path: root/doc/src/sgml/ddl.sgml

<!-- doc/src/sgml/ddl.sgml -->

<chapter id="ddl">
 <title>Data Definition</title>

 <para>
  This chapter covers how one creates the database structures that
  will hold one's data.  In a relational database, the raw data is
  stored in tables, so the majority of this chapter is devoted to
  explaining how tables are created and modified and what features are
  available to control what data is stored in the tables.
  Subsequently, we discuss how tables can be organized into
  schemas, and how privileges can be assigned to tables.  Finally,
  we will briefly look at other features that affect the data storage,
  such as inheritance, table partitioning, views, functions, and
  triggers.
 </para>

 <sect1 id="ddl-basics">
  <title>Table Basics</title>

  <indexterm zone="ddl-basics">
   <primary>table</primary>
  </indexterm>

  <indexterm>
   <primary>row</primary>
  </indexterm>

  <indexterm>
   <primary>column</primary>
  </indexterm>

  <para>
   A table in a relational database is much like a table on paper: It
   consists of rows and columns.  The number and order of the columns
   is fixed, and each column has a name.  The number of rows is
   variable &mdash; it reflects how much data is stored at a given moment.
   SQL does not make any guarantees about the order of the rows in a
   table.  When a table is read, the rows will appear in an unspecified order,
   unless sorting is explicitly requested.  This is covered in <xref
   linkend="queries">.  Furthermore, SQL does not assign unique
   identifiers to rows, so it is possible to have several completely
   identical rows in a table.  This is a consequence of the
   mathematical model that underlies SQL but is usually not desirable.
   Later in this chapter we will see how to deal with this issue.
  </para>

  <para>
   Each column has a data type.  The data type constrains the set of
   possible values that can be assigned to a column and assigns
   semantics to the data stored in the column so that it can be used
   for computations.  For instance, a column declared to be of a
   numerical type will not accept arbitrary text strings, and the data
   stored in such a column can be used for mathematical computations.
   By contrast, a column declared to be of a character string type
   will accept almost any kind of data but it does not lend itself to
   mathematical calculations, although other operations such as string
   concatenation are available.
  </para>

  <para>
   <productname>PostgreSQL</productname> includes a sizable set of
   built-in data types that fit many applications.  Users can also
   define their own data types.  Most built-in data types have obvious
   names and semantics, so we defer a detailed explanation to <xref
   linkend="datatype">.  Some of the frequently used data types are
   <type>integer</type> for whole numbers, <type>numeric</type> for
   possibly fractional numbers, <type>text</type> for character
   strings, <type>date</type> for dates, <type>time</type> for
   time-of-day values, and <type>timestamp</type> for values
   containing both date and time.
  </para>

  <indexterm>
   <primary>table</primary>
   <secondary>creating</secondary>
  </indexterm>

  <para>
   To create a table, you use the aptly named <xref
   linkend="sql-createtable"> command.
   In this command you specify at least a name for the new table, the
   names of the columns and the data type of each column.  For
   example:
<programlisting>
CREATE TABLE my_first_table (
    first_column text,
    second_column integer
);
</programlisting>
   This creates a table named <literal>my_first_table</literal> with
   two columns.  The first column is named
   <literal>first_column</literal> and has a data type of
   <type>text</type>; the second column has the name
   <literal>second_column</literal> and the type <type>integer</type>.
   The table and column names follow the identifier syntax explained
   in <xref linkend="sql-syntax-identifiers">.  The type names are
   usually also identifiers, but there are some exceptions.  Note that the
   column list is comma-separated and surrounded by parentheses.
  </para>

  <para>
   Of course, the previous example was heavily contrived.  Normally,
   you would give names to your tables and columns that convey what
   kind of data they store.  So let's look at a more realistic
   example:
<programlisting>
CREATE TABLE products (
    product_no integer,
    name text,
    price numeric
);
</programlisting>
   (The <type>numeric</type> type can store fractional components, as
   would be typical of monetary amounts.)
  </para>

  <tip>
   <para>
    When you create many interrelated tables it is wise to choose a
    consistent naming pattern for the tables and columns.  For
    instance, there is a choice of using singular or plural nouns for
    table names, both of which are favored by some theorist or other.
   </para>
  </tip>

  <para>
   There is a limit on how many columns a table can contain.
   Depending on the column types, it is between 250 and 1600.
   However, defining a table with anywhere near this many columns is
   highly unusual and often a questionable design.
  </para>

  <indexterm>
   <primary>table</primary>
   <secondary>removing</secondary>
  </indexterm>

  <para>
   If you no longer need a table, you can remove it using the <xref
   linkend="sql-droptable"> command.
   For example:
<programlisting>
DROP TABLE my_first_table;
DROP TABLE products;
</programlisting>
   Attempting to drop a table that does not exist is an error.
   Nevertheless, it is common in SQL script files to unconditionally
   try to drop each table before creating it, ignoring any error
   messages, so that the script works whether or not the table exists.
   (If you like, you can use the <literal>DROP TABLE IF EXISTS</> variant
   to avoid the error messages, but this is not standard SQL.)
  </para>

  <para>
   If you need to modify a table that already exists, see <xref
   linkend="ddl-alter"> later in this chapter.
  </para>

  <para>
   With the tools discussed so far you can create fully functional
   tables.  The remainder of this chapter is concerned with adding
   features to the table definition to ensure data integrity,
   security, or convenience.  If you are eager to fill your tables with
   data now you can skip ahead to <xref linkend="dml"> and read the
   rest of this chapter later.
  </para>
 </sect1>

 <sect1 id="ddl-default">
  <title>Default Values</title>

  <indexterm zone="ddl-default">
   <primary>default value</primary>
  </indexterm>

  <para>
   A column can be assigned a default value.  When a new row is
   created and no values are specified for some of the columns, those
   columns will be filled with their respective default values.  A
   data manipulation command can also request explicitly that a column
   be set to its default value, without having to know what that value is.
   (Details about data manipulation commands are in <xref linkend="dml">.)
  </para>

  <para>
   <indexterm><primary>null value</primary><secondary>default value</secondary></indexterm>
   If no default value is declared explicitly, the default value is the
   null value.  This usually makes sense because a null value can
   be considered to represent unknown data.
  </para>

  <para>
   In a table definition, default values are listed after the column
   data type.  For example:
<programlisting>
CREATE TABLE products (
    product_no integer,
    name text,
    price numeric <emphasis>DEFAULT 9.99</emphasis>
);
</programlisting>
  </para>

  <para>
   The default value can be an expression, which will be
   evaluated whenever the default value is inserted
   (<emphasis>not</emphasis> when the table is created).  A common example
   is for a <type>timestamp</type> column to have a default of <literal>CURRENT_TIMESTAMP</>,
   so that it gets set to the time of row insertion.  Another common
   example is generating a <quote>serial number</> for each row.
   In <productname>PostgreSQL</productname> this is typically done by
   something like:
<programlisting>
CREATE TABLE products (
    product_no integer <emphasis>DEFAULT nextval('products_product_no_seq')</emphasis>,
    ...
);
</programlisting>
   where the <literal>nextval()</> function supplies successive values
   from a <firstterm>sequence object</> (see <xref
   linkend="functions-sequence">). This arrangement is sufficiently common
   that there's a special shorthand for it:
<programlisting>
CREATE TABLE products (
    product_no <emphasis>SERIAL</emphasis>,
    ...
);
</programlisting>
   The <literal>SERIAL</> shorthand is discussed further in <xref
   linkend="datatype-serial">.
  </para>
 </sect1>

 <sect1 id="ddl-constraints">
  <title>Constraints</title>

  <indexterm zone="ddl-constraints">
   <primary>constraint</primary>
  </indexterm>

  <para>
   Data types are a way to limit the kind of data that can be stored
   in a table.  For many applications, however, the constraint they
   provide is too coarse.  For example, a column containing a product
   price should probably only accept positive values.  But there is no
   standard data type that accepts only positive numbers.  Another issue is
   that you might want to constrain column data with respect to other
   columns or rows.  For example, in a table containing product
   information, there should be only one row for each product number.
  </para>

  <para>
   To that end, SQL allows you to define constraints on columns and
   tables.  Constraints give you as much control over the data in your
   tables as you wish.  If a user attempts to store data in a column
   that would violate a constraint, an error is raised.  This applies
   even if the value came from the default value definition.
  </para>

  <sect2 id="ddl-constraints-check-constraints">
   <title>Check Constraints</title>

   <indexterm>
    <primary>check constraint</primary>
   </indexterm>

   <indexterm>
    <primary>constraint</primary>
    <secondary>check</secondary>
   </indexterm>

   <para>
    A check constraint is the most generic constraint type.  It allows
    you to specify that the value in a certain column must satisfy a
    Boolean (truth-value) expression.  For instance, to require positive
    product prices, you could use:
<programlisting>
CREATE TABLE products (
    product_no integer,
    name text,
    price numeric <emphasis>CHECK (price &gt; 0)</emphasis>
);
</programlisting>
   </para>

   <para>
    As you see, the constraint definition comes after the data type,
    just like default value definitions.  Default values and
    constraints can be listed in any order.  A check constraint
    consists of the key word <literal>CHECK</literal> followed by an
    expression in parentheses.  The check constraint expression should
    involve the column thus constrained, otherwise the constraint
    would not make too much sense.
   </para>

   <indexterm>
    <primary>constraint</primary>
    <secondary>name</secondary>
   </indexterm>

   <para>
    You can also give the constraint a separate name.  This clarifies
    error messages and allows you to refer to the constraint when you
    need to change it.  The syntax is:
<programlisting>
CREATE TABLE products (
    product_no integer,
    name text,
    price numeric <emphasis>CONSTRAINT positive_price</emphasis> CHECK (price &gt; 0)
);
</programlisting>
    So, to specify a named constraint, use the key word
    <literal>CONSTRAINT</literal> followed by an identifier followed
    by the constraint definition.  (If you don't specify a constraint
    name in this way, the system chooses a name for you.)
   </para>

   <para>
    A check constraint can also refer to several columns.  Say you
    store a regular price and a discounted price, and you want to
    ensure that the discounted price is lower than the regular price:
<programlisting>
CREATE TABLE products (
    product_no integer,
    name text,
    price numeric CHECK (price &gt; 0),
    discounted_price numeric CHECK (discounted_price &gt; 0),
    <emphasis>CHECK (price &gt; discounted_price)</emphasis>
);
</programlisting>
   </para>

   <para>
    The first two constraints should look familiar.  The third one
    uses a new syntax.  It is not attached to a particular column,
    instead it appears as a separate item in the comma-separated
    column list.  Column definitions and these constraint
    definitions can be listed in mixed order.
   </para>

   <para>
    We say that the first two constraints are column constraints, whereas the
    third one is a table constraint because it is written separately
    from any one column definition.  Column constraints can also be
    written as table constraints, while the reverse is not necessarily
    possible, since a column constraint is supposed to refer to only the
    column it is attached to.  (<productname>PostgreSQL</productname> doesn't
    enforce that rule, but you should follow it if you want your table
    definitions to work with other database systems.)  The above example could
    also be written as:
<programlisting>
CREATE TABLE products (
    product_no integer,
    name text,
    price numeric,
    CHECK (price &gt; 0),
    discounted_price numeric,
    CHECK (discounted_price &gt; 0),
    CHECK (price &gt; discounted_price)
);
</programlisting>
    or even:
<programlisting>
CREATE TABLE products (
    product_no integer,
    name text,
    price numeric CHECK (price &gt; 0),
    discounted_price numeric,
    CHECK (discounted_price &gt; 0 AND price &gt; discounted_price)
);
</programlisting>
    It's a matter of taste.
   </para>

   <para>
    Names can be assigned to table constraints in the same way as
    column constraints:
<programlisting>
CREATE TABLE products (
    product_no integer,
    name text,
    price numeric,
    CHECK (price &gt; 0),
    discounted_price numeric,
    CHECK (discounted_price &gt; 0),
    <emphasis>CONSTRAINT valid_discount</> CHECK (price &gt; discounted_price)
);
</programlisting>
   </para>

   <indexterm>
    <primary>null value</primary>
    <secondary sortas="check constraints">with check constraints</secondary>
   </indexterm>

   <para>
    It should be noted that a check constraint is satisfied if the
    check expression evaluates to true or the null value.  Since most
    expressions will evaluate to the null value if any operand is null,
    they will not prevent null values in the constrained columns.  To
    ensure that a column does not contain null values, the not-null
    constraint described in the next section can be used.
   </para>
  </sect2>

  <sect2>
   <title>Not-Null Constraints</title>

   <indexterm>
    <primary>not-null constraint</primary>
   </indexterm>

   <indexterm>
    <primary>constraint</primary>
    <secondary>NOT NULL</secondary>
   </indexterm>

   <para>
    A not-null constraint simply specifies that a column must not
    assume the null value.  A syntax example:
<programlisting>
CREATE TABLE products (
    product_no integer <emphasis>NOT NULL</emphasis>,
    name text <emphasis>NOT NULL</emphasis>,
    price numeric
);
</programlisting>
   </para>

   <para>
    A not-null constraint is always written as a column constraint.  A
    not-null constraint is functionally equivalent to creating a check
    constraint <literal>CHECK (<replaceable>column_name</replaceable>
    IS NOT NULL)</literal>, but in
    <productname>PostgreSQL</productname> creating an explicit
    not-null constraint is more efficient.  The drawback is that you
    cannot give explicit names to not-null constraints created this
    way.
   </para>

   <para>
    Of course, a column can have more than one constraint.  Just write
    the constraints one after another:
<programlisting>
CREATE TABLE products (
    product_no integer NOT NULL,
    name text NOT NULL,
    price numeric NOT NULL CHECK (price &gt; 0)
);
</programlisting>
    The order doesn't matter.  It does not necessarily determine in which
    order the constraints are checked.
   </para>

   <para>
    The <literal>NOT NULL</literal> constraint has an inverse: the
    <literal>NULL</literal> constraint.  This does not mean that the
    column must be null, which would surely be useless.  Instead, this
    simply selects the default behavior that the column might be null.
    The <literal>NULL</literal> constraint is not present in the SQL
    standard and should not be used in portable applications.  (It was
    only added to <productname>PostgreSQL</productname> to be
    compatible with some other database systems.)  Some users, however,
    like it because it makes it easy to toggle the constraint in a
    script file.  For example, you could start with:
<programlisting>
CREATE TABLE products (
    product_no integer NULL,
    name text NULL,
    price numeric NULL
);
</programlisting>
    and then insert the <literal>NOT</literal> key word where desired.
   </para>

   <tip>
    <para>
     In most database designs the majority of columns should be marked
     not null.
    </para>
   </tip>
  </sect2>

  <sect2 id="ddl-constraints-unique-constraints">
   <title>Unique Constraints</title>

   <indexterm>
    <primary>unique constraint</primary>
   </indexterm>

   <indexterm>
    <primary>constraint</primary>
    <secondary>unique</secondary>
   </indexterm>

   <para>
    Unique constraints ensure that the data contained in a column, or a
    group of columns, is unique among all the rows in the
    table.  The syntax is:
<programlisting>
CREATE TABLE products (
    product_no integer <emphasis>UNIQUE</emphasis>,
    name text,
    price numeric
);
</programlisting>
    when written as a column constraint, and:
<programlisting>
CREATE TABLE products (
    product_no integer,
    name text,
    price numeric,
    <emphasis>UNIQUE (product_no)</emphasis>
);
</programlisting>
    when written as a table constraint.
   </para>

   <para>
    To define a unique constraint for a group of columns, write it as a
    table constraint with the column names separated by commas:
<programlisting>
CREATE TABLE example (
    a integer,
    b integer,
    c integer,
    <emphasis>UNIQUE (a, c)</emphasis>
);
</programlisting>
    This specifies that the combination of values in the indicated columns
    is unique across the whole table, though any one of the columns
    need not be (and ordinarily isn't) unique.
   </para>

   <para>
    You can assign your own name for a unique constraint, in the usual way:
<programlisting>
CREATE TABLE products (
    product_no integer <emphasis>CONSTRAINT must_be_different</emphasis> UNIQUE,
    name text,
    price numeric
);
</programlisting>
   </para>

   <para>
    Adding a unique constraint will automatically create a unique B-tree
    index on the column or group of columns listed in the constraint.
    A uniqueness restriction covering only some rows cannot be written as
    a unique constraint, but it is possible to enforce such a restriction by
    creating a unique <link linkend="indexes-partial">partial index</link>.
   </para>

   <indexterm>
    <primary>null value</primary>
    <secondary sortas="unique constraints">with unique constraints</secondary>
   </indexterm>

   <para>
    In general, a unique constraint is violated if there is more than
    one row in the table where the values of all of the
    columns included in the constraint are equal.
    However, two null values are never considered equal in this
    comparison.  That means even in the presence of a
    unique constraint it is possible to store duplicate
    rows that contain a null value in at least one of the constrained
    columns.  This behavior conforms to the SQL standard, but we have
    heard that other SQL databases might not follow this rule.  So be
    careful when developing applications that are intended to be
    portable.
   </para>
  </sect2>

  <sect2 id="ddl-constraints-primary-keys">
   <title>Primary Keys</title>

   <indexterm>
    <primary>primary key</primary>
   </indexterm>

   <indexterm>
    <primary>constraint</primary>
    <secondary>primary key</secondary>
   </indexterm>

   <para>
    A primary key constraint indicates that a column, or group of columns,
    can be used as a unique identifier for rows in the table.  This
    requires that the values be both unique and not null.  So, the following
    two table definitions accept the same data:
<programlisting>
CREATE TABLE products (
    product_no integer UNIQUE NOT NULL,
    name text,
    price numeric
);
</programlisting>

<programlisting>
CREATE TABLE products (
    product_no integer <emphasis>PRIMARY KEY</emphasis>,
    name text,
    price numeric
);
</programlisting>
   </para>

   <para>
    Primary keys can span more than one column; the syntax
    is similar to unique constraints:
<programlisting>
CREATE TABLE example (
    a integer,
    b integer,
    c integer,
    <emphasis>PRIMARY KEY (a, c)</emphasis>
);
</programlisting>
   </para>

   <para>
    Adding a primary key will automatically create a unique B-tree index
    on the column or group of columns listed in the primary key, and will
    force the column(s) to be marked <literal>NOT NULL</>.
   </para>

   <para>
    A table can have at most one primary key.  (There can be any number
    of unique and not-null constraints, which are functionally almost the
    same thing, but only one can be identified as the primary key.)
    Relational database theory
    dictates that every table must have a primary key.  This rule is
    not enforced by <productname>PostgreSQL</productname>, but it is
    usually best to follow it.
   </para>

   <para>
    Primary keys are useful both for
    documentation purposes and for client applications.  For example,
    a GUI application that allows modifying row values probably needs
    to know the primary key of a table to be able to identify rows
    uniquely.  There are also various ways in which the database system
    makes use of a primary key if one has been declared; for example,
    the primary key defines the default target column(s) for foreign keys
    referencing its table.
   </para>
  </sect2>

  <sect2 id="ddl-constraints-fk">
   <title>Foreign Keys</title>

   <indexterm>
    <primary>foreign key</primary>
   </indexterm>

   <indexterm>
    <primary>constraint</primary>
    <secondary>foreign key</secondary>
   </indexterm>

   <indexterm>
    <primary>referential integrity</primary>
   </indexterm>

   <para>
    A foreign key constraint specifies that the values in a column (or
    a group of columns) must match the values appearing in some row
    of another table.
    We say this maintains the <firstterm>referential
    integrity</firstterm> between two related tables.
   </para>

   <para>
    Say you have the product table that we have used several times already:
<programlisting>
CREATE TABLE products (
    product_no integer PRIMARY KEY,
    name text,
    price numeric
);
</programlisting>
    Let's also assume you have a table storing orders of those
    products.  We want to ensure that the orders table only contains
    orders of products that actually exist.  So we define a foreign
    key constraint in the orders table that references the products
    table:
<programlisting>
CREATE TABLE orders (
    order_id integer PRIMARY KEY,
    product_no integer <emphasis>REFERENCES products (product_no)</emphasis>,
    quantity integer
);
</programlisting>
    Now it is impossible to create orders with non-NULL
    <structfield>product_no</structfield> entries that do not appear in the
    products table.
   </para>

   <para>
    We say that in this situation the orders table is the
    <firstterm>referencing</firstterm> table and the products table is
    the <firstterm>referenced</firstterm> table.  Similarly, there are
    referencing and referenced columns.
   </para>

   <para>
    You can also shorten the above command to:
<programlisting>
CREATE TABLE orders (
    order_id integer PRIMARY KEY,
    product_no integer <emphasis>REFERENCES products</emphasis>,
    quantity integer
);
</programlisting>
    because in absence of a column list the primary key of the
    referenced table is used as the referenced column(s).
   </para>

   <para>
    A foreign key can also constrain and reference a group of columns.
    As usual, it then needs to be written in table constraint form.
    Here is a contrived syntax example:
<programlisting>
CREATE TABLE t1 (
  a integer PRIMARY KEY,
  b integer,
  c integer,
  <emphasis>FOREIGN KEY (b, c) REFERENCES other_table (c1, c2)</emphasis>
);
</programlisting>
    Of course, the number and type of the constrained columns need to
    match the number and type of the referenced columns.
   </para>

   <para>
    You can assign your own name for a foreign key constraint,
    in the usual way.
   </para>

   <para>
    A table can have more than one foreign key constraint.  This is
    used to implement many-to-many relationships between tables.  Say
    you have tables about products and orders, but now you want to
    allow one order to contain possibly many products (which the
    structure above did not allow).  You could use this table structure:
<programlisting>
CREATE TABLE products (
    product_no integer PRIMARY KEY,
    name text,
    price numeric
);

CREATE TABLE orders (
    order_id integer PRIMARY KEY,
    shipping_address text,
    ...
);

CREATE TABLE order_items (
    product_no integer REFERENCES products,
    order_id integer REFERENCES orders,
    quantity integer,
    PRIMARY KEY (product_no, order_id)
);
</programlisting>
    Notice that the primary key overlaps with the foreign keys in
    the last table.
   </para>

   <indexterm>
    <primary>CASCADE</primary>
    <secondary>foreign key action</secondary>
   </indexterm>

   <indexterm>
    <primary>RESTRICT</primary>
    <secondary>foreign key action</secondary>
   </indexterm>

   <para>
    We know that the foreign keys disallow creation of orders that
    do not relate to any products.  But what if a product is removed
    after an order is created that references it?  SQL allows you to
    handle that as well.  Intuitively, we have a few options:
    <itemizedlist spacing="compact">
     <listitem><para>Disallow deleting a referenced product</para></listitem>
     <listitem><para>Delete the orders as well</para></listitem>
     <listitem><para>Something else?</para></listitem>
    </itemizedlist>
   </para>

   <para>
    To illustrate this, let's implement the following policy on the
    many-to-many relationship example above: when someone wants to
    remove a product that is still referenced by an order (via
    <literal>order_items</literal>), we disallow it.  If someone
    removes an order, the order items are removed as well:
<programlisting>
CREATE TABLE products (
    product_no integer PRIMARY KEY,
    name text,
    price numeric
);

CREATE TABLE orders (
    order_id integer PRIMARY KEY,
    shipping_address text,
    ...
);

CREATE TABLE order_items (
    product_no integer REFERENCES products <emphasis>ON DELETE RESTRICT</emphasis>,
    order_id integer REFERENCES orders <emphasis>ON DELETE CASCADE</emphasis>,
    quantity integer,
    PRIMARY KEY (product_no, order_id)
);
</programlisting>
   </para>

   <para>
    Restricting and cascading deletes are the two most common options.
    <literal>RESTRICT</literal> prevents deletion of a
    referenced row. <literal>NO ACTION</literal> means that if any
    referencing rows still exist when the constraint is checked, an error
    is raised; this is the default behavior if you do not specify anything.
    (The essential difference between these two choices is that
    <literal>NO ACTION</literal> allows the check to be deferred until
    later in the transaction, whereas <literal>RESTRICT</literal> does not.)
    <literal>CASCADE</> specifies that when a referenced row is deleted,
    row(s) referencing it should be automatically deleted as well.
    There are two other options:
    <literal>SET NULL</literal> and <literal>SET DEFAULT</literal>.
    These cause the referencing column(s) in the referencing row(s)
    to be set to nulls or their default
    values, respectively, when the referenced row is deleted.
    Note that these do not excuse you from observing any constraints.
    For example, if an action specifies <literal>SET DEFAULT</literal>
    but the default value would not satisfy the foreign key constraint, the
    operation will fail.
   </para>

   <para>
    Analogous to <literal>ON DELETE</literal> there is also
    <literal>ON UPDATE</literal> which is invoked when a referenced
    column is changed (updated).  The possible actions are the same.
    In this case, <literal>CASCADE</> means that the updated values of the
    referenced column(s) should be copied into the referencing row(s).
   </para>

   <para>
    Normally, a referencing row need not satisfy the foreign key constraint
    if any of its referencing columns are null.  If <literal>MATCH FULL</>
    is added to the foreign key declaration, a referencing row escapes
    satisfying the constraint only if all its referencing columns are null
    (so a mix of null and non-null values is guaranteed to fail a
    <literal>MATCH FULL</> constraint).  If you don't want referencing rows
    to be able to avoid satisfying the foreign key constraint, declare the
    referencing column(s) as <literal>NOT NULL</>.
   </para>

   <para>
    A foreign key must reference columns that either are a primary key or
    form a unique constraint.  This means that the referenced columns always
    have an index (the one underlying the primary key or unique constraint);
    so checks on whether a referencing row has a match will be efficient.
    Since a <command>DELETE</command> of a row from the referenced table
    or an <command>UPDATE</command> of a referenced column will require
    a scan of the referencing table for rows matching the old value, it
    is often a good idea to index the referencing columns too.  Because this
    is not always needed, and there are many choices available on how
    to index, declaration of a foreign key constraint does not
    automatically create an index on the referencing columns.
   </para>

   <para>
    More information about updating and deleting data is in <xref
    linkend="dml">.  Also see the description of foreign key constraint
    syntax in the reference documentation for
    <xref linkend="sql-createtable">.
   </para>
  </sect2>

  <sect2 id="ddl-constraints-exclusion">
   <title>Exclusion Constraints</title>

   <indexterm>
    <primary>exclusion constraint</primary>
   </indexterm>

   <indexterm>
    <primary>constraint</primary>
    <secondary>exclusion</secondary>
   </indexterm>

   <para>
    Exclusion constraints ensure that if any two rows are compared on
    the specified columns or expressions using the specified operators,
    at least one of these operator comparisons will return false or null.
    The syntax is:
<programlisting>
CREATE TABLE circles (
    c circle,
    EXCLUDE USING gist (c WITH &amp;&amp;)
);
</programlisting>
   </para>

   <para>
    See also <link linkend="SQL-CREATETABLE-EXCLUDE"><command>CREATE
    TABLE ... CONSTRAINT ... EXCLUDE</></link> for details.
   </para>

   <para>
    Adding an exclusion constraint will automatically create an index
    of the type specified in the constraint declaration.
   </para>
  </sect2>
 </sect1>

 <sect1 id="ddl-system-columns">
  <title>System Columns</title>

  <para>
   Every table has several <firstterm>system columns</> that are
   implicitly defined by the system.  Therefore, these names cannot be
   used as names of user-defined columns.  (Note that these
   restrictions are separate from whether the name is a key word or
   not; quoting a name will not allow you to escape these
   restrictions.)  You do not really need to be concerned about these
   columns; just know they exist.
  </para>

  <indexterm>
   <primary>column</primary>
   <secondary>system column</secondary>
  </indexterm>

  <variablelist>
   <varlistentry>
    <term><structfield>oid</></term>
    <listitem>
     <para>
      <indexterm>
       <primary>OID</primary>
       <secondary>column</secondary>
      </indexterm>
      The object identifier (object ID) of a row. This column is only
      present if the table was created using <literal>WITH
      OIDS</literal>, or if the <xref linkend="guc-default-with-oids">
      configuration variable was set at the time. This column is of type
      <type>oid</type> (same name as the column); see <xref
      linkend="datatype-oid"> for more information about the type.
     </para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term><structfield>tableoid</></term>
    <listitem>
     <indexterm>
      <primary>tableoid</primary>
     </indexterm>

     <para>
      The OID of the table containing this row.  This column is
      particularly handy for queries that select from inheritance
      hierarchies (see <xref linkend="ddl-inherit">), since without it,
      it's difficult to tell which individual table a row came from.  The
      <structfield>tableoid</structfield> can be joined against the
      <structfield>oid</structfield> column of
      <structname>pg_class</structname> to obtain the table name.
     </para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term><structfield>xmin</></term>
    <listitem>
     <indexterm>
      <primary>xmin</primary>
     </indexterm>

     <para>
      The identity (transaction ID) of the inserting transaction for
      this row version.  (A row version is an individual state of a
      row; each update of a row creates a new row version for the same
      logical row.)
     </para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term><structfield>cmin</></term>
    <listitem>
     <indexterm>
      <primary>cmin</primary>
     </indexterm>

     <para>
      The command identifier (starting at zero) within the inserting
      transaction.
     </para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term><structfield>xmax</></term>
    <listitem>
     <indexterm>
      <primary>xmax</primary>
     </indexterm>

     <para>
      The identity (transaction ID) of the deleting transaction, or
      zero for an undeleted row version.  It is possible for this column to
      be nonzero in a visible row version. That usually indicates that the
      deleting transaction hasn't committed yet, or that an attempted
      deletion was rolled back.
     </para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term><structfield>cmax</></term>
    <listitem>
     <indexterm>
      <primary>cmax</primary>
     </indexterm>

     <para>
      The command identifier within the deleting transaction, or zero.
     </para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term><structfield>ctid</></term>
    <listitem>
     <indexterm>
      <primary>ctid</primary>
     </indexterm>

     <para>
      The physical location of the row version within its table.  Note that
      although the <structfield>ctid</structfield> can be used to
      locate the row version very quickly, a row's
      <structfield>ctid</structfield> will change if it is
      updated or moved by <command>VACUUM FULL</>.  Therefore
      <structfield>ctid</structfield> is useless as a long-term row
      identifier.  The OID, or even better a user-defined serial
      number, should be used to identify logical rows.
     </para>
    </listitem>
   </varlistentry>
  </variablelist>

   <para>
    OIDs are 32-bit quantities and are assigned from a single
    cluster-wide counter.  In a large or long-lived database, it is
    possible for the counter to wrap around.  Hence, it is bad
    practice to assume that OIDs are unique, unless you take steps to
    ensure that this is the case.  If you need to identify the rows in
    a table, using a sequence generator is strongly recommended.
    However, OIDs can be used as well, provided that a few additional
    precautions are taken:

    <itemizedlist>
     <listitem>
      <para>
       A unique constraint should be created on the OID column of each
       table for which the OID will be used to identify rows.  When such
       a unique constraint (or unique index) exists, the system takes
       care not to generate an OID matching an already-existing row.
       (Of course, this is only possible if the table contains fewer
       than 2<superscript>32</> (4 billion) rows, and in practice the
       table size had better be much less than that, or performance
       might suffer.)
      </para>
     </listitem>
     <listitem>
      <para>
       OIDs should never be assumed to be unique across tables; use
       the combination of <structfield>tableoid</> and row OID if you
       need a database-wide identifier.
      </para>
     </listitem>
     <listitem>
      <para>
       Of course, the tables in question must be created <literal>WITH
       OIDS</literal>.  As of <productname>PostgreSQL</productname> 8.1,
       <literal>WITHOUT OIDS</> is the default.
      </para>
     </listitem>
    </itemizedlist>
   </para>

   <para>
    Transaction identifiers are also 32-bit quantities.  In a
    long-lived database it is possible for transaction IDs to wrap
    around.  This is not a fatal problem given appropriate maintenance
    procedures; see <xref linkend="maintenance"> for details.  It is
    unwise, however, to depend on the uniqueness of transaction IDs
    over the long term (more than one billion transactions).
   </para>

   <para>
    Command identifiers are also 32-bit quantities.  This creates a hard limit
    of 2<superscript>32</> (4 billion) <acronym>SQL</acronym> commands
    within a single transaction.  In practice this limit is not a
    problem &mdash; note that the limit is on the number of
    <acronym>SQL</acronym> commands, not the number of rows processed.
    Also, only commands that actually modify the database contents will
    consume a command identifier.
   </para>
 </sect1>

 <sect1 id="ddl-alter">
  <title>Modifying Tables</title>

  <indexterm zone="ddl-alter">
   <primary>table</primary>
   <secondary>modifying</secondary>
  </indexterm>

  <para>
   When you create a table and you realize that you made a mistake, or
   the requirements of the application change, you can drop the
   table and create it again.  But this is not a convenient option if
   the table is already filled with data, or if the table is
   referenced by other database objects (for instance a foreign key
   constraint).  Therefore <productname>PostgreSQL</productname>
   provides a family of commands to make modifications to existing
   tables.  Note that this is conceptually distinct from altering
   the data contained in the table: here we are interested in altering
   the definition, or structure, of the table.
  </para>

  <para>
   You can:
   <itemizedlist spacing="compact">
    <listitem>
     <para>Add columns</para>
    </listitem>
    <listitem>
     <para>Remove columns</para>
    </listitem>
    <listitem>
     <para>Add constraints</para>
    </listitem>
    <listitem>
     <para>Remove constraints</para>
    </listitem>
    <listitem>
     <para>Change default values</para>
    </listitem>
    <listitem>
     <para>Change column data types</para>
    </listitem>
    <listitem>
     <para>Rename columns</para>
    </listitem>
    <listitem>
     <para>Rename tables</para>
    </listitem>
   </itemizedlist>

   All these actions are performed using the
   <xref linkend="sql-altertable">
   command, whose reference page contains details beyond those given
   here.
  </para>

  <sect2 id="ddl-alter-adding-a-column">
   <title>Adding a Column</title>

   <indexterm>
    <primary>column</primary>
    <secondary>adding</secondary>
   </indexterm>

   <para>
    To add a column, use a command like:
<programlisting>
ALTER TABLE products ADD COLUMN description text;
</programlisting>
    The new column is initially filled with whatever default
    value is given (null if you don't specify a <literal>DEFAULT</> clause).
   </para>

   <para>
    You can also define constraints on the column at the same time,
    using the usual syntax:
<programlisting>
ALTER TABLE products ADD COLUMN description text CHECK (description &lt;&gt; '');
</programlisting>
    In fact all the options that can be applied to a column description
    in <command>CREATE TABLE</> can be used here.  Keep in mind however
    that the default value must satisfy the given constraints, or the
    <literal>ADD</> will fail.  Alternatively, you can add
    constraints later (see below) after you've filled in the new column
    correctly.
   </para>

  <tip>
   <para>
    Adding a column with a default requires updating each row of the
    table (to store the new column value).  However, if no default is
    specified, <productname>PostgreSQL</productname> is able to avoid
    the physical update.  So if you intend to fill the column with
    mostly nondefault values, it's best to add the column with no default,
    insert the correct values using <command>UPDATE</>, and then add any
    desired default as described below.
   </para>
  </tip>
  </sect2>

  <sect2 id="ddl-alter-removing-a-column">
   <title>Removing a Column</title>

   <indexterm>
    <primary>column</primary>
    <secondary>removing</secondary>
   </indexterm>

   <para>
    To remove a column, use a command like:
<programlisting>
ALTER TABLE products DROP COLUMN description;
</programlisting>
    Whatever data was in the column disappears.  Table constraints involving
    the column are dropped, too.  However, if the column is referenced by a
    foreign key constraint of another table,
    <productname>PostgreSQL</productname> will not silently drop that
    constraint.  You can authorize dropping everything that depends on
    the column by adding <literal>CASCADE</>:
<programlisting>
ALTER TABLE products DROP COLUMN description CASCADE;
</programlisting>
    See <xref linkend="ddl-depend"> for a description of the general
    mechanism behind this.
   </para>
  </sect2>

  <sect2 id="ddl-alter-adding-a-constraint">
   <title>Adding a Constraint</title>

   <indexterm>
    <primary>constraint</primary>
    <secondary>adding</secondary>
   </indexterm>

   <para>
    To add a constraint, the table constraint syntax is used.  For example:
<programlisting>
ALTER TABLE products ADD CHECK (name &lt;&gt; '');
ALTER TABLE products ADD CONSTRAINT some_name UNIQUE (product_no);
ALTER TABLE products ADD FOREIGN KEY (product_group_id) REFERENCES product_groups;
</programlisting>
    To add a not-null constraint, which cannot be written as a table
    constraint, use this syntax:
<programlisting>
ALTER TABLE products ALTER COLUMN product_no SET NOT NULL;
</programlisting>
   </para>

   <para>
    The constraint will be checked immediately, so the table data must
    satisfy the constraint before it can be added.
   </para>
  </sect2>

  <sect2 id="ddl-alter-removing-a-constraint">
   <title>Removing a Constraint</title>

   <indexterm>
    <primary>constraint</primary>
    <secondary>removing</secondary>
   </indexterm>

   <para>
    To remove a constraint you need to know its name.  If you gave it
    a name then that's easy.  Otherwise the system assigned a
    generated name, which you need to find out.  The
    <application>psql</application> command <literal>\d
    <replaceable>tablename</replaceable></literal> can be helpful
    here; other interfaces might also provide a way to inspect table
    details.  Then the command is:
<programlisting>
ALTER TABLE products DROP CONSTRAINT some_name;
</programlisting>
    (If you are dealing with a generated constraint name like <literal>$2</>,
    don't forget that you'll need to double-quote it to make it a valid
    identifier.)
   </para>

   <para>
    As with dropping a column, you need to add <literal>CASCADE</> if you
    want to drop a constraint that something else depends on.  An example
    is that a foreign key constraint depends on a unique or primary key
    constraint on the referenced column(s).
   </para>

   <para>
    This works the same for all constraint types except not-null
    constraints. To drop a not null constraint use:
<programlisting>
ALTER TABLE products ALTER COLUMN product_no DROP NOT NULL;
</programlisting>
    (Recall that not-null constraints do not have names.)
   </para>
  </sect2>

  <sect2>
   <title>Changing a Column's Default Value</title>

   <indexterm>
    <primary>default value</primary>
    <secondary>changing</secondary>
   </indexterm>

   <para>
    To set a new default for a column, use a command like:
<programlisting>
ALTER TABLE products ALTER COLUMN price SET DEFAULT 7.77;
</programlisting>
    Note that this doesn't affect any existing rows in the table, it
    just changes the default for future <command>INSERT</> commands.
   </para>

   <para>
    To remove any default value, use:
<programlisting>
ALTER TABLE products ALTER COLUMN price DROP DEFAULT;
</programlisting>
    This is effectively the same as setting the default to null.
    As a consequence, it is not an error
    to drop a default where one hadn't been defined, because the
    default is implicitly the null value.
   </para>
  </sect2>

  <sect2>
   <title>Changing a Column's Data Type</title>

   <indexterm>
    <primary>column data type</primary>
    <secondary>changing</secondary>
   </indexterm>

   <para>
    To convert a column to a different data type, use a command like:
<programlisting>
ALTER TABLE products ALTER COLUMN price TYPE numeric(10,2);
</programlisting>
    This will succeed only if each existing entry in the column can be
    converted to the new type by an implicit cast.  If a more complex
    conversion is needed, you can add a <literal>USING</> clause that
    specifies how to compute the new values from the old.
   </para>

   <para>
    <productname>PostgreSQL</> will attempt to convert the column's
    default value (if any) to the new type, as well as any constraints
    that involve the column.  But these conversions might fail, or might
    produce surprising results.  It's often best to drop any constraints
    on the column before altering its type, and then add back suitably
    modified constraints afterwards.
   </para>
  </sect2>

  <sect2>
   <title>Renaming a Column</title>

   <indexterm>
    <primary>column</primary>
    <secondary>renaming</secondary>
   </indexterm>

   <para>
    To rename a column:
<programlisting>
ALTER TABLE products RENAME COLUMN product_no TO product_number;
</programlisting>
   </para>
  </sect2>

  <sect2>
   <title>Renaming a Table</title>

   <indexterm>
    <primary>table</primary>
    <secondary>renaming</secondary>
   </indexterm>

   <para>
    To rename a table:
<programlisting>
ALTER TABLE products RENAME TO items;
</programlisting>
   </para>
  </sect2>
 </sect1>

 <sect1 id="ddl-priv">
  <title>Privileges</title>

  <indexterm zone="ddl-priv">
   <primary>privilege</primary>
  </indexterm>

  <indexterm>
   <primary>permission</primary>
   <see>privilege</see>
  </indexterm>

  <indexterm zone="ddl-priv">
   <primary>owner</primary>
  </indexterm>

  <indexterm zone="ddl-priv">
   <primary>GRANT</primary>
  </indexterm>

  <indexterm zone="ddl-priv">
   <primary>REVOKE</primary>
  </indexterm>

  <para>
   When an object is created, it is assigned an owner. The
   owner is normally the role that executed the creation statement.
   For most kinds of objects, the initial state is that only the owner
   (or a superuser) can do anything with the object. To allow
   other roles to use it, <firstterm>privileges</firstterm> must be
   granted.
  </para>

  <para>
   There are different kinds of privileges: <literal>SELECT</>,
   <literal>INSERT</>, <literal>UPDATE</>, <literal>DELETE</>,
   <literal>TRUNCATE</>, <literal>REFERENCES</>, <literal>TRIGGER</>,
   <literal>CREATE</>, <literal>CONNECT</>, <literal>TEMPORARY</>,
   <literal>EXECUTE</>, and <literal>USAGE</>.
   The privileges applicable to a particular
   object vary depending on the object's type (table, function, etc).
   For complete information on the different types of privileges
   supported by <productname>PostgreSQL</productname>, refer to the
   <xref linkend="sql-grant"> reference
   page.  The following sections and chapters will also show you how
   those privileges are used.
  </para>

  <para>
   The right to modify or destroy an object is always the privilege of
   the owner only.
  </para>

  <para>
   An object can be assigned to a new owner with an <command>ALTER</command>
   command of the appropriate kind for the object, e.g. <xref
   linkend="sql-altertable">.  Superusers can always do
   this; ordinary roles can only do it if they are both the current owner
   of the object (or a member of the owning role) and a member of the new
   owning role.
  </para>

  <para>
   To assign privileges, the <command>GRANT</command> command is
   used. For example, if <literal>joe</literal> is an existing role, and
   <literal>accounts</literal> is an existing table, the privilege to
   update the table can be granted with:
<programlisting>
GRANT UPDATE ON accounts TO joe;
</programlisting>
   Writing <literal>ALL</literal> in place of a specific privilege grants all
   privileges that are relevant for the object type.
  </para>

  <para>
   The special <quote>role</quote> name <literal>PUBLIC</literal> can
   be used to grant a privilege to every role on the system.  Also,
   <quote>group</> roles can be set up to help manage privileges when
   there are many users of a database &mdash; for details see
   <xref linkend="user-manag">.
  </para>

  <para>
   To revoke a privilege, use the fittingly named
   <command>REVOKE</command> command:
<programlisting>
REVOKE ALL ON accounts FROM PUBLIC;
</programlisting>
   The special privileges of the object owner (i.e., the right to do
   <command>DROP</>, <command>GRANT</>, <command>REVOKE</>, etc.)
   are always implicit in being the owner,
   and cannot be granted or revoked.  But the object owner can choose
   to revoke their own ordinary privileges, for example to make a
   table read-only for themselves as well as others.
  </para>

  <para>
   Ordinarily, only the object's owner (or a superuser) can grant or
   revoke privileges on an object.  However, it is possible to grant a
   privilege <quote>with grant option</>, which gives the recipient
   the right to grant it in turn to others.  If the grant option is
   subsequently revoked then all who received the privilege from that
   recipient (directly or through a chain of grants) will lose the
   privilege.  For details see the <xref linkend="sql-grant"> and
   <xref linkend="sql-revoke"> reference pages.
  </para>
 </sect1>

 <sect1 id="ddl-rowsecurity">
  <title>Row Security Policies</title>

  <indexterm zone="ddl-rowsecurity">
   <primary>row-level security</primary>
  </indexterm>

  <indexterm zone="ddl-rowsecurity">
   <primary>policy</primary>
  </indexterm>

  <para>
   In addition to the SQL-standard <link linkend="ddl-priv">privilege
   system</link> available through <xref linkend="sql-grant">,
   tables can have <firstterm>row security policies</> that restrict,
   on a per-user basis, which rows can be returned by normal queries
   or inserted, updated, or deleted by data modification commands.
   This feature is also known as <firstterm>Row-Level Security</>.
   By default, tables do not have any policies, so that if a user has
   access privileges to a table according to the SQL privilege system,
   all rows within it are equally available for querying or updating.
  </para>

  <para>
   When row security is enabled on a table (with
   <link linkend="sql-altertable">ALTER TABLE ... ENABLE ROW LEVEL
   SECURITY</>), all normal access to the table for selecting rows or
   modifying rows must be allowed by a row security policy.  (However, the
   table's owner is typically not subject to row security policies.)  If no
   policy exists for the table, a default-deny policy is used, meaning that
   no rows are visible or can be modified.  Operations that apply to the
   whole table, such as <command>TRUNCATE</> and <literal>REFERENCES</>,
   are not subject to row security.
  </para>

  <para>
   Row security policies can be specific to commands, or to roles, or to
   both.  A policy can be specified to apply to <literal>ALL</literal>
   commands, or to <literal>SELECT</>, <literal>INSERT</>, <literal>UPDATE</>,
   or <literal>DELETE</>.  Multiple roles can be assigned to a given
   policy, and normal role membership and inheritance rules apply.
  </para>

  <para>
   To specify which rows are visible or modifiable according to a policy,
   an expression is required that returns a Boolean result.  This
   expression will be evaluated for each row prior to any conditions or
   functions coming from the user's query.  (The only exceptions to this
   rule are <literal>leakproof</literal> functions, which are guaranteed to
   not leak information; the optimizer may choose to apply such functions
   ahead of the row-security check.)  Rows for which the expression does
   not return <literal>true</> will not be processed.  Separate expressions
   may be specified to provide independent control over the rows which are
   visible and the rows which are allowed to be modified.  Policy
   expressions are run as part of the query and with the privileges of the
   user running the query, although security-definer functions can be used
   to access data not available to the calling user.
  </para>

  <para>
   Superusers and roles with the <literal>BYPASSRLS</> attribute always
   bypass the row security system when accessing a table.  Table owners
   normally bypass row security as well, though a table owner can choose to
   be subject to row security with <link linkend="sql-altertable">ALTER
   TABLE ... FORCE ROW LEVEL SECURITY</>.
  </para>

  <para>
   Enabling and disabling row security, as well as adding policies to a
   table, is always the privilege of the table owner only.
  </para>

  <para>
   Policies are created using the <xref linkend="sql-createpolicy">
   command, altered using the <xref linkend="sql-alterpolicy"> command,
   and dropped using the <xref linkend="sql-droppolicy"> command.  To
   enable and disable row security for a given table, use the
   <xref linkend="sql-altertable"> command.
  </para>

  <para>
   Each policy has a name and multiple policies can be defined for a
   table.  As policies are table-specific, each policy for a table must
   have a unique name.  Different tables may have policies with the
   same name.
  </para>

  <para>
   When multiple policies apply to a given query, they are combined using
   either <literal>OR</literal> (for permissive policies, which are the
   default) or using <literal>AND</literal> (for restrictive policies).
   This is similar to the rule that a given role has the privileges
   of all roles that they are a member of.  Permissive vs. restrictive
   policies are discussed further below.
  </para>

  <para>
   As a simple example, here is how to create a policy on
   the <literal>account</> relation to allow only members of
   the <literal>managers</> role to access rows, and only rows of their
   accounts:
  </para>

<programlisting>
CREATE TABLE accounts (manager text, company text, contact_email text);

ALTER TABLE accounts ENABLE ROW LEVEL SECURITY;

CREATE POLICY account_managers ON accounts TO managers
    USING (manager = current_user);
</programlisting>

  <para>
   If no role is specified, or the special user name
   <literal>PUBLIC</literal> is used, then the policy applies to all
   users on the system.  To allow all users to access their own row in
   a <literal>users</> table, a simple policy can be used:
  </para>

<programlisting>
CREATE POLICY user_policy ON users
    USING (user_name = current_user);
</programlisting>

  <para>
   To use a different policy for rows that are being added to the table
   compared to those rows that are visible, the <literal>WITH CHECK</>
   clause can be used.  This policy would allow all users to view all rows
   in the <literal>users</> table, but only modify their own:
  </para>

<programlisting>
CREATE POLICY user_policy ON users
    USING (true)
    WITH CHECK (user_name = current_user);
</programlisting>

  <para>
   Row security can also be disabled with the <command>ALTER TABLE</>
   command.  Disabling row security does not remove any policies that are
   defined on the table; they are simply ignored.  Then all rows in the
   table are visible and modifiable, subject to the standard SQL privileges
   system.
  </para>

  <para>
   Below is a larger example of how this feature can be used in production
   environments.  The table <literal>passwd</> emulates a Unix password
   file:
  </para>

<programlisting>
-- Simple passwd-file based example
CREATE TABLE passwd (
  user_name             text UNIQUE NOT NULL,
  pwhash                text,
  uid                   int  PRIMARY KEY,
  gid                   int  NOT NULL,
  real_name             text NOT NULL,
  home_phone            text,
  extra_info            text,
  home_dir              text NOT NULL,
  shell                 text NOT NULL
);

CREATE ROLE admin;  -- Administrator
CREATE ROLE bob;    -- Normal user
CREATE ROLE alice;  -- Normal user

-- Populate the table
INSERT INTO passwd VALUES
  ('admin','xxx',0,0,'Admin','111-222-3333',null,'/root','/bin/dash');
INSERT INTO passwd VALUES
  ('bob','xxx',1,1,'Bob','123-456-7890',null,'/home/bob','/bin/zsh');
INSERT INTO passwd VALUES
  ('alice','xxx',2,1,'Alice','098-765-4321',null,'/home/alice','/bin/zsh');

-- Be sure to enable row level security on the table
ALTER TABLE passwd ENABLE ROW LEVEL SECURITY;

-- Create policies
-- Administrator can see all rows and add any rows
CREATE POLICY admin_all ON passwd TO admin USING (true) WITH CHECK (true);
-- Normal users can view all rows
CREATE POLICY all_view ON passwd FOR SELECT USING (true);
-- Normal users can update their own records, but
-- limit which shells a normal user is allowed to set
CREATE POLICY user_mod ON passwd FOR UPDATE
  USING (current_user = user_name)
  WITH CHECK (
    current_user = user_name AND
    shell IN ('/bin/bash','/bin/sh','/bin/dash','/bin/zsh','/bin/tcsh')
  );

-- Allow admin all normal rights
GRANT SELECT, INSERT, UPDATE, DELETE ON passwd TO admin;
-- Users only get select access on public columns
GRANT SELECT
  (user_name, uid, gid, real_name, home_phone, extra_info, home_dir, shell)
  ON passwd TO public;
-- Allow users to update certain columns
GRANT UPDATE
  (pwhash, real_name, home_phone, extra_info, shell)
  ON passwd TO public;
</programlisting>

  <para>
   As with any security settings, it's important to test and ensure that
   the system is behaving as expected.  Using the example above, this
   demonstrates that the permission system is working properly.
  </para>

<programlisting>
-- admin can view all rows and fields
postgres=&gt; set role admin;
SET
postgres=&gt; table passwd;
 user_name | pwhash | uid | gid | real_name |  home_phone  | extra_info | home_dir    |   shell
-----------+--------+-----+-----+-----------+--------------+------------+-------------+-----------
 admin     | xxx    |   0 |   0 | Admin     | 111-222-3333 |            | /root       | /bin/dash
 bob       | xxx    |   1 |   1 | Bob       | 123-456-7890 |            | /home/bob   | /bin/zsh
 alice     | xxx    |   2 |   1 | Alice     | 098-765-4321 |            | /home/alice | /bin/zsh
(3 rows)

-- Test what Alice is able to do
postgres=&gt; set role alice;
SET
postgres=&gt; table passwd;
ERROR:  permission denied for relation passwd
postgres=&gt; select user_name,real_name,home_phone,extra_info,home_dir,shell from passwd;
 user_name | real_name |  home_phone  | extra_info | home_dir    |   shell
-----------+-----------+--------------+------------+-------------+-----------
 admin     | Admin     | 111-222-3333 |            | /root       | /bin/dash
 bob       | Bob       | 123-456-7890 |            | /home/bob   | /bin/zsh
 alice     | Alice     | 098-765-4321 |            | /home/alice | /bin/zsh
(3 rows)

postgres=&gt; update passwd set user_name = 'joe';
ERROR:  permission denied for relation passwd
-- Alice is allowed to change her own real_name, but no others
postgres=&gt; update passwd set real_name = 'Alice Doe';
UPDATE 1
postgres=&gt; update passwd set real_name = 'John Doe' where user_name = 'admin';
UPDATE 0
postgres=&gt; update passwd set shell = '/bin/xx';
ERROR:  new row violates WITH CHECK OPTION for "passwd"
postgres=&gt; delete from passwd;
ERROR:  permission denied for relation passwd
postgres=&gt; insert into passwd (user_name) values ('xxx');
ERROR:  permission denied for relation passwd
-- Alice can change her own password; RLS silently prevents updating other rows
postgres=&gt; update passwd set pwhash = 'abc';
UPDATE 1
</programlisting>

  <para>
   All of the policies constructed thus far have been permissive policies,
   meaning that when multiple policies are applied they are combined using
   the <quote>OR</quote> Boolean operator.  While permissive policies can be constructed
   to only allow access to rows in the intended cases, it can be simpler to
   combine permissive policies with restrictive policies (which the records
   must pass and which are combined using the <quote>AND</quote> Boolean operator).
   Building on the example above, we add a restrictive policy to require
   the administrator to be connected over a local Unix socket to access the
   records of the <literal>passwd</literal> table:
  </para>

<programlisting>
CREATE POLICY admin_local_only ON passwd AS RESTRICTIVE TO admin
    USING (pg_catalog.inet_client_addr() IS NULL);
</programlisting>

  <para>
   We can then see that an administrator connecting over a network will not
   see any records, due to the restrictive policy:
  </para>

<programlisting>
=&gt; SELECT current_user;
 current_user 
--------------
 admin
(1 row)

=&gt; select inet_client_addr();
 inet_client_addr 
------------------
 127.0.0.1
(1 row)

=&gt; SELECT current_user;
 current_user 
--------------
 admin
(1 row)

=&gt; TABLE passwd;
 user_name | pwhash | uid | gid | real_name | home_phone | extra_info | home_dir | shell
-----------+--------+-----+-----+-----------+------------+------------+----------+-------
(0 rows)

=&gt; UPDATE passwd set pwhash = NULL;
UPDATE 0
</programlisting>

  <para>
   Referential integrity checks, such as unique or primary key constraints
   and foreign key references, always bypass row security to ensure that
   data integrity is maintained.  Care must be taken when developing
   schemas and row level policies to avoid <quote>covert channel</> leaks of
   information through such referential integrity checks.
  </para>

  <para>
   In some contexts it is important to be sure that row security is
   not being applied.  For example, when taking a backup, it could be
   disastrous if row security silently caused some rows to be omitted
   from the backup.  In such a situation, you can set the
   <xref linkend="guc-row-security"> configuration parameter
   to <literal>off</>.  This does not in itself bypass row security;
   what it does is throw an error if any query's results would get filtered
   by a policy.  The reason for the error can then be investigated and
   fixed.
  </para>

  <para>
   In the examples above, the policy expressions consider only the current
   values in the row to be accessed or updated.  This is the simplest and
   best-performing case; when possible, it's best to design row security
   applications to work this way.  If it is necessary to consult other rows
   or other tables to make a policy decision, that can be accomplished using
   sub-<command>SELECT</>s, or functions that contain <command>SELECT</>s,
   in the policy expressions.  Be aware however that such accesses can
   create race conditions that could allow information leakage if care is
   not taken.  As an example, consider the following table design:
  </para>

<programlisting>
-- definition of privilege groups
CREATE TABLE groups (group_id int PRIMARY KEY,
                     group_name text NOT NULL);

INSERT INTO groups VALUES
  (1, 'low'),
  (2, 'medium'),
  (5, 'high');

GRANT ALL ON groups TO alice;  -- alice is the administrator
GRANT SELECT ON groups TO public;

-- definition of users' privilege levels
CREATE TABLE users (user_name text PRIMARY KEY,
                    group_id int NOT NULL REFERENCES groups);

INSERT INTO users VALUES
  ('alice', 5),
  ('bob', 2),
  ('mallory', 2);

GRANT ALL ON users TO alice;
GRANT SELECT ON users TO public;

-- table holding the information to be protected
CREATE TABLE information (info text,
                          group_id int NOT NULL REFERENCES groups);

INSERT INTO information VALUES
  ('barely secret', 1),
  ('slightly secret', 2),
  ('very secret', 5);

ALTER TABLE information ENABLE ROW LEVEL SECURITY;

-- a row should be visible to/updatable by users whose security group_id is
-- greater than or equal to the row's group_id
CREATE POLICY fp_s ON information FOR SELECT
  USING (group_id &lt;= (SELECT group_id FROM users WHERE user_name = current_user));
CREATE POLICY fp_u ON information FOR UPDATE
  USING (group_id &lt;= (SELECT group_id FROM users WHERE user_name = current_user));

-- we rely only on RLS to protect the information table
GRANT ALL ON information TO public;
</programlisting>

  <para>
   Now suppose that <literal>alice</> wishes to change the <quote>slightly
   secret</> information, but decides that <literal>mallory</> should not
   be trusted with the new content of that row, so she does:
  </para>

<programlisting>
BEGIN;
UPDATE users SET group_id = 1 WHERE user_name = 'mallory';
UPDATE information SET info = 'secret from mallory' WHERE group_id = 2;
COMMIT;
</programlisting>

  <para>
   That looks safe; there is no window wherein <literal>mallory</> should be
   able to see the <quote>secret from mallory</> string.  However, there is
   a race condition here.  If <literal>mallory</> is concurrently doing,
   say,
<programlisting>
SELECT * FROM information WHERE group_id = 2 FOR UPDATE;
</programlisting>
   and her transaction is in <literal>READ COMMITTED</> mode, it is possible
   for her to see <quote>secret from mallory</>.  That happens if her
   transaction reaches the <structname>information</> row just
   after <literal>alice</>'s does.  It blocks waiting
   for <literal>alice</>'s transaction to commit, then fetches the updated
   row contents thanks to the <literal>FOR UPDATE</> clause.  However, it
   does <emphasis>not</> fetch an updated row for the
   implicit <command>SELECT</> from <structname>users</>, because that
   sub-<command>SELECT</> did not have <literal>FOR UPDATE</>; instead
   the <structname>users</> row is read with the snapshot taken at the start
   of the query.  Therefore, the policy expression tests the old value
   of <literal>mallory</>'s privilege level and allows her to see the
   updated row.
  </para>

  <para>
   There are several ways around this problem.  One simple answer is to use
   <literal>SELECT ... FOR SHARE</> in sub-<command>SELECT</>s in row
   security policies.  However, that requires granting <literal>UPDATE</>
   privilege on the referenced table (here <structname>users</>) to the
   affected users, which might be undesirable.  (But another row security
   policy could be applied to prevent them from actually exercising that
   privilege; or the sub-<command>SELECT</> could be embedded into a security
   definer function.)  Also, heavy concurrent use of row share locks on the
   referenced table could pose a performance problem, especially if updates
   of it are frequent.  Another solution, practical if updates of the
   referenced table are infrequent, is to take an exclusive lock on the
   referenced table when updating it, so that no concurrent transactions
   could be examining old row values.  Or one could just wait for all
   concurrent transactions to end after committing an update of the
   referenced table and before making changes that rely on the new security
   situation.
  </para>

  <para>
   For additional details see <xref linkend="sql-createpolicy">
   and <xref linkend="sql-altertable">.
  </para>

 </sect1>

 <sect1 id="ddl-schemas">
  <title>Schemas</title>

  <indexterm zone="ddl-schemas">
   <primary>schema</primary>
  </indexterm>

  <para>
   A <productname>PostgreSQL</productname> database cluster
   contains one or more named databases.  Users and groups of users are
   shared across the entire cluster, but no other data is shared across
   databases.  Any given client connection to the server can access
   only the data in a single database, the one specified in the connection
   request.
  </para>

  <note>
   <para>
    Users of a cluster do not necessarily have the privilege to access every
    database in the cluster.  Sharing of user names means that there
    cannot be different users named, say, <literal>joe</> in two databases
    in the same cluster; but the system can be configured to allow
    <literal>joe</> access to only some of the databases.
   </para>
  </note>

  <para>
   A database contains one or more named <firstterm>schemas</>, which
   in turn contain tables.  Schemas also contain other kinds of named
   objects, including data types, functions, and operators.  The same
   object name can be used in different schemas without conflict; for
   example, both <literal>schema1</> and <literal>myschema</> can
   contain tables named <literal>mytable</>.  Unlike databases,
   schemas are not rigidly separated: a user can access objects in any
   of the schemas in the database they are connected to, if they have
   privileges to do so.
  </para>

  <para>
   There are several reasons why one might want to use schemas:

   <itemizedlist>
    <listitem>
     <para>
      To allow many users to use one database without interfering with
      each other.
     </para>
    </listitem>

    <listitem>
     <para>
      To organize database objects into logical groups to make them
      more manageable.
     </para>
    </listitem>

    <listitem>
     <para>
      Third-party applications can be put into separate schemas so
      they do not collide with the names of other objects.
     </para>
    </listitem>
   </itemizedlist>

   Schemas are analogous to directories at the operating system level,
   except that schemas cannot be nested.
  </para>

  <sect2 id="ddl-schemas-create">
   <title>Creating a Schema</title>

   <indexterm zone="ddl-schemas-create">
    <primary>schema</primary>
    <secondary>creating</secondary>
   </indexterm>

   <para>
    To create a schema, use the <xref linkend="sql-createschema">
    command.  Give the schema a name
    of your choice.  For example:
<programlisting>
CREATE SCHEMA myschema;
</programlisting>
   </para>

   <indexterm>
    <primary>qualified name</primary>
   </indexterm>

   <indexterm>
    <primary>name</primary>
    <secondary>qualified</secondary>
   </indexterm>

   <para>
    To create or access objects in a schema, write a
    <firstterm>qualified name</> consisting of the schema name and
    table name separated by a dot:
<synopsis>
<replaceable>schema</><literal>.</><replaceable>table</>
</synopsis>
    This works anywhere a table name is expected, including the table
    modification commands and the data access commands discussed in
    the following chapters.
    (For brevity we will speak of tables only, but the same ideas apply
    to other kinds of named objects, such as types and functions.)
   </para>

   <para>
    Actually, the even more general syntax
<synopsis>
<replaceable>database</><literal>.</><replaceable>schema</><literal>.</><replaceable>table</>
</synopsis>
    can be used too, but at present this is just for <foreignphrase>pro
    forma</> compliance with the SQL standard.  If you write a database name,
    it must be the same as the database you are connected to.
   </para>

   <para>
    So to create a table in the new schema, use:
<programlisting>
CREATE TABLE myschema.mytable (
 ...
);
</programlisting>
   </para>

   <indexterm>
    <primary>schema</primary>
    <secondary>removing</secondary>
   </indexterm>

   <para>
    To drop a schema if it's empty (all objects in it have been
    dropped), use:
<programlisting>
DROP SCHEMA myschema;
</programlisting>
    To drop a schema including all contained objects, use:
<programlisting>
DROP SCHEMA myschema CASCADE;
</programlisting>
    See <xref linkend="ddl-depend"> for a description of the general
    mechanism behind this.
   </para>

   <para>
    Often you will want to create a schema owned by someone else
    (since this is one of the ways to restrict the activities of your
    users to well-defined namespaces).  The syntax for that is:
<programlisting>
CREATE SCHEMA <replaceable>schema_name</replaceable> AUTHORIZATION <replaceable>user_name</replaceable>;
</programlisting>
    You can even omit the schema name, in which case the schema name
    will be the same as the user name.  See <xref
    linkend="ddl-schemas-patterns"> for how this can be useful.
   </para>

   <para>
    Schema names beginning with <literal>pg_</> are reserved for
    system purposes and cannot be created by users.
   </para>
  </sect2>

  <sect2 id="ddl-schemas-public">
   <title>The Public Schema</title>

   <indexterm zone="ddl-schemas-public">
    <primary>schema</primary>
    <secondary>public</secondary>
   </indexterm>

   <para>
    In the previous sections we created tables without specifying any
    schema names.  By default such tables (and other objects) are
    automatically put into a schema named <quote>public</quote>.  Every new
    database contains such a schema.  Thus, the following are equivalent:
<programlisting>
CREATE TABLE products ( ... );
</programlisting>
    and:
<programlisting>
CREATE TABLE public.products ( ... );
</programlisting>
   </para>
  </sect2>

  <sect2 id="ddl-schemas-path">
   <title>The Schema Search Path</title>

   <indexterm>
    <primary>search path</primary>
   </indexterm>

   <indexterm>
    <primary>unqualified name</primary>
   </indexterm>

   <indexterm>
    <primary>name</primary>
    <secondary>unqualified</secondary>
   </indexterm>

   <para>
    Qualified names are tedious to write, and it's often best not to
    wire a particular schema name into applications anyway.  Therefore
    tables are often referred to by <firstterm>unqualified names</>,
    which consist of just the table name.  The system determines which table
    is meant by following a <firstterm>search path</>, which is a list
    of schemas to look in.  The first matching table in the search path
    is taken to be the one wanted.  If there is no match in the search
    path, an error is reported, even if matching table names exist
    in other schemas in the database.
   </para>

   <indexterm>
    <primary>schema</primary>
    <secondary>current</secondary>
   </indexterm>

   <para>
    The first schema named in the search path is called the current schema.
    Aside from being the first schema searched, it is also the schema in
    which new tables will be created if the <command>CREATE TABLE</>
    command does not specify a schema name.
   </para>

   <indexterm>
    <primary><varname>search_path</varname> configuration parameter</primary>
   </indexterm>

   <para>
    To show the current search path, use the following command:
<programlisting>
SHOW search_path;
</programlisting>
    In the default setup this returns:
<screen>
 search_path
--------------
 "$user", public
</screen>
    The first element specifies that a schema with the same name as
    the current user is to be searched.  If no such schema exists,
    the entry is ignored.  The second element refers to the
    public schema that we have seen already.
   </para>

   <para>
    The first schema in the search path that exists is the default
    location for creating new objects.  That is the reason that by
    default objects are created in the public schema.  When objects
    are referenced in any other context without schema qualification
    (table modification, data modification, or query commands) the
    search path is traversed until a matching object is found.
    Therefore, in the default configuration, any unqualified access
    again can only refer to the public schema.
   </para>

   <para>
    To put our new schema in the path, we use:
<programlisting>
SET search_path TO myschema,public;
</programlisting>
    (We omit the <literal>$user</literal> here because we have no
    immediate need for it.)  And then we can access the table without
    schema qualification:
<programlisting>
DROP TABLE mytable;
</programlisting>
    Also, since <literal>myschema</literal> is the first element in
    the path, new objects would by default be created in it.
   </para>

   <para>
    We could also have written:
<programlisting>
SET search_path TO myschema;
</programlisting>
    Then we no longer have access to the public schema without
    explicit qualification.  There is nothing special about the public
    schema except that it exists by default.  It can be dropped, too.
   </para>

   <para>
    See also <xref linkend="functions-info"> for other ways to manipulate
    the schema search path.
   </para>

   <para>
    The search path works in the same way for data type names, function names,
    and operator names as it does for table names.  Data type and function
    names can be qualified in exactly the same way as table names.  If you
    need to write a qualified operator name in an expression, there is a
    special provision: you must write
<synopsis>
<literal>OPERATOR(</><replaceable>schema</><literal>.</><replaceable>operator</><literal>)</>
</synopsis>
    This is needed to avoid syntactic ambiguity.  An example is:
<programlisting>
SELECT 3 OPERATOR(pg_catalog.+) 4;
</programlisting>
    In practice one usually relies on the search path for operators,
    so as not to have to write anything so ugly as that.
   </para>
  </sect2>

  <sect2 id="ddl-schemas-priv">
   <title>Schemas and Privileges</title>

   <indexterm zone="ddl-schemas-priv">
    <primary>privilege</primary>
    <secondary sortas="schemas">for schemas</secondary>
   </indexterm>

   <para>
    By default, users cannot access any objects in schemas they do not
    own.  To allow that, the owner of the schema must grant the
    <literal>USAGE</literal> privilege on the schema.  To allow users
    to make use of the objects in the schema, additional privileges
    might need to be granted, as appropriate for the object.
   </para>

   <para>
    A user can also be allowed to create objects in someone else's
    schema.  To allow that, the <literal>CREATE</literal> privilege on
    the schema needs to be granted.  Note that by default, everyone
    has <literal>CREATE</literal> and <literal>USAGE</literal> privileges on
    the schema
    <literal>public</literal>.  This allows all users that are able to
    connect to a given database to create objects in its
    <literal>public</literal> schema.  If you do
    not want to allow that, you can revoke that privilege:
<programlisting>
REVOKE CREATE ON SCHEMA public FROM PUBLIC;
</programlisting>
    (The first <quote>public</quote> is the schema, the second
    <quote>public</quote> means <quote>every user</quote>.  In the
    first sense it is an identifier, in the second sense it is a
    key word, hence the different capitalization; recall the
    guidelines from <xref linkend="sql-syntax-identifiers">.)
   </para>
  </sect2>

  <sect2 id="ddl-schemas-catalog">
   <title>The System Catalog Schema</title>

   <indexterm zone="ddl-schemas-catalog">
    <primary>system catalog</primary>
    <secondary>schema</secondary>
   </indexterm>

   <para>
    In addition to <literal>public</> and user-created schemas, each
    database contains a <literal>pg_catalog</> schema, which contains
    the system tables and all the built-in data types, functions, and
    operators.  <literal>pg_catalog</> is always effectively part of
    the search path.  If it is not named explicitly in the path then
    it is implicitly searched <emphasis>before</> searching the path's
    schemas.  This ensures that built-in names will always be
    findable.  However, you can explicitly place
    <literal>pg_catalog</> at the end of your search path if you
    prefer to have user-defined names override built-in names.
   </para>

   <para>
    Since system table names begin with <literal>pg_</>, it is best to
    avoid such names to ensure that you won't suffer a conflict if some
    future version defines a system table named the same as your
    table.  (With the default search path, an unqualified reference to
    your table name would then be resolved as the system table instead.)
    System tables will continue to follow the convention of having
    names beginning with <literal>pg_</>, so that they will not
    conflict with unqualified user-table names so long as users avoid
    the <literal>pg_</> prefix.
   </para>
  </sect2>

  <sect2 id="ddl-schemas-patterns">
   <title>Usage Patterns</title>

   <para>
    Schemas can be used to organize your data in many ways.  There are
    a few usage patterns that are recommended and are easily supported by
    the default configuration:
    <itemizedlist>
     <listitem>
      <para>
       If you do not create any schemas then all users access the
       public schema implicitly.  This simulates the situation where
       schemas are not available at all.  This setup is mainly
       recommended when there is only a single user or a few cooperating
       users in a database.  This setup also allows smooth transition
       from the non-schema-aware world.
      </para>
     </listitem>

     <listitem>
      <para>
       You can create a schema for each user with the same name as
       that user.  Recall that the default search path starts with
       <literal>$user</literal>, which resolves to the user name.
       Therefore, if each user has a separate schema, they access their
       own schemas by default.
      </para>

      <para>
       If you use this setup then you might also want to revoke access
       to the public schema (or drop it altogether), so users are
       truly constrained to their own schemas.
      </para>
     </listitem>

     <listitem>
      <para>
       To install shared applications (tables to be used by everyone,
       additional functions provided by third parties, etc.), put them
       into separate schemas.  Remember to grant appropriate
       privileges to allow the other users to access them.  Users can
       then refer to these additional objects by qualifying the names
       with a schema name, or they can put the additional schemas into
       their search path, as they choose.
      </para>
     </listitem>
    </itemizedlist>
   </para>
  </sect2>

  <sect2 id="ddl-schemas-portability">
   <title>Portability</title>

   <para>
    In the SQL standard, the notion of objects in the same schema
    being owned by different users does not exist.  Moreover, some
    implementations do not allow you to create schemas that have a
    different name than their owner.  In fact, the concepts of schema
    and user are nearly equivalent in a database system that
    implements only the basic schema support specified in the
    standard.  Therefore, many users consider qualified names to
    really consist of
    <literal><replaceable>user_name</>.<replaceable>table_name</></literal>.
    This is how <productname>PostgreSQL</productname> will effectively
    behave if you create a per-user schema for every user.
   </para>

   <para>
    Also, there is no concept of a <literal>public</> schema in the
    SQL standard.  For maximum conformance to the standard, you should
    not use (perhaps even remove) the <literal>public</> schema.
   </para>

   <para>
    Of course, some SQL database systems might not implement schemas
    at all, or provide namespace support by allowing (possibly
    limited) cross-database access.  If you need to work with those
    systems, then maximum portability would be achieved by not using
    schemas at all.
   </para>
  </sect2>
 </sect1>

 <sect1 id="ddl-inherit">
  <title>Inheritance</title>

  <indexterm>
   <primary>inheritance</primary>
  </indexterm>

  <indexterm>
   <primary>table</primary>
   <secondary>inheritance</secondary>
  </indexterm>

  <para>
   <productname>PostgreSQL</productname> implements table inheritance,
   which can be a useful tool for database designers.  (SQL:1999 and
   later define a type inheritance feature, which differs in many
   respects from the features described here.)
  </para>

  <para>
   Let's start with an example: suppose we are trying to build a data
   model for cities.  Each state has many cities, but only one
   capital. We want to be able to quickly retrieve the capital city
   for any particular state. This can be done by creating two tables,
   one for state capitals and one for cities that are not
   capitals. However, what happens when we want to ask for data about
   a city, regardless of whether it is a capital or not? The
   inheritance feature can help to resolve this problem. We define the
   <structname>capitals</structname> table so that it inherits from
   <structname>cities</structname>:

<programlisting>
CREATE TABLE cities (
    name            text,
    population      float,
    altitude        int     -- in feet
);

CREATE TABLE capitals (
    state           char(2)
) INHERITS (cities);
</programlisting>

   In this case, the <structname>capitals</> table <firstterm>inherits</>
   all the columns of its parent table, <structname>cities</>. State
   capitals also have an extra column, <structfield>state</>, that shows
   their state.
  </para>

  <para>
   In <productname>PostgreSQL</productname>, a table can inherit from
   zero or more other tables, and a query can reference either all
   rows of a table or all rows of a table plus all of its descendant tables.
   The latter behavior is the default.
   For example, the following query finds the names of all cities,
   including state capitals, that are located at an altitude over
   500 feet:

<programlisting>
SELECT name, altitude
    FROM cities
    WHERE altitude &gt; 500;
</programlisting>

   Given the sample data from the <productname>PostgreSQL</productname>
   tutorial (see <xref linkend="tutorial-sql-intro">), this returns:

<programlisting>
   name    | altitude
-----------+----------
 Las Vegas |     2174
 Mariposa  |     1953
 Madison   |      845
</programlisting>
  </para>

  <para>
   On the other hand, the following query finds all the cities that
   are not state capitals and are situated at an altitude over 500 feet:

<programlisting>
SELECT name, altitude
    FROM ONLY cities
    WHERE altitude &gt; 500;

   name    | altitude
-----------+----------
 Las Vegas |     2174
 Mariposa  |     1953
</programlisting>
  </para>

  <para>
   Here the <literal>ONLY</literal> keyword indicates that the query
   should apply only to <structname>cities</structname>, and not any tables
   below <structname>cities</structname> in the inheritance hierarchy.  Many
   of the commands that we have already discussed &mdash;
   <command>SELECT</command>, <command>UPDATE</command> and
   <command>DELETE</command> &mdash; support the
   <literal>ONLY</literal> keyword.
  </para>

  <para>
   You can also write the table name with a trailing <literal>*</>
   to explicitly specify that descendant tables are included:

<programlisting>
SELECT name, altitude
    FROM cities*
    WHERE altitude &gt; 500;
</programlisting>

   Writing <literal>*</> is not necessary, since this behavior is always
   the default.  However, this syntax is still supported for
   compatibility with older releases where the default could be changed.
  </para>

  <para>
   In some cases you might wish to know which table a particular row
   originated from. There is a system column called
   <structfield>tableoid</structfield> in each table which can tell you the
   originating table:

<programlisting>
SELECT c.tableoid, c.name, c.altitude
FROM cities c
WHERE c.altitude &gt; 500;
</programlisting>

   which returns:

<programlisting>
 tableoid |   name    | altitude
----------+-----------+----------
   139793 | Las Vegas |     2174
   139793 | Mariposa  |     1953
   139798 | Madison   |      845
</programlisting>

   (If you try to reproduce this example, you will probably get
   different numeric OIDs.)  By doing a join with
   <structname>pg_class</> you can see the actual table names:

<programlisting>
SELECT p.relname, c.name, c.altitude
FROM cities c, pg_class p
WHERE c.altitude &gt; 500 AND c.tableoid = p.oid;
</programlisting>

   which returns:

<programlisting>
 relname  |   name    | altitude
----------+-----------+----------
 cities   | Las Vegas |     2174
 cities   | Mariposa  |     1953
 capitals | Madison   |      845
</programlisting>
  </para>

  <para>
   Another way to get the same effect is to use the <type>regclass</>
   alias type, which will print the table OID symbolically:

<programlisting>
SELECT c.tableoid::regclass, c.name, c.altitude
FROM cities c
WHERE c.altitude &gt; 500;
</programlisting>
  </para>

  <para>
   Inheritance does not automatically propagate data from
   <command>INSERT</command> or <command>COPY</command> commands to
   other tables in the inheritance hierarchy. In our example, the
   following <command>INSERT</command> statement will fail:
<programlisting>
INSERT INTO cities (name, population, altitude, state)
VALUES ('Albany', NULL, NULL, 'NY');
</programlisting>
   We might hope that the data would somehow be routed to the
   <structname>capitals</structname> table, but this does not happen:
   <command>INSERT</command> always inserts into exactly the table
   specified.  In some cases it is possible to redirect the insertion
   using a rule (see <xref linkend="rules">).  However that does not
   help for the above case because the <structname>cities</> table
   does not contain the column <structfield>state</>, and so the
   command will be rejected before the rule can be applied.
  </para>

  <para>
   All check constraints and not-null constraints on a parent table are
   automatically inherited by its children, unless explicitly specified
   otherwise with <literal>NO INHERIT</> clauses.  Other types of constraints
   (unique, primary key, and foreign key constraints) are not inherited.
  </para>

  <para>
   A table can inherit from more than one parent table, in which case it has
   the union of the columns defined by the parent tables.  Any columns
   declared in the child table's definition are added to these.  If the
   same column name appears in multiple parent tables, or in both a parent
   table and the child's definition, then these columns are <quote>merged</>
   so that there is only one such column in the child table.  To be merged,
   columns must have the same data types, else an error is raised.
   Inheritable check constraints and not-null constraints are merged in a
   similar fashion.  Thus, for example, a merged column will be marked
   not-null if any one of the column definitions it came from is marked
   not-null.  Check constraints are merged if they have the same name,
   and the merge will fail if their conditions are different.
  </para>

  <para>
   Table inheritance is typically established when the child table is
   created, using the <literal>INHERITS</> clause of the
   <xref linkend="sql-createtable">
   statement.
   Alternatively, a table which is already defined in a compatible way can
   have a new parent relationship added, using the <literal>INHERIT</literal>
   variant of <xref linkend="sql-altertable">.
   To do this the new child table must already include columns with
   the same names and types as the columns of the parent. It must also include
   check constraints with the same names and check expressions as those of the
   parent. Similarly an inheritance link can be removed from a child using the
   <literal>NO INHERIT</literal> variant of <command>ALTER TABLE</>.
   Dynamically adding and removing inheritance links like this can be useful
   when the inheritance relationship is being used for table
   partitioning (see <xref linkend="ddl-partitioning">).
  </para>

  <para>
   One convenient way to create a compatible table that will later be made
   a new child is to use the <literal>LIKE</literal> clause in <command>CREATE
   TABLE</command>. This creates a new table with the same columns as
   the source table. If there are any <literal>CHECK</literal>
   constraints defined on the source table, the <literal>INCLUDING
   CONSTRAINTS</literal> option to <literal>LIKE</literal> should be
   specified, as the new child must have constraints matching the parent
   to be considered compatible.
  </para>

  <para>
   A parent table cannot be dropped while any of its children remain. Neither
   can columns or check constraints of child tables be dropped or altered
   if they are inherited
   from any parent tables. If you wish to remove a table and all of its
   descendants, one easy way is to drop the parent table with the
   <literal>CASCADE</literal> option (see <xref linkend="ddl-depend">).
  </para>

  <para>
   <xref linkend="sql-altertable"> will
   propagate any changes in column data definitions and check
   constraints down the inheritance hierarchy.  Again, dropping
   columns that are depended on by other tables is only possible when using
   the <literal>CASCADE</literal> option. <command>ALTER
   TABLE</command> follows the same rules for duplicate column merging
   and rejection that apply during <command>CREATE TABLE</command>.
  </para>

  <para>
   Inherited queries perform access permission checks on the parent table
   only.  Thus, for example, granting <literal>UPDATE</> permission on
   the <structname>cities</> table implies permission to update rows in
   the <structname>capitals</structname> table as well, when they are
   accessed through <structname>cities</>.  This preserves the appearance
   that the data is (also) in the parent table.  But
   the <structname>capitals</structname> table could not be updated directly
   without an additional grant.  In a similar way, the parent table's row
   security policies (see <xref linkend="ddl-rowsecurity">) are applied to
   rows coming from child tables during an inherited query.  A child table's
   policies, if any, are applied only when it is the table explicitly named
   in the query; and in that case, any policies attached to its parent(s) are
   ignored.
  </para>

  <para>
   Foreign tables (see <xref linkend="ddl-foreign-data">) can also
   be part of inheritance hierarchies, either as parent or child
   tables, just as regular tables can be.  If a foreign table is part
   of an inheritance hierarchy then any operations not supported by
   the foreign table are not supported on the whole hierarchy either.
  </para>

 <sect2 id="ddl-inherit-caveats">
  <title>Caveats</title>

  <para>
   Note that not all SQL commands are able to work on
   inheritance hierarchies.  Commands that are used for data querying,
   data modification, or schema modification
   (e.g., <literal>SELECT</literal>, <literal>UPDATE</literal>, <literal>DELETE</literal>,
   most variants of <literal>ALTER TABLE</literal>, but
   not <literal>INSERT</literal> or <literal>ALTER TABLE ...
   RENAME</literal>) typically default to including child tables and
   support the <literal>ONLY</literal> notation to exclude them.
   Commands that do database maintenance and tuning
   (e.g., <literal>REINDEX</literal>, <literal>VACUUM</literal>)
   typically only work on individual, physical tables and do not
   support recursing over inheritance hierarchies.  The respective
   behavior of each individual command is documented in its reference
   page (<xref linkend="sql-commands">).
  </para>

  <para>
   A serious limitation of the inheritance feature is that indexes (including
   unique constraints) and foreign key constraints only apply to single
   tables, not to their inheritance children. This is true on both the
   referencing and referenced sides of a foreign key constraint. Thus,
   in the terms of the above example:

   <itemizedlist>
    <listitem>
     <para>
      If we declared <structname>cities</>.<structfield>name</> to be
      <literal>UNIQUE</> or a <literal>PRIMARY KEY</>, this would not stop the
      <structname>capitals</> table from having rows with names duplicating
      rows in <structname>cities</>.  And those duplicate rows would by
      default show up in queries from <structname>cities</>.  In fact, by
      default <structname>capitals</> would have no unique constraint at all,
      and so could contain multiple rows with the same name.
      You could add a unique constraint to <structname>capitals</>, but this
      would not prevent duplication compared to <structname>cities</>.
     </para>
    </listitem>

    <listitem>
     <para>
      Similarly, if we were to specify that
      <structname>cities</>.<structfield>name</> <literal>REFERENCES</> some
      other table, this constraint would not automatically propagate to
      <structname>capitals</>.  In this case you could work around it by
      manually adding the same <literal>REFERENCES</> constraint to
      <structname>capitals</>.
     </para>
    </listitem>

    <listitem>
     <para>
      Specifying that another table's column <literal>REFERENCES
      cities(name)</> would allow the other table to contain city names, but
      not capital names.  There is no good workaround for this case.
     </para>
    </listitem>
   </itemizedlist>

   These deficiencies will probably be fixed in some future release,
   but in the meantime considerable care is needed in deciding whether
   inheritance is useful for your application.
  </para>

   </sect2>
  </sect1>

  <sect1 id="ddl-partitioning">
   <title>Table Partitioning</title>

   <indexterm>
    <primary>partitioning</primary>
   </indexterm>

   <indexterm>
    <primary>table</primary>
    <secondary>partitioning</secondary>
   </indexterm>

   <indexterm>
    <primary>partitioned table</primary>
   </indexterm>

   <para>
    <productname>PostgreSQL</productname> supports basic table
    partitioning. This section describes why and how to implement
    partitioning as part of your database design.
   </para>

   <sect2 id="ddl-partitioning-overview">
     <title>Overview</title>

    <para>
     Partitioning refers to splitting what is logically one large table into
     smaller physical pieces.  Partitioning can provide several benefits:
    <itemizedlist>
     <listitem>
      <para>
       Query performance can be improved dramatically in certain situations,
       particularly when most of the heavily accessed rows of the table are in a
       single partition or a small number of partitions.  The partitioning
       substitutes for leading columns of indexes, reducing index size and
       making it more likely that the heavily-used parts of the indexes
       fit in memory.
      </para>
     </listitem>

     <listitem>
      <para>
       When queries or updates access a large percentage of a single
       partition, performance can be improved by taking advantage
       of sequential scan of that partition instead of using an
       index and random access reads scattered across the whole table.
      </para>
     </listitem>

     <listitem>
      <para>
       Bulk loads and deletes can be accomplished by adding or removing
       partitions, if that requirement is planned into the partitioning design.
       Doing <command>ALTER TABLE DETACH PARTITION</> or dropping an individual
       partition using <command>DROP TABLE</> is far faster than a bulk
       operation.  These commands also entirely avoid the
       <command>VACUUM</command> overhead caused by a bulk <command>DELETE</>.
      </para>
     </listitem>

     <listitem>
      <para>
       Seldom-used data can be migrated to cheaper and slower storage media.
      </para>
     </listitem>
    </itemizedlist>

     The benefits will normally be worthwhile only when a table would
     otherwise be very large. The exact point at which a table will
     benefit from partitioning depends on the application, although a
     rule of thumb is that the size of the table should exceed the physical
     memory of the database server.
    </para>

    <para>
     <productname>PostgreSQL</productname> offers built-in support for the
     following forms of partitioning:

     <variablelist>
      <varlistentry>
       <term>Range Partitioning</term>

       <listitem>
        <para>
         The table is partitioned into <quote>ranges</quote> defined
         by a key column or set of columns, with no overlap between
         the ranges of values assigned to different partitions.  For
         example, one might partition by date ranges, or by ranges of
         identifiers for particular business objects.
        </para>
       </listitem>
      </varlistentry>

      <varlistentry>
       <term>List Partitioning</term>

       <listitem>
        <para>
         The table is partitioned by explicitly listing which key values
         appear in each partition.
        </para>
       </listitem>
      </varlistentry>
     </variablelist>

     If your application needs to use other forms of partitioning not listed
     above, alternative methods such as inheritance and
     <literal>UNION ALL</literal> views can be used instead.  Such methods
     offer flexibility but do not have some of the performance benefits
     of built-in declarative partitioning.
    </para>
   </sect2>

  <sect2 id="ddl-partitioning-declarative">
   <title>Declarative Partitioning</title>

   <para>
    <productname>PostgreSQL</productname> offers a way to specify how to
    divide a table into pieces called partitions.  The table that is divided
    is referred to as a <firstterm>partitioned table</firstterm>.  The
    specification consists of the <firstterm>partitioning method</firstterm>
    and a list of columns or expressions to be used as the 
    <firstterm>partition key</firstterm>.
   </para>

   <para>
    All rows inserted into a partitioned table will be routed to one of the
    <firstterm>partitions</firstterm> based on the value of the partition
    key.  Each partition has a subset of the data defined by its
    <firstterm>partition bounds</firstterm>.  Currently supported
    partitioning methods include range and list, where each partition is
    assigned a range of keys and a list of keys, respectively.
   </para>

   <para>
    Partitions may themselves be defined as partitioned tables, using what is
    called <firstterm>sub-partitioning</firstterm>.  Partitions may have their
    own indexes, constraints and default values, distinct from those of other
    partitions.  Indexes must be created separately for each partition.  See
    <xref linkend="sql-createtable"> for more details on creating partitioned
    tables and partitions.
   </para>

   <para>
    It is not possible to turn a regular table into a partitioned table or
    vice versa.  However, it is possible to add a regular or partitioned table
    containing data as a partition of a partitioned table, or remove a
    partition from a partitioned table turning it into a standalone table;
    see <xref linkend="sql-altertable"> to learn more about the
    <command>ATTACH PARTITION</> and <command>DETACH PARTITION</>
    sub-commands.
   </para>

   <para>
    Individual partitions are linked to the partitioned table with inheritance
    behind-the-scenes; however, it is not possible to use some of the
    inheritance features discussed in the previous section with partitioned
    tables and partitions.  For example, a partition cannot have any parents
    other than the partitioned table it is a partition of, nor can a regular
    table inherit from a partitioned table making the latter its parent.
    That means partitioned tables and partitions do not participate in
    inheritance with regular tables.  Since a partition hierarchy consisting
    of the partitioned table and its partitions is still an inheritance
    hierarchy, all the normal rules of inheritance apply as described in
    <xref linkend="ddl-inherit"> with some exceptions, most notably:

    <itemizedlist>
     <listitem>
      <para>
       Both <literal>CHECK</literal> and <literal>NOT NULL</literal>
       constraints of a partitioned table are always inherited by all its
       partitions.  <literal>CHECK</literal> constraints that are marked
       <literal>NO INHERIT</literal> are not allowed to be created on
       partitioned tables.
      </para>
     </listitem>

     <listitem>
      <para>
       Using <literal>ONLY</literal> to add or drop a constraint on only the
       partitioned table is supported when there are no partitions.  Once
       partitions exist, using <literal>ONLY</literal> will result in an error
       as adding or dropping constraints on only the partitioned table, when
       partitions exist, is not supported.  Instead, constraints can be added
       or dropped, when they are not present in the parent table, directly on
       the partitions.  As a partitioned table does not have any data
       directly, attempts to use <command>TRUNCATE</command>
       <literal>ONLY</literal> on a partitioned table will always return an
       error.
      </para>
     </listitem>

     <listitem>
      <para>
       Partitions cannot have columns that are not present in the parent.  It
       is neither possible to specify columns when creating partitions with
       <command>CREATE TABLE</> nor is it possible to add columns to
       partitions after-the-fact using <command>ALTER TABLE</>.  Tables may be
       added as a partition with <command>ALTER TABLE ... ATTACH PARTITION</>
       only if their columns exactly match the parent, including any
       <literal>oid</literal> column.
      </para>
     </listitem>

     <listitem>
      <para>
       You cannot drop the <literal>NOT NULL</literal> constraint on a
       partition's column if the constraint is present in the parent table.
      </para>
     </listitem>
    </itemizedlist>
   </para>

   <para>
    Partitions can also be foreign tables
    (see <xref linkend="sql-createforeigntable">),
    although these have some limitations that normal tables do not.  For
    example, data inserted into the partitioned table is not routed to
    foreign table partitions.
   </para>

   <sect3 id="ddl-partitioning-declarative-example">
    <title>Example</title>

   <para>
    Suppose we are constructing a database for a large ice cream company.
    The company measures peak temperatures every day as well as ice cream
    sales in each region. Conceptually, we want a table like:

<programlisting>
CREATE TABLE measurement (
    city_id         int not null,
    logdate         date not null,
    peaktemp        int,
    unitsales       int
);
</programlisting>

    We know that most queries will access just the last week's, month's or
    quarter's data, since the main use of this table will be to prepare
    online reports for management.  To reduce the amount of old data that
    needs to be stored, we decide to only keep the most recent 3 years
    worth of data. At the beginning of each month we will remove the oldest
    month's data.  In this situation we can use partitioning to help us meet
    all of our different requirements for the measurements table.
   </para>

   <para>
    To use declarative partitioning in this case, use the following steps:

    <orderedlist spacing="compact">
     <listitem>
      <para>
       Create <structname>measurement</structname> table as a partitioned
       table by specifying the <literal>PARTITION BY</literal> clause, which
       includes the partitioning method (<literal>RANGE</literal> in this
       case) and the list of column(s) to use as the partition key.

<programlisting>
CREATE TABLE measurement (
    city_id         int not null,
    logdate         date not null,
    peaktemp        int,
    unitsales       int
) PARTITION BY RANGE (logdate);
</programlisting>
      </para>

      <para>
       You may decide to use multiple columns in the partition key for range
       partitioning, if desired.  Of course, this will often result in a larger
       number of partitions, each of which is individually smaller.  On the
       other hand, using fewer columns may lead to a coarser-grained
       partitioning criteria with smaller number of partitions.  A query
       accessing the partitioned table will have to scan fewer partitions if
       the conditions involve some or all of these columns.
       For example, consider a table range partitioned using columns
       <structfield>lastname</> and <structfield>firstname</> (in that order)
       as the partition key.
      </para>
     </listitem>

     <listitem>
      <para>
       Create partitions.  Each partition's definition must specify the bounds
       that correspond to the partitioning method and partition key of the
       parent.  Note that specifying bounds such that the new partition's
       values will overlap with those in one or more existing partitions will
       cause an error.  Inserting data into the parent table that does not map
       to one of the existing partitions will cause an error; appropriate
       partition must be added manually.
      </para>

      <para>
       Partitions thus created are in every way normal
       <productname>PostgreSQL</>
       tables (or, possibly, foreign tables).  It is possible to specify a
       tablespace and storage parameters for each partition separately.
      </para>

      <para>
       It is not necessary to create table constraints describing partition
       boundary condition for partitions.  Instead, partition constraints are
       generated implicitly from the partition bound specification whenever
       there is need to refer to them.

<programlisting>
CREATE TABLE measurement_y2006m02 PARTITION OF measurement
    FOR VALUES FROM ('2006-02-01') TO ('2006-03-01')

CREATE TABLE measurement_y2006m03 PARTITION OF measurement
    FOR VALUES FROM ('2006-03-01') TO ('2006-04-01')

...
CREATE TABLE measurement_y2007m11 PARTITION OF measurement
    FOR VALUES FROM ('2007-11-01') TO ('2007-12-01')

CREATE TABLE measurement_y2007m12 PARTITION OF measurement
    FOR VALUES FROM ('2007-12-01') TO ('2008-01-01')
    TABLESPACE fasttablespace;

CREATE TABLE measurement_y2008m01 PARTITION OF measurement
    FOR VALUES FROM ('2008-01-01') TO ('2008-02-01')
    TABLESPACE fasttablespace
    WITH (parallel_workers = 4);
</programlisting>
      </para>

      <para>
       To implement sub-partitioning, specify the
       <literal>PARTITION BY</literal> clause in the commands used to create
       individual partitions, for example:

<programlisting>
CREATE TABLE measurement_y2006m02 PARTITION OF measurement
    FOR VALUES FROM ('2006-02-01') TO ('2006-03-01')
    PARTITION BY RANGE (peaktemp);
</programlisting>

       After creating partitions of <structname>measurement_y2006m02</>,
       any data inserted into <structname>measurement</> that is mapped to
       <structname>measurement_y2006m02</> (or data that is directly inserted
       into <structname>measurement_y2006m02</>, provided it satisfies its
       partition constraint) will be further redirected to one of its
       partitions based on the <structfield>peaktemp</> column.  The partition
       key specified may overlap with the parent's partition key, although
       care should be taken when specifying the bounds of a sub-partition
       such that the set of data it accepts constitutes a subset of what
       the partition's own bounds allows; the system does not try to check
       whether that's really the case.
      </para>
     </listitem>

     <listitem>
      <para>
       Create an index on the key column(s), as well as any other indexes you
       might want for every partition.  (The key index is not strictly
       necessary, but in most scenarios it is helpful.  If you intend the key
       values to be unique then you should always create a unique or
       primary-key constraint for each partition.)

<programlisting>
CREATE INDEX ON measurement_y2006m02 (logdate);
CREATE INDEX ON measurement_y2006m03 (logdate);
...
CREATE INDEX ON measurement_y2007m11 (logdate);
CREATE INDEX ON measurement_y2007m12 (logdate);
CREATE INDEX ON measurement_y2008m01 (logdate);
</programlisting>
      </para>
     </listitem>

      <listitem>
       <para>
        Ensure that the <xref linkend="guc-constraint-exclusion">
        configuration parameter is not disabled in <filename>postgresql.conf</>.
        If it is, queries will not be optimized as desired.
       </para>
      </listitem>
    </orderedlist>
   </para>

   <para>
    In the above example we would be creating a new partition each month, so
    it might be wise to write a script that generates the required DDL
    automatically.
   </para>
   </sect3>

   <sect3 id="ddl-partitioning-declarative-maintenance">
    <title>Partition Maintenance</title>

    <para>
      Normally the set of partitions established when initially defining the
      table are not intended to remain static.  It is common to want to
      remove old partitions of data and periodically add new partitions for
      new data. One of the most important advantages of partitioning is
      precisely that it allows this otherwise painful task to be executed
      nearly instantaneously by manipulating the partition structure, rather
      than physically moving large amounts of data around.
    </para>

    <para>
     The simplest option for removing old data is to drop the partition that
     is no longer necessary:
<programlisting>
DROP TABLE measurement_y2006m02;
</programlisting>
     This can very quickly delete millions of records because it doesn't have
     to individually delete every record.  Note however that the above command
     requires taking an <literal>ACCESS EXCLUSIVE</literal> lock on the parent
     table.
    </para>

   <para>
     Another option that is often preferable is to remove the partition from
     the partitioned table but retain access to it as a table in its own
     right:

<programlisting>
ALTER TABLE measurement DETACH PARTITION measurement_y2006m02;
</programlisting>

     This allows further operations to be performed on the data before
     it is dropped. For example, this is often a useful time to back up
     the data using <command>COPY</>, <application>pg_dump</>, or
     similar tools. It might also be a useful time to aggregate data
     into smaller formats, perform other data manipulations, or run
     reports.
   </para>

   <para>
     Similarly we can add a new partition to handle new data. We can create an
     empty partition in the partitioned table just as the original partitions
     were created above:

<programlisting>
CREATE TABLE measurement_y2008m02 PARTITION OF measurement
    FOR VALUES FROM ('2008-02-01') TO ('2008-03-01')
    TABLESPACE fasttablespace;
</programlisting>

     As an alternative, it is sometimes more convenient to create the
     new table outside the partition structure, and make it a proper
     partition later. This allows the data to be loaded, checked, and
     transformed prior to it appearing in the partitioned table:

<programlisting>
CREATE TABLE measurement_y2008m02
  (LIKE measurement INCLUDING DEFAULTS INCLUDING CONSTRAINTS)
  TABLESPACE fasttablespace;

ALTER TABLE measurement_y2008m02 ADD CONSTRAINT y2008m02
   CHECK ( logdate &gt;= DATE '2008-02-01' AND logdate &lt; DATE '2008-03-01' );

\copy measurement_y2008m02 from 'measurement_y2008m02'
-- possibly some other data preparation work

ALTER TABLE measurement ATTACH PARTITION measurement_y2008m02
    FOR VALUES FROM ('2008-02-01') TO ('2008-03-01' );
</programlisting>
    </para>

    <para>
     Before running the <command>ATTACH PARTITION</> command, it is
     recommended to create a <literal>CHECK</> constraint on the table to
     be attached describing the desired partition constraint.  That way,
     the system will be able to skip the scan to validate the implicit
     partition constraint. Without such a constraint, the table will be
     scanned to validate the partition constraint while holding an
     <literal>ACCESS EXCLUSIVE</literal> lock on the parent table.
     One may then drop the constraint after <command>ATTACH PARTITION</>
     is finished, because it is no longer necessary.
    </para>
   </sect3>

   <sect3 id="ddl-partitioning-declarative-limitations">
    <title>Limitations</title>

   <para>
    The following limitations apply to partitioned tables:
    <itemizedlist>
     <listitem>
      <para>
       There is no facility available to create the matching indexes on all
       partitions automatically.  Indexes must be added to each partition with
       separate commands.  This also means that there is no way to create a
       primary key, unique constraint, or exclusion constraint spanning all
       partitions; it is only possible to constrain each leaf partition
       individually.
      </para>
     </listitem>

     <listitem>
      <para>
       Since primary keys are not supported on partitioned tables, foreign
       keys referencing partitioned tables are not supported, nor are foreign
       key references from a partitioned table to some other table.
      </para>
     </listitem>

     <listitem>
      <para>
       Using the <literal>ON CONFLICT</literal> clause with partitioned tables
       will cause an error, because unique or exclusion constraints can only be
       created on individual partitions.  There is no support for enforcing
       uniqueness (or an exclusion constraint) across an entire partitioning
       hierarchy.
      </para>
     </listitem>

     <listitem>
      <para>
       An <command>UPDATE</> that causes a row to move from one partition to
       another fails, because the new value of the row fails to satisfy the
       implicit partition constraint of the original partition.
      </para>
     </listitem>

     <listitem>
      <para>
       Row triggers, if necessary, must be defined on individual partitions,
       not the partitioned table.
      </para>
     </listitem>
    </itemizedlist>
    </para>
    </sect3>
   </sect2>

   <sect2 id="ddl-partitioning-implementation-inheritance">
    <title>Implementation Using Inheritance</title>
    <para>
     While the built-in declarative partitioning is suitable for most
     common use cases, there are some circumstances where a more flexible
     approach may be useful.  Partitioning can be implemented using table
     inheritance, which allows for several features which are not supported
     by declarative partitioning, such as:

     <itemizedlist>
      <listitem>
       <para>
        Partitioning enforces a rule that all partitions must have exactly
        the same set of columns as the parent, but table inheritance allows
        children to have extra columns not present in the parent.
       </para>
      </listitem>

      <listitem>
       <para>
        Table inheritance allows for multiple inheritance.
       </para>
      </listitem>

      <listitem>
       <para>
        Declarative partitioning only supports list and range partitioning,
        whereas table inheritance allows data to be divided in a manner of
        the user's choosing.  (Note, however, that if constraint exclusion is
        unable to prune partitions effectively, query performance will be very
        poor.)
       </para>
      </listitem>

      <listitem>
       <para>
        Some operations require a stronger lock when using declarative
        partitioning than when using table inheritance.  For example, adding
        or removing a partition to or from a partitioned table requires taking
        an <literal>ACCESS EXCLUSIVE</literal> lock on the parent table,
        whereas a <literal>SHARE UPDATE EXCLUSIVE</literal> lock is enough
        in the case of regular inheritance.
       </para>
      </listitem>
     </itemizedlist>
    </para>

    <sect3 id="ddl-partitioning-inheritance-example">
     <title>Example</title>

     <para>
      We use the same <structname>measurement</structname> table we used
      above.  To implement it as a partitioned table using inheritance, use
      the following steps:

      <orderedlist spacing="compact">
       <listitem>
        <para>
         Create the <quote>master</quote> table, from which all of the
         partitions will inherit.  This table will contain no data.  Do not
         define any check constraints on this table, unless you intend them
         to be applied equally to all partitions.  There is no point in
         defining any indexes or unique constraints on it, either.  For our
         example, master table is the <structname>measurement</structname>
         table as originally defined.
        </para>
       </listitem>

       <listitem>
        <para>
         Create several <quote>child</quote> tables that each inherit from
         the master table.  Normally, these tables will not add any columns
         to the set inherited from the master.  Just as with declarative
         partitioning, these partitions are in every way normal
         <productname>PostgreSQL</> tables (or foreign tables).
        </para>

        <para>
<programlisting>
CREATE TABLE measurement_y2006m02 () INHERITS (measurement);
CREATE TABLE measurement_y2006m03 () INHERITS (measurement);
...
CREATE TABLE measurement_y2007m11 () INHERITS (measurement);
CREATE TABLE measurement_y2007m12 () INHERITS (measurement);
CREATE TABLE measurement_y2008m01 () INHERITS (measurement);
</programlisting>
        </para>
       </listitem>

       <listitem>
        <para>
         Add non-overlapping table constraints to the partition tables to
         define the allowed key values in each partition.
        </para>

        <para>
         Typical examples would be:
<programlisting>
CHECK ( x = 1 )
CHECK ( county IN ( 'Oxfordshire', 'Buckinghamshire', 'Warwickshire' ))
CHECK ( outletID &gt;= 100 AND outletID &lt; 200 )
</programlisting>
         Ensure that the constraints guarantee that there is no overlap
         between the key values permitted in different partitions.  A common
         mistake is to set up range constraints like:
<programlisting>
CHECK ( outletID BETWEEN 100 AND 200 )
CHECK ( outletID BETWEEN 200 AND 300 )
</programlisting>
         This is wrong since it is not clear which partition the key value
         200 belongs in.
        </para>

        <para>
         It would be better to instead create partitions as follows:

<programlisting>
CREATE TABLE measurement_y2006m02 (
    CHECK ( logdate &gt;= DATE '2006-02-01' AND logdate &lt; DATE '2006-03-01' )
) INHERITS (measurement);

CREATE TABLE measurement_y2006m03 (
    CHECK ( logdate &gt;= DATE '2006-03-01' AND logdate &lt; DATE '2006-04-01' )
) INHERITS (measurement);

...
CREATE TABLE measurement_y2007m11 (
    CHECK ( logdate &gt;= DATE '2007-11-01' AND logdate &lt; DATE '2007-12-01' )
) INHERITS (measurement);

CREATE TABLE measurement_y2007m12 (
    CHECK ( logdate &gt;= DATE '2007-12-01' AND logdate &lt; DATE '2008-01-01' )
) INHERITS (measurement);

CREATE TABLE measurement_y2008m01 (
    CHECK ( logdate &gt;= DATE '2008-01-01' AND logdate &lt; DATE '2008-02-01' )
) INHERITS (measurement);
</programlisting>
        </para>
       </listitem>

       <listitem>
        <para>
         For each partition, create an index on the key column(s),
         as well as any other indexes you might want.
<programlisting>
CREATE INDEX measurement_y2006m02_logdate ON measurement_y2006m02 (logdate);
CREATE INDEX measurement_y2006m03_logdate ON measurement_y2006m03 (logdate);
CREATE INDEX measurement_y2007m11_logdate ON measurement_y2007m11 (logdate);
CREATE INDEX measurement_y2007m12_logdate ON measurement_y2007m12 (logdate);
CREATE INDEX measurement_y2008m01_logdate ON measurement_y2008m01 (logdate);
</programlisting>
        </para>
       </listitem>

       <listitem>
        <para>
         We want our application to be able to say <literal>INSERT INTO
         measurement ...</> and have the data be redirected into the
         appropriate partition table.  We can arrange that by attaching
         a suitable trigger function to the master table.
         If data will be added only to the latest partition, we can
         use a very simple trigger function:

<programlisting>
CREATE OR REPLACE FUNCTION measurement_insert_trigger()
RETURNS TRIGGER AS $$
BEGIN
    INSERT INTO measurement_y2008m01 VALUES (NEW.*);
    RETURN NULL;
END;
$$
LANGUAGE plpgsql;
</programlisting>
        </para>

        <para>
         After creating the function, we create a trigger which
         calls the trigger function:

<programlisting>
CREATE TRIGGER insert_measurement_trigger
    BEFORE INSERT ON measurement
    FOR EACH ROW EXECUTE PROCEDURE measurement_insert_trigger();
</programlisting>

         We must redefine the trigger function each month so that it always
         points to the current partition.  The trigger definition does
         not need to be updated, however.
        </para>

        <para>
         We might want to insert data and have the server automatically
         locate the partition into which the row should be added. We
         could do this with a more complex trigger function, for example:

<programlisting>
CREATE OR REPLACE FUNCTION measurement_insert_trigger()
RETURNS TRIGGER AS $$
BEGIN
    IF ( NEW.logdate &gt;= DATE '2006-02-01' AND
         NEW.logdate &lt; DATE '2006-03-01' ) THEN
        INSERT INTO measurement_y2006m02 VALUES (NEW.*);
    ELSIF ( NEW.logdate &gt;= DATE '2006-03-01' AND
            NEW.logdate &lt; DATE '2006-04-01' ) THEN
        INSERT INTO measurement_y2006m03 VALUES (NEW.*);
    ...
    ELSIF ( NEW.logdate &gt;= DATE '2008-01-01' AND
            NEW.logdate &lt; DATE '2008-02-01' ) THEN
        INSERT INTO measurement_y2008m01 VALUES (NEW.*);
    ELSE
        RAISE EXCEPTION 'Date out of range.  Fix the measurement_insert_trigger() function!';
    END IF;
    RETURN NULL;
END;
$$
LANGUAGE plpgsql;
</programlisting>

         The trigger definition is the same as before.
         Note that each <literal>IF</literal> test must exactly match the
         <literal>CHECK</literal> constraint for its partition.
        </para>

        <para>
         While this function is more complex than the single-month case,
         it doesn't need to be updated as often, since branches can be
         added in advance of being needed.
        </para>

        <note>
         <para>
          In practice it might be best to check the newest partition first,
          if most inserts go into that partition.  For simplicity we have
          shown the trigger's tests in the same order as in other parts
          of this example.
         </para>
        </note>

        <para>
         A different approach to redirecting inserts into the appropriate
         partition table is to set up rules, instead of a trigger, on the
         master table.  For example:

<programlisting>
CREATE RULE measurement_insert_y2006m02 AS
ON INSERT TO measurement WHERE
    ( logdate &gt;= DATE '2006-02-01' AND logdate &lt; DATE '2006-03-01' )
DO INSTEAD
    INSERT INTO measurement_y2006m02 VALUES (NEW.*);
...
CREATE RULE measurement_insert_y2008m01 AS
ON INSERT TO measurement WHERE
    ( logdate &gt;= DATE '2008-01-01' AND logdate &lt; DATE '2008-02-01' )
DO INSTEAD
    INSERT INTO measurement_y2008m01 VALUES (NEW.*);
</programlisting>

         A rule has significantly more overhead than a trigger, but the
         overhead is paid once per query rather than once per row, so this
         method might be advantageous for bulk-insert situations.  In most
         cases, however, the trigger method will offer better performance.
        </para>

        <para>
         Be aware that <command>COPY</> ignores rules.  If you want to
         use <command>COPY</> to insert data, you'll need to copy into the
         correct partition table rather than into the master. <command>COPY</>
         does fire triggers, so you can use it normally if you use the trigger
         approach.
        </para>

        <para>
         Another disadvantage of the rule approach is that there is no simple
         way to force an error if the set of rules doesn't cover the insertion
         date; the data will silently go into the master table instead.
        </para>
       </listitem>

       <listitem>
        <para>
         Ensure that the <xref linkend="guc-constraint-exclusion">
         configuration parameter is not disabled in
         <filename>postgresql.conf</>.
         If it is, queries will not be optimized as desired.
        </para>
       </listitem>
      </orderedlist>
     </para>

     <para>
      As we can see, a complex partitioning scheme could require a
      substantial amount of DDL.  In the above example we would be creating
      a new partition each month, so it might be wise to write a script that
      generates the required DDL automatically.
     </para>
    </sect3>

    <sect3 id="ddl-partitioning-inheritance-maintenance">
     <title>Partition Maintenance</title>
     <para>
      To remove old data quickly, simply drop the partition that is no longer
      necessary:
<programlisting>
DROP TABLE measurement_y2006m02;
</programlisting>
     </para>

    <para>
     To remove the partition from the partitioned table but retain access to
     it as a table in its own right:

<programlisting>
ALTER TABLE measurement_y2006m02 NO INHERIT measurement;
</programlisting>
    </para>

    <para>
     To add a new partition to handle new data, create an empty partition
     just as the original partitions were created above:

<programlisting>
CREATE TABLE measurement_y2008m02 (
    CHECK ( logdate &gt;= DATE '2008-02-01' AND logdate &lt; DATE '2008-03-01' )
) INHERITS (measurement);
</programlisting>

     Alternatively, one may want to create the new table outside the partition
     structure, and make it a partition after the data is loaded, checked,
     and transformed.

<programlisting>
CREATE TABLE measurement_y2008m02
  (LIKE measurement INCLUDING DEFAULTS INCLUDING CONSTRAINTS);
ALTER TABLE measurement_y2008m02 ADD CONSTRAINT y2008m02
   CHECK ( logdate &gt;= DATE '2008-02-01' AND logdate &lt; DATE '2008-03-01' );
\copy measurement_y2008m02 from 'measurement_y2008m02'
-- possibly some other data preparation work
ALTER TABLE measurement_y2008m02 INHERIT measurement;
</programlisting>
    </para>
   </sect3>

   <sect3 id="ddl-partitioning-inheritance-caveats">
    <title>Caveats</title>

    <para>
     The following caveats apply to partitioned tables implemented using
     inheritance:
     <itemizedlist>
      <listitem>
       <para>
        There is no automatic way to verify that all of the
        <literal>CHECK</literal> constraints are mutually
        exclusive.  It is safer to create code that generates
        partitions and creates and/or modifies associated objects than
        to write each by hand.
       </para>
      </listitem>

      <listitem>
       <para>
        The schemes shown here assume that the partition key column(s)
        of a row never change, or at least do not change enough to require
        it to move to another partition.  An <command>UPDATE</> that attempts
        to do that will fail because of the <literal>CHECK</> constraints.
        If you need to handle such cases, you can put suitable update triggers
        on the partition tables, but it makes management of the structure
        much more complicated.
       </para>
      </listitem>

      <listitem>
       <para>
        If you are using manual <command>VACUUM</command> or
        <command>ANALYZE</command> commands, don't forget that
        you need to run them on each partition individually. A command like:
<programlisting>
ANALYZE measurement;
</programlisting>
        will only process the master table.
       </para>
      </listitem>

      <listitem>
       <para>
        <command>INSERT</command> statements with <literal>ON CONFLICT</>
        clauses are unlikely to work as expected, as the <literal>ON CONFLICT</>
        action is only taken in case of unique violations on the specified
        target relation, not its child relations.
       </para>
      </listitem>

      <listitem>
       <para>
        Triggers or rules will be needed to route rows to the desired
        partition, unless the application is explicitly aware of the
        partitioning scheme.  Triggers may be complicated to write, and will
        be much slower than the tuple routing performed internally by
        declarative partitioning.
       </para>
      </listitem>
     </itemizedlist>
    </para>
   </sect3>
  </sect2>

  <sect2 id="ddl-partitioning-constraint-exclusion">
   <title>Partitioning and Constraint Exclusion</title>

   <indexterm>
    <primary>constraint exclusion</primary>
   </indexterm>

   <para>
    <firstterm>Constraint exclusion</> is a query optimization technique
    that improves performance for partitioned tables defined in the
    fashion described above (both declaratively partitioned tables and those
    implemented using inheritance).  As an example:

<programlisting>
SET constraint_exclusion = on;
SELECT count(*) FROM measurement WHERE logdate &gt;= DATE '2008-01-01';
</programlisting>

    Without constraint exclusion, the above query would scan each of
    the partitions of the <structname>measurement</> table. With constraint
    exclusion enabled, the planner will examine the constraints of each
    partition and try to prove that the partition need not
    be scanned because it could not contain any rows meeting the query's
    <literal>WHERE</> clause.  When the planner can prove this, it
    excludes the partition from the query plan.
   </para>

   <para>
    You can use the <command>EXPLAIN</> command to show the difference
    between a plan with <varname>constraint_exclusion</> on and a plan
    with it off.  A typical unoptimized plan for this type of table setup is:

<programlisting>
SET constraint_exclusion = off;
EXPLAIN SELECT count(*) FROM measurement WHERE logdate &gt;= DATE '2008-01-01';

                                          QUERY PLAN
-----------------------------------------------------------------------------------------------
 Aggregate  (cost=158.66..158.68 rows=1 width=0)
   -&gt;  Append  (cost=0.00..151.88 rows=2715 width=0)
         -&gt;  Seq Scan on measurement  (cost=0.00..30.38 rows=543 width=0)
               Filter: (logdate &gt;= '2008-01-01'::date)
         -&gt;  Seq Scan on measurement_y2006m02 measurement  (cost=0.00..30.38 rows=543 width=0)
               Filter: (logdate &gt;= '2008-01-01'::date)
         -&gt;  Seq Scan on measurement_y2006m03 measurement  (cost=0.00..30.38 rows=543 width=0)
               Filter: (logdate &gt;= '2008-01-01'::date)
...
         -&gt;  Seq Scan on measurement_y2007m12 measurement  (cost=0.00..30.38 rows=543 width=0)
               Filter: (logdate &gt;= '2008-01-01'::date)
         -&gt;  Seq Scan on measurement_y2008m01 measurement  (cost=0.00..30.38 rows=543 width=0)
               Filter: (logdate &gt;= '2008-01-01'::date)
</programlisting>

    Some or all of the partitions might use index scans instead of
    full-table sequential scans, but the point here is that there
    is no need to scan the older partitions at all to answer this query.
    When we enable constraint exclusion, we get a significantly
    cheaper plan that will deliver the same answer:

<programlisting>
SET constraint_exclusion = on;
EXPLAIN SELECT count(*) FROM measurement WHERE logdate &gt;= DATE '2008-01-01';
                                          QUERY PLAN
-----------------------------------------------------------------------------------------------
 Aggregate  (cost=63.47..63.48 rows=1 width=0)
   -&gt;  Append  (cost=0.00..60.75 rows=1086 width=0)
         -&gt;  Seq Scan on measurement  (cost=0.00..30.38 rows=543 width=0)
               Filter: (logdate &gt;= '2008-01-01'::date)
         -&gt;  Seq Scan on measurement_y2008m01 measurement  (cost=0.00..30.38 rows=543 width=0)
               Filter: (logdate &gt;= '2008-01-01'::date)
</programlisting>
   </para>

   <para>
    Note that constraint exclusion is driven only by <literal>CHECK</>
    constraints, not by the presence of indexes.  Therefore it isn't
    necessary to define indexes on the key columns.  Whether an index
    needs to be created for a given partition depends on whether you
    expect that queries that scan the partition will generally scan
    a large part of the partition or just a small part.  An index will
    be helpful in the latter case but not the former.
   </para>

   <para>
    The default (and recommended) setting of
    <xref linkend="guc-constraint-exclusion"> is actually neither
    <literal>on</> nor <literal>off</>, but an intermediate setting
    called <literal>partition</>, which causes the technique to be
    applied only to queries that are likely to be working on partitioned
    tables.  The <literal>on</> setting causes the planner to examine
    <literal>CHECK</> constraints in all queries, even simple ones that
    are unlikely to benefit.
   </para>

   <para>
    The following caveats apply to constraint exclusion, which is used by
    both inheritance and partitioned tables:

   <itemizedlist>
    <listitem>
     <para>
      Constraint exclusion only works when the query's <literal>WHERE</>
      clause contains constants (or externally supplied parameters).
      For example, a comparison against a non-immutable function such as
      <function>CURRENT_TIMESTAMP</function> cannot be optimized, since the
      planner cannot know which partition the function value might fall
      into at run time.
     </para>
    </listitem>

    <listitem>
     <para>
      Keep the partitioning constraints simple, else the planner may not be
      able to prove that partitions don't need to be visited.  Use simple
      equality conditions for list partitioning, or simple
      range tests for range partitioning, as illustrated in the preceding
      examples.  A good rule of thumb is that partitioning constraints should
      contain only comparisons of the partitioning column(s) to constants
      using B-tree-indexable operators, which applies even to partitioned
      tables, because only B-tree-indexable column(s) are allowed in the
      partition key.  (This is not a problem when using declarative
      partitioning, since the automatically generated constraints are simple
      enough to be understood by the planner.)
     </para>
    </listitem>

    <listitem>
     <para>
      All constraints on all partitions of the master table are examined
      during constraint exclusion, so large numbers of partitions are likely
      to increase query planning time considerably.  Partitioning using
      these techniques will work well with up to perhaps a hundred partitions;
      don't try to use many thousands of partitions.
     </para>
    </listitem>

   </itemizedlist>
   </para>
  </sect2>
 </sect1>

 <sect1 id="ddl-foreign-data">
  <title>Foreign Data</title>

   <indexterm>
    <primary>foreign data</primary>
   </indexterm>
   <indexterm>
    <primary>foreign table</primary>
   </indexterm>
   <indexterm>
    <primary>user mapping</primary>
   </indexterm>

   <para>
    <productname>PostgreSQL</productname> implements portions of the SQL/MED
    specification, allowing you to access data that resides outside
    PostgreSQL using regular SQL queries.  Such data is referred to as
    <firstterm>foreign data</>.  (Note that this usage is not to be confused
    with foreign keys, which are a type of constraint within the database.)
   </para>

   <para>
    Foreign data is accessed with help from a
    <firstterm>foreign data wrapper</firstterm>. A foreign data wrapper is a
    library that can communicate with an external data source, hiding the
    details of connecting to the data source and obtaining data from it.
    There are some foreign data wrappers available as <filename>contrib</>
    modules; see <xref linkend="contrib">.  Other kinds of foreign data
    wrappers might be found as third party products.  If none of the existing
    foreign data wrappers suit your needs, you can write your own; see <xref
    linkend="fdwhandler">.
   </para>

   <para>
    To access foreign data, you need to create a <firstterm>foreign server</>
    object, which defines how to connect to a particular external data source
    according to the set of options used by its supporting foreign data
    wrapper. Then you need to create one or more <firstterm>foreign
    tables</firstterm>, which define the structure of the remote data. A
    foreign table can be used in queries just like a normal table, but a
    foreign table has no storage in the PostgreSQL server.  Whenever it is
    used, <productname>PostgreSQL</productname> asks the foreign data wrapper
    to fetch data from the external source, or transmit data to the external
    source in the case of update commands.
   </para>

   <para>
    Accessing remote data may require authenticating to the external
    data source.  This information can be provided by a
    <firstterm>user mapping</>, which can provide additional data
    such as user names and passwords based
    on the current <productname>PostgreSQL</productname> role.
   </para>

   <para>
    For additional information, see
    <xref linkend="sql-createforeigndatawrapper">,
    <xref linkend="sql-createserver">,
    <xref linkend="sql-createusermapping">,
    <xref linkend="sql-createforeigntable">, and
    <xref linkend="sql-importforeignschema">.
   </para>
 </sect1>

 <sect1 id="ddl-others">
  <title>Other Database Objects</title>

  <para>
   Tables are the central objects in a relational database structure,
   because they hold your data.  But they are not the only objects
   that exist in a database.  Many other kinds of objects can be
   created to make the use and management of the data more efficient
   or convenient.  They are not discussed in this chapter, but we give
   you a list here so that you are aware of what is possible:
  </para>

  <itemizedlist>
   <listitem>
    <para>
     Views
    </para>
   </listitem>

   <listitem>
    <para>
     Functions and operators
    </para>
   </listitem>

   <listitem>
    <para>
     Data types and domains
    </para>
   </listitem>

   <listitem>
    <para>
     Triggers and rewrite rules
    </para>
   </listitem>
  </itemizedlist>

  <para>
   Detailed information on
   these topics appears in <xref linkend="server-programming">.
  </para>
 </sect1>

 <sect1 id="ddl-depend">
  <title>Dependency Tracking</title>

  <indexterm zone="ddl-depend">
   <primary>CASCADE</primary>
   <secondary sortas="DROP">with DROP</secondary>
  </indexterm>

  <indexterm zone="ddl-depend">
   <primary>RESTRICT</primary>
   <secondary sortas="DROP">with DROP</secondary>
  </indexterm>

  <para>
   When you create complex database structures involving many tables
   with foreign key constraints, views, triggers, functions, etc. you
   implicitly create a net of dependencies between the objects.
   For instance, a table with a foreign key constraint depends on the
   table it references.
  </para>

  <para>
   To ensure the integrity of the entire database structure,
   <productname>PostgreSQL</productname> makes sure that you cannot
   drop objects that other objects still depend on.  For example,
   attempting to drop the products table we considered in <xref
   linkend="ddl-constraints-fk">, with the orders table depending on
   it, would result in an error message like this:
<screen>
DROP TABLE products;

ERROR:  cannot drop table products because other objects depend on it
DETAIL:  constraint orders_product_no_fkey on table orders depends on table products
HINT:  Use DROP ... CASCADE to drop the dependent objects too.
</screen>
   The error message contains a useful hint: if you do not want to
   bother deleting all the dependent objects individually, you can run:
<screen>
DROP TABLE products CASCADE;
</screen>
   and all the dependent objects will be removed, as will any objects
   that depend on them, recursively.  In this case, it doesn't remove
   the orders table, it only removes the foreign key constraint.
   It stops there because nothing depends on the foreign key constraint.
   (If you want to check what <command>DROP ... CASCADE</> will do,
   run <command>DROP</> without <literal>CASCADE</> and read the
   <literal>DETAIL</> output.)
  </para>

  <para>
   Almost all <command>DROP</> commands in <productname>PostgreSQL</> support
   specifying <literal>CASCADE</literal>.  Of course, the nature of
   the possible dependencies varies with the type of the object.  You
   can also write <literal>RESTRICT</literal> instead of
   <literal>CASCADE</literal> to get the default behavior, which is to
   prevent dropping objects that any other objects depend on.
  </para>

  <note>
   <para>
    According to the SQL standard, specifying either
    <literal>RESTRICT</literal> or <literal>CASCADE</literal> is
    required in a <command>DROP</> command.  No database system actually
    enforces that rule, but whether the default behavior
    is <literal>RESTRICT</literal> or <literal>CASCADE</literal> varies
    across systems.
   </para>
  </note>

  <para>
   If a <command>DROP</> command lists multiple
   objects, <literal>CASCADE</literal> is only required when there are
   dependencies outside the specified group.  For example, when saying
   <literal>DROP TABLE tab1, tab2</literal> the existence of a foreign
   key referencing <literal>tab1</> from <literal>tab2</> would not mean
   that <literal>CASCADE</literal> is needed to succeed.
  </para>

  <para>
   For user-defined functions, <productname>PostgreSQL</productname> tracks
   dependencies associated with a function's externally-visible properties,
   such as its argument and result types, but <emphasis>not</> dependencies
   that could only be known by examining the function body.  As an example,
   consider this situation:

<programlisting>
CREATE TYPE rainbow AS ENUM ('red', 'orange', 'yellow',
                             'green', 'blue', 'purple');

CREATE TABLE my_colors (color rainbow, note text);

CREATE FUNCTION get_color_note (rainbow) RETURNS text AS
  'SELECT note FROM my_colors WHERE color = $1'
  LANGUAGE SQL;
</programlisting>

   (See <xref linkend="xfunc-sql"> for an explanation of SQL-language
   functions.)  <productname>PostgreSQL</productname> will be aware that
   the <function>get_color_note</> function depends on the <type>rainbow</>
   type: dropping the type would force dropping the function, because its
   argument type would no longer be defined.  But <productname>PostgreSQL</>
   will not consider <function>get_color_note</> to depend on
   the <structname>my_colors</> table, and so will not drop the function if
   the table is dropped.  While there are disadvantages to this approach,
   there are also benefits.  The function is still valid in some sense if the
   table is missing, though executing it would cause an error; creating a new
   table of the same name would allow the function to work again.
  </para>
 </sect1>

</chapter>