Linux SRV-16 6.18.44-paas #1 SMP PREEMPT_DYNAMIC Thu Aug 13 10:08:12 UTC 2026 x86_64
/proc/self/root/opt/postgresql-11/doc/html/
//proc/self/root/opt/postgresql-11/doc/html/release-8-4.html
 <?xml version="1.0" encoding="UTF-8" standalone="no"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"><html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>E.181. Release 8.4</title><link rel="stylesheet" type="text/css" href="stylesheet.css" /><link rev="made" href="pgsql-docs@postgresql.org" /><meta name="generator" content="DocBook XSL Stylesheets V1.79.1" /><link rel="prev" href="release-8-4-1.html" title="E.180. Release 8.4.1" /><link rel="next" href="release-8-3-23.html" title="E.182. Release 8.3.23" /></head><body><div xmlns="http://www.w3.org/TR/xhtml1/transitional" class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="5" align="center">E.181. Release 8.4</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="release-8-4-1.html" title="E.180. Release 8.4.1">Prev</a> </td><td width="10%" align="left"><a accesskey="u" href="release.html" title="Appendix E. Release Notes">Up</a></td><th width="60%" align="center">Appendix E. Release Notes</th><td width="10%" align="right"><a accesskey="h" href="index.html" title="PostgreSQL 11.1 Documentation">Home</a></td><td width="10%" align="right"> <a accesskey="n" href="release-8-3-23.html" title="E.182. Release 8.3.23">Next</a></td></tr></table><hr></hr></div><div class="sect1" id="RELEASE-8-4"><div class="titlepage"><div><div><h2 class="title" style="clear: both">E.181. Release 8.4</h2></div></div></div><div class="toc"><dl class="toc"><dt><span class="sect2"><a href="release-8-4.html#id-1.11.6.185.3">E.181.1. Overview</a></span></dt><dt><span class="sect2"><a href="release-8-4.html#id-1.11.6.185.4">E.181.2. Migration to Version 8.4</a></span></dt><dt><span class="sect2"><a href="release-8-4.html#id-1.11.6.185.5">E.181.3. Changes</a></span></dt></dl></div><p><strong>Release date: </strong>2009-07-01</p><div class="sect2" id="id-1.11.6.185.3"><div class="titlepage"><div><div><h3 class="title">E.181.1. Overview</h3></div></div></div><p>
    After many years of development, <span class="productname">PostgreSQL</span> has
    become feature-complete in many areas.   This release shows a
    targeted approach to adding features (e.g., authentication,
    monitoring, space reuse), and adds capabilities defined in the
    later SQL standards.  The major areas of enhancement are:
   </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
      Windowing Functions
     </p></li><li class="listitem"><p>
      Common Table Expressions and Recursive Queries
     </p></li><li class="listitem"><p>
      Default and variadic parameters for functions
     </p></li><li class="listitem"><p>
      Parallel Restore
     </p></li><li class="listitem"><p>
      Column Permissions
     </p></li><li class="listitem"><p>
      Per-database locale settings
     </p></li><li class="listitem"><p>
      Improved hash indexes
     </p></li><li class="listitem"><p>
      Improved join performance for <code class="literal">EXISTS</code> and <code class="literal">NOT EXISTS</code> queries
     </p></li><li class="listitem"><p>
      Easier-to-use Warm Standby
     </p></li><li class="listitem"><p>
      Automatic sizing of the Free Space Map
     </p></li><li class="listitem"><p>
      Visibility Map (greatly reduces vacuum overhead for slowly-changing tables)
     </p></li><li class="listitem"><p>
      Version-aware psql (backslash commands work against older servers)
     </p></li><li class="listitem"><p>
      Support SSL certificates for user authentication
     </p></li><li class="listitem"><p>
      Per-function runtime statistics
     </p></li><li class="listitem"><p>
      Easy editing of functions in psql
     </p></li><li class="listitem"><p>
      New contrib modules: pg_stat_statements, auto_explain, citext, btree_gin
     </p></li></ul></div><p>
    The above items are explained in more detail in the sections below.
   </p></div><div class="sect2" id="id-1.11.6.185.4"><div class="titlepage"><div><div><h3 class="title">E.181.2. Migration to Version 8.4</h3></div></div></div><p>
    A dump/restore using <span class="application">pg_dump</span> is
    required for those wishing to migrate data from any previous
    release.
   </p><p>
    Observe the following incompatibilities:
   </p><div class="sect3" id="id-1.11.6.185.4.4"><div class="titlepage"><div><div><h4 class="title">E.181.2.1. General</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Use 64-bit integer datetimes by default (Neil Conway)
      </p><p>
       Previously this was selected by <span class="application">configure</span>'s
       <code class="option">--enable-integer-datetimes</code> option.  To retain
       the old behavior, build with <code class="option">--disable-integer-datetimes</code>.
      </p></li><li class="listitem"><p>
       Remove <span class="application">ipcclean</span> utility command (Bruce)
      </p><p>
       The utility only worked on a few platforms.  Users should use
       their operating system tools instead.
      </p></li></ul></div></div><div class="sect3" id="id-1.11.6.185.4.5"><div class="titlepage"><div><div><h4 class="title">E.181.2.2. Server Settings</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Change default setting for
       <code class="literal">log_min_messages</code> to <code class="literal">warning</code> (previously
       it was <code class="literal">notice</code>) to reduce log file volume (Tom)
      </p></li><li class="listitem"><p>
       Change default setting for <code class="literal">max_prepared_transactions</code> to
       zero (previously it was 5) (Tom)
      </p></li><li class="listitem"><p>
       Make <code class="literal">debug_print_parse</code>, <code class="literal">debug_print_rewritten</code>,
       and <code class="literal">debug_print_plan</code>
       output appear at <code class="literal">LOG</code> message level, not
       <code class="literal">DEBUG1</code> as formerly (Tom)
      </p></li><li class="listitem"><p>
       Make <code class="literal">debug_pretty_print</code> default to <code class="literal">on</code> (Tom)
      </p></li><li class="listitem"><p>
       Remove <code class="varname">explain_pretty_print</code> parameter (no longer needed) (Tom)
      </p></li><li class="listitem"><p>
       Make <code class="varname">log_temp_files</code> settable by superusers only, like other
       logging options (Simon Riggs)
      </p></li><li class="listitem"><p>
       Remove automatic appending of the epoch timestamp when no <code class="literal">%</code>
       escapes are present in <code class="literal">log_filename</code> (Robert Haas)
      </p><p>
       This change was made because some users wanted a fixed log filename,
       for use with an external log rotation tool.
      </p></li><li class="listitem"><p>
       Remove <code class="varname">log_restartpoints</code> from <code class="filename">recovery.conf</code>;
       instead use <code class="varname">log_checkpoints</code> (Simon)
      </p></li><li class="listitem"><p>
       Remove <code class="varname">krb_realm</code> and <code class="varname">krb_server_hostname</code>;
       these are now set in <code class="filename">pg_hba.conf</code> instead (Magnus)
      </p></li><li class="listitem"><p>
       There are also significant changes in <a class="link" href="release-8-4.html#RELEASE-8-4-PG-HBA-CONF" title="E.181.3.2.3. pg_hba.conf"><code class="filename">pg_hba.conf</code></a>,
       as described below.
      </p></li></ul></div></div><div class="sect3" id="id-1.11.6.185.4.6"><div class="titlepage"><div><div><h4 class="title">E.181.2.3. Queries</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Change <code class="command">TRUNCATE</code> and <code class="command">LOCK</code> to
       apply to child tables of the specified table(s) (Peter)
      </p><p>
       These commands now accept an <code class="literal">ONLY</code> option that prevents
       processing child tables; this option must be used if the old
       behavior is needed.
      </p></li><li class="listitem"><p>
       <code class="command">SELECT DISTINCT</code> and
       <code class="literal">UNION</code>/<code class="literal">INTERSECT</code>/<code class="literal">EXCEPT</code>
       no longer always produce sorted output (Tom)
      </p><p>
       Previously, these types of queries always removed duplicate rows
       by means of Sort/Unique processing (i.e., sort then remove adjacent
       duplicates).  Now they can be implemented by hashing, which will not
       produce sorted output.  If an application relied on the output being
       in sorted order, the recommended fix is to add an <code class="literal">ORDER BY</code>
       clause.  As a short-term workaround, the previous behavior can be
       restored by disabling <code class="literal">enable_hashagg</code>, but that is a very
       performance-expensive fix.  <code class="literal">SELECT DISTINCT ON</code> never uses
       hashing, however, so its behavior is unchanged.
      </p></li><li class="listitem"><p>
       Force child tables to inherit <code class="literal">CHECK</code> constraints from parents
       (Alex Hunsaker, Nikhil Sontakke, Tom)
      </p><p>
       Formerly it was possible to drop such a constraint from a child
       table, allowing rows that violate the constraint to be visible
       when scanning the parent table.  This was deemed inconsistent,
       as well as contrary to SQL standard.
      </p></li><li class="listitem"><p>
       Disallow negative <code class="literal">LIMIT</code> or <code class="literal">OFFSET</code>
       values, rather than treating them as zero (Simon)
      </p></li><li class="listitem"><p>
       Disallow <code class="command">LOCK TABLE</code> outside a transaction block
       (Tom)
      </p><p>
       Such an operation is useless because the lock would be released
       immediately.
      </p></li><li class="listitem"><p>
       Sequences now contain an additional <code class="structfield">start_value</code> column
       (Zoltan Boszormenyi)
      </p><p>
       This supports <code class="command">ALTER SEQUENCE ... RESTART</code>.
      </p></li></ul></div></div><div class="sect3" id="id-1.11.6.185.4.7"><div class="titlepage"><div><div><h4 class="title">E.181.2.4. Functions and Operators</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Make <code class="type">numeric</code> zero raised to a fractional power return
       <code class="literal">0</code>, rather than throwing an error, and make
       <code class="type">numeric</code> zero raised to the zero power return <code class="literal">1</code>,
       rather than error (Bruce)
      </p><p>
       This matches the longstanding <code class="type">float8</code> behavior.
      </p></li><li class="listitem"><p>
       Allow unary minus of floating-point values to produce minus zero (Tom)
      </p><p>
       The changed behavior is more <acronym class="acronym">IEEE</acronym>-standard
       compliant.
      </p></li><li class="listitem"><p>
       Throw an error if an escape character is the last character in
       a <code class="literal">LIKE</code> pattern (i.e., it has nothing to escape) (Tom)
      </p><p>
       Previously, such an escape character was silently ignored,
       thus possibly masking application logic errors.
      </p></li><li class="listitem"><p>
       Remove <code class="literal">~=~</code> and <code class="literal">~&lt;&gt;~</code> operators
       formerly used for <code class="literal">LIKE</code> index comparisons (Tom)
      </p><p>
       Pattern indexes now use the regular equality operator.
      </p></li><li class="listitem"><p>
       <code class="function">xpath()</code> now passes its arguments to <span class="application">libxml</span>
       without any changes (Andrew)
      </p><p>
       This means that the XML argument must be a well-formed XML document.
       The previous coding attempted to allow XML fragments, but it did not
       work well.
      </p></li><li class="listitem"><p>
       Make <code class="function">xmlelement()</code> format attribute values just like
       content values (Peter)
      </p><p>
       Previously, attribute values were formatted according to the
       normal SQL output behavior, which is sometimes at odds with
       XML rules.
      </p></li><li class="listitem"><p>
       Rewrite memory management for <span class="application">libxml</span>-using functions
       (Tom)
      </p><p>
       This change should avoid some compatibility problems with use of
       <span class="application">libxml</span> in PL/Perl and other add-on code.
      </p></li><li class="listitem"><p>
       Adopt a faster algorithm for hash functions (Kenneth Marshall,
       based on work of Bob Jenkins)
      </p><p>
       Many of the built-in hash functions now deliver different results on
       little-endian and big-endian platforms.
      </p></li></ul></div><div class="sect4" id="id-1.11.6.185.4.7.3"><div class="titlepage"><div><div><h5 class="title">E.181.2.4.1. Temporal Functions and Operators</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        <code class="varname">DateStyle</code> no longer controls <code class="type">interval</code> output
        formatting; instead there is a new variable <code class="varname">IntervalStyle</code>
        (Ron Mayer)
       </p></li><li class="listitem"><p>
        Improve consistency of handling of fractional seconds in
        <code class="type">timestamp</code> and <code class="type">interval</code> output (Ron Mayer)
       </p><p>
        This may result in displaying a different number of fractional
        digits than before, or rounding instead of truncating.
       </p></li><li class="listitem"><p>
        Make <code class="function">to_char()</code>'s localized month/day names depend
        on <code class="varname">LC_TIME</code>, not <code class="varname">LC_MESSAGES</code> (Euler
        Taveira de Oliveira)
       </p></li><li class="listitem"><p>
        Cause <code class="function">to_date()</code> and <code class="function">to_timestamp()</code>
        to more consistently report errors for invalid input (Brendan
        Jurd)
       </p><p>
        Previous versions would often ignore or silently misread input
        that did not match the format string.  Such cases will now
        result in an error.
       </p></li><li class="listitem"><p>
        Fix <code class="function">to_timestamp()</code> to not require upper/lower case
        matching for meridian (<code class="literal">AM</code>/<code class="literal">PM</code>) and era
        (<code class="literal">BC</code>/<code class="literal">AD</code>) format designations  (Brendan
        Jurd)
       </p><p>
        For example, input value <code class="literal">ad</code> now matches the format
        string <code class="literal">AD</code>.
       </p></li></ul></div></div></div></div><div class="sect2" id="id-1.11.6.185.5"><div class="titlepage"><div><div><h3 class="title">E.181.3. Changes</h3></div></div></div><p>
    Below you will find a detailed account of the changes between
    <span class="productname">PostgreSQL</span> 8.4 and the previous major
    release.
   </p><div class="sect3" id="id-1.11.6.185.5.3"><div class="titlepage"><div><div><h4 class="title">E.181.3.1. Performance</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Improve optimizer statistics calculations (Jan Urbanski, Tom)
      </p><p>
       In particular, estimates for full-text-search operators are
       greatly improved.
      </p></li><li class="listitem"><p>
       Allow <code class="command">SELECT DISTINCT</code> and
       <code class="literal">UNION</code>/<code class="literal">INTERSECT</code>/<code class="literal">EXCEPT</code> to
       use hashing (Tom)
      </p><p>
       This means that these types of queries no longer automatically
       produce sorted output.
      </p></li><li class="listitem"><p>
       Create explicit concepts of semi-joins and anti-joins (Tom)
      </p><p>
       This work formalizes our previous ad-hoc treatment of <code class="literal">IN
       (SELECT ...)</code> clauses, and extends it to <code class="literal">EXISTS</code> and
       <code class="literal">NOT EXISTS</code> clauses.  It should result in significantly
       better planning of <code class="literal">EXISTS</code> and <code class="literal">NOT EXISTS</code>
       queries.  In general, logically equivalent <code class="literal">IN</code> and
       <code class="literal">EXISTS</code> clauses should now have similar performance,
       whereas previously <code class="literal">IN</code> often won.
      </p></li><li class="listitem"><p>
       Improve optimization of sub-selects beneath outer joins (Tom)
      </p><p>
       Formerly, a sub-select or view could not be optimized very well if it
       appeared within the nullable side of an outer join and contained
       non-strict expressions (for instance, constants) in its result list.
      </p></li><li class="listitem"><p>
       Improve the performance of <code class="function">text_position()</code> and
       related functions by using Boyer-Moore-Horspool searching (David
       Rowley)
      </p><p>
       This is particularly helpful for long search patterns.
      </p></li><li class="listitem"><p>
       Reduce I/O load of writing the statistics collection file
       by writing the file only when requested (Martin Pihlak)
      </p></li><li class="listitem"><p>
       Improve performance for bulk inserts (Robert Haas, Simon)
      </p></li><li class="listitem"><p>
       Increase the default value of <code class="varname">default_statistics_target</code>
       from <code class="literal">10</code> to <code class="literal">100</code> (Greg Sabino Mullane,
       Tom)
      </p><p>
       The maximum value was also increased from <code class="literal">1000</code> to
       <code class="literal">10000</code>.
      </p></li><li class="listitem"><p>
       Perform <code class="varname">constraint_exclusion</code> checking by default
       in queries involving inheritance or <code class="literal">UNION ALL</code> (Tom)
      </p><p>
       A new <code class="varname">constraint_exclusion</code> setting,
       <code class="literal">partition</code>, was added to specify this behavior.
      </p></li><li class="listitem"><p>
       Allow I/O read-ahead for bitmap index scans (Greg Stark)
      </p><p>
       The amount of read-ahead is controlled by
       <code class="varname">effective_io_concurrency</code>.  This feature is available only
       if the kernel has <code class="function">posix_fadvise()</code> support.
      </p></li><li class="listitem"><p>
       Inline simple set-returning <acronym class="acronym">SQL</acronym> functions in
       <code class="literal">FROM</code> clauses (Richard Rowell)
      </p></li><li class="listitem"><p>
       Improve performance of multi-batch hash joins by providing a special
       case for join key values that are especially common in the outer
       relation (Bryce Cutt, Ramon Lawrence)
      </p></li><li class="listitem"><p>
       Reduce volume of temporary data in multi-batch hash joins
       by suppressing <span class="quote">“<span class="quote">physical tlist</span>”</span> optimization (Michael
       Henderson, Ramon Lawrence)
      </p></li><li class="listitem"><p>
       Avoid waiting for idle-in-transaction sessions during
       <code class="command">CREATE INDEX CONCURRENTLY</code> (Simon)
      </p></li><li class="listitem"><p>
       Improve performance of shared cache invalidation (Tom)
      </p></li></ul></div></div><div class="sect3" id="id-1.11.6.185.5.4"><div class="titlepage"><div><div><h4 class="title">E.181.3.2. Server</h4></div></div></div><div class="sect4" id="id-1.11.6.185.5.4.2"><div class="titlepage"><div><div><h5 class="title">E.181.3.2.1. Settings</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Convert many <code class="filename">postgresql.conf</code> settings to enumerated
        values so that <code class="literal">pg_settings</code> can display the valid
        values (Magnus)
       </p></li><li class="listitem"><p>
        Add <code class="varname">cursor_tuple_fraction</code> parameter to control the
        fraction of a cursor's rows that the planner assumes will be
        fetched (Robert Hell)
       </p></li><li class="listitem"><p>
        Allow underscores in the names of custom variable
        classes in <code class="filename">postgresql.conf</code> (Tom)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.4.3"><div class="titlepage"><div><div><h5 class="title">E.181.3.2.2. Authentication and security</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Remove support for the (insecure) <code class="literal">crypt</code> authentication method
        (Magnus)
       </p><p>
        This effectively obsoletes pre-<span class="productname">PostgreSQL</span> 7.2 client
        libraries, as there is no longer any non-plaintext password method that
        they can use.
       </p></li><li class="listitem"><p>
        Support regular expressions in <code class="filename">pg_ident.conf</code>
        (Magnus)
       </p></li><li class="listitem"><p>
        Allow <span class="productname">Kerberos</span>/<acronym class="acronym">GSSAPI</acronym> parameters
        to be changed without restarting the postmaster (Magnus)
       </p></li><li class="listitem"><p>
        Support <acronym class="acronym">SSL</acronym> certificate chains in server certificate
        file (Andrew Gierth)
       </p><p>
        Including the full certificate chain makes the client able
        to verify the certificate without having all intermediate CA
        certificates present in the local store, which is often the case for
        commercial CAs.
       </p></li><li class="listitem"><p>
        Report appropriate error message for combination of <code class="literal">MD5</code>
        authentication and <code class="varname">db_user_namespace</code> enabled (Bruce)
       </p></li></ul></div></div><div class="sect4" id="RELEASE-8-4-PG-HBA-CONF"><div class="titlepage"><div><div><h5 class="title">E.181.3.2.3. <code class="filename">pg_hba.conf</code></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Change all authentication options to use <code class="literal">name=value</code>
        syntax (Magnus)
       </p><p>
        This makes incompatible changes to the <code class="literal">ldap</code>,
        <code class="literal">pam</code> and <code class="literal">ident</code> authentication methods. All
        <code class="filename">pg_hba.conf</code> entries with these methods need to be
        rewritten using the new format.
       </p></li><li class="listitem"><p>
        Remove the <code class="literal">ident sameuser</code> option, instead making that
        behavior the default if no usermap is specified (Magnus)
       </p></li><li class="listitem"><p>
        Allow a usermap parameter for all external authentication methods
        (Magnus)
       </p><p>
        Previously a usermap was only supported for <code class="literal">ident</code>
        authentication.
       </p></li><li class="listitem"><p>
        Add <code class="literal">clientcert</code> option to control requesting of a
        client certificate (Magnus)
       </p><p>
        Previously this was controlled by the presence of a root
        certificate file in the server's data directory.
       </p></li><li class="listitem"><p>
        Add <code class="literal">cert</code> authentication method to allow
        <span class="emphasis"><em>user</em></span> authentication via <acronym class="acronym">SSL</acronym> certificates
        (Magnus)
       </p><p>
        Previously <acronym class="acronym">SSL</acronym> certificates could only verify that
        the client had access to a certificate, not authenticate a
        user.
       </p></li><li class="listitem"><p>
        Allow <code class="literal">krb5</code>, <code class="literal">gssapi</code> and <code class="literal">sspi</code>
        realm and <code class="literal">krb5</code> host settings to be specified in
        <code class="filename">pg_hba.conf</code> (Magnus)
       </p><p>
        These override the settings in <code class="filename">postgresql.conf</code>.
       </p></li><li class="listitem"><p>
        Add <code class="varname">include_realm</code> parameter for <code class="literal">krb5</code>,
        <code class="literal">gssapi</code>, and <code class="literal">sspi</code> methods (Magnus)
       </p><p>
        This allows identical usernames from different realms to be
        authenticated as different database users using usermaps.
       </p></li><li class="listitem"><p>
        Parse <code class="filename">pg_hba.conf</code> fully when it is loaded,
        so that errors are reported immediately (Magnus)
       </p><p>
        Previously, most errors in the file wouldn't be detected until clients
        tried to connect, so an erroneous file could render the system
        unusable.  With the new behavior, if an error is detected during
        reload then the bad file is rejected and the postmaster continues
        to use its old copy.
       </p></li><li class="listitem"><p>
        Show all parsing errors in <code class="filename">pg_hba.conf</code> instead of
        aborting after the first one (Selena Deckelmann)
       </p></li><li class="listitem"><p>
        Support <code class="literal">ident</code> authentication over Unix-domain sockets
        on <span class="productname">Solaris</span> (Garick Hamlin)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.4.5"><div class="titlepage"><div><div><h5 class="title">E.181.3.2.4. Continuous Archiving</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Provide an option to <code class="function">pg_start_backup()</code> to force its
        implied checkpoint to finish as quickly as possible (Tom)
       </p><p>
        The default behavior avoids excess I/O consumption, but that is
        pointless if no concurrent query activity is going on.
       </p></li><li class="listitem"><p>
        Make <code class="function">pg_stop_backup()</code> wait for modified <acronym class="acronym">WAL</acronym>
        files to be archived (Simon)
       </p><p>
        This guarantees that the backup is valid at the time
        <code class="function">pg_stop_backup()</code> completes.
       </p></li><li class="listitem"><p>
        When archiving is enabled, rotate the last WAL segment at shutdown
        so that all transactions can be archived immediately
        (Guillaume Smet, Heikki)
       </p></li><li class="listitem"><p>
        Delay <span class="quote">“<span class="quote">smart</span>”</span> shutdown while a continuous archiving base backup
        is in progress (Laurenz Albe)
       </p></li><li class="listitem"><p>
        Cancel a continuous archiving base backup if <span class="quote">“<span class="quote">fast</span>”</span> shutdown
        is requested (Laurenz Albe)
       </p></li><li class="listitem"><p>
        Allow <code class="filename">recovery.conf</code> boolean variables to take the
        same range of string values as <code class="filename">postgresql.conf</code>
        boolean variables
        (Bruce)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.4.6"><div class="titlepage"><div><div><h5 class="title">E.181.3.2.5. Monitoring</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add <code class="function">pg_conf_load_time()</code> to report when
        the <span class="productname">PostgreSQL</span> configuration files were last loaded
        (George Gensure)
       </p></li><li class="listitem"><p>
        Add <code class="function">pg_terminate_backend()</code> to safely terminate a
        backend (the <code class="literal">SIGTERM</code> signal works also) (Tom, Bruce)
       </p><p>
        While it's always been possible to <code class="literal">SIGTERM</code> a single
        backend, this was previously considered unsupported; and testing
        of the case found some bugs that are now fixed.
       </p></li><li class="listitem"><p>
        Add ability to track user-defined functions' call counts and
        runtimes (Martin Pihlak)
       </p><p>
        Function statistics appear in a new system view,
        <code class="literal">pg_stat_user_functions</code>.  Tracking is controlled
        by the new parameter <code class="varname">track_functions</code>.
       </p></li><li class="listitem"><p>
        Allow specification of the maximum query string size in
        <code class="literal">pg_stat_activity</code> via new
        <code class="varname">track_activity_query_size</code> parameter (Thomas Lee)
       </p></li><li class="listitem"><p>
        Increase the maximum line length sent to <span class="application">syslog</span>, in
        hopes of improving performance (Tom)
       </p></li><li class="listitem"><p>
        Add read-only configuration variables <code class="varname">segment_size</code>,
        <code class="varname">wal_block_size</code>, and <code class="varname">wal_segment_size</code>
        (Bernd Helmle)
       </p></li><li class="listitem"><p>
        When reporting a deadlock, report the text of all queries involved
        in the deadlock to the server log  (Itagaki Takahiro)
       </p></li><li class="listitem"><p>
        Add <code class="function">pg_stat_get_activity(pid)</code> function to return
        information about a specific process id (Magnus)
       </p></li><li class="listitem"><p>
        Allow the location of the server's statistics file to be specified
        via <code class="varname">stats_temp_directory</code> (Magnus)
       </p><p>
        This allows the statistics file to be placed in a
        <acronym class="acronym">RAM</acronym>-resident directory to reduce I/O requirements.
        On startup/shutdown, the file is copied to its traditional location
        (<code class="literal">$PGDATA/global/</code>) so it is preserved across restarts.
       </p></li></ul></div></div></div><div class="sect3" id="id-1.11.6.185.5.5"><div class="titlepage"><div><div><h4 class="title">E.181.3.3. Queries</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Add support for <code class="literal">WINDOW</code> functions (Hitoshi Harada)
      </p></li><li class="listitem"><p>
       Add support for <code class="literal">WITH</code> clauses (CTEs), including <code class="literal">WITH
       RECURSIVE</code> (Yoshiyuki Asaba, Tatsuo Ishii, Tom)
      </p></li><li class="listitem"><p>
       Add <code class="command">TABLE</code> command (Peter)
      </p><p>
       <code class="literal">TABLE tablename</code> is a SQL standard short-hand for
       <code class="literal">SELECT * FROM tablename</code>.
      </p></li><li class="listitem"><p>
       Allow <code class="literal">AS</code> to be optional when specifying a
       <code class="command">SELECT</code> (or <code class="literal">RETURNING</code>) column output
       label (Hiroshi Saito)
      </p><p>
       This works so long as the column label is not any
       <span class="productname">PostgreSQL</span> keyword; otherwise <code class="literal">AS</code> is still
       needed.
      </p></li><li class="listitem"><p>
       Support set-returning functions in <code class="command">SELECT</code> result lists
       even for functions that return their result via a tuplestore (Tom)
      </p><p>
       In particular, this means that functions written in PL/pgSQL
       and other PL languages can now be called this way.
      </p></li><li class="listitem"><p>
       Support set-returning functions in the output of aggregation
       and grouping queries (Tom)
      </p></li><li class="listitem"><p>
       Allow <code class="command">SELECT FOR UPDATE</code>/<code class="literal">SHARE</code> to work
       on inheritance trees (Tom)
      </p></li><li class="listitem"><p>
       Add infrastructure for <acronym class="acronym">SQL/MED</acronym> (Martin Pihlak,
       Peter)
      </p><p>
       There are no remote or external <acronym class="acronym">SQL/MED</acronym> capabilities
       yet, but this change provides a standardized and future-proof
       system for managing connection information for modules like
       <code class="filename">dblink</code> and <code class="filename">plproxy</code>.
      </p></li><li class="listitem"><p>
       Invalidate cached plans when referenced schemas, functions, operators,
       or operator classes are modified (Martin Pihlak, Tom)
      </p><p>
       This improves the system's ability to respond to on-the-fly
       DDL changes.
      </p></li><li class="listitem"><p>
       Allow comparison of composite types and allow arrays of
       anonymous composite types (Tom)
      </p><p>
       This allows constructs such as
       <code class="literal">row(1, 1.1) = any (array[row(7, 7.7), row(1, 1.0)])</code>.
       This is particularly useful in recursive queries.
      </p></li><li class="listitem"><p>
       Add support for Unicode string literal and identifier specifications
       using code points, e.g. <code class="literal">U&amp;'d\0061t\+000061'</code>
       (Peter)
      </p></li><li class="listitem"><p>
       Reject <code class="literal">\000</code> in string literals and <code class="command">COPY</code> data
       (Tom)
      </p><p>
       Previously, this was accepted but had the effect of terminating
       the string contents.
      </p></li><li class="listitem"><p>
       Improve the parser's ability to report error locations (Tom)
      </p><p>
       An error location is now reported for many semantic errors,
       such as mismatched datatypes, that previously could not be localized.
      </p></li></ul></div><div class="sect4" id="id-1.11.6.185.5.5.3"><div class="titlepage"><div><div><h5 class="title">E.181.3.3.1. <code class="command">TRUNCATE</code></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Support statement-level <code class="literal">ON TRUNCATE</code> triggers (Simon)
       </p></li><li class="listitem"><p>
        Add <code class="literal">RESTART</code>/<code class="literal">CONTINUE IDENTITY</code> options
        for <code class="command">TRUNCATE TABLE</code>
        (Zoltan Boszormenyi)
       </p><p>
        The start value of a sequence can be changed by <code class="command">ALTER
        SEQUENCE START WITH</code>.
       </p></li><li class="listitem"><p>
        Allow <code class="command">TRUNCATE tab1, tab1</code> to succeed (Bruce)
       </p></li><li class="listitem"><p>
        Add a separate <code class="command">TRUNCATE</code> permission (Robert Haas)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.5.4"><div class="titlepage"><div><div><h5 class="title">E.181.3.3.2. <code class="command">EXPLAIN</code></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Make <code class="command">EXPLAIN VERBOSE</code> show the output columns of each
        plan node (Tom)
       </p><p>
        Previously <code class="command">EXPLAIN VERBOSE</code> output an internal
        representation of the query plan.  (That behavior is now
        available via <code class="varname">debug_print_plan</code>.)
       </p></li><li class="listitem"><p>
        Make <code class="command">EXPLAIN</code> identify subplans and initplans with
        individual labels (Tom)
       </p></li><li class="listitem"><p>
        Make <code class="command">EXPLAIN</code> honor <code class="varname">debug_print_plan</code> (Tom)
       </p></li><li class="listitem"><p>
        Allow <code class="command">EXPLAIN</code> on <code class="command">CREATE TABLE AS</code> (Peter)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.5.5"><div class="titlepage"><div><div><h5 class="title">E.181.3.3.3. <code class="literal">LIMIT</code>/<code class="literal">OFFSET</code></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Allow sub-selects in <code class="literal">LIMIT</code> and <code class="literal">OFFSET</code> (Tom)
       </p></li><li class="listitem"><p>
        Add <acronym class="acronym">SQL</acronym>-standard syntax for
        <code class="literal">LIMIT</code>/<code class="literal">OFFSET</code> capabilities (Peter)
       </p><p>
        To wit,
        <code class="literal">OFFSET num {ROW|ROWS} FETCH {FIRST|NEXT} [num] {ROW|ROWS}
        ONLY</code>.
       </p></li></ul></div></div></div><div class="sect3" id="id-1.11.6.185.5.6"><div class="titlepage"><div><div><h4 class="title">E.181.3.4. Object Manipulation</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Add support for column-level privileges (Stephen Frost, KaiGai
       Kohei)
      </p></li><li class="listitem"><p>
       Refactor multi-object <code class="command">DROP</code> operations to reduce the
       need for <code class="literal">CASCADE</code> (Alex Hunsaker)
      </p><p>
       For example, if table <code class="literal">B</code> has a dependency on table
       <code class="literal">A</code>, the command <code class="literal">DROP TABLE A, B</code> no longer
       requires the <code class="literal">CASCADE</code> option.
      </p></li><li class="listitem"><p>
       Fix various problems with concurrent <code class="command">DROP</code> commands
       by ensuring that locks are taken before we begin to drop dependencies
       of an object (Tom)
      </p></li><li class="listitem"><p>
       Improve reporting of dependencies during <code class="command">DROP</code>
       commands (Tom)
      </p></li><li class="listitem"><p>
       Add <code class="literal">WITH [NO] DATA</code> clause to <code class="command">CREATE TABLE
       AS</code>, per the <acronym class="acronym">SQL</acronym> standard (Peter, Tom)
      </p></li><li class="listitem"><p>
       Add support for user-defined I/O conversion casts (Heikki)
      </p></li><li class="listitem"><p>
       Allow <code class="command">CREATE AGGREGATE</code> to use an <code class="type">internal</code>
       transition datatype (Tom)
      </p></li><li class="listitem"><p>
       Add <code class="literal">LIKE</code> clause to <code class="command">CREATE TYPE</code> (Tom)
      </p><p>
       This simplifies creation of data types that use the same internal
       representation as an existing type.
      </p></li><li class="listitem"><p>
       Allow specification of the type category and <span class="quote">“<span class="quote">preferred</span>”</span>
       status for user-defined base types (Tom)
      </p><p>
       This allows more control over the coercion behavior of user-defined
       types.
      </p></li><li class="listitem"><p>
       Allow <code class="command">CREATE OR REPLACE VIEW</code> to add columns to the
       end of a view (Robert Haas)
      </p></li></ul></div><div class="sect4" id="id-1.11.6.185.5.6.3"><div class="titlepage"><div><div><h5 class="title">E.181.3.4.1. <code class="command">ALTER</code></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add <code class="command">ALTER TYPE RENAME</code> (Petr Jelinek)
       </p></li><li class="listitem"><p>
        Add <code class="command">ALTER SEQUENCE ... RESTART</code> (with no parameter) to
        reset a sequence to its initial value (Zoltan Boszormenyi)
       </p></li><li class="listitem"><p>
        Modify the <code class="command">ALTER TABLE</code> syntax to allow all reasonable
        combinations for tables, indexes, sequences, and views (Tom)
       </p><p>
        This change allows the following new syntaxes:

        </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: circle; "><li class="listitem"><p>
           <code class="command">ALTER SEQUENCE OWNER TO</code>
          </p></li><li class="listitem"><p>
           <code class="command">ALTER VIEW ALTER COLUMN SET/DROP DEFAULT</code>
          </p></li><li class="listitem"><p>
           <code class="command">ALTER VIEW OWNER TO</code>
          </p></li><li class="listitem"><p>
           <code class="command">ALTER VIEW SET SCHEMA</code>
          </p></li></ul></div><p>

        There is no actual new functionality here, but formerly
        you had to say <code class="command">ALTER TABLE</code> to do these things,
        which was confusing.
       </p></li><li class="listitem"><p>
        Add support for the syntax <code class="command">ALTER TABLE ... ALTER COLUMN
        ... SET DATA TYPE</code> (Peter)
       </p><p>
        This is <acronym class="acronym">SQL</acronym>-standard syntax for functionality that
        was already supported.
       </p></li><li class="listitem"><p>
        Make <code class="command">ALTER TABLE SET WITHOUT OIDS</code> rewrite the table
        to physically remove <code class="type">OID</code> values (Tom)
       </p><p>
        Also, add <code class="command">ALTER TABLE SET WITH OIDS</code> to rewrite the
        table to add <code class="type">OID</code>s.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.6.4"><div class="titlepage"><div><div><h5 class="title">E.181.3.4.2. Database Manipulation</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Improve reporting of
        <code class="command">CREATE</code>/<code class="command">DROP</code>/<code class="command">RENAME DATABASE</code>
        failure when uncommitted prepared transactions are the cause
        (Tom)
       </p></li><li class="listitem"><p>
        Make <code class="varname">LC_COLLATE</code> and <code class="varname">LC_CTYPE</code> into
        per-database settings (Radek Strnad, Heikki)
       </p><p>
        This makes collation similar to encoding, which was always
        configurable per database.
       </p></li><li class="listitem"><p>
        Improve checks that the database encoding, collation
        (<code class="varname">LC_COLLATE</code>), and character classes
        (<code class="varname">LC_CTYPE</code>) match (Heikki, Tom)
       </p><p>
        Note in particular that a new database's encoding and locale
        settings can be changed only when copying from <code class="literal">template0</code>.
        This prevents possibly copying data that doesn't match the settings.
       </p></li><li class="listitem"><p>
        Add <code class="command">ALTER DATABASE SET TABLESPACE</code> to move a database
        to a new tablespace (Guillaume Lelarge, Bernd Helmle)
       </p></li></ul></div></div></div><div class="sect3" id="id-1.11.6.185.5.7"><div class="titlepage"><div><div><h4 class="title">E.181.3.5. Utility Operations</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Add a <code class="literal">VERBOSE</code> option to the <code class="command">CLUSTER</code> command and
       <span class="application">clusterdb</span> (Jim Cox)
      </p></li><li class="listitem"><p>
       Decrease memory requirements for recording pending trigger
       events (Tom)
      </p></li></ul></div><div class="sect4" id="id-1.11.6.185.5.7.3"><div class="titlepage"><div><div><h5 class="title">E.181.3.5.1. Indexes</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Dramatically improve the speed of building and accessing hash
        indexes (Tom Raney, Shreya Bhargava)
       </p><p>
        This allows hash indexes to be sometimes faster than btree
        indexes.  However, hash indexes are still not crash-safe.
       </p></li><li class="listitem"><p>
        Make hash indexes store only the hash code, not the full value of
        the indexed column (Xiao Meng)
       </p><p>
        This greatly reduces the size of hash indexes for long indexed
        values, improving performance.
       </p></li><li class="listitem"><p>
        Implement fast update option for GIN indexes (Teodor, Oleg)
       </p><p>
        This option greatly improves update speed at a small penalty in search
        speed.
       </p></li><li class="listitem"><p>
        <code class="literal">xxx_pattern_ops</code> indexes can now be used for simple
        equality comparisons, not only for <code class="literal">LIKE</code> (Tom)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.7.4"><div class="titlepage"><div><div><h5 class="title">E.181.3.5.2. Full Text Indexes</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Remove the requirement to use <code class="literal">@@@</code> when doing
        <acronym class="acronym">GIN</acronym> weighted lookups on full text indexes (Tom, Teodor)
       </p><p>
        The normal <code class="literal">@@</code> text search operator can be used
        instead.
       </p></li><li class="listitem"><p>
        Add an optimizer selectivity function for <code class="literal">@@</code> text
        search operations (Jan Urbanski)
       </p></li><li class="listitem"><p>
        Allow prefix matching in full text searches (Teodor Sigaev,
        Oleg Bartunov)
       </p></li><li class="listitem"><p>
        Support multi-column <acronym class="acronym">GIN</acronym> indexes (Teodor Sigaev)
       </p></li><li class="listitem"><p>
        Improve support for Nepali language and Devanagari alphabet (Teodor)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.7.5"><div class="titlepage"><div><div><h5 class="title">E.181.3.5.3. <code class="command">VACUUM</code></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Track free space in separate per-relation <span class="quote">“<span class="quote">fork</span>”</span> files (Heikki)
       </p><p>
        Free space discovered by <code class="command">VACUUM</code> is now recorded in
        <code class="filename">*_fsm</code> files, rather than in a fixed-sized shared memory
        area.  The <code class="varname">max_fsm_pages</code> and <code class="varname">max_fsm_relations</code>
        settings have been removed, greatly simplifying administration of
        free space management.
       </p></li><li class="listitem"><p>
        Add a visibility map to track pages that do not require
        vacuuming (Heikki)
       </p><p>
        This allows <code class="command">VACUUM</code> to avoid scanning all of
        a table when only a portion of the table needs vacuuming.
        The visibility map is stored in per-relation <span class="quote">“<span class="quote">fork</span>”</span> files.
       </p></li><li class="listitem"><p>
        Add <code class="varname">vacuum_freeze_table_age</code> parameter to control
        when <code class="command">VACUUM</code> should ignore the visibility map and
        do a full table scan to freeze tuples (Heikki)
       </p></li><li class="listitem"><p>
        Track transaction snapshots more carefully (Alvaro)
       </p><p>
        This improves <code class="command">VACUUM</code>'s ability to reclaim space
        in the presence of long-running transactions.
       </p></li><li class="listitem"><p>
        Add ability to specify per-relation autovacuum and <acronym class="acronym">TOAST</acronym>
        parameters in <code class="command">CREATE TABLE</code> (Alvaro, Euler Taveira de
        Oliveira)
       </p><p>
        Autovacuum options used to be stored in a system table.
       </p></li><li class="listitem"><p>
        Add <code class="literal">--freeze</code> option to <span class="application">vacuumdb</span>
        (Bruce)
       </p></li></ul></div></div></div><div class="sect3" id="id-1.11.6.185.5.8"><div class="titlepage"><div><div><h4 class="title">E.181.3.6. Data Types</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Add a <code class="literal">CaseSensitive</code> option for text search synonym
       dictionaries (Simon)
      </p></li><li class="listitem"><p>
       Improve the precision of <code class="type">NUMERIC</code> division (Tom)
      </p></li><li class="listitem"><p>
       Add basic arithmetic operators for <code class="type">int2</code> with <code class="type">int8</code>
       (Tom)
      </p><p>
       This eliminates the need for explicit casting in some situations.
      </p></li><li class="listitem"><p>
       Allow <code class="type">UUID</code> input to accept an optional hyphen after
       every fourth digit (Robert Haas)
      </p></li><li class="listitem"><p>
       Allow <code class="literal">on</code>/<code class="literal">off</code> as input for the boolean data type
       (Itagaki Takahiro)
      </p></li><li class="listitem"><p>
       Allow spaces around <code class="literal">NaN</code> in the input string for
       type <code class="type">numeric</code> (Sam Mason)
      </p></li></ul></div><div class="sect4" id="id-1.11.6.185.5.8.3"><div class="titlepage"><div><div><h5 class="title">E.181.3.6.1. Temporal Data Types</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Reject year <code class="literal">0 BC</code> and years <code class="literal">000</code> and
        <code class="literal">0000</code> (Tom)
       </p><p>
        Previously these were interpreted as <code class="literal">1 BC</code>.
        (Note: years <code class="literal">0</code> and <code class="literal">00</code> are still assumed to be
        the year 2000.)
       </p></li><li class="listitem"><p>
        Include <code class="literal">SGT</code> (Singapore time) in the default list of
        known time zone abbreviations (Tom)
       </p></li><li class="listitem"><p>
        Support <code class="literal">infinity</code> and <code class="literal">-infinity</code> as
        values of type <code class="type">date</code> (Tom)
       </p></li><li class="listitem"><p>
        Make parsing of <code class="type">interval</code> literals more standard-compliant
        (Tom, Ron Mayer)
       </p><p>
        For example, <code class="literal">INTERVAL '1' YEAR</code> now does what it's
        supposed to.
       </p></li><li class="listitem"><p>
        Allow <code class="type">interval</code> fractional-seconds precision to be specified
        after the <code class="literal">second</code> keyword, for <acronym class="acronym">SQL</acronym> standard
        compliance (Tom)
       </p><p>
        Formerly the precision had to be specified after the keyword
        <code class="type">interval</code>.  (For backwards compatibility, this syntax is still
        supported, though deprecated.)  Data type definitions will now be
        output using the standard format.
       </p></li><li class="listitem"><p>
        Support the <acronym class="acronym">IS0 8601</acronym> <code class="type">interval</code> syntax (Ron
        Mayer, Kevin Grittner)
       </p><p>
        For example, <code class="literal">INTERVAL 'P1Y2M3DT4H5M6.7S'</code> is now
        supported.
       </p></li><li class="listitem"><p>
        Add <code class="varname">IntervalStyle</code> parameter
        which controls how <code class="type">interval</code> values are output (Ron Mayer)
       </p><p>
        Valid values are:  <code class="literal">postgres</code>, <code class="literal">postgres_verbose</code>,
        <code class="literal">sql_standard</code>, <code class="literal">iso_8601</code>.  This setting also
        controls the handling of negative <code class="type">interval</code> input when only
        some fields have positive/negative designations.
       </p></li><li class="listitem"><p>
        Improve consistency of handling of fractional seconds in
        <code class="type">timestamp</code> and <code class="type">interval</code> output (Ron Mayer)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.8.4"><div class="titlepage"><div><div><h5 class="title">E.181.3.6.2. Arrays</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Improve the handling of casts applied to <code class="literal">ARRAY[]</code>
        constructs, such as <code class="literal">ARRAY[...]::integer[]</code>
        (Brendan Jurd)
       </p><p>
        Formerly <span class="productname">PostgreSQL</span> attempted to determine a data type
        for the <code class="literal">ARRAY[]</code> construct without reference to the ensuing
        cast.  This could fail unnecessarily in many cases, in particular when
        the <code class="literal">ARRAY[]</code> construct was empty or contained only
        ambiguous entries such as <code class="literal">NULL</code>.  Now the cast is consulted
        to determine the type that the array elements must be.
       </p></li><li class="listitem"><p>
        Make <acronym class="acronym">SQL</acronym>-syntax <code class="type">ARRAY</code> dimensions optional
        to match the <acronym class="acronym">SQL</acronym> standard (Peter)
       </p></li><li class="listitem"><p>
        Add <code class="function">array_ndims()</code> to return the number
        of dimensions of an array (Robert Haas)
       </p></li><li class="listitem"><p>
        Add <code class="function">array_length()</code> to return the length
        of an array for a specified dimension (Jim Nasby, Robert
        Haas, Peter Eisentraut)
       </p></li><li class="listitem"><p>
        Add aggregate function <code class="function">array_agg()</code>, which
        returns all aggregated values as a single array (Robert Haas,
        Jeff Davis, Peter)
       </p></li><li class="listitem"><p>
        Add <code class="function">unnest()</code>, which converts an array to
        individual row values (Tom)
       </p><p>
        This is the opposite of <code class="function">array_agg()</code>.
       </p></li><li class="listitem"><p>
        Add <code class="function">array_fill()</code> to create arrays initialized with
        a value (Pavel Stehule)
       </p></li><li class="listitem"><p>
        Add <code class="function">generate_subscripts()</code> to simplify generating
        the range of an array's subscripts (Pavel Stehule)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.8.5"><div class="titlepage"><div><div><h5 class="title">E.181.3.6.3. Wide-Value Storage (<acronym class="acronym">TOAST</acronym>)</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Consider <acronym class="acronym">TOAST</acronym> compression on values as short as
        32 bytes (previously 256 bytes) (Greg Stark)
       </p></li><li class="listitem"><p>
        Require 25% minimum space savings before using <acronym class="acronym">TOAST</acronym>
        compression (previously 20% for small values and any-savings-at-all
        for large values) (Greg)
       </p></li><li class="listitem"><p>
        Improve <acronym class="acronym">TOAST</acronym> heuristics for rows that have a mix of large
        and small toastable fields, so that we prefer to push large values out
        of line and don't compress small values unnecessarily (Greg, Tom)
       </p></li></ul></div></div></div><div class="sect3" id="id-1.11.6.185.5.9"><div class="titlepage"><div><div><h4 class="title">E.181.3.7. Functions</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Document that <code class="function">setseed()</code> allows values from
       <code class="literal">-1</code> to <code class="literal">1</code> (not just <code class="literal">0</code> to
       <code class="literal">1</code>), and enforce the valid range (Kris Jurka)
      </p></li><li class="listitem"><p>
       Add server-side function <code class="function">lo_import(filename, oid)</code>
       (Tatsuo)
      </p></li><li class="listitem"><p>
       Add <code class="function">quote_nullable()</code>, which behaves like
       <code class="function">quote_literal()</code> but returns the string <code class="literal">NULL</code> for
       a null argument (Brendan Jurd)
      </p></li><li class="listitem"><p>
       Improve full text search <code class="function">headline()</code> function to
       allow extracting several fragments of text (Sushant Sinha)
      </p></li><li class="listitem"><p>
       Add <code class="function">suppress_redundant_updates_trigger()</code> trigger
       function to avoid overhead for non-data-changing updates (Andrew)
      </p></li><li class="listitem"><p>
       Add <code class="function">div(numeric, numeric)</code> to perform <code class="type">numeric</code>
       division without rounding (Tom)
      </p></li><li class="listitem"><p>
       Add <code class="type">timestamp</code> and <code class="type">timestamptz</code> versions of
       <code class="function">generate_series()</code> (Hitoshi Harada)
      </p></li></ul></div><div class="sect4" id="id-1.11.6.185.5.9.3"><div class="titlepage"><div><div><h5 class="title">E.181.3.7.1. Object Information Functions</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Implement <code class="function">current_query()</code> for use by functions
        that need to know the currently running query (Tomas Doran)
       </p></li><li class="listitem"><p>
        Add <code class="function">pg_get_keywords()</code> to return a list of the
        parser keywords (Dave Page)
       </p></li><li class="listitem"><p>
        Add <code class="function">pg_get_functiondef()</code> to see a function's
        definition (Abhijit Menon-Sen)
       </p></li><li class="listitem"><p>
        Allow the second argument of <code class="function">pg_get_expr()</code> to be zero
        when deparsing an expression that does not contain variables (Tom)
       </p></li><li class="listitem"><p>
        Modify <code class="function">pg_relation_size()</code> to use <code class="literal">regclass</code>
        (Heikki)
       </p><p>
        <code class="function">pg_relation_size(data_type_name)</code> no longer works.
       </p></li><li class="listitem"><p>
        Add <code class="literal">boot_val</code> and <code class="literal">reset_val</code> columns to
        <code class="literal">pg_settings</code> output (Greg Smith)
       </p></li><li class="listitem"><p>
        Add source file name and line number columns to
        <code class="literal">pg_settings</code> output for variables set in a configuration
        file (Magnus, Alvaro)
       </p><p>
        For security reasons, these columns are only visible to superusers.
       </p></li><li class="listitem"><p>
        Add support for <code class="varname">CURRENT_CATALOG</code>,
        <code class="varname">CURRENT_SCHEMA</code>, <code class="varname">SET CATALOG</code>, <code class="varname">SET
        SCHEMA</code> (Peter)
       </p><p>
        These provide <acronym class="acronym">SQL</acronym>-standard syntax for existing features.
       </p></li><li class="listitem"><p>
        Add <code class="function">pg_typeof()</code> which returns the data type
        of any value (Brendan Jurd)
       </p></li><li class="listitem"><p>
        Make <code class="function">version()</code> return information about whether
        the server is a 32- or 64-bit binary (Bruce)
       </p></li><li class="listitem"><p>
        Fix the behavior of information schema columns
        <code class="structfield">is_insertable_into</code> and <code class="structfield">is_updatable</code> to
        be consistent (Peter)
       </p></li><li class="listitem"><p>
        Improve the behavior of information schema
        <code class="structfield">datetime_precision</code> columns (Peter)
       </p><p>
        These columns now show zero for <code class="type">date</code> columns, and 6
        (the default precision) for <code class="type">time</code>, <code class="type">timestamp</code>, and
        <code class="type">interval</code> without a declared precision, rather than showing
        null as formerly.
       </p></li><li class="listitem"><p>
        Convert remaining builtin set-returning functions to use
        <code class="literal">OUT</code> parameters (Jaime Casanova)
       </p><p>
        This makes it possible to call these functions without specifying
        a column list:  <code class="function">pg_show_all_settings()</code>,
        <code class="function">pg_lock_status()</code>, <code class="function">pg_prepared_xact()</code>,
        <code class="function">pg_prepared_statement()</code>, <code class="function">pg_cursor()</code>
       </p></li><li class="listitem"><p>
        Make <code class="function">pg_*_is_visible()</code> and
        <code class="function">has_*_privilege()</code> functions return <code class="literal">NULL</code>
        for invalid OIDs, rather than reporting an error (Tom)
       </p></li><li class="listitem"><p>
        Extend <code class="function">has_*_privilege()</code> functions to allow inquiring
        about the OR of multiple privileges in one call (Stephen
        Frost, Tom)
       </p></li><li class="listitem"><p>
        Add <code class="function">has_column_privilege()</code> and
        <code class="function">has_any_column_privilege()</code> functions (Stephen
        Frost, Tom)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.9.4"><div class="titlepage"><div><div><h5 class="title">E.181.3.7.2. Function Creation</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Support variadic functions (functions with a variable number
        of arguments) (Pavel Stehule)
       </p><p>
        Only trailing arguments can be optional, and they all must be
        of the same data type.
       </p></li><li class="listitem"><p>
        Support default values for function arguments (Pavel Stehule)
       </p></li><li class="listitem"><p>
        Add <code class="command">CREATE FUNCTION ... RETURNS TABLE</code> clause (Pavel
        Stehule)
       </p></li><li class="listitem"><p>
        Allow <acronym class="acronym">SQL</acronym>-language functions to return the output
        of an <code class="command">INSERT</code>/<code class="command">UPDATE</code>/<code class="command">DELETE</code>
        <code class="literal">RETURNING</code> clause (Tom)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.9.5"><div class="titlepage"><div><div><h5 class="title">E.181.3.7.3. PL/pgSQL Server-Side Language</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Support <code class="literal">EXECUTE USING</code> for easier insertion of data
        values into a dynamic query string (Pavel Stehule)
       </p></li><li class="listitem"><p>
        Allow looping over the results of a cursor using a <code class="literal">FOR</code>
        loop (Pavel Stehule)
       </p></li><li class="listitem"><p>
        Support <code class="literal">RETURN QUERY EXECUTE</code> (Pavel
        Stehule)
       </p></li><li class="listitem"><p>
        Improve the <code class="literal">RAISE</code> command (Pavel Stehule)

        </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: circle; "><li class="listitem"><p>
           Support <code class="literal">DETAIL</code> and <code class="literal">HINT</code> fields
          </p></li><li class="listitem"><p>
           Support specification of the <code class="literal">SQLSTATE</code> error code
          </p></li><li class="listitem"><p>
           Support an exception name parameter
          </p></li><li class="listitem"><p>
           Allow <code class="literal">RAISE</code> without parameters in an exception
           block to re-throw the current error
          </p></li></ul></div><p>
       </p></li><li class="listitem"><p>
        Allow specification of <code class="varname">SQLSTATE</code> codes
        in <code class="literal">EXCEPTION</code> lists (Pavel Stehule)
       </p><p>
        This is useful for handling custom <code class="varname">SQLSTATE</code> codes.
       </p></li><li class="listitem"><p>
        Support the <code class="literal">CASE</code> statement (Pavel Stehule)
       </p></li><li class="listitem"><p>
        Make <code class="command">RETURN QUERY</code> set the special <code class="literal">FOUND</code> and
        <code class="command">GET DIAGNOSTICS</code> <code class="literal">ROW_COUNT</code> variables
        (Pavel Stehule)
       </p></li><li class="listitem"><p>
        Make <code class="command">FETCH</code> and <code class="command">MOVE</code> set the
        <code class="command">GET DIAGNOSTICS</code> <code class="literal">ROW_COUNT</code> variable
        (Andrew Gierth)
       </p></li><li class="listitem"><p>
        Make <code class="command">EXIT</code> without a label always exit the innermost
        loop (Tom)
       </p><p>
        Formerly, if there were a <code class="literal">BEGIN</code> block more closely nested
        than any loop, it would exit that block instead.  The new behavior
        matches Oracle(TM) and is also what was previously stated by our own
        documentation.
       </p></li><li class="listitem"><p>
        Make processing of string literals and nested block comments
        match the main SQL parser's processing (Tom)
       </p><p>
        In particular, the format string in <code class="command">RAISE</code> now works
        the same as any other string literal, including being subject
        to <code class="varname">standard_conforming_strings</code>.  This change also
        fixes other cases in which valid commands would fail when
        <code class="varname">standard_conforming_strings</code> is on.
       </p></li><li class="listitem"><p>
        Avoid memory leakage when the same function is called at varying
        exception-block nesting depths (Tom)
       </p></li></ul></div></div></div><div class="sect3" id="id-1.11.6.185.5.10"><div class="titlepage"><div><div><h4 class="title">E.181.3.8. Client Applications</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Fix <code class="literal">pg_ctl restart</code> to preserve command-line arguments
       (Bruce)
      </p></li><li class="listitem"><p>
       Add <code class="literal">-w</code>/<code class="literal">--no-password</code> option that
       prevents password prompting in all utilities that have a
       <code class="literal">-W</code>/<code class="literal">--password</code> option (Peter)
      </p></li><li class="listitem"><p>
       Remove <code class="option">-q</code> (quiet) option of <span class="application">createdb</span>,
       <span class="application">createuser</span>, <span class="application">dropdb</span>,
       <span class="application">dropuser</span> (Peter)
      </p><p>
       These options have had no effect since <span class="productname">PostgreSQL</span>
       8.3.
      </p></li></ul></div><div class="sect4" id="id-1.11.6.185.5.10.3"><div class="titlepage"><div><div><h5 class="title">E.181.3.8.1. <span class="application">psql</span></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Remove verbose startup banner; now just suggest <code class="literal">help</code>
        (Joshua Drake)
       </p></li><li class="listitem"><p>
        Make <code class="literal">help</code> show common backslash commands (Greg
        Sabino Mullane)
       </p></li><li class="listitem"><p>
        Add <code class="literal">\pset format wrapped</code> mode to wrap output to the
        screen width, or file/pipe output too if <code class="literal">\pset columns</code>
        is set (Bryce Nesbitt)
       </p></li><li class="listitem"><p>
        Allow all supported spellings of boolean values in <code class="command">\pset</code>,
        rather than just <code class="literal">on</code> and <code class="literal">off</code> (Bruce)
       </p><p>
        Formerly, any string other than <span class="quote">“<span class="quote">off</span>”</span> was silently taken
        to mean <code class="literal">true</code>.  <span class="application">psql</span> will now complain
        about unrecognized spellings (but still take them as <code class="literal">true</code>).
       </p></li><li class="listitem"><p>
        Use the pager for wide output (Bruce)
       </p></li><li class="listitem"><p>
        Require a space between a one-letter backslash command and its first
        argument (Bernd Helmle)
       </p><p>
        This removes a historical source of ambiguity.
       </p></li><li class="listitem"><p>
        Improve tab completion support for schema-qualified and
        quoted identifiers (Greg Sabino Mullane)
       </p></li><li class="listitem"><p>
        Add optional <code class="literal">on</code>/<code class="literal">off</code> argument for
        <code class="command">\timing</code> (David Fetter)
       </p></li><li class="listitem"><p>
        Display access control rights on multiple lines (Brendan
        Jurd, Andreas Scherbaum)
       </p></li><li class="listitem"><p>
        Make <code class="command">\l</code> show database access privileges (Andrew Gilligan)
       </p></li><li class="listitem"><p>
        Make <code class="command">\l+</code> show database sizes, if permissions
        allow (Andrew Gilligan)
       </p></li><li class="listitem"><p>
        Add the <code class="command">\ef</code> command to edit function definitions
        (Abhijit Menon-Sen)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.10.4"><div class="titlepage"><div><div><h5 class="title">E.181.3.8.2. <span class="application">psql</span> \d* commands</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Make <code class="command">\d*</code> commands that do not have a pattern argument
        show system objects only if the <code class="literal">S</code> modifier is specified
        (Greg Sabino Mullane, Bruce)
       </p><p>
        The former behavior was inconsistent across different variants
        of <code class="command">\d</code>, and in most cases it provided no easy way to see
        just user objects.
       </p></li><li class="listitem"><p>
        Improve <code class="command">\d*</code> commands to work with older
        <span class="productname">PostgreSQL</span> server versions (back to 7.4),
        not only the current server version
        (Guillaume Lelarge)
       </p></li><li class="listitem"><p>
        Make <code class="command">\d</code> show foreign-key constraints that reference
        the selected table (Kenneth D'Souza)
       </p></li><li class="listitem"><p>
        Make <code class="command">\d</code> on a sequence show its column values
        (Euler Taveira de Oliveira)
       </p></li><li class="listitem"><p>
        Add column storage type and other relation options to the
        <code class="command">\d+</code> display (Gregory Stark, Euler Taveira de
        Oliveira)
       </p></li><li class="listitem"><p>
        Show relation size in <code class="command">\dt+</code> output (Dickson S.
        Guedes)
       </p></li><li class="listitem"><p>
        Show the possible values of <code class="literal">enum</code> types in <code class="command">\dT+</code>
        (David Fetter)
       </p></li><li class="listitem"><p>
        Allow <code class="command">\dC</code> to accept a wildcard pattern, which matches
        either datatype involved in the cast (Tom)
       </p></li><li class="listitem"><p>
        Add a function type column to <code class="command">\df</code>'s output, and add
        options to list only selected types of functions (David Fetter)
       </p></li><li class="listitem"><p>
        Make <code class="command">\df</code> not hide functions that take or return
        type <code class="type">cstring</code> (Tom)
       </p><p>
        Previously, such functions were hidden because most of them are
        datatype I/O functions, which were deemed uninteresting.  The new
        policy about hiding system functions by default makes this wart
        unnecessary.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.10.5"><div class="titlepage"><div><div><h5 class="title">E.181.3.8.3. <span class="application">pg_dump</span></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add a <code class="literal">--no-tablespaces</code> option to
        <span class="application">pg_dump</span>/<span class="application">pg_dumpall</span>/<span class="application">pg_restore</span>
        so that dumps can be restored to clusters that have non-matching
        tablespace layouts (Gavin Roy)
       </p></li><li class="listitem"><p>
        Remove <code class="option">-d</code> and <code class="option">-D</code> options from
        <span class="application">pg_dump</span> and <span class="application">pg_dumpall</span> (Tom)
       </p><p>
        These options were too frequently confused with the option to
        select a database name in other <span class="productname">PostgreSQL</span>
        client applications.  The functionality is still available,
        but you must now spell out the long option name
        <code class="option">--inserts</code> or <code class="option">--column-inserts</code>.
       </p></li><li class="listitem"><p>
        Remove <code class="option">-i</code>/<code class="option">--ignore-version</code> option from
        <span class="application">pg_dump</span> and <span class="application">pg_dumpall</span> (Tom)
       </p><p>
        Use of this option does not throw an error, but it has no
        effect.  This option was removed because the version checks
        are necessary for safety.
       </p></li><li class="listitem"><p>
        Disable <code class="varname">statement_timeout</code> during dump and restore
        (Joshua Drake)
       </p></li><li class="listitem"><p>
        Add <span class="application">pg_dump</span>/<span class="application">pg_dumpall</span> option
        <code class="option">--lock-wait-timeout</code> (David Gould)
       </p><p>
        This allows dumps to fail if unable to acquire a shared lock
        within the specified amount of time.
       </p></li><li class="listitem"><p>
        Reorder <span class="application">pg_dump</span> <code class="literal">--data-only</code> output
        to dump tables referenced by foreign keys before
        the referencing tables (Tom)
       </p><p>
        This allows data loads when foreign keys are already present.
        If circular references make a safe ordering impossible, a
        <code class="literal">NOTICE</code> is issued.
       </p></li><li class="listitem"><p>
        Allow <span class="application">pg_dump</span>, <span class="application">pg_dumpall</span>, and
        <span class="application">pg_restore</span> to use a specified role (Benedek
        László)
       </p></li><li class="listitem"><p>
        Allow <span class="application">pg_restore</span> to use multiple concurrent
        connections to do the restore (Andrew)
       </p><p>
        The number of concurrent connections is controlled by the option
        <code class="literal">--jobs</code>.  This is supported only for custom-format archives.
       </p></li></ul></div></div></div><div class="sect3" id="id-1.11.6.185.5.11"><div class="titlepage"><div><div><h4 class="title">E.181.3.9. Programming Tools</h4></div></div></div><div class="sect4" id="id-1.11.6.185.5.11.2"><div class="titlepage"><div><div><h5 class="title">E.181.3.9.1. <span class="application">libpq</span></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Allow the <code class="type">OID</code> to be specified when importing a large
        object, via new function <code class="function">lo_import_with_oid()</code> (Tatsuo)
       </p></li><li class="listitem"><p>
        Add <span class="quote">“<span class="quote">events</span>”</span> support (Andrew Chernow, Merlin Moncure)
       </p><p>
        This adds the ability to register callbacks to manage private
        data associated with <code class="structname">PGconn</code> and <code class="structname">PGresult</code>
        objects.
       </p></li><li class="listitem"><p>
        Improve error handling to allow the return of multiple
        error messages as multi-line error reports (Magnus)
       </p></li><li class="listitem"><p>
        Make <code class="function">PQexecParams()</code> and related functions return
        <code class="varname">PGRES_EMPTY_QUERY</code> for an empty query (Tom)
       </p><p>
        They previously returned <code class="varname">PGRES_COMMAND_OK</code>.
       </p></li><li class="listitem"><p>
        Document how to avoid the overhead of <code class="function">WSACleanup()</code>
        on Windows (Andrew Chernow)
       </p></li><li class="listitem"><p>
        Do not rely on Kerberos tickets to determine the default database
        username (Magnus)
       </p><p>
        Previously, a Kerberos-capable build of libpq would use the
        principal name from any available Kerberos ticket as default
        database username, even if the connection wasn't using Kerberos
        authentication.  This was deemed inconsistent and confusing.
        The default username is now determined the same way with or
        without Kerberos.  Note however that the database username must still
        match the ticket when Kerberos authentication is used.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.11.3"><div class="titlepage"><div><div><h5 class="title">E.181.3.9.2. <span class="application">libpq</span> <acronym class="acronym">SSL</acronym> (Secure Sockets Layer)
      support</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Fix certificate validation for <acronym class="acronym">SSL</acronym> connections
        (Magnus)
       </p><p>
        <span class="application">libpq</span> now supports verifying both the certificate
        and the name of the server when making <acronym class="acronym">SSL</acronym>
        connections. If a root certificate is not available to use for
        verification, <acronym class="acronym">SSL</acronym> connections will fail. The
        <code class="literal">sslmode</code> parameter is used to enable certificate
        verification and set the level of checking.
        The default is still not to do any verification, allowing connections
        to SSL-enabled servers without requiring a root certificate on the
        client.
       </p></li><li class="listitem"><p>
        Support wildcard server certificates (Magnus)
       </p><p>
        If a certificate <acronym class="acronym">CN</acronym> starts with <code class="literal">*</code>, it will
        be treated as a wildcard when matching the hostname, allowing the
        use of the same certificate for multiple servers.
       </p></li><li class="listitem"><p>
        Allow the file locations for client certificates to be specified
        (Mark Woodward, Alvaro, Magnus)
       </p></li><li class="listitem"><p>
        Add a <code class="function">PQinitOpenSSL</code> function to allow greater control
        over OpenSSL/libcrypto initialization (Andrew Chernow)
       </p></li><li class="listitem"><p>
        Make <span class="application">libpq</span> unregister its <span class="application">OpenSSL</span>
        callbacks when no database connections remain open
        (Bruce, Magnus, Russell Smith)
       </p><p>
        This is required for applications that unload the libpq library,
        otherwise invalid <span class="application">OpenSSL</span> callbacks will remain.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.11.4"><div class="titlepage"><div><div><h5 class="title">E.181.3.9.3. <span class="application">ecpg</span></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add localization support for messages (Euler Taveira de
        Oliveira)
       </p></li><li class="listitem"><p>
        ecpg parser is now automatically generated from the server
        parser (Michael)
       </p><p>
        Previously the ecpg parser was hand-maintained.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.185.5.11.5"><div class="titlepage"><div><div><h5 class="title">E.181.3.9.4. Server Programming Interface (<acronym class="acronym">SPI</acronym>)</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add support for single-use plans with out-of-line
        parameters (Tom)
       </p></li><li class="listitem"><p>
        Add new <code class="varname">SPI_OK_REWRITTEN</code> return code for
        <code class="function">SPI_execute()</code> (Heikki)
       </p><p>
        This is used when a command is rewritten to another type of
        command.
       </p></li><li class="listitem"><p>
        Remove unnecessary inclusions from <code class="filename">executor/spi.h</code> (Tom)
       </p><p>
        SPI-using modules might need to add some <code class="literal">#include</code>
        lines if they were depending on <code class="filename">spi.h</code> to include
        things for them.
       </p></li></ul></div></div></div><div class="sect3" id="id-1.11.6.185.5.12"><div class="titlepage"><div><div><h4 class="title">E.181.3.10. Build Options</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Update build system to use <span class="productname">Autoconf</span> 2.61 (Peter)
      </p></li><li class="listitem"><p>
       Require <span class="productname">GNU bison</span> for source code builds (Peter)
      </p><p>
       This has effectively been required for several years, but now there
       is no infrastructure claiming to support other parser tools.
      </p></li><li class="listitem"><p>
       Add <span class="application">pg_config</span> <code class="literal">--htmldir</code> option
       (Peter)
      </p></li><li class="listitem"><p>
       Pass <code class="type">float4</code> by value inside the server (Zoltan
       Boszormenyi)
      </p><p>
       Add <span class="application">configure</span> option
       <code class="literal">--disable-float4-byval</code> to use the old behavior.
       External C functions that use old-style (version 0) call convention
       and pass or return <code class="type">float4</code> values will be broken by this
       change, so you may need the <span class="application">configure</span> option if you
       have such functions and don't want to update them.
      </p></li><li class="listitem"><p>
       Pass <code class="type">float8</code>, <code class="type">int8</code>, and related datatypes
       by value inside the server on 64-bit platforms (Zoltan Boszormenyi)
      </p><p>
       Add <span class="application">configure</span> option
       <code class="literal">--disable-float8-byval</code> to use the old behavior.
       As above, this change might break old-style external C functions.
      </p></li><li class="listitem"><p>
       Add configure options <code class="literal">--with-segsize</code>,
       <code class="literal">--with-blocksize</code>, <code class="literal">--with-wal-blocksize</code>,
       <code class="literal">--with-wal-segsize</code> (Zdenek Kotala, Tom)
      </p><p>
       This simplifies build-time control over several constants that
       previously could only be changed by editing
       <code class="filename">pg_config_manual.h</code>.
      </p></li><li class="listitem"><p>
       Allow threaded builds on <span class="productname">Solaris</span> 2.5 (Bruce)
      </p></li><li class="listitem"><p>
       Use the system's <code class="function">getopt_long()</code> on <span class="productname">Solaris</span>
       (Zdenek Kotala, Tom)
      </p><p>
       This makes option processing more consistent with what Solaris users
       expect.
      </p></li><li class="listitem"><p>
       Add support for the <span class="productname">Sun Studio</span> compiler on
       <span class="productname">Linux</span> (Julius Stroffek)
      </p></li><li class="listitem"><p>
       Append the major version number to the backend <span class="application">gettext</span>
       domain, and the <code class="literal">soname</code> major version number to
       libraries' <span class="application">gettext</span> domain (Peter)
      </p><p>
       This simplifies parallel installations of multiple versions.
      </p></li><li class="listitem"><p>
       Add support for code coverage testing with <span class="application">gcov</span>
       (Michelle Caisse)
      </p></li><li class="listitem"><p>
       Allow out-of-tree builds on <span class="productname">Mingw</span> and
       <span class="productname">Cygwin</span> (Richard Evans)
      </p></li><li class="listitem"><p>
       Fix the use of <span class="productname">Mingw</span> as a cross-compiling source
       platform (Peter)
      </p></li></ul></div></div><div class="sect3" id="id-1.11.6.185.5.13"><div class="titlepage"><div><div><h4 class="title">E.181.3.11. Source Code</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Support 64-bit time zone data files (Heikki)
      </p><p>
       This adds support for daylight saving time (<acronym class="acronym">DST</acronym>)
       calculations beyond the year 2038.
      </p></li><li class="listitem"><p>
       Deprecate use of platform's <code class="type">time_t</code> data type (Tom)
      </p><p>
       Some platforms have migrated to 64-bit <code class="type">time_t</code>, some have
       not, and Windows can't make up its mind what it's doing.  Define
       <code class="type">pg_time_t</code> to have the same meaning as <code class="type">time_t</code>,
       but always be 64 bits (unless the platform has no 64-bit integer type),
       and use that type in all module APIs and on-disk data formats.
      </p></li><li class="listitem"><p>
       Fix bug in handling of the time zone database when cross-compiling
       (Richard Evans)
      </p></li><li class="listitem"><p>
       Link backend object files in one step, rather than in stages
       (Peter)
      </p></li><li class="listitem"><p>
       Improve <span class="application">gettext</span> support to allow better translation
       of plurals (Peter)
      </p></li><li class="listitem"><p>
       Add message translation support to the PL languages (Alvaro, Peter)
      </p></li><li class="listitem"><p>
       Add more <span class="application">DTrace</span> probes (Robert Lor)
      </p></li><li class="listitem"><p>
       Enable <span class="application">DTrace</span> support on <span class="application">macOS
       Leopard</span> and other non-Solaris platforms (Robert Lor)
      </p></li><li class="listitem"><p>
       Simplify and standardize conversions between C strings and
       <code class="type">text</code> datums, by providing common functions for the purpose
       (Brendan Jurd, Tom)
      </p></li><li class="listitem"><p>
       Clean up the <code class="filename">include/catalog/</code> header files so that
       frontend programs can include them without including
       <code class="filename">postgres.h</code>
       (Zdenek Kotala)
      </p></li><li class="listitem"><p>
       Make <code class="type">name</code> char-aligned, and suppress zero-padding of
       <code class="type">name</code> entries in indexes (Tom)
      </p></li><li class="listitem"><p>
       Recover better if dynamically-loaded code executes <code class="function">exit()</code>
       (Tom)
      </p></li><li class="listitem"><p>
       Add a hook to let plug-ins monitor the executor (Itagaki
       Takahiro)
      </p></li><li class="listitem"><p>
       Add a hook to allow the planner's statistics lookup behavior to
       be overridden (Simon Riggs)
      </p></li><li class="listitem"><p>
       Add <code class="function">shmem_startup_hook()</code> for custom shared memory
       requirements (Tom)
      </p></li><li class="listitem"><p>
       Replace the index access method <code class="function">amgetmulti</code> entry point
       with <code class="function">amgetbitmap</code>, and extend the API for
       <code class="function">amgettuple</code> to support run-time determination of
       operator lossiness (Heikki, Tom, Teodor)
      </p><p>
       The API for GIN and GiST opclass <code class="function">consistent</code> functions
       has been extended as well.
      </p></li><li class="listitem"><p>
       Add support for partial-match searches in <acronym class="acronym">GIN</acronym> indexes
       (Teodor Sigaev, Oleg Bartunov)
      </p></li><li class="listitem"><p>
       Replace <code class="structname">pg_class</code> column <code class="structfield">reltriggers</code>
       with boolean <code class="structfield">relhastriggers</code> (Simon)
      </p><p>
       Also remove unused <code class="structname">pg_class</code> columns
       <code class="structfield">relukeys</code>, <code class="structfield">relfkeys</code>, and
       <code class="structfield">relrefs</code>.
      </p></li><li class="listitem"><p>
       Add a <code class="structfield">relistemp</code> column to <code class="structname">pg_class</code>
       to ease identification of temporary tables (Tom)
      </p></li><li class="listitem"><p>
       Move platform <acronym class="acronym">FAQ</acronym>s into the main documentation
       (Peter)
      </p></li><li class="listitem"><p>
       Prevent parser input files from being built with any conflicts
       (Peter)
      </p></li><li class="listitem"><p>
       Add support for the <code class="literal">KOI8U</code> (Ukrainian) encoding
       (Peter)
      </p></li><li class="listitem"><p>
       Add Japanese message translations (Japan PostgreSQL Users Group)
      </p><p>
       This used to be maintained as a separate project.
      </p></li><li class="listitem"><p>
       Fix problem when setting <code class="varname">LC_MESSAGES</code> on
       <span class="application">MSVC</span>-built systems (Hiroshi Inoue, Hiroshi
       Saito, Magnus)
      </p></li></ul></div></div><div class="sect3" id="id-1.11.6.185.5.14"><div class="titlepage"><div><div><h4 class="title">E.181.3.12. Contrib</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
       Add <code class="filename">contrib/auto_explain</code> to automatically run
       <code class="command">EXPLAIN</code> on queries exceeding a specified duration
       (Itagaki Takahiro, Tom)
      </p></li><li class="listitem"><p>
       Add <code class="filename">contrib/btree_gin</code> to allow GIN indexes to
       handle more datatypes (Oleg, Teodor)
      </p></li><li class="listitem"><p>
       Add <code class="filename">contrib/citext</code> to provide a case-insensitive,
       multibyte-aware text data type (David Wheeler)
      </p></li><li class="listitem"><p>
       Add <code class="filename">contrib/pg_stat_statements</code> for server-wide
       tracking of statement execution statistics (Itagaki Takahiro)
      </p></li><li class="listitem"><p>
       Add duration and query mode options to <code class="filename">contrib/pgbench</code>
       (Itagaki Takahiro)
      </p></li><li class="listitem"><p>
       Make <code class="filename">contrib/pgbench</code> use table names
       <code class="structname">pgbench_accounts</code>, <code class="structname">pgbench_branches</code>,
       <code class="structname">pgbench_history</code>, and <code class="structname">pgbench_tellers</code>,
       rather than just <code class="structname">accounts</code>, <code class="structname">branches</code>,
       <code class="structname">history</code>, and <code class="structname">tellers</code> (Tom)
      </p><p>
       This is to reduce the risk of accidentally destroying real data
       by running <span class="application">pgbench</span>.
      </p></li><li class="listitem"><p>
       Fix <code class="filename">contrib/pgstattuple</code> to handle tables and
       indexes with over 2 billion pages (Tatsuhito Kasahara)
      </p></li><li class="listitem"><p>
       In <code class="filename">contrib/fuzzystrmatch</code>, add a version of the
       Levenshtein string-distance function that allows the user to
       specify the costs of insertion, deletion, and substitution
       (Volkan Yazici)
      </p></li><li class="listitem"><p>
       Make <code class="filename">contrib/ltree</code> support multibyte encodings
       (laser)
      </p></li><li class="listitem"><p>
       Enable <code class="filename">contrib/dblink</code> to use connection information
       stored in the SQL/MED catalogs (Joe Conway)
      </p></li><li class="listitem"><p>
       Improve <code class="filename">contrib/dblink</code>'s reporting of errors from
       the remote server (Joe Conway)
      </p></li><li class="listitem"><p>
       Make <code class="filename">contrib/dblink</code> set <code class="varname">client_encoding</code>
       to match the local database's encoding (Joe Conway)
      </p><p>
       This prevents encoding problems when communicating with a remote
       database that uses a different encoding.
      </p></li><li class="listitem"><p>
       Make sure <code class="filename">contrib/dblink</code> uses a password supplied
       by the user, and not accidentally taken from the server's
       <code class="filename">.pgpass</code> file (Joe Conway)
      </p><p>
       This is a minor security enhancement.
      </p></li><li class="listitem"><p>
       Add <code class="function">fsm_page_contents()</code>
       to <code class="filename">contrib/pageinspect</code> (Heikki)
      </p></li><li class="listitem"><p>
       Modify <code class="function">get_raw_page()</code> to support free space map
       (<code class="filename">*_fsm</code>) files.  Also update
       <code class="filename">contrib/pg_freespacemap</code>.
      </p></li><li class="listitem"><p>
       Add support for multibyte encodings to <code class="filename">contrib/pg_trgm</code>
       (Teodor)
      </p></li><li class="listitem"><p>
       Rewrite <code class="filename">contrib/intagg</code> to use new
       functions <code class="function">array_agg()</code> and <code class="function">unnest()</code>
       (Tom)
      </p></li><li class="listitem"><p>
       Make <code class="filename">contrib/pg_standby</code> recover all available WAL before
       failover (Fujii Masao, Simon, Heikki)
      </p><p>
       To make this work safely, you now need to set the new
       <code class="literal">recovery_end_command</code> option in <code class="filename">recovery.conf</code>
       to clean up the trigger file after failover. <span class="application">pg_standby</span>
       will no longer remove the trigger file itself.
      </p></li><li class="listitem"><p>
       <code class="filename">contrib/pg_standby</code>'s <code class="option">-l</code> option is now a no-op,
       because it is unsafe to use a symlink (Simon)
      </p></li></ul></div></div></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="release-8-4-1.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="release.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="release-8-3-23.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">E.180. Release 8.4.1 </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> E.182. Release 8.3.23</td></tr></table></div></body></html>