  </sect1>

  <sect1 id="plpgsql-transactions">
   <title>Transaction Management</title>

   <para>
    In procedures invoked by the <command>CALL</command> command
    as well as in anonymous code blocks (<command>DO</command> command),
    it is possible to end transactions using the
    commands <command>COMMIT</command> and <command>ROLLBACK</command>.  A new
    transaction is started automatically after a transaction is ended using
    these commands, so there is no separate <command>START
    TRANSACTION</command> command.  (Note that <command>BEGIN</command> and
    <command>END</command> have different meanings in PL/pgSQL.)
   </para>

   <para>
    Here is a simple example:
<programlisting>
CREATE PROCEDURE transaction_test1()
LANGUAGE plpgsql
AS $$
BEGIN
    FOR i IN 0..9 LOOP
        INSERT INTO test1 (a) VALUES (i);
        IF i % 2 = 0 THEN
            COMMIT;
        ELSE
            ROLLBACK;
        END IF;
    END LOOP;
END;
$$;

CALL transaction_test1();
</programlisting>
   </para>

   <indexterm zone="plpgsql-transaction-chain">
    <primary>chained transactions</primary>
    <secondary>in PL/pgSQL</secondary>
   </indexterm>

   <para id="plpgsql-transaction-chain">
    A new transaction starts out with default transaction characteristics such
    as transaction isolation level.  In cases where transactions are committed
    in a loop, it might be desirable to start new transactions automatically
    with the same characteristics as the previous one.  The commands
    <command>COMMIT AND CHAIN</command> and <command>ROLLBACK AND
    CHAIN</command> accomplish this.
   </para>

   <para>
    Transaction control is only possible in <command>CALL</command> or
    <command>DO</command> invocations from the top level or nested
    <command>CALL</command> or <command>DO</command> invocations without any
    other intervening command.  For example, if the call stack is
    <command>CALL proc1()</command> &rarr; <command>CALL proc2()</command>
    &rarr; <command>CALL proc3()</command>, then the second and third
    procedures can perform transaction control actions.  But if the call stack
    is <command>CALL proc1()</command> &rarr; <command>SELECT
    func2()</command> &rarr; <command>CALL proc3()</command>, then the last
    procedure cannot do transaction control, because of the
    <command>SELECT</command> in between.
   </para>

   <para>
    <application>PL/pgSQL</application> does not support savepoints
    (<command>SAVEPOINT</command>/<command>ROLLBACK TO
    SAVEPOINT</command>/<command>RELEASE SAVEPOINT</command> commands).
    Typical usage patterns for savepoints can be replaced by blocks with
    exception handlers (see <xref linkend="plpgsql-error-trapping"/>).
    Under the hood, a block with exception handlers forms a
    subtransaction, which means that transactions cannot be ended inside
    such a block.
   </para>

   <para>
    Special considerations apply to cursor loops.  Consider this example:
<programlisting>
CREATE PROCEDURE transaction_test2()
LANGUAGE plpgsql
AS $$
DECLARE
    r RECORD;
BEGIN
    FOR r IN SELECT * FROM test2 ORDER BY x LOOP
        INSERT INTO test1 (a) VALUES (r.x);
        COMMIT;
    END LOOP;
END;
$$;

CALL transaction_test2();
</programlisting>
    Normally, cursors are automatically closed at transaction commit.
    However, a cursor created as part of a loop like this is automatically
    converted to a holdable cursor by the first <command>COMMIT</command> or
    <command>ROLLBACK</command>.  That means that the cursor is fully
    evaluated at the first <command>COMMIT</command> or
    <command>ROLLBACK</command> rather than row by row.  The cursor is still
    removed automatically after the loop, so this is mostly invisible to the
    user.  But one must keep in mind that any table or row locks taken by
    the cursor's query will no longer be held after the
    first <command>COMMIT</command> or
    <command>ROLLBACK</command>.
   </para>

   <para>
    Transaction commands are not allowed in cursor loops driven by commands
    that are not read-only (for example <command>UPDATE
    ... RETURNING</command>).
   </para>
  </sect1>

  <sect1 id="plpgsql-errors-and-messages">
   <title>Errors and Messages</title>

  <sect2 id="plpgsql-statements-raise">
   <title>Reporting Errors and Messages</title>

   <indexterm>
    <primary>RAISE</primary>
    <secondary>in PL/pgSQL</secondary>
   </indexterm>

   <indexterm>
    <primary>reporting errors</primary>
    <secondary>in PL/pgSQL</secondary>
   </indexterm>

   <para>
    Use the <command>RAISE</command> statement to report messages and
    raise errors.

<synopsis>
RAISE <optional> <replaceable class="parameter">level</replaceable> </optional> '<replaceable class="parameter">format</replaceable>' <optional>, <replaceable class="parameter">expression</replaceable> <optional>, ... </optional></optional> <optional> USING <replaceable class="parameter">option</replaceable> { = | := } <replaceable class="parameter">expression</replaceable> <optional>, ... </optional> </optional>;
RAISE <optional> <replaceable class="parameter">level</replaceable> </optional> <replaceable class="parameter">condition_name</replaceable> <optional> USING <replaceable class="parameter">option</replaceable> { = | := } <replaceable class="parameter">expression</replaceable> <optional>, ... </optional> </optional>;
RAISE <optional> <replaceable class="parameter">level</replaceable> </optional> SQLSTATE '<replaceable class="parameter">sqlstate</replaceable>' <optional> USING <replaceable class="parameter">option</replaceable> { = | := } <replaceable class="parameter">expression</replaceable> <optional>, ... </optional> </optional>;
RAISE <optional> <replaceable class="parameter">level</replaceable> </optional> USING <replaceable class="parameter">option</replaceable> { = | := } <replaceable class="parameter">expression</replaceable> <optional>, ... </optional>;
RAISE ;
</synopsis>

    The <replaceable class="parameter">level</replaceable> option specifies
    the error severity.  Allowed levels are <literal>DEBUG</literal>,
    <literal>LOG</literal>, <literal>INFO</literal>,
    <literal>NOTICE</literal>, <literal>WARNING</literal>,
    and <literal>EXCEPTION</literal>, with <literal>EXCEPTION</literal>
    being the default.
    <literal>EXCEPTION</literal> raises an error (which normally aborts the
    current transaction); the other levels only generate messages of different
    priority levels.
    Whether messages of a particular priority are reported to the client,
    written to the server log, or both is controlled by the
    <xref linkend="guc-log-min-messages"/> and
    <xref linkend="guc-client-min-messages"/> configuration
    variables. See <xref linkend="runtime-config"/> for more
    information.
   </para>

   <para>
    In the first syntax variant,
    after the <replaceable class="parameter">level</replaceable> if any,
    write a <replaceable class="parameter">format</replaceable> string
    (which must be a simple string literal, not an expression).  The
    format string specifies the error message text to be reported.
    The format string can be followed
    by optional argument expressions to be inserted into the message.
    Inside the format string, <literal>%</literal> is replaced by the
    string representation of the next optional argument's value. Write
    <literal>%%</literal> to emit a literal <literal>%</literal>.
    The number of arguments must match the number of <literal>%</literal>
    placeholders in the format string, or an error is raised during
    the compilation of the function.
   </para>

   <para>
    In this example, the value of <literal>v_job_id</literal> will replace the
    <literal>%</literal> in the string:
<programlisting>
RAISE NOTICE 'Calling cs_create_job(%)', v_job_id;
</programlisting>
   </para>

   <para>
    In the second and third syntax variants,
    <replaceable class="parameter">condition_name</replaceable> and
    <replaceable class="parameter">sqlstate</replaceable> specify an
    error condition name or a five-character SQLSTATE code, respectively.
    See <xref linkend="errcodes-appendix"/> for the valid error condition
    names and the predefined SQLSTATE codes.
   </para>

   <para>
    Here are examples
    of <replaceable class="parameter">condition_name</replaceable>
    and <replaceable class="parameter">sqlstate</replaceable> usage:
<programlisting>
RAISE division_by_zero;
RAISE WARNING SQLSTATE '22012';
</programlisting>
   </para>

   <para>
    In any of these syntax variants,
    you can attach additional information to the error report by writing
    <literal>USING</literal> followed by <replaceable
    class="parameter">option</replaceable> = <replaceable
    class="parameter">expression</replaceable> items.  Each
    <replaceable class="parameter">expression</replaceable> can be any
    string-valued expression.  The allowed <replaceable
    class="parameter">option</replaceable> key words are:

    <variablelist id="raise-using-options">
     <varlistentry id="raise-using-option-message">
      <term><literal>MESSAGE</literal></term>
      <listitem>
       <para>Sets the error message text.  This option can't be used in the
        first syntax variant, since the message is already supplied.</para>
      </listitem>
     </varlistentry>

     <varlistentry id="raise-using-option-detail">
      <term><literal>DETAIL</literal></term>
      <listitem>
       <para>Supplies an error detail message.</para>
      </listitem>
     </varlistentry>

     <varlistentry id="raise-using-option-hint">
      <term><literal>HINT</literal></term>
      <listitem>
       <para>Supplies a hint message.</para>
      </listitem>
     </varlistentry>

     <varlistentry id="raise-using-option-errcode">
      <term><literal>ERRCODE</literal></term>
      <listitem>
       <para>Specifies the error code (SQLSTATE) to report, either by condition
        name, as shown in <xref linkend="errcodes-appendix"/>, or directly as a
        five-character SQLSTATE code.  This option can't be used in the
        second or third syntax variant, since the error code is already
        supplied.</para>
      </listitem>
     </varlistentry>

     <varlistentry id="raise-using-option-column">
      <term><literal>COLUMN</literal></term>
      <term><literal>CONSTRAINT</literal></term>
      <term><literal>DATATYPE</literal></term>
      <term><literal>TABLE</literal></term>
      <term><literal>SCHEMA</literal></term>
      <listitem>
       <para>Supplies the name of a related object.</para>
      </listitem>
     </varlistentry>
    </variablelist>
   </para>

   <para>
    This example will abort the transaction with the given error message
    and hint:
<programlisting>
RAISE EXCEPTION 'Nonexistent ID --> %', user_id
      USING HINT = 'Please check your user ID';
</programlisting>
   </para>

   <para>
    These two examples show equivalent ways of setting the SQLSTATE:
<programlisting>
RAISE 'Duplicate user ID: %', user_id USING ERRCODE = 'unique_violation';
RAISE 'Duplicate user ID: %', user_id USING ERRCODE = '23505';
</programlisting>
    Another way to produce the same result is:
<programlisting>
RAISE unique_violation USING MESSAGE = 'Duplicate user ID: ' || user_id;
</programlisting>
   </para>

   <para>
    As shown in the fourth syntax variant, it is also possible to
    write <literal>RAISE USING</literal> or <literal>RAISE
    <replaceable class="parameter">level</replaceable> USING</literal> and put
    everything else into the <literal>USING</literal> list.
   </para>

   <para>
    The last variant of <command>RAISE</command> has no parameters at all.
    This form can only be used inside a <literal>BEGIN</literal> block's
    <literal>EXCEPTION</literal> clause;
    it causes the error currently being handled to be re-thrown.
   </para>

   <note>
    <para>
     Before <productname>PostgreSQL</productname> 9.1, <command>RAISE</command> without
     parameters was interpreted as re-throwing the error from the block
     containing the active exception handler.  Thus an <literal>EXCEPTION</literal>
     clause nested within that handler could not catch it, even if the
     <command>RAISE</command> was within the nested <literal>EXCEPTION</literal> clause's
     block. This was deemed surprising as well as being incompatible with
     Oracle's PL/SQL.
    </para>
   </note>

   <para>
    If no condition name nor SQLSTATE is specified in a
    <command>RAISE EXCEPTION</command> command, the default is to use
    <literal>raise_exception</literal> (<literal>P0001</literal>).
    If no message text is specified, the default is to use the condition
    name or SQLSTATE as message text.
   </para>

   <note>
    <para>
     When specifying an error code by SQLSTATE code, you are not
     limited to the predefined error codes, but can select any
     error code consisting of five digits and/or upper-case ASCII
     letters, other than <literal>00000</literal>.  It is recommended that
     you avoid throwing error codes that end in three zeroes, because
     these are category codes and can only be trapped by trapping
     the whole category.
    </para>
   </note>

  </sect2>

  <sect2 id="plpgsql-statements-assert">
   <title>Checking Assertions</title>

   <indexterm>
    <primary>ASSERT</primary>
    <secondary>in PL/pgSQL</secondary>
   </indexterm>

   <indexterm>
    <primary>assertions</primary>
    <secondary>in PL/pgSQL</secondary>
   </indexterm>

   <indexterm>
    <primary><varname>plpgsql.check_asserts</varname> configuration parameter</primary>
   </indexterm>

   <para>
    The <command>ASSERT</command> statement is a convenient shorthand for
    inserting debugging checks into <application>PL/pgSQL</application>
    functions.

<synopsis>
ASSERT <replaceable class="parameter">condition</replaceable> <optional> , <replaceable class="parameter">message</replaceable> </optional>;
</synopsis>

    The <replaceable class="parameter">condition</replaceable> is a Boolean
    expression that is expected to always evaluate to true; if it does,
    the <command>ASSERT</command> statement does nothing further.  If the
    result is false or null, then an <literal>ASSERT_FAILURE</literal> exception
    is raised.  (If an error occurs while evaluating
    the <replaceable class="parameter">condition</replaceable>, it is
    reported as a normal error.)
   </para>

   <para>
    If the optional <replaceable class="parameter">message</replaceable> is
    provided, it is an expression whose result (if not null) replaces the
    default error message text <quote>assertion failed</quote>, should
    the <replaceable class="parameter">condition</replaceable> fail.
    The <replaceable class="parameter">message</replaceable> expression is
    not evaluated in the normal case where the assertion succeeds.
   </para>

   <para>
    Testing of assertions can be enabled or disabled via the configuration
    parameter <literal>plpgsql.check_asserts</literal>, which takes a Boolean
    value; the default is <literal>on</literal>.  If this parameter
    is <literal>off</literal> then <command>ASSERT</command> statements do nothing.
   </para>

   <para>
    Note that <command>ASSERT</command> is meant for detecting program
    bugs, not for reporting ordinary error conditions.  Use
    the <command>RAISE</command> statement, described above, for that.
   </para>

  </sect2>

 </sect1>

 <sect1 id="plpgsql-trigger">
  <title>Trigger Functions</title>

  <indexterm zone="plpgsql-trigger">
   <primary>trigger</primary>
   <secondary>in PL/pgSQL</secondary>
  </indexterm>

  <para>
   <application>PL/pgSQL</application> can be used to define trigger
   functions on data changes or database events.
   A trigger function is created with the <command>CREATE FUNCTION</command>
   command, declaring it as a function with no arguments and a return type of
   <type>trigger</type> (for data change triggers) or
   <type>event_trigger</type> (for database event triggers).
   Special local variables named <varname>TG_<replaceable>something</replaceable></varname> are
   automatically defined to describe the condition that triggered the call.
  </para>

  <sect2 id="plpgsql-dml-trigger">
   <title>Triggers on Data Changes</title>

  <para>
   A <link linkend="triggers">data change trigger</link> is declared as a
   function with no arguments and a return type of <type>trigger</type>.
   Note that the function must be declared with no arguments even if it
   expects to receive some arguments specified in <command>CREATE TRIGGER</command>
   &mdash; such arguments are passed via <varname>TG_ARGV</varname>, as described
   below.
  </para>

  <para>
   When a <application>PL/pgSQL</application> function is called as a
   trigger, several special variables are created automatically in the
   top-level block. They are:

   <variablelist>
    <varlistentry id="plpgsql-dml-trigger-new">
     <term><varname>NEW</varname> <type>record</type></term>
     <listitem>
      <para>
       new database row for <command>INSERT</command>/<command>UPDATE</command> operations in row-level
       triggers. This variable is null in statement-level triggers
       and for <command>DELETE</command> operations.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-old">
     <term><varname>OLD</varname> <type>record</type></term>
     <listitem>
      <para>
       old database row for <command>UPDATE</command>/<command>DELETE</command> operations in row-level
       triggers. This variable is null in statement-level triggers
       and for <command>INSERT</command> operations.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-tg-name">
     <term><varname>TG_NAME</varname> <type>name</type></term>
     <listitem>
      <para>
       name of the trigger which fired.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-tg-when">
     <term><varname>TG_WHEN</varname> <type>text</type></term>
     <listitem>
      <para>
       <literal>BEFORE</literal>, <literal>AFTER</literal>, or
       <literal>INSTEAD OF</literal>, depending on the trigger's definition.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-tg-level">
     <term><varname>TG_LEVEL</varname> <type>text</type></term>
     <listitem>
      <para>
       <literal>ROW</literal> or <literal>STATEMENT</literal>,
       depending on the trigger's definition.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-tg-op">
     <term><varname>TG_OP</varname> <type>text</type></term>
     <listitem>
      <para>
       operation for which the trigger was fired:
       <literal>INSERT</literal>, <literal>UPDATE</literal>,
       <literal>DELETE</literal>, or <literal>TRUNCATE</literal>.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-tg-relid">
     <term><varname>TG_RELID</varname> <type>oid</type> (references <link linkend="catalog-pg-class"><structname>pg_class</structname></link>.<structfield>oid</structfield>)</term>
     <listitem>
      <para>
       object ID of the table that caused the trigger invocation.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-tg-relname">
     <term><varname>TG_RELNAME</varname> <type>name</type></term>
     <listitem>
      <para>
       table that caused the trigger
       invocation. This is now deprecated, and could disappear in a future
       release. Use <literal>TG_TABLE_NAME</literal> instead.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-tg-table-name">
     <term><varname>TG_TABLE_NAME</varname> <type>name</type></term>
     <listitem>
      <para>
       table that caused the trigger invocation.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-tg-table-schema">
     <term><varname>TG_TABLE_SCHEMA</varname> <type>name</type></term>
     <listitem>
      <para>
       schema of the table that caused the trigger invocation.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-tg-nargs">
     <term><varname>TG_NARGS</varname> <type>integer</type></term>
     <listitem>
      <para>
       number of arguments given to the trigger
       function in the <command>CREATE TRIGGER</command> statement.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-dml-trigger-tg-argv">
     <term><varname>TG_ARGV</varname> <type>text[]</type></term>
     <listitem>
      <para>
       arguments from
       the <command>CREATE TRIGGER</command> statement.
       The index counts from 0. Invalid
       indexes (less than 0 or greater than or equal to <varname>tg_nargs</varname>)
       result in a null value.
      </para>
     </listitem>
    </varlistentry>
   </variablelist>
  </para>

   <para>
    A trigger function must return either <symbol>NULL</symbol> or a
    record/row value having exactly the structure of the table the
    trigger was fired for.
   </para>

   <para>
    Row-level triggers fired <literal>BEFORE</literal> can return null to signal the
    trigger manager to skip the rest of the operation for this row
    (i.e., subsequent triggers are not fired, and the
    <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> does not occur
    for this row).  If a nonnull
    value is returned then the operation proceeds with that row value.
    Returning a row value different from the original value
    of <varname>NEW</varname> alters the row that will be inserted or
    updated.  Thus, if the trigger function wants the triggering
    action to succeed normally without altering the row
    value, <varname>NEW</varname> (or a value equal thereto) has to be
    returned.  To alter the row to be stored, it is possible to
    replace single values directly in <varname>NEW</varname> and return the
    modified <varname>NEW</varname>, or to build a complete new record/row to
    return.  In the case of a before-trigger
    on <command>DELETE</command>, the returned value has no direct
    effect, but it has to be nonnull to allow the trigger action to
    proceed.  Note that <varname>NEW</varname> is null
    in <command>DELETE</command> triggers, so returning that is
    usually not sensible.  The usual idiom in <command>DELETE</command>
    triggers is to return <varname>OLD</varname>.
   </para>

   <para>
    <literal>INSTEAD OF</literal> triggers (which are always row-level triggers,
    and may only be used on views) can return null to signal that they did
    not perform any updates, and that the rest of the operation for this
    row should be skipped (i.e., subsequent triggers are not fired, and the
    row is not counted in the rows-affected status for the surrounding
    <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>).
    Otherwise a nonnull value should be returned, to signal
    that the trigger performed the requested operation. For
    <command>INSERT</command> and <command>UPDATE</command> operations, the return value
    should be <varname>NEW</varname>, which the trigger function may modify to
    support <command>INSERT RETURNING</command> and <command>UPDATE RETURNING</command>
    (this will also affect the row value passed to any subsequent triggers,
    or passed to a special <varname>EXCLUDED</varname> alias reference within
    an <command>INSERT</command> statement with an <literal>ON CONFLICT DO
    UPDATE</literal> clause).  For <command>DELETE</command> operations, the return
    value should be <varname>OLD</varname>.
   </para>

   <para>
    The return value of a row-level trigger
    fired <literal>AFTER</literal> or a statement-level trigger
    fired <literal>BEFORE</literal> or <literal>AFTER</literal> is
    always ignored; it might as well be null. However, any of these types of
    triggers might still abort the entire operation by raising an error.
   </para>

   <para>
    <xref linkend="plpgsql-trigger-example"/> shows an example of a
    trigger function in <application>PL/pgSQL</application>.
   </para>

   <example id="plpgsql-trigger-example">
    <title>A <application>PL/pgSQL</application> Trigger Function</title>

    <para>
     This example trigger ensures that any time a row is inserted or updated
     in the table, the current user name and time are stamped into the
     row. And it checks that an employee's name is given and that the
     salary is a positive value.
    </para>

<programlisting>
CREATE TABLE emp (
    empname           text,
    salary            integer,
    last_date         timestamp,
    last_user         text
);

CREATE FUNCTION emp_stamp() RETURNS trigger AS $emp_stamp$
    BEGIN
        -- Check that empname and salary are given
        IF NEW.empname IS NULL THEN
            RAISE EXCEPTION 'empname cannot be null';
        END IF;
        IF NEW.salary IS NULL THEN
            RAISE EXCEPTION '% cannot have null salary', NEW.empname;
        END IF;

        -- Who works for us when they must pay for it?
        IF NEW.salary &lt; 0 THEN
            RAISE EXCEPTION '% cannot have a negative salary', NEW.empname;
        END IF;

        -- Remember who changed the payroll when
        NEW.last_date := current_timestamp;
        NEW.last_user := current_user;
        RETURN NEW;
    END;
$emp_stamp$ LANGUAGE plpgsql;

CREATE TRIGGER emp_stamp BEFORE INSERT OR UPDATE ON emp
    FOR EACH ROW EXECUTE FUNCTION emp_stamp();
</programlisting>
   </example>

   <para>
    Another way to log changes to a table involves creating a new table that
    holds a row for each insert, update, or delete that occurs. This approach
    can be thought of as auditing changes to a table.
    <xref linkend="plpgsql-trigger-audit-example"/> shows an example of an
    audit trigger function in <application>PL/pgSQL</application>.
   </para>

   <example id="plpgsql-trigger-audit-example">
    <title>A <application>PL/pgSQL</application> Trigger Function for Auditing</title>

    <para>
     This example trigger ensures that any insert, update or delete of a row
     in the <literal>emp</literal> table is recorded (i.e., audited) in the <literal>emp_audit</literal> table.
     The current time and user name are stamped into the row, together with
     the type of operation performed on it.
    </para>

<programlisting>
CREATE TABLE emp (
    empname           text NOT NULL,
    salary            integer
);

CREATE TABLE emp_audit(
    operation         char(1)   NOT NULL,
    stamp             timestamp NOT NULL,
    userid            text      NOT NULL,
    empname           text      NOT NULL,
    salary            integer
);

CREATE OR REPLACE FUNCTION process_emp_audit() RETURNS TRIGGER AS $emp_audit$
    BEGIN
        --
        -- Create a row in emp_audit to reflect the operation performed on emp,
        -- making use of the special variable TG_OP to work out the operation.
        --
        IF (TG_OP = 'DELETE') THEN
            INSERT INTO emp_audit SELECT 'D', now(), current_user, OLD.*;
        ELSIF (TG_OP = 'UPDATE') THEN
            INSERT INTO emp_audit SELECT 'U', now(), current_user, NEW.*;
        ELSIF (TG_OP = 'INSERT') THEN
            INSERT INTO emp_audit SELECT 'I', now(), current_user, NEW.*;
        END IF;
        RETURN NULL; -- result is ignored since this is an AFTER trigger
    END;
$emp_audit$ LANGUAGE plpgsql;

CREATE TRIGGER emp_audit
AFTER INSERT OR UPDATE OR DELETE ON emp
    FOR EACH ROW EXECUTE FUNCTION process_emp_audit();
</programlisting>
   </example>

   <para>
    A variation of the previous example uses a view joining the main table
    to the audit table, to show when each entry was last modified. This
    approach still records the full audit trail of changes to the table,
    but also presents a simplified view of the audit trail, showing just
    the last modified timestamp derived from the audit trail for each entry.
    <xref linkend="plpgsql-view-trigger-audit-example"/> shows an example
    of an audit trigger on a view in <application>PL/pgSQL</application>.
   </para>

   <example id="plpgsql-view-trigger-audit-example">
    <title>A <application>PL/pgSQL</application> View Trigger Function for Auditing</title>

    <para>
     This example uses a trigger on the view to make it updatable, and
     ensure that any insert, update or delete of a row in the view is
     recorded (i.e., audited) in the <literal>emp_audit</literal> table. The current time
     and user name are recorded, together with the type of operation
     performed, and the view displays the last modified time of each row.
    </para>

<programlisting>
CREATE TABLE emp (
    empname           text PRIMARY KEY,
    salary            integer
);

CREATE TABLE emp_audit(
    operation         char(1)   NOT NULL,
    userid            text      NOT NULL,
    empname           text      NOT NULL,
    salary            integer,
    stamp             timestamp NOT NULL
);

CREATE VIEW emp_view AS
    SELECT e.empname,
           e.salary,
           max(ea.stamp) AS last_updated
      FROM emp e
      LEFT JOIN emp_audit ea ON ea.empname = e.empname
     GROUP BY 1, 2;

CREATE OR REPLACE FUNCTION update_emp_view() RETURNS TRIGGER AS $$
    BEGIN
        --
        -- Perform the required operation on emp, and create a row in emp_audit
        -- to reflect the change made to emp.
        --
        IF (TG_OP = 'DELETE') THEN
            DELETE FROM emp WHERE empname = OLD.empname;
            IF NOT FOUND THEN RETURN NULL; END IF;

            OLD.last_updated = now();
            INSERT INTO emp_audit VALUES('D', current_user, OLD.*);
            RETURN OLD;
        ELSIF (TG_OP = 'UPDATE') THEN
            UPDATE emp SET salary = NEW.salary WHERE empname = OLD.empname;
            IF NOT FOUND THEN RETURN NULL; END IF;

            NEW.last_updated = now();
            INSERT INTO emp_audit VALUES('U', current_user, NEW.*);
            RETURN NEW;
        ELSIF (TG_OP = 'INSERT') THEN
            INSERT INTO emp VALUES(NEW.empname, NEW.salary);

            NEW.last_updated = now();
            INSERT INTO emp_audit VALUES('I', current_user, NEW.*);
            RETURN NEW;
        END IF;
    END;
$$ LANGUAGE plpgsql;

CREATE TRIGGER emp_audit
INSTEAD OF INSERT OR UPDATE OR DELETE ON emp_view
    FOR EACH ROW EXECUTE FUNCTION update_emp_view();
</programlisting>
   </example>

   <para>
    One use of triggers is to maintain a summary table
    of another table. The resulting summary can be used in place of the
    original table for certain queries &mdash; often with vastly reduced run
    times.
    This technique is commonly used in Data Warehousing, where the tables
    of measured or observed data (called fact tables) might be extremely large.
    <xref linkend="plpgsql-trigger-summary-example"/> shows an example of a
    trigger function in <application>PL/pgSQL</application> that maintains
    a summary table for a fact table in a data warehouse.
   </para>


   <example id="plpgsql-trigger-summary-example">
    <title>A <application>PL/pgSQL</application> Trigger Function for Maintaining a Summary Table</title>

    <para>
     The schema detailed here is partly based on the <emphasis>Grocery Store
     </emphasis> example from <emphasis>The Data Warehouse Toolkit</emphasis>
     by Ralph Kimball.
    </para>

<programlisting>
--
-- Main tables - time dimension and sales fact.
--
CREATE TABLE time_dimension (
    time_key                    integer NOT NULL,
    day_of_week                 integer NOT NULL,
    day_of_month                integer NOT NULL,
    month                       integer NOT NULL,
    quarter                     integer NOT NULL,
    year                        integer NOT NULL
);
CREATE UNIQUE INDEX time_dimension_key ON time_dimension(time_key);

CREATE TABLE sales_fact (
    time_key                    integer NOT NULL,
    product_key                 integer NOT NULL,
    store_key                   integer NOT NULL,
    amount_sold                 numeric(12,2) NOT NULL,
    units_sold                  integer NOT NULL,
    amount_cost                 numeric(12,2) NOT NULL
);
CREATE INDEX sales_fact_time ON sales_fact(time_key);

--
-- Summary table - sales by time.
--
CREATE TABLE sales_summary_bytime (
    time_key                    integer NOT NULL,
    amount_sold                 numeric(15,2) NOT NULL,
    units_sold                  numeric(12) NOT NULL,
    amount_cost                 numeric(15,2) NOT NULL
);
CREATE UNIQUE INDEX sales_summary_bytime_key ON sales_summary_bytime(time_key);

--
-- Function and trigger to amend summarized column(s) on UPDATE, INSERT, DELETE.
--
CREATE OR REPLACE FUNCTION maint_sales_summary_bytime() RETURNS TRIGGER
AS $maint_sales_summary_bytime$
    DECLARE
        delta_time_key          integer;
        delta_amount_sold       numeric(15,2);
        delta_units_sold        numeric(12);
        delta_amount_cost       numeric(15,2);
    BEGIN

        -- Work out the increment/decrement amount(s).
        IF (TG_OP = 'DELETE') THEN

            delta_time_key = OLD.time_key;
            delta_amount_sold = -1 * OLD.amount_sold;
            delta_units_sold = -1 * OLD.units_sold;
            delta_amount_cost = -1 * OLD.amount_cost;

        ELSIF (TG_OP = 'UPDATE') THEN

            -- forbid updates that change the time_key -
            -- (probably not too onerous, as DELETE + INSERT is how most
            -- changes will be made).
            IF ( OLD.time_key != NEW.time_key) THEN
                RAISE EXCEPTION 'Update of time_key : % -&gt; % not allowed',
                                                      OLD.time_key, NEW.time_key;
            END IF;

            delta_time_key = OLD.time_key;
            delta_amount_sold = NEW.amount_sold - OLD.amount_sold;
            delta_units_sold = NEW.units_sold - OLD.units_sold;
            delta_amount_cost = NEW.amount_cost - OLD.amount_cost;

        ELSIF (TG_OP = 'INSERT') THEN

            delta_time_key = NEW.time_key;
            delta_amount_sold = NEW.amount_sold;
            delta_units_sold = NEW.units_sold;
            delta_amount_cost = NEW.amount_cost;

        END IF;


        -- Insert or update the summary row with the new values.
        &lt;&lt;insert_update&gt;&gt;
        LOOP
            UPDATE sales_summary_bytime
                SET amount_sold = amount_sold + delta_amount_sold,
                    units_sold = units_sold + delta_units_sold,
                    amount_cost = amount_cost + delta_amount_cost
                WHERE time_key = delta_time_key;

            EXIT insert_update WHEN found;

            BEGIN
                INSERT INTO sales_summary_bytime (
                            time_key,
                            amount_sold,
                            units_sold,
                            amount_cost)
                    VALUES (
                            delta_time_key,
                            delta_amount_sold,
                            delta_units_sold,
                            delta_amount_cost
                           );

                EXIT insert_update;

            EXCEPTION
                WHEN UNIQUE_VIOLATION THEN
                    -- do nothing
            END;
        END LOOP insert_update;

        RETURN NULL;

    END;
$maint_sales_summary_bytime$ LANGUAGE plpgsql;

CREATE TRIGGER maint_sales_summary_bytime
AFTER INSERT OR UPDATE OR DELETE ON sales_fact
    FOR EACH ROW EXECUTE FUNCTION maint_sales_summary_bytime();

INSERT INTO sales_fact VALUES(1,1,1,10,3,15);
INSERT INTO sales_fact VALUES(1,2,1,20,5,35);
INSERT INTO sales_fact VALUES(2,2,1,40,15,135);
INSERT INTO sales_fact VALUES(2,3,1,10,1,13);
SELECT * FROM sales_summary_bytime;
DELETE FROM sales_fact WHERE product_key = 1;
SELECT * FROM sales_summary_bytime;
UPDATE sales_fact SET units_sold = units_sold * 2;
SELECT * FROM sales_summary_bytime;
</programlisting>
   </example>

   <para>
    <literal>AFTER</literal> triggers can also make use of <firstterm>transition
    tables</firstterm> to inspect the entire set of rows changed by the triggering
    statement.  The <command>CREATE TRIGGER</command> command assigns names to one
    or both transition tables, and then the function can refer to those names
    as though they were read-only temporary tables.
    <xref linkend="plpgsql-trigger-audit-transition-example"/> shows an example.
   </para>

   <example id="plpgsql-trigger-audit-transition-example">
    <title>Auditing with Transition Tables</title>

    <para>
     This example produces the same results as
     <xref linkend="plpgsql-trigger-audit-example"/>, but instead of using a
     trigger that fires for every row, it uses a trigger that fires once
     per statement, after collecting the relevant information in a transition
     table.  This can be significantly faster than the row-trigger approach
     when the invoking statement has modified many rows.  Notice that we must
     make a separate trigger declaration for each kind of event, since the
     <literal>REFERENCING</literal> clauses must be different for each case.  But
     this does not stop us from using a single trigger function if we choose.
     (In practice, it might be better to use three separate functions and
     avoid the run-time tests on <varname>TG_OP</varname>.)
    </para>

<programlisting>
CREATE TABLE emp (
    empname           text NOT NULL,
    salary            integer
);

CREATE TABLE emp_audit(
    operation         char(1)   NOT NULL,
    stamp             timestamp NOT NULL,
    userid            text      NOT NULL,
    empname           text      NOT NULL,
    salary            integer
);

CREATE OR REPLACE FUNCTION process_emp_audit() RETURNS TRIGGER AS $emp_audit$
    BEGIN
        --
        -- Create rows in emp_audit to reflect the operations performed on emp,
        -- making use of the special variable TG_OP to work out the operation.
        --
        IF (TG_OP = 'DELETE') THEN
            INSERT INTO emp_audit
                SELECT 'D', now(), current_user, o.* FROM old_table o;
        ELSIF (TG_OP = 'UPDATE') THEN
            INSERT INTO emp_audit
                SELECT 'U', now(), current_user, n.* FROM new_table n;
        ELSIF (TG_OP = 'INSERT') THEN
            INSERT INTO emp_audit
                SELECT 'I', now(), current_user, n.* FROM new_table n;
        END IF;
        RETURN NULL; -- result is ignored since this is an AFTER trigger
    END;
$emp_audit$ LANGUAGE plpgsql;

CREATE TRIGGER emp_audit_ins
    AFTER INSERT ON emp
    REFERENCING NEW TABLE AS new_table
    FOR EACH STATEMENT EXECUTE FUNCTION process_emp_audit();
CREATE TRIGGER emp_audit_upd
    AFTER UPDATE ON emp
    REFERENCING OLD TABLE AS old_table NEW TABLE AS new_table
    FOR EACH STATEMENT EXECUTE FUNCTION process_emp_audit();
CREATE TRIGGER emp_audit_del
    AFTER DELETE ON emp
    REFERENCING OLD TABLE AS old_table
    FOR EACH STATEMENT EXECUTE FUNCTION process_emp_audit();
</programlisting>
   </example>

</sect2>

  <sect2 id="plpgsql-event-trigger">
   <title>Triggers on Events</title>

   <para>
    <application>PL/pgSQL</application> can be used to define
    <link linkend="event-triggers">event triggers</link>.
    <productname>PostgreSQL</productname> requires that a function that
    is to be called as an event trigger must be declared as a function with
    no arguments and a return type of <literal>event_trigger</literal>.
   </para>

   <para>
    When a <application>PL/pgSQL</application> function is called as an
    event trigger, several special variables are created automatically
    in the top-level block. They are:

   <variablelist>
    <varlistentry id="plpgsql-event-trigger-tg-event">
     <term><varname>TG_EVENT</varname> <type>text</type></term>
     <listitem>
      <para>
       event the trigger is fired for.
      </para>
     </listitem>
    </varlistentry>

    <varlistentry id="plpgsql-event-trigger-tg-tag">
     <term><varname>TG_TAG</varname> <type>text</type></term>
     <listitem>
      <para>
       command tag for which the trigger is fired.
      </para>
     </listitem>
    </varlistentry>
   </variablelist>
  </para>

   <para>
    <xref linkend="plpgsql-event-trigger-example"/> shows an example of an
    event trigger function in <application>PL/pgSQL</application>.
   </para>

   <example id="plpgsql-event-trigger-example">
    <title>A <application>PL/pgSQL</application> Event Trigger Function</title>

    <para>
     This example trigger simply raises a <literal>NOTICE</literal> message
     each time a supported command is executed.
    </para>

<programlisting>
CREATE OR REPLACE FUNCTION snitch() RETURNS event_trigger AS $$
BEGIN
    RAISE NOTICE 'snitch: % %', tg_event, tg_tag;
END;
$$ LANGUAGE plpgsql;

CREATE EVENT TRIGGER snitch ON ddl_command_start EXECUTE FUNCTION snitch();
</programlisting>
   </example>
  </sect2>

  </sect1>

  <sect1 id="plpgsql-implementation">
   <title><application>PL/pgSQL</application> under the Hood</title>

   <para>
    This section discusses some implementation details that are
    frequently important for <application>PL/pgSQL</application> users to know.
   </para>

  <sect2 id="plpgsql-var-subst">
   <title>Variable Substitution</title>

   <para>
    SQL statements and expressions within a <application>PL/pgSQL</application> function
    can refer to variables and parameters of the function.  Behind the scenes,
    <application>PL/pgSQL</application> substitutes query parameters for such references.
    Query parameters will only be substituted in places where they are
    syntactically permissible.  As an extreme case, consider
    this example of poor programming style:
<programlisting>
INSERT INTO foo (foo) VALUES (foo(foo));
</programlisting>
    The first occurrence of <literal>foo</literal> must syntactically be a table
    name, so it will not be substituted, even if the function has a variable
    named <literal>foo</literal>.  The second occurrence must be the name of a
    column of that table, so it will not be substituted either.  Likewise
    the third occurrence must be a function name, so it also will not be
    substituted for.  Only the last occurrence is a candidate to be a
    reference to a variable of the <application>PL/pgSQL</application>
    function.
   </para>

   <para>
    Another way to understand this is that variable substitution can only
    insert data values into an SQL command; it cannot dynamically change which
    database objects are referenced by the command.  (If you want to do
    that, you must build a command string dynamically, as explained in
    <xref linkend="plpgsql-statements-executing-dyn"/>.)
   </para>

   <para>
    Since the names of variables are syntactically no different from the names
    of table columns, there can be ambiguity in statements that also refer to
    tables: is a given name meant to refer to a table column, or a variable?
    Let's change the previous example to
<programlisting>
INSERT INTO dest (col) SELECT foo + bar FROM src;
</programlisting>
    Here, <literal>dest</literal> and <literal>src</literal> must be table names, and
    <literal>col</literal> must be a column of <literal>dest</literal>, but <literal>foo</literal>
    and <literal>bar</literal> might reasonably be either variables of the function
    or columns of <literal>src</literal>.
   </para>

   <para>
    By default, <application>PL/pgSQL</application> will report an error if a name
    in an SQL statement could refer to either a variable or a table column.
    You can fix such a problem by renaming the variable or column,
    or by qualifying the ambiguous reference, or by telling
    <application>PL/pgSQL</application> which interpretation to prefer.
   </para>

   <para>
    The simplest solution is to rename the variable or column.
    A common coding rule is to use a
    different naming convention for <application>PL/pgSQL</application>
    variables than you use for column names.  For example,
    if you consistently name function variables
    <literal>v_<replaceable>something</replaceable></literal> while none of your
    column names start with <literal>v_</literal>, no conflicts will occur.
   </para>

   <para>
    Alternatively you can qualify ambiguous references to make them clear.
    In the above example, <literal>src.foo</literal> would be an unambiguous reference
    to the table column.  To create an unambiguous reference to a variable,
    declare it in a labeled block and use the block's label
    (see <xref linkend="plpgsql-structure"/>).  For example,
<programlisting>
&lt;&lt;block&gt;&gt;
DECLARE
    foo int;
BEGIN
    foo := ...;
    INSERT INTO dest (col) SELECT block.foo + bar FROM src;
</programlisting>
    Here <literal>block.foo</literal> means the variable even if there is a column
    <literal>foo</literal> in <literal>src</literal>.  Function parameters, as well as
    special variables such as <literal>FOUND</literal>, can be qualified by the
    function's name, because they are implicitly declared in an outer block
    labeled with the function's name.
   </para>

   <para>
    Sometimes it is impractical to fix all the ambiguous references in a
    large body of <application>PL/pgSQL</application> code.  In such cases you can
    specify that <application>PL/pgSQL</application> should resolve ambiguous references
    as the variable (which is compatible with <application>PL/pgSQL</application>'s
    behavior before <productname>PostgreSQL</productname> 9.0), or as the
    table column (which is compatible with some other systems such as
    <productname>Oracle</productname>).
   </para>

   <indexterm>
     <primary><varname>plpgsql.variable_conflict</varname> configuration parameter</primary>
   </indexterm>

   <para>
    To change this behavior on a system-wide basis, set the configuration
    parameter <literal>plpgsql.variable_conflict</literal> to one of
    <literal>error</literal>, <literal>use_variable</literal>, or
    <literal>use_column</literal> (where <literal>error</literal> is the factory default).
    This parameter affects subsequent compilations
    of statements in <application>PL/pgSQL</application> functions, but not statements
    already compiled in the current session.
    Because changing this setting
    can cause unexpected changes in the behavior of <application>PL/pgSQL</application>
    functions, it can only be changed by a superuser.
   </para>

   <para>
    You can also set the behavior on a function-by-function basis, by
    inserting one of these special commands at the start of the function
    text:
<programlisting>
#variable_conflict error
#variable_conflict use_variable
#variable_conflict use_column
</programlisting>
    These commands affect only the function they are written in, and override
    the setting of <literal>plpgsql.variable_conflict</literal>.  An example is
<programlisting>
CREATE FUNCTION stamp_user(id int, comment text) RETURNS void AS $$
    #variable_conflict use_variable
    DECLARE
        curtime timestamp := now();
    BEGIN
        UPDATE users SET last_modified = curtime, comment = comment
          WHERE users.id = id;
    END;
$$ LANGUAGE plpgsql;
</programlisting>
    In the <literal>UPDATE</literal> command, <literal>curtime</literal>, <literal>comment</literal>,
    and <literal>id</literal> will refer to the function's variable and parameters
    whether or not <literal>users</literal> has columns of those names.  Notice
    that we had to qualify the reference to <literal>users.id</literal> in the
    <literal>WHERE</literal> clause to make it refer to the table column.
    But we did not have to qualify the reference to <literal>comment</literal>
    as a target in the <literal>UPDATE</literal> list, because syntactically
    that must be a column of <literal>users</literal>.  We could write the same
    function without depending on the <literal>variable_conflict</literal> setting
    in this way:
<programlisting>
CREATE FUNCTION stamp_user(id int, comment text) RETURNS void AS $$
    &lt;&lt;fn&gt;&gt;
    DECLARE
        curtime timestamp := now();
    BEGIN
        UPDATE users SET last_modified = fn.curtime, comment = stamp_user.comment
          WHERE users.id = stamp_user.id;
    END;
$$ LANGUAGE plpgsql;
</programlisting>
   </para>

   <para>
    Variable substitution does not happen in a command string given
    to <command>EXECUTE</command> or one of its variants.  If you need to
    insert a varying value into such a command, do so as part of
    constructing the string value, or use <literal>USING</literal>, as illustrated in
    <xref linkend="plpgsql-statements-executing-dyn"/>.
   </para>

   <para>
    Variable substitution currently works only in <command>SELECT</command>,
    <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, <command>MERGE</command> and commands
    containing one of these (such as <command>EXPLAIN</command> and
    <command>CREATE TABLE ... AS SELECT</command>), because the main SQL
    engine allows query parameters only in these commands.  To use a
    non-constant name or value in other statement types (generically called
    utility statements), you must construct the utility statement as a string
    and <command>EXECUTE</command> it.
   </para>

  </sect2>

  <sect2 id="plpgsql-plan-caching">
   <title>Plan Caching</title>

   <para>
    The <application>PL/pgSQL</application> interpreter parses the function's source
    text and produces an internal binary instruction tree the first time the
    function is called (within each session).  The instruction tree
    fully translates the
    <application>PL/pgSQL</application> statement structure, but individual
    <acronym>SQL</acronym> expressions and <acronym>SQL</acronym> commands
    used in the function are not translated immediately.
   </para>

   <para>
    <indexterm>
     <primary>preparing a query</primary>
     <secondary>in PL/pgSQL</secondary>
    </indexterm>
    As each expression and <acronym>SQL</acronym> command is first
    executed in the function, the <application>PL/pgSQL</application> interpreter
    parses and analyzes the command to create a prepared statement,
    using the <acronym>SPI</acronym> manager's
    <function>SPI_prepare</function> function.
    Subsequent visits to that expression or command
    reuse the prepared statement.  Thus, a function with conditional code
    paths that are seldom visited will never incur the overhead of
    analyzing those commands that are never executed within the current
    session.  A disadvantage is that errors
    in a specific expression or command cannot be detected until that
    part of the function is reached in execution.  (Trivial syntax
    errors will be detected during the initial parsing pass, but
    anything deeper will not be detected until execution.)
   </para>

   <para>
    <application>PL/pgSQL</application> (or more precisely, the SPI manager) can
    furthermore attempt to cache the execution plan associated with any
    particular prepared statement.  If a cached plan is not used, then
    a fresh execution plan is generated on each visit to the statement,
    and the current parameter values (that is, <application>PL/pgSQL</application>
    variable values) can be used to optimize the selected plan.  If the
    statement has no parameters, or is executed many times, the SPI manager
    will consider creating a <firstterm>generic</firstterm> plan that is not dependent
    on specific parameter values, and caching that for re-use.  Typically
    this will happen only if the execution plan is not very sensitive to
    the values of the <application>PL/pgSQL</application> variables referenced in it.
    If it is, generating a plan each time is a net win.  See <xref
    linkend="sql-prepare"/> for more information about the behavior of
    prepared statements.
   </para>

   <para>
    Because <application>PL/pgSQL</application> saves prepared statements
    and sometimes execution plans in this way,
    SQL commands that appear directly in a
    <application>PL/pgSQL</application> function must refer to the
    same tables and columns on every execution; that is, you cannot use
    a parameter as the name of a table or column in an SQL command.  To get
    around this restriction, you can construct dynamic commands using
    the <application>PL/pgSQL</application> <command>EXECUTE</command>
    statement &mdash; at the price of performing new parse analysis and
    constructing a new execution plan on every execution.
   </para>

    <para>
     The mutable nature of record variables presents another problem in this
     connection.  When fields of a record variable are used in
     expressions or statements, the data types of the fields must not
     change from one call of the function to the next, since each
     expression will be analyzed using the data type that is present
     when the expression is first reached.  <command>EXECUTE</command> can be
     used to get around this problem when necessary.
    </para>

    <para>
     If the same function is used as a trigger for more than one table,
     <application>PL/pgSQL</application> prepares and caches statements
     independently for each such table &mdash; that is, there is a cache
     for each trigger function and table combination, not just for each
     function.  This alleviates some of the problems with varying
     data types; for instance, a trigger function will be able to work
     successfully with a column named <literal>key</literal> even if it happens
     to have different types in different tables.
    </para>

    <para>
     Likewise, functions having polymorphic argument types have a separate
     statement cache for each combination of actual argument types they have
     been invoked for, so that data type differences do not cause unexpected
     failures.
    </para>

   <para>
    Statement caching can sometimes have surprising effects on the
    interpretation of time-sensitive values.  For example there
    is a difference between what these two functions do:

<programlisting>
CREATE FUNCTION logfunc1(logtxt text) RETURNS void AS $$
    BEGIN
        INSERT INTO logtable VALUES (logtxt, 'now');
    END;
$$ LANGUAGE plpgsql;
</programlisting>

     and:

<programlisting>
CREATE FUNCTION logfunc2(logtxt text) RETURNS void AS $$
    DECLARE
        curtime timestamp;
    BEGIN
        curtime := 'now';
        INSERT INTO logtable VALUES (logtxt, curtime);
    END;
$$ LANGUAGE plpgsql;
</programlisting>
    </para>

    <para>
     In the case of <function>logfunc1</function>, the
     <productname>PostgreSQL</productname> main parser knows when
     analyzing the <command>INSERT</command> that the
     string <literal>'now'</literal> should be interpreted as
     <type>timestamp</type>, because the target column of
     <classname>logtable</classname> is of that type. Thus,
     <literal>'now'</literal> will be converted to a <type>timestamp</type>
     constant when the
     <command>INSERT</command> is analyzed, and then used in all
     invocations of <function>logfunc1</function> during the lifetime
     of the session. Needless to say, this isn't what the programmer
     wanted.  A better idea is to use the <literal>now()</literal> or
     <literal>current_timestamp</literal> function.
    </para>

    <para>
     In the case of <function>logfunc2</function>, the
     <productname>PostgreSQL</productname> main parser does not know
     what type <literal>'now'</literal> should become and therefore
     it returns a data value of type <type>text</type> containing the string
     <literal>now</literal>. During the ensuing assignment
     to the local variable <varname>curtime</varname>, the
     <application>PL/pgSQL</application> interpreter casts this
     string to the <type>timestamp</type> type by calling the
     <function>textout</function> and <function>timestamp_in</function>
     functions for the conversion.  So, the computed time stamp is updated
     on each execution as the programmer expects.  Even though this
     happens to work as expected, it's not terribly efficient, so
     use of the <literal>now()</literal> function would still be a better idea.
    </para>

  </sect2>

  </sect1>

 <sect1 id="plpgsql-development-tips">
  <title>Tips for Developing in <application>PL/pgSQL</application></title>

   <para>
    One good way to develop in
    <application>PL/pgSQL</application> is to use the text editor of your
    choice to create your functions, and in another window, use
    <application>psql</application> to load and test those functions.
    If you are doing it this way, it
    is a good idea to write the function using <command>CREATE OR
    REPLACE FUNCTION</command>. That way you can just reload the file to update
    the function definition.  For example:
<programlisting>
CREATE OR REPLACE FUNCTION testfunc(integer) RETURNS integer AS $$
          ....
$$ LANGUAGE plpgsql;
</programlisting>
   </para>

   <para>
    While running <application>psql</application>, you can load or reload such
    a function definition file with:
<programlisting>
\i filename.sql
</programlisting>
    and then immediately issue SQL commands to test the function.
   </para>

   <para>
    Another good way to develop in <application>PL/pgSQL</application> is with a
    GUI database access tool that facilitates development in a
    procedural language. One example of such a tool is
    <application>pgAdmin</application>, although others exist. These tools often
    provide convenient features such as escaping single quotes and
    making it easier to recreate and debug functions.
   </para>

  <sect2 id="plpgsql-quote-tips">
   <title>Handling of Quotation Marks</title>

   <para>
    The code of a <application>PL/pgSQL</application> function is specified in
    <command>CREATE FUNCTION</command> as a string literal.  If you
    write the string literal in the ordinary way with surrounding
    single quotes, then any single quotes inside the function body
    must be doubled; likewise any backslashes must be doubled (assuming
    escape string syntax is used).
    Doubling quotes is at best tedious, and in more complicated cases
    the code can become downright incomprehensible, because you can
    easily find yourself needing half a dozen or more adjacent quote marks.
    It's recommended that you instead write the function body as a
    <quote>dollar-quoted</quote> string literal (see <xref
    linkend="sql-syntax-dollar-quoting"/>).  In the dollar-quoting
    approach, you never double any quote marks, but instead take care to
    choose a different dollar-quoting delimiter for each level of
    nesting you need.  For example, you might write the <command>CREATE
    FUNCTION</command> command as:
<programlisting>
CREATE OR REPLACE FUNCTION testfunc(integer) RETURNS integer AS $PROC$
          ....
$PROC$ LANGUAGE plpgsql;
</programlisting>
    Within this, you might use quote marks for simple literal strings in
    SQL commands and <literal>$$</literal> to delimit fragments of SQL commands
    that you are assembling as strings.  If you need to quote text that
    includes <literal>$$</literal>, you could use <literal>$Q$</literal>, and so on.
   </para>

   <para>
    The following chart shows what you have to do when writing quote
    marks without dollar quoting.  It might be useful when translating
    pre-dollar quoting code into something more comprehensible.
  </para>

  <variablelist>
   <varlistentry id="plpgsql-quote-tips-1-quot">
    <term>1 quotation mark</term>
    <listitem>
     <para>
      To begin and end the function body, for example:
<programlisting>
CREATE FUNCTION foo() RETURNS integer AS '
          ....
' LANGUAGE plpgsql;
</programlisting>
      Anywhere within a single-quoted function body, quote marks
      <emphasis>must</emphasis> appear in pairs.
     </para>
    </listitem>
   </varlistentry>

   <varlistentry id="plpgsql-quote-tips-2-quot">
    <term>2 quotation marks</term>
    <listitem>
     <para>
      For string literals inside the function body, for example:
<programlisting>
a_output := ''Blah'';
SELECT * FROM users WHERE f_name=''foobar'';
</programlisting>
      In the dollar-quoting approach, you'd just write:
<programlisting>
a_output := 'Blah';
SELECT * FROM users WHERE f_name='foobar';
</programlisting>
      which is exactly what the <application>PL/pgSQL</application> parser would see
      in either case.
     </para>
    </listitem>
   </varlistentry>

   <varlistentry id="plpgsql-quote-tips-4-quot">
    <term>4 quotation marks</term>
    <listitem>
     <para>
      When you need a single quotation mark in a string constant inside the
      function body, for example:
<programlisting>
a_output := a_output || '' AND name LIKE ''''foobar'''' AND xyz''
</programlisting>
      The value actually appended to <literal>a_output</literal> would be:
      <literal> AND name LIKE 'foobar' AND xyz</literal>.
     </para>
     <para>
      In the dollar-quoting approach, you'd write:
<programlisting>
a_output := a_output || $$ AND name LIKE 'foobar' AND xyz$$
</programlisting>
      being careful that any dollar-quote delimiters around this are not
      just <literal>$$</literal>.
     </para>
    </listitem>
   </varlistentry>

   <varlistentry id="plpgsql-quote-tips-6-quot">
    <term>6 quotation marks</term>
    <listitem>
     <para>
      When a single quotation mark in a string inside the function body is
      adjacent to the end of that string constant, for example:
<programlisting>
a_output := a_output || '' AND name LIKE ''''foobar''''''
</programlisting>
      The value appended to <literal>a_output</literal> would then be:
      <literal> AND name LIKE 'foobar'</literal>.
     </para>
     <para>
      In the dollar-quoting approach, this becomes:
<programlisting>
a_output := a_output || $$ AND name LIKE 'foobar'$$
</programlisting>
     </para>
    </listitem>
   </varlistentry>

   <varlistentry id="plpgsql-quote-tips-10-quot">
    <term>10 quotation marks</term>
    <listitem>
     <para>
      When you want two single quotation marks in a string constant (which
      accounts for 8 quotation marks) and this is adjacent to the end of that
      string constant (2 more).  You will probably only need that if
      you are writing a function that generates other functions, as in
      <xref linkend="plpgsql-porting-ex2"/>.
      For example:
<programlisting>
a_output := a_output || '' IF v_'' ||
    referrer_keys.kind || '' LIKE ''''''''''
    || referrer_keys.key_string || ''''''''''
    THEN RETURN '''''' || referrer_keys.referrer_type
    || ''''''; END IF;'';
</programlisting>
      The value of <literal>a_output</literal> would then be:
<programlisting>
IF v_... LIKE ''...'' THEN RETURN ''...''; END IF;
</programlisting>
     </para>
     <para>
      In the dollar-quoting approach, this becomes:
<programlisting>
a_output := a_output || $$ IF v_$$ || referrer_keys.kind || $$ LIKE '$$
    || referrer_keys.key_string || $$'
    THEN RETURN '$$ || referrer_keys.referrer_type
    || $$'; END IF;$$;
</programlisting>
      where we assume we only need to put single quote marks into
      <literal>a_output</literal>, because it will be re-quoted before use.
     </para>
    </listitem>
   </varlistentry>
  </variablelist>

  </sect2>
  <sect2 id="plpgsql-extra-checks">
   <title>Additional Compile-Time and Run-Time Checks</title>

   <para>
    To aid the user in finding instances of simple but common problems before
    they cause harm, <application>PL/pgSQL</application> provides additional
    <replaceable>checks</replaceable>. When enabled, depending on the configuration, they
    can be used to emit either a <literal>WARNING</literal> or an <literal>ERROR</literal>
    during the compilation of a function. A function which has received
    a <literal>WARNING</literal> can be executed without producing further messages,
    so you are advised to test in a separate development environment.
   </para>

   <para>
    Setting <varname>plpgsql.extra_warnings</varname>, or
    <varname>plpgsql.extra_errors</varname>, as appropriate, to <literal>"all"</literal>
    is encouraged in development and/or testing environments.
   </para>

   <para>
    These additional checks are enabled through the configuration variables
    <varname>plpgsql.extra_warnings</varname> for warnings and
    <varname>plpgsql.extra_errors</varname> for errors. Both can be set either to
    a comma-separated list of checks, <literal>"none"</literal> or
    <literal>"all"</literal>. The default is <literal>"none"</literal>. Currently
    the list of available checks includes:
    <variablelist>
     <varlistentry id="plpgsql-extra-checks-shadowed-variables">
      <term><varname>shadowed_variables</varname></term>
      <listitem>
       <para>
        Checks if a declaration shadows a previously defined variable.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry id="plpgsql-extra-checks-strict-multi-assignment">
      <term><varname>strict_multi_assignment</varname></term>
      <listitem>
       <para>
        Some <application>PL/pgSQL</application> commands allow assigning
        values to more than one variable at a time, such as
        <command>SELECT INTO</command>.  Typically, the number of target
        variables and the number of source variables should match, though
        <application>PL/pgSQL</application> will use <literal>NULL</literal>
        for missing values and extra variables are ignored.  Enabling this
        check will cause <application>PL/pgSQL</application> to throw a
        <literal>WARNING</literal> or <literal>ERROR</literal> whenever the
        number of target variables and the number of source variables are
        different.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry id="plpgsql-extra-checks-too-many-rows">
      <term><varname>too_many_rows</varname></term>
      <listitem>
       <para>
        Enabling this check will cause <application>PL/pgSQL</application> to
        check if a given query returns more than one row when an
        <literal>INTO</literal> clause is used.  As an <literal>INTO</literal>
        statement will only ever use one row, having a query return multiple
        rows is generally either inefficient and/or nondeterministic and
        therefore is likely an error.
       </para>
      </listitem>
     </varlistentry>
    </variablelist>

    The following example shows the effect of <varname>plpgsql.extra_warnings</varname>
    set to <varname>shadowed_variables</varname>:
<programlisting>
SET plpgsql.extra_warnings TO 'shadowed_variables';

CREATE FUNCTION foo(f1 int) RETURNS int AS $$
DECLARE
f1 int;
BEGIN
RETURN f1;
END;
$$ LANGUAGE plpgsql;
WARNING:  variable "f1" shadows a previously defined variable
LINE 3: f1 int;
        ^
CREATE FUNCTION
</programlisting>
    The below example shows the effects of setting
    <varname>plpgsql.extra_warnings</varname> to
    <varname>strict_multi_assignment</varname>:
<programlisting>
SET plpgsql.extra_warnings TO 'strict_multi_assignment';

CREATE OR REPLACE FUNCTION public.foo()
 RETURNS void
 LANGUAGE plpgsql
AS $$
DECLARE
  x int;
  y int;
BEGIN
  SELECT 1 INTO x, y;
  SELECT 1, 2 INTO x, y;
  SELECT 1, 2, 3 INTO x, y;
END;
$$;

SELECT foo();
WARNING:  number of source and target fields in assignment does not match
DETAIL:  strict_multi_assignment check of extra_warnings is active.
HINT:  Make sure the query returns the exact list of columns.
WARNING:  number of source and target fields in assignment does not match
DETAIL:  strict_multi_assignment check of extra_warnings is active.
HINT:  Make sure the query returns the exact list of columns.

 foo
-----

(1 row)
</programlisting>
   </para>
  </sect2>
 </sect1>

  <!-- **** Porting from Oracle PL/SQL **** -->

 <sect1 id="plpgsql-porting">
  <title>Porting from <productname>Oracle</productname> PL/SQL</title>

  <indexterm zone="plpgsql-porting">
   <primary>Oracle</primary>
   <secondary>porting from PL/SQL to PL/pgSQL</secondary>
  </indexterm>

  <indexterm zone="plpgsql-porting">
   <primary>PL/SQL (Oracle)</primary>
   <secondary>porting to PL/pgSQL</secondary>
  </indexterm>

  <para>
   This section explains differences between
   <productname>PostgreSQL</productname>'s <application>PL/pgSQL</application>
   language and Oracle's <application>PL/SQL</application> language,
   to help developers who port applications from
   <trademark class="registered">Oracle</trademark> to <productname>PostgreSQL</productname>.
  </para>

  <para>
   <application>PL/pgSQL</application> is similar to PL/SQL in many
   aspects. It is a block-structured, imperative language, and all
   variables have to be declared.  Assignments, loops, and conditionals
   are similar.  The main differences you should keep in mind when
   porting from <application>PL/SQL</application> to
   <application>PL/pgSQL</application> are:

    <itemizedlist>
     <listitem>
      <para>
       If a name used in an SQL command could be either a column name of a
       table used in the command or a reference to a variable of the function,
       <application>PL/SQL</application> treats it as a column name.
       By default, <application>PL/pgSQL</application> will throw an error
       complaining that the name is ambiguous.  You can specify
       <literal>plpgsql.variable_conflict</literal> = <literal>use_column</literal>
       to change this behavior to match <application>PL/SQL</application>,
       as explained in <xref linkend="plpgsql-var-subst"/>.
       It's often best to avoid such ambiguities in the first place,
       but if you have to port a large amount of code that depends on
       this behavior, setting <literal>variable_conflict</literal> may be the
       best solution.
      </para>
     </listitem>

     <listitem>
      <para>
       In <productname>PostgreSQL</productname> the function body must be written as
       a string literal.  Therefore you need to use dollar quoting or escape
       single quotes in the function body. (See <xref
       linkend="plpgsql-quote-tips"/>.)
      </para>
     </listitem>

     <listitem>
      <para>
       Data type names often need translation.  For example, in Oracle string
       values are commonly declared as being of type <type>varchar2</type>, which
       is a non-SQL-standard type.  In <productname>PostgreSQL</productname>,
       use type <type>varchar</type> or <type>text</type> instead.  Similarly, replace
       type <type>number</type> with <type>numeric</type>, or use some other numeric
       data type if there's a more appropriate one.
      </para>
     </listitem>

     <listitem>
      <para>
       Instead of packages, use schemas to organize your functions
       into groups.
      </para>
     </listitem>

     <listitem>
      <para>
       Since there are no packages, there are no package-level variables
       either. This is somewhat annoying.  You can keep per-session state
       in temporary tables instead.
      </para>
     </listitem>

     <listitem>
      <para>
       Integer <command>FOR</command> loops with <literal>REVERSE</literal> work
       differently: <application>PL/SQL</application> counts down from the second
       number to the first, while <application>PL/pgSQL</application> counts down
       from the first number to the second, requiring the loop bounds
       to be swapped when porting.  This incompatibility is unfortunate
       but is unlikely to be changed. (See <xref
       linkend="plpgsql-integer-for"/>.)
      </para>
     </listitem>

     <listitem>
      <para>
       <command>FOR</command> loops over queries (other than cursors) also work
       differently: the target variable(s) must have been declared,
       whereas <application>PL/SQL</application> always declares them implicitly.
       An advantage of this is that the variable values are still accessible
       after the loop exits.
      </para>
     </listitem>

     <listitem>
      <para>
       There are various notational differences for the use of cursor
       variables.
      </para>
     </listitem>

    </itemizedlist>
   </para>

  <sect2 id="plpgsql-porting-examples">
   <title>Porting Examples</title>

   <para>
    <xref linkend="pgsql-porting-ex1"/> shows how to port a simple
    function from <application>PL/SQL</application> to <application>PL/pgSQL</application>.
   </para>

   <example id="pgsql-porting-ex1">
    <title>Porting a Simple Function from <application>PL/SQL</application> to <application>PL/pgSQL</application></title>

    <para>
     Here is an <productname>Oracle</productname> <application>PL/SQL</application> function:
<programlisting>
CREATE OR REPLACE FUNCTION cs_fmt_browser_version(v_name varchar2,
                                                  v_version varchar2)
RETURN varchar2 IS
BEGIN
    IF v_version IS NULL THEN
        RETURN v_name;
    END IF;
    RETURN v_name || '/' || v_version;
END;
/
show errors;
</programlisting>
    </para>

    <para>
     Let's go through this function and see the differences compared to
     <application>PL/pgSQL</application>:

     <itemizedlist>
      <listitem>
       <para>
        The type name <type>varchar2</type> has to be changed to <type>varchar</type>
        or <type>text</type>.  In the examples in this section, we'll
        use <type>varchar</type>, but <type>text</type> is often a better choice if
        you do not need specific string length limits.
       </para>
      </listitem>

      <listitem>
       <para>
        The <literal>RETURN</literal> key word in the function
        prototype (not the function body) becomes
        <literal>RETURNS</literal> in
        <productname>PostgreSQL</productname>.
        Also, <literal>IS</literal> becomes <literal>AS</literal>, and you need to
        add a <literal>LANGUAGE</literal> clause because <application>PL/pgSQL</application>
        is not the only possible function language.
       </para>
      </listitem>

      <listitem>
       <para>
        In <productname>PostgreSQL</productname>, the function body is considered
        to be a string literal, so you need to use quote marks or dollar
        quotes around it.  This substitutes for the terminating <literal>/</literal>
        in the Oracle approach.
       </para>
      </listitem>

      <listitem>
       <para>
        The <literal>show errors</literal> command does not exist in
        <productname>PostgreSQL</productname>, and is not needed since errors are
        reported automatically.
       </para>
      </listitem>
     </itemizedlist>
    </para>

    <para>
     This is how this function would look when ported to
     <productname>PostgreSQL</productname>:

<programlisting>
CREATE OR REPLACE FUNCTION cs_fmt_browser_version(v_name varchar,
                                                  v_version varchar)
RETURNS varchar AS $$
BEGIN
    IF v_version IS NULL THEN
        RETURN v_name;
    END IF;
    RETURN v_name || '/' || v_version;
END;
$$ LANGUAGE plpgsql;
</programlisting>
    </para>
   </example>

   <para>
    <xref linkend="plpgsql-porting-ex2"/> shows how to port a
    function that creates another function and how to handle the
    ensuing quoting problems.
   </para>

   <example id="plpgsql-porting-ex2">
    <title>Porting a Function that Creates Another Function from <application>PL/SQL</application> to <application>PL/pgSQL</application></title>

    <para>
     The following procedure grabs rows from a
     <command>SELECT</command> statement and builds a large function
     with the results in <literal>IF</literal> statements, for the
     sake of efficiency.
    </para>

    <para>
     This is the Oracle version:
<programlisting>
CREATE OR REPLACE PROCEDURE cs_update_referrer_type_proc IS
    CURSOR referrer_keys IS
        SELECT * FROM cs_referrer_keys
        ORDER BY try_order;
    func_cmd VARCHAR(4000);
BEGIN
    func_cmd := 'CREATE OR REPLACE FUNCTION cs_find_referrer_type(v_host IN VARCHAR2,
                 v_domain IN VARCHAR2, v_url IN VARCHAR2) RETURN VARCHAR2 IS BEGIN';

    FOR referrer_key IN referrer_keys LOOP
        func_cmd := func_cmd ||
          ' IF v_' || referrer_key.kind
          || ' LIKE ''' || referrer_key.key_string
          || ''' THEN RETURN ''' || referrer_key.referrer_type
          || '''; END IF;';
    END LOOP;

    func_cmd := func_cmd || ' RETURN NULL; END;';

    EXECUTE IMMEDIATE func_cmd;
END;
/
show errors;
</programlisting>
    </para>

    <para>
     Here is how this function would end up in <productname>PostgreSQL</productname>:
<programlisting>
CREATE OR REPLACE PROCEDURE cs_update_referrer_type_proc() AS $func$
DECLARE
    referrer_keys CURSOR IS
        SELECT * FROM cs_referrer_keys
        ORDER BY try_order;
    func_body text;
    func_cmd text;
BEGIN
    func_body := 'BEGIN';

    FOR referrer_key IN referrer_keys LOOP
        func_body := func_body ||
          ' IF v_' || referrer_key.kind
          || ' LIKE ' || quote_literal(referrer_key.key_string)
          || ' THEN RETURN ' || quote_literal(referrer_key.referrer_type)
          || '; END IF;' ;
    END LOOP;

    func_body := func_body || ' RETURN NULL; END;';

    func_cmd :=
      'CREATE OR REPLACE FUNCTION cs_find_referrer_type(v_host varchar,
                                                        v_domain varchar,
                                                        v_url varchar)
        RETURNS varchar AS '
      || quote_literal(func_body)
      || ' LANGUAGE plpgsql;' ;

    EXECUTE func_cmd;
END;
$func$ LANGUAGE plpgsql;
</programlisting>
     Notice how the body of the function is built separately and passed
     through <literal>quote_literal</literal> to double any quote marks in it.  This
     technique is needed because we cannot safely use dollar quoting for
     defining the new function: we do not know for sure what strings will
     be interpolated from the <structfield>referrer_key.key_string</structfield> field.
     (We are assuming here that <structfield>referrer_key.kind</structfield> can be
     trusted to always be <literal>host</literal>, <literal>domain</literal>, or
     <literal>url</literal>, but <structfield>referrer_key.key_string</structfield> might be
     anything, in particular it might contain dollar signs.) This function
     is actually an improvement on the Oracle original, because it will
     not generate broken code when <structfield>referrer_key.key_string</structfield> or
     <structfield>referrer_key.referrer_type</structfield> contain quote marks.
    </para>
   </example>

   <para>
    <xref linkend="plpgsql-porting-ex3"/> shows how to port a function
    with <literal>OUT</literal> parameters and string manipulation.
    <productname>PostgreSQL</productname> does not have a built-in
    <function>instr</function> function, but you can create one
    using a combination of other
    functions. In <xref linkend="plpgsql-porting-appendix"/> there is a
    <application>PL/pgSQL</application> implementation of
    <function>instr</function> that you can use to make your porting
    easier.
   </para>

   <example id="plpgsql-porting-ex3">
    <title>Porting a Procedure With String Manipulation and
    <literal>OUT</literal> Parameters from <application>PL/SQL</application> to
    <application>PL/pgSQL</application></title>

    <para>
     The following <productname>Oracle</productname> PL/SQL procedure is used
     to parse a URL and return several elements (host, path, and query).
    </para>

    <para>
     This is the Oracle version:
<programlisting>
CREATE OR REPLACE PROCEDURE cs_parse_url(
    v_url IN VARCHAR2,
    v_host OUT VARCHAR2,  -- This will be passed back
    v_path OUT VARCHAR2,  -- This one too
    v_query OUT VARCHAR2) -- And this one
IS
    a_pos1 INTEGER;
    a_pos2 INTEGER;
BEGIN
    v_host := NULL;
    v_path := NULL;
    v_query := NULL;
    a_pos1 := instr(v_url, '//');

    IF a_pos1 = 0 THEN
        RETURN;
    END IF;
    a_pos2 := instr(v_url, '/', a_pos1 + 2);
    IF a_pos2 = 0 THEN
        v_host := substr(v_url, a_pos1 + 2);
        v_path := '/';
        RETURN;
    END IF;

    v_host := substr(v_url, a_pos1 + 2, a_pos2 - a_pos1 - 2);
    a_pos1 := instr(v_url, '?', a_pos2 + 1);

    IF a_pos1 = 0 THEN
        v_path := substr(v_url, a_pos2);
        RETURN;
    END IF;

    v_path := substr(v_url, a_pos2, a_pos1 - a_pos2);
    v_query := substr(v_url, a_pos1 + 1);
END;
/
show errors;
</programlisting>
    </para>

    <para>
     Here is a possible translation into <application>PL/pgSQL</application>:
<programlisting>
CREATE OR REPLACE FUNCTION cs_parse_url(
    v_url IN VARCHAR,
    v_host OUT VARCHAR,  -- This will be passed back
    v_path OUT VARCHAR,  -- This one too
    v_query OUT VARCHAR) -- And this one
AS $$
DECLARE
    a_pos1 INTEGER;
    a_pos2 INTEGER;
BEGIN
    v_host := NULL;
    v_path := NULL;
    v_query := NULL;
    a_pos1 := instr(v_url, '//');

    IF a_pos1 = 0 THEN
        RETURN;
    END IF;
    a_pos2 := instr(v_url, '/', a_pos1 + 2);
    IF a_pos2 = 0 THEN
        v_host := substr(v_url, a_pos1 + 2);
        v_path := '/';
        RETURN;
    END IF;

    v_host := substr(v_url, a_pos1 + 2, a_pos2 - a_pos1 - 2);
    a_pos1 := instr(v_url, '?', a_pos2 + 1);

    IF a_pos1 = 0 THEN
        v_path := substr(v_url, a_pos2);
        RETURN;
    END IF;

    v_path := substr(v_url, a_pos2, a_pos1 - a_pos2);
    v_query := substr(v_url, a_pos1 + 1);
END;
$$ LANGUAGE plpgsql;
</programlisting>

     This function could be used like this:
<programlisting>
SELECT * FROM cs_parse_url('http://foobar.com/query.cgi?baz');
</programlisting>
    </para>
   </example>

   <para>
    <xref linkend="plpgsql-porting-ex4"/> shows how to port a procedure
    that uses numerous features that are specific to Oracle.
   </para>

   <example id="plpgsql-porting-ex4">
    <title>Porting a Procedure from <application>PL/SQL</application> to <application>PL/pgSQL</application></title>

    <para>
     The Oracle version:

<programlisting>
CREATE OR REPLACE PROCEDURE cs_create_job(v_job_id IN INTEGER) IS
    a_running_job_count INTEGER;
BEGIN
    LOCK TABLE cs_jobs IN EXCLUSIVE MODE;

    SELECT count(*) INTO a_running_job_count FROM cs_jobs WHERE end_stamp IS NULL;

    IF a_running_job_count &gt; 0 THEN
        COMMIT; -- free lock
        raise_application_error(-20000,
                 'Unable to create a new job: a job is currently running.');
    END IF;

    DELETE FROM cs_active_job;
    INSERT INTO cs_active_job(job_id) VALUES (v_job_id);

    BEGIN
        INSERT INTO cs_jobs (job_id, start_stamp) VALUES (v_job_id, now());
    EXCEPTION
        WHEN dup_val_on_index THEN NULL; -- don't worry if it already exists
    END;
    COMMIT;
END;
/
show errors
</programlisting>
   </para>

   <para>
    This is how we could port this procedure to <application>PL/pgSQL</application>:

<programlisting>
CREATE OR REPLACE PROCEDURE cs_create_job(v_job_id integer) AS $$
DECLARE
    a_running_job_count integer;
BEGIN
    LOCK TABLE cs_jobs IN EXCLUSIVE MODE;

    SELECT count(*) INTO a_running_job_count FROM cs_jobs WHERE end_stamp IS NULL;

    IF a_running_job_count &gt; 0 THEN
        COMMIT; -- free lock
        RAISE EXCEPTION 'Unable to create a new job: a job is currently running'; -- <co id="co.plpgsql-porting-raise"/>
    END IF;

    DELETE FROM cs_active_job;
    INSERT INTO cs_active_job(job_id) VALUES (v_job_id);

    BEGIN
        INSERT INTO cs_jobs (job_id, start_stamp) VALUES (v_job_id, now());
    EXCEPTION
        WHEN unique_violation THEN -- <co id="co.plpgsql-porting-exception"/>
            -- don't worry if it already exists
    END;
    COMMIT;
END;
$$ LANGUAGE plpgsql;
</programlisting>

    <calloutlist>
     <callout arearefs="co.plpgsql-porting-raise">
      <para>
       The syntax of <literal>RAISE</literal> is considerably different from
       Oracle's statement, although the basic case <literal>RAISE</literal>
       <replaceable class="parameter">exception_name</replaceable> works
       similarly.
      </para>
     </callout>
     <callout arearefs="co.plpgsql-porting-exception">
      <para>
       The exception names supported by <application>PL/pgSQL</application> are
       different from Oracle's.  The set of built-in exception names
       is much larger (see <xref linkend="errcodes-appendix"/>).  There
       is not currently a way to declare user-defined exception names,
       although you can throw user-chosen SQLSTATE values instead.
      </para>
     </callout>
    </calloutlist>
   </para>
   </example>
  </sect2>

  <sect2 id="plpgsql-porting-other">
   <title>Other Things to Watch For</title>

   <para>
    This section explains a few other things to watch for when porting
    Oracle <application>PL/SQL</application> functions to
    <productname>PostgreSQL</productname>.
   </para>

   <sect3 id="plpgsql-porting-exceptions">
    <title>Implicit Rollback after Exceptions</title>

    <para>
     In <application>PL/pgSQL</application>, when an exception is caught by an
     <literal>EXCEPTION</literal> clause, all database changes since the block's
     <literal>BEGIN</literal> are automatically rolled back.  That is, the behavior
     is equivalent to what you'd get in Oracle with:

<programlisting>
BEGIN
    SAVEPOINT s1;
    ... code here ...
EXCEPTION
    WHEN ... THEN
        ROLLBACK TO s1;
        ... code here ...
    WHEN ... THEN
        ROLLBACK TO s1;
        ... code here ...
END;
</programlisting>

     If you are translating an Oracle procedure that uses
     <command>SAVEPOINT</command> and <command>ROLLBACK TO</command> in this style,
     your task is easy: just omit the <command>SAVEPOINT</command> and
     <command>ROLLBACK TO</command>.  If you have a procedure that uses
     <command>SAVEPOINT</command> and <command>ROLLBACK TO</command> in a different way
     then some actual thought will be required.
    </para>
   </sect3>

   <sect3 id="plpgsql-porting-other-execute">
    <title><command>EXECUTE</command></title>

    <para>
     The <application>PL/pgSQL</application> version of
     <command>EXECUTE</command> works similarly to the
     <application>PL/SQL</application> version, but you have to remember to use
     <function>quote_literal</function> and
     <function>quote_ident</function> as described in <xref
     linkend="plpgsql-statements-executing-dyn"/>.  Constructs of the
     type <literal>EXECUTE 'SELECT * FROM $1';</literal> will not work
     reliably unless you use these functions.
    </para>
   </sect3>

   <sect3 id="plpgsql-porting-optimization">
    <title>Optimizing <application>PL/pgSQL</application> Functions</title>

    <para>
     <productname>PostgreSQL</productname> gives you two function creation
     modifiers to optimize execution: <quote>volatility</quote> (whether
     the function always returns the same result when given the same
     arguments) and <quote>strictness</quote> (whether the function
     returns null if any argument is null).  Consult the <xref
     linkend="sql-createfunction"/>
     reference page for details.
    </para>

    <para>
     When making use of these optimization attributes, your
     <command>CREATE FUNCTION</command> statement might look something
     like this:

<programlisting>
CREATE FUNCTION foo(...) RETURNS integer AS $$
...
$$ LANGUAGE plpgsql STRICT IMMUTABLE;
</programlisting>
    </para>
   </sect3>
  </sect2>

  <sect2 id="plpgsql-porting-appendix">
   <title>Appendix</title>

   <para>
    This section contains the code for a set of Oracle-compatible
    <function>instr</function> functions that you can use to simplify
    your porting efforts.
   </para>

   <indexterm>
    <primary><function>instr</function> function</primary>
   </indexterm>

<programlisting><![CDATA[
--
-- instr functions that mimic Oracle's counterpart
-- Syntax: instr(string1, string2 [, n [, m]])
-- where [] denotes optional parameters.
--
-- Search string1, beginning at the nth character, for the mth occurrence
-- of string2.  If n is negative, search backwards, starting at the abs(n)'th
-- character from the end of string1.
-- If n is not passed, assume 1 (search starts at first character).
-- If m is not passed, assume 1 (find first occurrence).
-- Returns starting index of string2 in string1, or 0 if string2 is not found.
--

CREATE FUNCTION instr(varchar, varchar) RETURNS integer AS $$
BEGIN
    RETURN instr($1, $2, 1);
END;
$$ LANGUAGE plpgsql STRICT IMMUTABLE;


CREATE FUNCTION instr(string varchar, string_to_search_for varchar,
                      beg_index integer)
RETURNS integer AS $$
DECLARE
    pos integer NOT NULL DEFAULT 0;
    temp_str varchar;
    beg integer;
    length integer;
    ss_length integer;
BEGIN
    IF beg_index > 0 THEN
        temp_str := substring(string FROM beg_index);
        pos := position(string_to_search_for IN temp_str);

        IF pos = 0 THEN
            RETURN 0;
        ELSE
            RETURN pos + beg_index - 1;
        END IF;
    ELSIF beg_index < 0 THEN
        ss_length := char_length(string_to_search_for);
        length := char_length(string);
        beg := length + 1 + beg_index;

        WHILE beg > 0 LOOP
            temp_str := substring(string FROM beg FOR ss_length);
            IF string_to_search_for = temp_str THEN
                RETURN beg;
            END IF;

            beg := beg - 1;
        END LOOP;

        RETURN 0;
    ELSE
        RETURN 0;
    END IF;
END;
$$ LANGUAGE plpgsql STRICT IMMUTABLE;


CREATE FUNCTION instr(string varchar, string_to_search_for varchar,
                      beg_index integer, occur_index integer)
RETURNS integer AS $$
DECLARE
    pos integer NOT NULL DEFAULT 0;
    occur_number integer NOT NULL DEFAULT 0;
    temp_str varchar;
    beg integer;
    i integer;
    length integer;
    ss_length integer;
BEGIN
    IF occur_index <= 0 THEN
        RAISE 'argument ''%'' is out of range', occur_index
          USING ERRCODE = '22003';
    END IF;

    IF beg_index > 0 THEN
        beg := beg_index - 1;
        FOR i IN 1..occur_index LOOP
            temp_str := substring(string FROM beg + 1);
            pos := position(string_to_search_for IN temp_str);
            IF pos = 0 THEN
                RETURN 0;
            END IF;
            beg := beg + pos;
        END LOOP;

        RETURN beg;
    ELSIF beg_index < 0 THEN
        ss_length := char_length(string_to_search_for);
        length := char_length(string);
        beg := length + 1 + beg_index;

        WHILE beg > 0 LOOP
            temp_str := substring(string FROM beg FOR ss_length);
            IF string_to_search_for = temp_str THEN
                occur_number := occur_number + 1;
                IF occur_number = occur_index THEN
                    RETURN beg;
                END IF;
            END IF;

            beg := beg - 1;
        END LOOP;

        RETURN 0;
    ELSE
        RETURN 0;
    END IF;
END;
$$ LANGUAGE plpgsql STRICT IMMUTABLE;
]]>
</programlisting>
  </sect2>

 </sect1>

</chapter>