Linux SRV-16 6.18.44-paas #1 SMP PREEMPT_DYNAMIC Thu Aug 13 10:08:12 UTC 2026 x86_64
//opt/postgresql-11/doc/html/release-9-6.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.21. Release 9.6</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-9-6-1.html" title="E.20. Release 9.6.1" /><link rel="next" href="release-9-5-15.html" title="E.22. Release 9.5.15" /></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.21. Release 9.6</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="release-9-6-1.html" title="E.20. Release 9.6.1">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-9-5-15.html" title="E.22. Release 9.5.15">Next</a></td></tr></table><hr></hr></div><div class="sect1" id="RELEASE-9-6"><div class="titlepage"><div><div><h2 class="title" style="clear: both">E.21. Release 9.6</h2></div></div></div><div class="toc"><dl class="toc"><dt><span class="sect2"><a href="release-9-6.html#id-1.11.6.25.3">E.21.1. Overview</a></span></dt><dt><span class="sect2"><a href="release-9-6.html#id-1.11.6.25.4">E.21.2. Migration to Version 9.6</a></span></dt><dt><span class="sect2"><a href="release-9-6.html#id-1.11.6.25.5">E.21.3. Changes</a></span></dt></dl></div><p><strong>Release date: </strong>2016-09-29</p><div class="sect2" id="id-1.11.6.25.3"><div class="titlepage"><div><div><h3 class="title">E.21.1. Overview</h3></div></div></div><p>
    Major enhancements in <span class="productname">PostgreSQL</span> 9.6 include:
   </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
      Parallel execution of sequential scans, joins and aggregates
     </p></li><li class="listitem"><p>
      Avoid scanning pages unnecessarily during vacuum freeze operations
     </p></li><li class="listitem"><p>
      Synchronous replication now allows multiple standby servers for
      increased reliability
     </p></li><li class="listitem"><p>
      Full-text search can now search for phrases (multiple adjacent words)
     </p></li><li class="listitem"><p>
      <code class="filename">postgres_fdw</code> now supports remote joins, sorts,
      <code class="command">UPDATE</code>s, and <code class="command">DELETE</code>s
     </p></li><li class="listitem"><p>
      Substantial performance improvements, especially in the area of
      scalability on multi-<acronym class="acronym">CPU</acronym>-socket servers
     </p></li></ul></div><p>
    The above items are explained in more detail in the sections below.
   </p></div><div class="sect2" id="id-1.11.6.25.4"><div class="titlepage"><div><div><h3 class="title">E.21.2. Migration to Version 9.6</h3></div></div></div><p>
    A dump/restore using <a class="xref" href="app-pg-dumpall.html" title="pg_dumpall"><span class="refentrytitle"><span class="application">pg_dumpall</span></span></a>, or use of <a class="xref" href="pgupgrade.html" title="pg_upgrade"><span class="refentrytitle"><span class="application">pg_upgrade</span></span></a>, is required for those wishing to migrate data
    from any previous release.
   </p><p>
    Version 9.6 contains a number of changes that may affect compatibility
    with previous releases.  Observe the following incompatibilities:
   </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Improve the <a class="link" href="monitoring-stats.html#PG-STAT-ACTIVITY-VIEW" title="Table 28.3. pg_stat_activity View"><code class="structname">pg_stat_activity</code></a>
        view's information about what a process is waiting for (Amit
        Kapila, Ildus Kurbangaliev)
       </p><p>
        Historically a process has only been shown as waiting if it was
        waiting for a heavyweight lock.  Now waits for lightweight locks
        and buffer pins are also shown in <code class="structname">pg_stat_activity</code>.
        Also, the type of lock being waited for is now visible.
        These changes replace the <code class="structfield">waiting</code> column with
        <code class="structfield">wait_event_type</code> and <code class="structfield">wait_event</code>.
       </p></li><li class="listitem"><p>
        In <a class="link" href="functions-formatting.html#FUNCTIONS-FORMATTING-TABLE" title="Table 9.23. Formatting Functions"><code class="function">to_char()</code></a>,
        do not count a minus sign (when needed) as part of the field
        width for time-related fields (Bruce Momjian)
       </p><p>
        For example, <code class="literal">to_char('-4 years'::interval, 'YY')</code>
        now returns <code class="literal">-04</code>, rather than <code class="literal">-4</code>.
       </p></li><li class="listitem"><p>
        Make <a class="link" href="functions-datetime.html#FUNCTIONS-DATETIME-TABLE" title="Table 9.30. Date/Time Functions"><code class="function">extract()</code></a> behave
        more reasonably with infinite inputs (Vitaly Burovoy)
       </p><p>
        Historically the <code class="function">extract()</code> function just returned
        zero given an infinite timestamp, regardless of the given
        field name.  Make it return <code class="literal">infinity</code>
        or <code class="literal">-infinity</code> as appropriate when the
        requested field is one that is monotonically increasing (e.g,
        <code class="literal">year</code>, <code class="literal">epoch</code>), or <code class="literal">NULL</code> when
        it is not (e.g., <code class="literal">day</code>, <code class="literal">hour</code>).  Also,
        throw the expected error for bad field names.
       </p></li><li class="listitem"><p>
        Remove PL/pgSQL's <span class="quote">“<span class="quote">feature</span>”</span> that suppressed the
        innermost line of <code class="literal">CONTEXT</code> for messages emitted by
        <code class="command">RAISE</code> commands (Pavel Stehule)
       </p><p>
        This ancient backwards-compatibility hack was agreed to have
        outlived its usefulness.
       </p></li><li class="listitem"><p>
        Fix the default text search parser to allow leading digits
        in <code class="literal">email</code> and <code class="literal">host</code> tokens (Artur Zakirov)
       </p><p>
        In most cases this will result in few changes in the parsing of
        text.  But if you have data where such addresses occur frequently,
        it may be worth rebuilding dependent <code class="type">tsvector</code> columns
        and indexes so that addresses of this form will be found properly
        by text searches.
       </p></li><li class="listitem"><p>
        Extend <a class="link" href="unaccent.html" title="F.43. unaccent"><code class="filename">contrib/unaccent</code></a>'s
        standard <code class="filename">unaccent.rules</code> file to handle all diacritics
        known to Unicode, and to expand ligatures correctly (Thomas Munro,
        Léonard Benedetti)
       </p><p>
        The previous version neglected to convert some less-common letters
        with diacritic marks.  Also, ligatures are now expanded into
        separate letters.  Installations that use this rules file may wish
        to rebuild <code class="type">tsvector</code> columns and indexes that depend on the
        result.
       </p></li><li class="listitem"><p>
        Remove the long-deprecated
        <code class="literal">CREATEUSER</code>/<code class="literal">NOCREATEUSER</code> options from
        <code class="command">CREATE ROLE</code> and allied commands (Tom Lane)
       </p><p>
        <code class="literal">CREATEUSER</code> actually meant <code class="literal">SUPERUSER</code>,
        for ancient backwards-compatibility reasons.  This has been a
        constant source of confusion for people who (reasonably) expect
        it to mean <code class="literal">CREATEROLE</code>.  It has been deprecated for
        ten years now, so fix the problem by removing it.
       </p></li><li class="listitem"><p>
        Treat role names beginning with <code class="literal">pg_</code> as reserved
        (Stephen Frost)
       </p><p>
        User creation of such role names is now disallowed.  This prevents
        conflicts with built-in roles created by <span class="application">initdb</span>.
       </p></li><li class="listitem"><p>
        Change a column name in the
        <code class="structname">information_schema</code>.<code class="structname">routines</code>
        view from <code class="structfield">result_cast_character_set_name</code>
        to <code class="structfield">result_cast_char_set_name</code> (Clément
        Prévost)
       </p><p>
        The SQL:2011 standard specifies the longer name, but that appears
        to be a mistake, because adjacent column names use the shorter
        style, as do other <code class="structname">information_schema</code> views.
       </p></li><li class="listitem"><p>
        <span class="application">psql</span>'s <code class="option">-c</code> option no longer implies
        <code class="option">--no-psqlrc</code>
        (Pavel Stehule, Catalin Iacob)
       </p><p>
        Write <code class="option">--no-psqlrc</code> (or its
        abbreviation <code class="option">-X</code>) explicitly to obtain the old
        behavior.  Scripts so modified will still work with old
        versions of <span class="application">psql</span>.
       </p></li><li class="listitem"><p>
        Improve <span class="application">pg_restore</span>'s <code class="option">-t</code> option to
        match all types of relations, not only plain tables (Craig Ringer)
       </p></li><li class="listitem"><p>
        Change the display format used for <code class="literal">NextXID</code> in
        <span class="application">pg_controldata</span> and related places (Joe Conway,
        Bruce Momjian)
       </p><p>
        Display epoch-and-transaction-ID values in the format
        <em class="replaceable"><code>number</code></em><code class="literal">:</code><em class="replaceable"><code>number</code></em>.
        The previous format
        <em class="replaceable"><code>number</code></em><code class="literal">/</code><em class="replaceable"><code>number</code></em> was
        confusingly similar to that used for <acronym class="acronym">LSN</acronym>s.
       </p></li><li class="listitem"><p>
        Update extension functions to be marked parallel-safe where
        appropriate (Andreas Karlsson)
       </p><p>
        Many of the standard extensions have been updated to allow their
        functions to be executed within parallel query worker processes.
        These changes will not take effect in
        databases <span class="application">pg_upgrade</span>'d from prior versions unless
        you apply <code class="command">ALTER EXTENSION UPDATE</code> to each such extension
        (in each database of a cluster).
       </p></li></ul></div></div><div class="sect2" id="id-1.11.6.25.5"><div class="titlepage"><div><div><h3 class="title">E.21.3. Changes</h3></div></div></div><p>
    Below you will find a detailed account of the changes between
    <span class="productname">PostgreSQL</span> 9.6 and the previous major
    release.
   </p><div class="sect3" id="id-1.11.6.25.5.3"><div class="titlepage"><div><div><h4 class="title">E.21.3.1. Server</h4></div></div></div><div class="sect4" id="id-1.11.6.25.5.3.2"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.1. Parallel Queries</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Parallel queries (Robert Haas, Amit Kapila, David Rowley,
        many others)
       </p><p>
        With 9.6, <span class="productname">PostgreSQL</span> introduces initial support
        for parallel execution of large queries.  Only strictly read-only
        queries where the driving table is accessed via a sequential scan
        can be parallelized.  Hash joins and nested loops can be performed
        in parallel, as can aggregation (for supported aggregates).
        Much remains to be done, but this is already a useful set of
        features.
       </p><p>
        Parallel query execution is not (yet) enabled by default.
        To allow it, set the new configuration
        parameter <a class="xref" href="runtime-config-resource.html#GUC-MAX-PARALLEL-WORKERS-PER-GATHER">max_parallel_workers_per_gather</a> to a
        value larger than zero.  Additional control over use of parallelism
        is available through other new configuration parameters
        <a class="xref" href="runtime-config-query.html#GUC-FORCE-PARALLEL-MODE">force_parallel_mode</a>,
        <a class="xref" href="runtime-config-query.html#GUC-PARALLEL-SETUP-COST">parallel_setup_cost</a>, <a class="xref" href="runtime-config-query.html#GUC-PARALLEL-TUPLE-COST">parallel_tuple_cost</a>, and
        <code class="literal">min_parallel_relation_size</code>.
       </p></li><li class="listitem"><p>
        Provide infrastructure for marking the parallel-safety status of
        functions (Robert Haas, Amit Kapila)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.3.3"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.2. Indexes</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Allow <a class="link" href="gin.html" title="Chapter 66. GIN Indexes"><acronym class="acronym">GIN</acronym></a> index builds to
        make effective use of <a class="xref" href="runtime-config-resource.html#GUC-MAINTENANCE-WORK-MEM">maintenance_work_mem</a>
        settings larger than 1 GB (Robert Abraham, Teodor Sigaev)
       </p></li><li class="listitem"><p>
        Add pages deleted from a GIN index's pending list to the free space
        map immediately
        (Jeff Janes, Teodor Sigaev)
       </p><p>
        This reduces bloat if the table is not vacuumed often.
       </p></li><li class="listitem"><p>
        Add <a class="link" href="functions-admin.html#FUNCTIONS-ADMIN-INDEX" title="9.26.8. Index Maintenance Functions"><code class="function">gin_clean_pending_list()</code></a>
        function to allow manual invocation of pending-list cleanup for a
        GIN index (Jeff Janes)
       </p><p>
        Formerly, such cleanup happened only as a byproduct of vacuuming or
        analyzing the parent table.
       </p></li><li class="listitem"><p>
        Improve handling of dead index tuples in <a class="link" href="gist.html" title="Chapter 64. GiST Indexes">GiST</a> indexes (Anastasia Lubennikova)
       </p><p>
        Dead index tuples are now marked as such when an index scan notices
        that the corresponding heap tuple is dead.  When inserting tuples,
        marked-dead tuples will be removed if needed to make space on
        the page.
       </p></li><li class="listitem"><p>
        Add an <a class="link" href="spgist.html" title="Chapter 65. SP-GiST Indexes">SP-GiST</a> operator class for
        type <code class="type">box</code> (Alexander Lebedev)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.3.4"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.3. Sorting</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Improve sorting performance by using quicksort, not replacement
        selection sort, when performing external sort steps (Peter
        Geoghegan)
       </p><p>
        The new approach makes better use of the <acronym class="acronym">CPU</acronym> cache
        for typical cache sizes and data volumes.  Where necessary,
        the behavior can be adjusted via the new configuration parameter
        <code class="literal">replacement_sort_tuples</code>.
       </p></li><li class="listitem"><p>
        Speed up text sorts where the same string occurs multiple times
        (Peter Geoghegan)
       </p></li><li class="listitem"><p>
        Speed up sorting of <code class="type">uuid</code>, <code class="type">bytea</code>, and
        <code class="type">char(n)</code> fields by using <span class="quote">“<span class="quote">abbreviated</span>”</span> keys
        (Peter Geoghegan)
       </p><p>
        Support for abbreviated keys has also been
        added to the non-default operator classes <a class="link" href="indexes-opclass.html" title="11.9. Operator Classes and Operator Families"><code class="literal">text_pattern_ops</code></a>,
        <code class="literal">varchar_pattern_ops</code>, and
        <code class="literal">bpchar_pattern_ops</code>.  Processing of ordered-set
        aggregates can also now exploit abbreviated keys.
       </p></li><li class="listitem"><p>
        Speed up <code class="command">CREATE INDEX CONCURRENTLY</code> by treating
        <acronym class="acronym">TID</acronym>s as 64-bit integers during sorting (Peter
        Geoghegan)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.3.5"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.4. Locking</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Reduce contention for the <code class="literal">ProcArrayLock</code> (Amit Kapila,
        Robert Haas)
       </p></li><li class="listitem"><p>
        Improve performance by moving buffer content locks into the buffer
        descriptors (Andres Freund, Simon Riggs)
       </p></li><li class="listitem"><p>
        Replace shared-buffer header spinlocks with atomic operations to
        improve scalability (Alexander Korotkov, Andres Freund)
       </p></li><li class="listitem"><p>
        Use atomic operations, rather than a spinlock, to protect an
        <code class="literal">LWLock</code>'s wait queue (Andres Freund)
       </p></li><li class="listitem"><p>
        Partition the shared hash table freelist to reduce contention on
        multi-<acronym class="acronym">CPU</acronym>-socket servers (Aleksander Alekseev)
       </p></li><li class="listitem"><p>
        Reduce interlocking on standby servers during the replay of btree
        index vacuuming operations (Simon Riggs)
       </p><p>
        This change avoids substantial replication delays that sometimes
        occurred while replaying such operations.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.3.6"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.5. Optimizer Statistics</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Improve <code class="command">ANALYZE</code>'s estimates for columns with many nulls
        (Tomas Vondra, Alex Shulgin)
       </p><p>
        Previously <code class="command">ANALYZE</code> tended to underestimate the number
        of non-<code class="literal">NULL</code> distinct values in a column with many
        <code class="literal">NULL</code>s, and was also inaccurate in computing the
        most-common values.
       </p></li><li class="listitem"><p>
        Improve planner's estimate of the number of distinct values in
        a query result (Tomas Vondra)
       </p></li><li class="listitem"><p>
        Use foreign key relationships to infer selectivity for join
        predicates (Tomas Vondra, David Rowley)
       </p><p>
        If a table <code class="literal">t</code> has a foreign key restriction, say
        <code class="literal">(a,b) REFERENCES r (x,y)</code>, then a <code class="literal">WHERE</code>
        condition such as <code class="literal">t.a = r.x AND t.b = r.y</code> cannot
        select more than one <code class="literal">r</code> row per <code class="literal">t</code> row.
        The planner formerly considered these <code class="literal">AND</code> conditions
        to be independent and would often drastically misestimate
        selectivity as a result.  Now it compares the <code class="literal">WHERE</code>
        conditions to applicable foreign key constraints and produces
        better estimates.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.3.7"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.6. <code class="command">VACUUM</code></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Avoid re-vacuuming pages containing only frozen tuples (Masahiko
        Sawada, Robert Haas, Andres Freund)
       </p><p>
        Formerly, anti-wraparound vacuum had to visit every page of
        a table, even pages where there was nothing to do.  Now, pages
        containing only already-frozen tuples are identified in the table's
        visibility map, and can be skipped by vacuum even when doing
        transaction wraparound prevention.  This should greatly reduce the
        cost of maintaining large tables containing mostly-unchanging data.
       </p><p>
        If necessary, vacuum can be forced to process all-frozen
        pages using the new <code class="literal">DISABLE_PAGE_SKIPPING</code> option.
        Normally this should never be needed, but it might help in
        recovering from visibility-map corruption.
       </p></li><li class="listitem"><p>
        Avoid useless heap-truncation attempts during <code class="command">VACUUM</code>
        (Jeff Janes, Tom Lane)
       </p><p>
        This change avoids taking an exclusive table lock in some cases
        where no truncation is possible.  The main benefit comes from
        avoiding unnecessary query cancellations on standby servers.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.3.8"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.7. General Performance</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Allow old <acronym class="acronym">MVCC</acronym> snapshots to be invalidated after a
        configurable timeout (Kevin Grittner)
       </p><p>
        Normally, deleted tuples cannot be physically removed by
        vacuuming until the last transaction that could <span class="quote">“<span class="quote">see</span>”</span>
        them is gone.  A transaction that stays open for a long
        time can thus cause considerable table bloat because
        space cannot be recycled.  This feature allows setting
        a time-based limit, via the new configuration parameter
        <a class="xref" href="runtime-config-resource.html#GUC-OLD-SNAPSHOT-THRESHOLD">old_snapshot_threshold</a>, on how long an
        <acronym class="acronym">MVCC</acronym> snapshot is guaranteed to be valid.  After that,
        dead tuples are candidates for removal.  A transaction using an
        outdated snapshot will get an error if it attempts to read a page
        that potentially could have contained such data.
       </p></li><li class="listitem"><p>
        Ignore <code class="literal">GROUP BY</code> columns that are
        functionally dependent on other columns (David Rowley)
       </p><p>
        If a <code class="literal">GROUP BY</code> clause includes all columns of a
        non-deferred primary key, as well as other columns of the same
        table, those other columns are redundant and can be dropped
        from the grouping.  This saves computation in many common cases.
       </p></li><li class="listitem"><p>
        Allow use of an <a class="link" href="indexes-index-only-scans.html" title="11.11. Index-Only Scans">index-only
        scan</a> on a partial index when the index's <code class="literal">WHERE</code>
        clause references columns that are not indexed (Tomas Vondra,
        Kyotaro Horiguchi)
       </p><p>
        For example, an index defined by <code class="command">CREATE INDEX tidx_partial
        ON t(b) WHERE a &gt; 0</code> can now be used for an index-only scan by
        a query that specifies <code class="literal">WHERE a &gt; 0</code> and does not
        otherwise use <code class="literal">a</code>.  Previously this was disallowed
        because <code class="literal">a</code> is not listed as an index column.
       </p></li><li class="listitem"><p>
        Perform checkpoint writes in sorted order (Fabien Coelho,
        Andres Freund)
       </p><p>
        Previously, checkpoints wrote out dirty pages in whatever order
        they happen to appear in shared buffers, which usually is nearly
        random.  That performs poorly, especially on rotating media.
        This change causes checkpoint-driven writes to be done in order
        by file and block number, and to be balanced across tablespaces.
       </p></li><li class="listitem"><p>
        Where feasible, trigger kernel writeback after a configurable
        number of writes, to prevent accumulation of dirty data in kernel
        disk buffers (Fabien Coelho, Andres Freund)
       </p><p>
        <span class="productname">PostgreSQL</span> writes data to the kernel's disk cache,
        from where it will be flushed to physical storage in due time.
        Many operating systems are not smart about managing this and allow
        large amounts of dirty data to accumulate before deciding to flush
        it all at once, causing long delays for new I/O requests until the
        flushing finishes.
        This change attempts to alleviate this problem by explicitly
        requesting data flushes after a configurable interval.
       </p><p>
        On Linux, <code class="function">sync_file_range()</code> is used for this purpose,
        and the feature is on by default on Linux because that function has
        few downsides.  This flushing capability is also available on other
        platforms if they have <code class="function">msync()</code>
        or <code class="function">posix_fadvise()</code>, but those interfaces have some
        undesirable side-effects so the feature is disabled by default on
        non-Linux platforms.
       </p><p>
        The new configuration parameters <a class="xref" href="runtime-config-resource.html#GUC-BACKEND-FLUSH-AFTER">backend_flush_after</a>, <a class="xref" href="runtime-config-resource.html#GUC-BGWRITER-FLUSH-AFTER">bgwriter_flush_after</a>, <a class="xref" href="runtime-config-wal.html#GUC-CHECKPOINT-FLUSH-AFTER">checkpoint_flush_after</a>, and <a class="xref" href="runtime-config-wal.html#GUC-WAL-WRITER-FLUSH-AFTER">wal_writer_flush_after</a> control this behavior.
       </p></li><li class="listitem"><p>
        Improve aggregate-function performance by sharing calculations
        across multiple aggregates if they have the same arguments and
        transition functions (David Rowley)
       </p><p>
        For example, <code class="command">SELECT AVG(x), VARIANCE(x) FROM tab</code> can use
        a single per-row computation for both aggregates.
       </p></li><li class="listitem"><p>
        Speed up visibility tests for recently-created tuples by checking
        the current transaction's snapshot, not <code class="structname">pg_clog</code>, to
        decide if the source transaction should be considered committed
        (Jeff Janes, Tom Lane)
       </p></li><li class="listitem"><p>
        Allow tuple hint bits to be set sooner than before (Andres Freund)
       </p></li><li class="listitem"><p>
        Improve performance of short-lived prepared transactions (Stas
        Kelvich, Simon Riggs, Pavan Deolasee)
       </p><p>
        Two-phase commit information is now written only to <acronym class="acronym">WAL</acronym>
        during <code class="command">PREPARE TRANSACTION</code>, and will be read back from
        <acronym class="acronym">WAL</acronym> during <code class="command">COMMIT PREPARED</code> if that happens
        soon thereafter.  A separate state file is created only if the
        pending transaction does not get committed or aborted by the time
        of the next checkpoint.
       </p></li><li class="listitem"><p>
        Improve performance of memory context destruction (Jan Wieck)
       </p></li><li class="listitem"><p>
        Improve performance of resource owners with many tracked objects
        (Aleksander Alekseev)
       </p></li><li class="listitem"><p>
        Improve speed of the output functions for <code class="type">timestamp</code>,
        <code class="type">time</code>, and <code class="type">date</code> data types (David Rowley,
        Andres Freund)
       </p></li><li class="listitem"><p>
        Avoid some unnecessary cancellations of hot-standby queries
        during replay of actions that take <code class="literal">AccessExclusive</code>
        locks (Jeff Janes)
       </p></li><li class="listitem"><p>
        Extend relations multiple blocks at a time when there is contention
        for the relation's extension lock (Dilip Kumar)
       </p><p>
        This improves scalability by decreasing contention.
       </p></li><li class="listitem"><p>
        Increase the number of clog buffers for better scalability (Amit
        Kapila, Andres Freund)
       </p></li><li class="listitem"><p>
        Speed up expression evaluation in <span class="application">PL/pgSQL</span> by
        keeping <code class="literal">ParamListInfo</code> entries for simple variables
        valid at all times (Tom Lane)
       </p></li><li class="listitem"><p>
        Avoid reducing the <code class="literal">SO_SNDBUF</code> setting below its default
        on recent Windows versions (Chen Huajun)
       </p></li><li class="listitem"><p>
        Disable <a class="xref" href="runtime-config-logging.html#GUC-UPDATE-PROCESS-TITLE">update_process_title</a> by default on
        Windows (Takayuki Tsunakawa)
       </p><p>
        The overhead of updating the process title is much larger on Windows
        than most other platforms, and it is also less useful to do it since
        most Windows users do not have tools that can display process titles.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.3.9"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.8. Monitoring</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add <a class="link" href="progress-reporting.html#PG-STAT-PROGRESS-VACUUM-VIEW" title="Table 28.21. pg_stat_progress_vacuum View"><code class="structname">pg_stat_progress_vacuum</code></a>
        system view to provide progress reporting for <code class="command">VACUUM</code>
        operations (Amit Langote, Robert Haas, Vinayak Pokale, Rahila Syed)
       </p></li><li class="listitem"><p>
        Add <a class="link" href="functions-info.html#FUNCTIONS-CONTROLDATA" title="Table 9.72. Control Data Functions"><code class="function">pg_control_system()</code></a>,
        <code class="function">pg_control_checkpoint()</code>,
        <code class="function">pg_control_recovery()</code>, and
        <code class="function">pg_control_init()</code> functions to expose fields of
        <code class="filename">pg_control</code> to <acronym class="acronym">SQL</acronym> (Joe Conway, Michael
        Paquier)
       </p></li><li class="listitem"><p>
        Add <a class="link" href="view-pg-config.html" title="52.67. pg_config"><code class="structname">pg_config</code></a>
        system view (Joe Conway)
       </p><p>
        This view exposes the same information available from
        the <span class="application">pg_config</span> command-line utility,
        namely assorted compile-time configuration information for
        <span class="productname">PostgreSQL</span>.
       </p></li><li class="listitem"><p>
        Add a <code class="structfield">confirmed_flush_lsn</code> column to the <a class="link" href="view-pg-replication-slots.html" title="52.80. pg_replication_slots"><code class="structname">pg_replication_slots</code></a>
        system view (Marko Tiikkaja)
       </p></li><li class="listitem"><p>
        Add <a class="link" href="monitoring-stats.html#PG-STAT-WAL-RECEIVER-VIEW" title="Table 28.6. pg_stat_wal_receiver View"><code class="structname">pg_stat_wal_receiver</code></a>
        system view to provide information about the state of a hot-standby
        server's <acronym class="acronym">WAL</acronym> receiver process (Michael Paquier)
       </p></li><li class="listitem"><p>
        Add <a class="link" href="functions-info.html#FUNCTIONS-INFO-SESSION-TABLE" title="Table 9.60. Session Information Functions"><code class="function">pg_blocking_pids()</code></a>
        function to reliably identify which sessions block which others
        (Tom Lane)
       </p><p>
        This function returns an array of the process IDs of any
        sessions that are blocking the session with the given process ID.
        Historically users have obtained such information using a self-join
        on the <code class="structname">pg_locks</code> view.  However, it is unreasonably
        tedious to do it that way with any modicum of correctness, and
        the addition of parallel queries has made the old approach entirely
        impractical, since locks might be held or awaited by child worker
        processes rather than the session's main process.
       </p></li><li class="listitem"><p>
        Add function <a class="link" href="functions-admin.html#FUNCTIONS-ADMIN-BACKUP-TABLE" title="Table 9.79. Backup Control Functions"><code class="function">pg_current_xlog_flush_location()</code></a>
        to expose the current transaction log flush location (Tomas Vondra)
       </p></li><li class="listitem"><p>
        Add function <a class="link" href="functions-info.html#FUNCTIONS-INFO-SESSION-TABLE" title="Table 9.60. Session Information Functions"><code class="function">pg_notification_queue_usage()</code></a>
        to report how full the <code class="command">NOTIFY</code> queue is (Brendan Jurd)
       </p></li><li class="listitem"><p>
        Limit the verbosity of memory context statistics dumps (Tom Lane)
       </p><p>
        The memory usage dump that is output to the postmaster log during an
        out-of-memory failure now summarizes statistics when there are a
        large number of memory contexts, rather than possibly generating
        a very large report.  There is also a <span class="quote">“<span class="quote">grand total</span>”</span>
        summary line now.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.3.10"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.9. <acronym class="acronym">Authentication</acronym></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add a <a class="link" href="auth-bsd.html" title="20.14. BSD Authentication"><acronym class="acronym">BSD</acronym> authentication
        method</a> to allow use of
        the <span class="systemitem">BSD</span> Authentication service for
        <span class="productname">PostgreSQL</span> client authentication (Marisa Emerson)
       </p><p>
        BSD Authentication is currently only available on <span class="systemitem">OpenBSD</span>.
       </p></li><li class="listitem"><p>
        When using <a class="link" href="auth-pam.html" title="20.13. PAM Authentication"><acronym class="acronym">PAM</acronym>
        authentication</a>, provide the client IP address or host name
        to <acronym class="acronym">PAM</acronym> modules via the <code class="literal">PAM_RHOST</code> item
        (Grzegorz Sampolski)
       </p></li><li class="listitem"><p>
        Provide detail in the postmaster log for more types of password
        authentication failure (Tom Lane)
       </p><p>
        All ordinarily-reachable password authentication failure cases
        should now provide specific <code class="literal">DETAIL</code> fields in the log.
       </p></li><li class="listitem"><p>
        Support <a class="link" href="auth-radius.html" title="20.11. RADIUS Authentication"><acronym class="acronym">RADIUS</acronym> passwords</a>
        up to 128 characters long (Marko Tiikkaja)
       </p></li><li class="listitem"><p>
        Add new <a class="link" href="sspi-auth.html" title="20.7. SSPI Authentication"><acronym class="acronym">SSPI</acronym>
        authentication</a> parameters
        <code class="varname">compat_realm</code> and <code class="varname">upn_username</code> to control
        whether <span class="productname">NetBIOS</span> or <span class="productname">Kerberos</span>
        realm names and user names are used during <acronym class="acronym">SSPI</acronym>
        authentication (Christian Ullrich)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.3.11"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.10. Server Configuration</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Allow sessions to be terminated automatically if they are in
        idle-in-transaction state for too long (Vik Fearing)
       </p><p>
        This behavior is controlled by the new configuration parameter
        <a class="xref" href="runtime-config-client.html#GUC-IDLE-IN-TRANSACTION-SESSION-TIMEOUT">idle_in_transaction_session_timeout</a>.  It can
        be useful to prevent forgotten transactions from holding locks
        or preventing vacuum cleanup for too long.
       </p></li><li class="listitem"><p>
        Raise the maximum allowed value
        of <a class="xref" href="runtime-config-wal.html#GUC-CHECKPOINT-TIMEOUT">checkpoint_timeout</a> to 24 hours (Simon Riggs)
       </p></li><li class="listitem"><p>
        Allow <code class="varname">effective_io_concurrency</code> to be set per-tablespace
        to support cases where different tablespaces have different I/O
        characteristics (Julien Rouhaud)
       </p></li><li class="listitem"><p>
        Add <a class="xref" href="runtime-config-logging.html#GUC-LOG-LINE-PREFIX">log_line_prefix</a> option <code class="literal">%n</code> to
        print the current time in Unix epoch form, with milliseconds (Tomas
        Vondra, Jeff Davis)
       </p></li><li class="listitem"><p>
        Add <a class="xref" href="runtime-config-logging.html#GUC-SYSLOG-SEQUENCE-NUMBERS">syslog_sequence_numbers</a> and <a class="xref" href="runtime-config-logging.html#GUC-SYSLOG-SPLIT-MESSAGES">syslog_split_messages</a> configuration parameters
        to provide more control over the message format when logging to
        <span class="systemitem">syslog</span> (Peter Eisentraut)
       </p></li><li class="listitem"><p>
        Merge the <code class="literal">archive</code> and <code class="literal">hot_standby</code> values
        of the <a class="xref" href="runtime-config-wal.html#GUC-WAL-LEVEL">wal_level</a> configuration parameter
        into a single new value <code class="literal">replica</code> (Peter Eisentraut)
       </p><p>
        Making a distinction between these settings is no longer useful,
        and merging them is a step towards a planned future simplification
        of replication setup.  The old names are still accepted but are
        converted to <code class="literal">replica</code> internally.
       </p></li><li class="listitem"><p>
        Add configure option <code class="option">--with-systemd</code> to enable
        calling <code class="function">sd_notify()</code> at server start and stop (Peter
        Eisentraut)
       </p><p>
        This allows the use of <span class="application">systemd</span> service units of
        type <code class="literal">notify</code>, which greatly simplifies the management
        of <span class="productname">PostgreSQL</span> under <span class="application">systemd</span>.
       </p></li><li class="listitem"><p>
        Allow the server's <acronym class="acronym">SSL</acronym> key file to have group read
        access if it is owned by <code class="literal">root</code> (Christoph Berg)
       </p><p>
        Formerly, we insisted the key file be owned by the
        user running the <span class="productname">PostgreSQL</span> server, but
        that is inconvenient on some systems (such as <span class="systemitem">Debian</span>) that are configured to manage
        certificates centrally.  Therefore, allow the case where the key
        file is owned by <code class="literal">root</code> and has group read access.
        It is up to the operating system administrator to ensure that
        the group does not include any untrusted users.
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.3.12"><div class="titlepage"><div><div><h5 class="title">E.21.3.1.11. Reliability</h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Force backends to exit if the postmaster dies (Rajeev Rastogi,
        Robert Haas)
       </p><p>
        Under normal circumstances the postmaster should always outlive
        its child processes.  If for some reason the postmaster dies,
        force backend sessions to exit with an error.  Formerly, existing
        backends would continue to run until their clients disconnect,
        but that is unsafe and inefficient.  It also prevents a new
        postmaster from being started until the last old backend has
        exited.  Backends will detect postmaster death when waiting for
        client I/O, so the exit will not be instantaneous, but it should
        happen no later than the end of the current query.
       </p></li><li class="listitem"><p>
        Check for serializability conflicts before reporting
        constraint-violation failures (Thomas Munro)
       </p><p>
        When using serializable transaction isolation, it is desirable
        that any error due to concurrent transactions should manifest
        as a serialization failure, thereby cueing the application that
        a retry might succeed.  Unfortunately, this does not reliably
        happen for duplicate-key failures caused by concurrent insertions.
        This change ensures that such an error will be reported as a
        serialization error if the application explicitly checked for
        the presence of a conflicting key (and did not find it) earlier
        in the transaction.
       </p></li><li class="listitem"><p>
        Ensure that invalidation messages are recorded in <acronym class="acronym">WAL</acronym>
        even when issued by a transaction that has no <acronym class="acronym">XID</acronym>
        assigned (Andres Freund)
       </p><p>
        This fixes some corner cases in which transactions on standby
        servers failed to notice changes, such as new indexes.
       </p></li><li class="listitem"><p>
        Prevent multiple processes from trying to clean a <acronym class="acronym">GIN</acronym>
        index's pending list concurrently (Teodor Sigaev, Jeff Janes)
       </p><p>
        This had been intentionally allowed, but it causes race conditions
        that can result in vacuum missing index entries it needs to delete.
       </p></li></ul></div></div></div><div class="sect3" id="id-1.11.6.25.5.4"><div class="titlepage"><div><div><h4 class="title">E.21.3.2. Replication and Recovery</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Allow synchronous replication to support multiple simultaneous
        synchronous standby servers, not just one (Masahiko Sawada,
        Beena Emerson, Michael Paquier, Fujii Masao, Kyotaro Horiguchi)
       </p><p>
        The number of standby servers that must acknowledge a commit
        before it is considered complete is now configurable as part of
        the <a class="xref" href="runtime-config-replication.html#GUC-SYNCHRONOUS-STANDBY-NAMES">synchronous_standby_names</a> parameter.
       </p></li><li class="listitem"><p>
        Add new setting <code class="literal">remote_apply</code> for configuration
        parameter <a class="xref" href="runtime-config-wal.html#GUC-SYNCHRONOUS-COMMIT">synchronous_commit</a> (Thomas Munro)
       </p><p>
        In this mode, the master waits for the transaction to be
        <span class="emphasis"><em>applied</em></span> on the standby server, not just written
        to disk.  That means that you can count on a transaction started
        on the standby to see all commits previously acknowledged by
        the master.
       </p></li><li class="listitem"><p>
        Add a feature to the replication
        protocol, and a corresponding option to <a class="link" href="functions-admin.html#FUNCTIONS-REPLICATION-TABLE" title="Table 9.83. Replication SQL Functions"><code class="function">pg_create_physical_replication_slot()</code></a>,
        to allow reserving <acronym class="acronym">WAL</acronym> immediately when creating a
        replication slot (Gurjeet Singh, Michael Paquier)
       </p><p>
        This allows the creation of a replication slot to guarantee
        that all the <acronym class="acronym">WAL</acronym> needed for a base backup will be
        available.
       </p></li><li class="listitem"><p>
        Add a <code class="option">--slot</code> option to
        <a class="link" href="app-pgbasebackup.html" title="pg_basebackup"><span class="application">pg_basebackup</span></a>
        (Peter Eisentraut)
       </p><p>
        This lets <span class="application">pg_basebackup</span> use a replication
        slot defined for <acronym class="acronym">WAL</acronym> streaming.  After the base
        backup completes, selecting the same slot for regular streaming
        replication allows seamless startup of the new standby server.
       </p></li><li class="listitem"><p>
        Extend <a class="link" href="functions-admin.html#FUNCTIONS-ADMIN-BACKUP-TABLE" title="Table 9.79. Backup Control Functions"><code class="function">pg_start_backup()</code></a>
        and <code class="function">pg_stop_backup()</code> to support non-exclusive backups
        (Magnus Hagander)
       </p></li></ul></div></div><div class="sect3" id="id-1.11.6.25.5.5"><div class="titlepage"><div><div><h4 class="title">E.21.3.3. Queries</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Allow functions that return sets of tuples to return simple
        <code class="literal">NULL</code>s (Andrew Gierth, Tom Lane)
       </p><p>
        In the context of <code class="literal">SELECT FROM function(...)</code>, a function
        that returned a set of composite values was previously not allowed
        to return a plain <code class="literal">NULL</code> value as part of the set.
        Now that is allowed and interpreted as a row of <code class="literal">NULL</code>s.
        This avoids corner-case errors with, for example, unnesting an
        array of composite values.
       </p></li><li class="listitem"><p>
        Fully support array subscripts and field selections in the
        target column list of an <code class="command">INSERT</code> with multiple
        <code class="literal">VALUES</code> rows (Tom Lane)
       </p><p>
        Previously, such cases failed if the same target column was
        mentioned more than once, e.g., <code class="literal">INSERT INTO tab (x[1],
        x[2]) VALUES (...)</code>.
       </p></li><li class="listitem"><p>
        When appropriate, postpone evaluation of <code class="command">SELECT</code>
        output expressions until after an <code class="literal">ORDER BY</code> sort
        (Konstantin Knizhnik)
       </p><p>
        This change ensures that volatile or expensive functions in the
        output list are executed in the order suggested by <code class="literal">ORDER
        BY</code>, and that they are not evaluated more times than required
        when there is a <code class="literal">LIMIT</code> clause.  Previously, these
        properties held if the ordering was performed by an index scan or
        pre-merge-join sort, but not if it was performed by a top-level
        sort.
       </p></li><li class="listitem"><p>
        Widen counters recording the number of tuples processed to 64 bits
        (Andreas Scherbaum)
       </p><p>
        This change allows command tags, e.g. <code class="command">SELECT</code>, to
        correctly report tuple counts larger than 4 billion.  This also
        applies to PL/pgSQL's <code class="command">GET DIAGNOSTICS ... ROW_COUNT</code>
        command.
       </p></li><li class="listitem"><p>
        Avoid doing encoding conversions by converting through the
        <code class="literal">MULE_INTERNAL</code> encoding (Tom Lane)
       </p><p>
        Previously, many conversions for Cyrillic and Central
        European single-byte encodings were done by converting to a
        related <code class="literal">MULE_INTERNAL</code> coding scheme and then to the
        destination encoding.  Aside from being inefficient, this meant
        that when the conversion encountered an untranslatable character,
        the error message would confusingly complain about failure to
        convert to or from <code class="literal">MULE_INTERNAL</code>, rather than the
        user-visible encoding.
       </p></li><li class="listitem"><p>
        Consider performing joins of foreign tables remotely only when the
        tables will be accessed under the same role ID (Shigeru Hanada,
        Ashutosh Bapat, Etsuro Fujita)
       </p><p>
        Previously, the foreign join pushdown infrastructure left the
        question of security entirely up to individual foreign data
        wrappers, but that made it too easy for an <acronym class="acronym">FDW</acronym> to
        inadvertently create subtle security holes.  So, make it the core
        code's job to determine which role ID will access each table,
        and do not attempt join pushdown unless the role is the same for
        all relevant relations.
       </p></li></ul></div></div><div class="sect3" id="id-1.11.6.25.5.6"><div class="titlepage"><div><div><h4 class="title">E.21.3.4. Utility Commands</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Allow <code class="command">COPY</code> to copy the output of an
        <code class="literal">INSERT</code>/<code class="literal">UPDATE</code>/<code class="literal">DELETE</code>
        ... <code class="literal">RETURNING</code> query (Marko Tiikkaja)
       </p><p>
        Previously, an intermediate <acronym class="acronym">CTE</acronym> had to be written to
        get this result.
       </p></li><li class="listitem"><p>
        Introduce <code class="command">ALTER <em class="replaceable"><code>object</code></em> DEPENDS ON
        EXTENSION</code> (Abhijit Menon-Sen)
       </p><p>
        This command allows a database object to be marked as depending
        on an extension, so that it will be dropped automatically if
        the extension is dropped (without needing <code class="literal">CASCADE</code>).
        However, the object is not part of the extension, and thus will
        be dumped separately by <span class="application">pg_dump</span>.
       </p></li><li class="listitem"><p>
        Make <code class="command">ALTER <em class="replaceable"><code>object</code></em> SET SCHEMA</code> do nothing
        when the object is already in the requested schema, rather than
        throwing an error as it historically has for most object types
        (Marti Raudsepp)
       </p></li><li class="listitem"><p>
        Add options to <code class="command">ALTER OPERATOR</code> to allow changing
        the selectivity functions associated with an existing operator
        (Yury Zhuravlev)
       </p></li><li class="listitem"><p>
        Add an <code class="option">IF NOT EXISTS</code> option to <code class="command">ALTER TABLE
        ADD COLUMN</code> (Fabrízio de Royes Mello)
       </p></li><li class="listitem"><p>
        Reduce the lock strength needed by <code class="command">ALTER TABLE</code>
        when setting fillfactor and autovacuum-related relation options
        (Fabrízio de Royes Mello, Simon Riggs)
       </p></li><li class="listitem"><p>
        Introduce <a class="link" href="sql-create-access-method.html" title="CREATE ACCESS METHOD"><code class="command">CREATE
        ACCESS METHOD</code></a> to allow extensions to create index access
        methods (Alexander Korotkov, Petr Jelínek)
       </p></li><li class="listitem"><p>
        Add a <code class="literal">CASCADE</code> option to <code class="command">CREATE
        EXTENSION</code> to automatically create any extensions the
        requested one depends on (Petr Jelínek)
       </p></li><li class="listitem"><p>
        Make <code class="command">CREATE TABLE ... LIKE</code> include an <code class="type">OID</code>
        column if any source table has one (Bruce Momjian)
       </p></li><li class="listitem"><p>
        If a <code class="literal">CHECK</code> constraint is declared <code class="literal">NOT VALID</code>
        in a table creation command, automatically mark it as valid
        (Amit Langote, Amul Sul)
       </p><p>
        This is safe because the table has no existing rows.  This matches
        the longstanding behavior of <code class="literal">FOREIGN KEY</code> constraints.
       </p></li><li class="listitem"><p>
        Fix <code class="command">DROP OPERATOR</code> to clear
        <code class="structname">pg_operator</code>.<code class="structfield">oprcom</code> and
        <code class="structname">pg_operator</code>.<code class="structfield">oprnegate</code> links to
        the dropped operator (Roma Sokolov)
       </p><p>
        Formerly such links were left as-is, which could pose a problem
        in the somewhat unlikely event that the dropped operator's
        <code class="type">OID</code> was reused for another operator.
       </p></li><li class="listitem"><p>
        Do not show the same subplan twice in <code class="command">EXPLAIN</code> output
        (Tom Lane)
       </p><p>
        In certain cases, typically involving SubPlan nodes in index
        conditions, <code class="command">EXPLAIN</code> would print data for the same
        subplan twice.
       </p></li><li class="listitem"><p>
        Disallow creation of indexes on system columns, except for
        <code class="type">OID</code> columns (David Rowley)
       </p><p>
        Such indexes were never considered supported, and would very
        possibly misbehave since the system might change the system-column
        fields of a tuple without updating indexes.  However, previously
        there were no error checks to prevent them from being created.
       </p></li></ul></div></div><div class="sect3" id="id-1.11.6.25.5.7"><div class="titlepage"><div><div><h4 class="title">E.21.3.5. Permissions Management</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Use the privilege system to manage access to sensitive functions
        (Stephen Frost)
       </p><p>
        Formerly, many security-sensitive functions contained hard-wired
        checks that would throw an error if they were called by a
        non-superuser.  This forced the use of superuser roles for
        some relatively pedestrian tasks.  The hard-wired error checks
        are now gone in favor of making <span class="application">initdb</span> revoke the
        default public <code class="literal">EXECUTE</code> privilege on these functions.
        This allows installations to choose to grant usage of such
        functions to trusted roles that do not need all superuser
        privileges.
       </p></li><li class="listitem"><p>
        Create some <a class="link" href="default-roles.html" title="21.5. Default Roles">built-in roles</a>
        that can be used to grant access to what were previously
        superuser-only functions (Stephen Frost)
       </p><p>
        Currently the only such role is <code class="literal">pg_signal_backend</code>,
        but more are expected to be added in future.
       </p></li></ul></div></div><div class="sect3" id="id-1.11.6.25.5.8"><div class="titlepage"><div><div><h4 class="title">E.21.3.6. Data Types</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Improve <a class="link" href="textsearch.html" title="Chapter 12. Full Text Search">full-text search</a> to support
        searching for phrases, that is, lexemes appearing adjacent to each
        other in a specific order, or with a specified distance between
        them (Teodor Sigaev, Oleg Bartunov, Dmitry Ivanov)
       </p><p>
        A phrase-search query can be specified in <code class="type">tsquery</code>
        input using the new operators <code class="literal">&lt;-&gt;</code> and
        <code class="literal">&lt;<em class="replaceable"><code>N</code></em>&gt;</code>.  The former means
        that the lexemes before and after it must appear adjacent to
        each other in that order.  The latter means they must be exactly
        <em class="replaceable"><code>N</code></em> lexemes apart.
       </p></li><li class="listitem"><p>
        Allow omitting one or both boundaries in an array slice specifier,
        e.g. <code class="literal">array_col[3:]</code> (Yury Zhuravlev)
       </p><p>
        Omitted boundaries are taken as the upper or lower limit of the
        corresponding array subscript.  This allows simpler specification
        for many common use-cases.
       </p></li><li class="listitem"><p>
        Be more careful about out-of-range dates and timestamps (Vitaly
        Burovoy)
       </p><p>
        This change prevents unexpected out-of-range errors for
        <code class="type">timestamp with time zone</code> values very close to the
        implementation limits.  Previously, the <span class="quote">“<span class="quote">same</span>”</span> value might
        be accepted or not depending on the <code class="varname">timezone</code> setting,
        meaning that a dump and reload could fail on a value that had been
        accepted when presented.  Now the limits are enforced according
        to the equivalent <acronym class="acronym">UTC</acronym> time, not local time, so as to
        be independent of <code class="varname">timezone</code>.
       </p><p>
        Also, <span class="productname">PostgreSQL</span> is now more careful to detect
        overflow in operations that compute new date or timestamp values,
        such as <code class="type">date</code> <code class="literal">+</code> <code class="type">integer</code>.
       </p></li><li class="listitem"><p>
        For geometric data types, make sure <code class="literal">infinity</code> and
        <code class="literal">NaN</code> component values are treated consistently during
        input and output (Tom Lane)
       </p><p>
        Such values will now always print the same as they would in
        a simple <code class="type">float8</code> column, and be accepted the same way
        on input.  Previously the behavior was platform-dependent.
       </p></li><li class="listitem"><p>
        Upgrade
        the <a class="link" href="textsearch-dictionaries.html#TEXTSEARCH-ISPELL-DICTIONARY" title="12.6.5. Ispell Dictionary"><code class="literal">ispell</code></a>
        dictionary type to handle modern <span class="productname">Hunspell</span> files and
        support more languages (Artur Zakirov)
       </p></li><li class="listitem"><p>
        Implement look-behind constraints
        in <a class="link" href="functions-matching.html#FUNCTIONS-POSIX-REGEXP" title="9.7.3. POSIX Regular Expressions">regular expressions</a>
        (Tom Lane)
       </p><p>
        A look-behind constraint is like a lookahead constraint in that it
        consumes no text; but it checks for existence (or nonexistence)
        of a match ending at the current point in the string, rather
        than one starting at the current point.  Similar features exist
        in many other regular-expression engines.
       </p></li><li class="listitem"><p>
        In regular expressions, if an apparent three-digit octal escape
        <code class="literal">\</code><em class="replaceable"><code>nnn</code></em> would exceed 377 (255 decimal),
        assume it is a two-digit octal escape instead (Tom Lane)
       </p><p>
        This makes the behavior match current <span class="application">Tcl</span> releases.
       </p></li><li class="listitem"><p>
        Add transaction ID operators <code class="type">xid</code> <code class="literal">&lt;&gt;</code>
        <code class="type">xid</code> and <code class="type">xid</code> <code class="literal">&lt;&gt;</code> <code class="type">int4</code>,
        for consistency with the corresponding equality operators
        (Michael Paquier)
       </p></li></ul></div></div><div class="sect3" id="id-1.11.6.25.5.9"><div class="titlepage"><div><div><h4 class="title">E.21.3.7. Functions</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add <a class="link" href="functions-json.html#FUNCTIONS-JSON-PROCESSING-TABLE" title="Table 9.46. JSON Processing Functions"><code class="function">jsonb_insert()</code></a>
        function to insert a new element into a <code class="type">jsonb</code> array,
        or a not-previously-existing key into a <code class="type">jsonb</code> object
        (Dmitry Dolgov)
       </p></li><li class="listitem"><p>
        Improve the accuracy of the <code class="function">ln()</code>, <code class="function">log()</code>,
        <code class="function">exp()</code>, and <code class="function">pow()</code> functions for type
        <code class="type">numeric</code> (Dean Rasheed)
       </p></li><li class="listitem"><p>
        Add a <a class="link" href="functions-math.html#FUNCTIONS-MATH-FUNC-TABLE" title="Table 9.5. Mathematical Functions"><code class="function">scale(numeric)</code></a>
        function to extract the display scale of a <code class="type">numeric</code> value
        (Marko Tiikkaja)
       </p></li><li class="listitem"><p>
        Add trigonometric functions that work in degrees (Dean Rasheed)
       </p><p>
        For example, <a class="link" href="functions-math.html#FUNCTIONS-MATH-TRIG-TABLE" title="Table 9.7. Trigonometric Functions"><code class="function">sind()</code></a>
        measures its argument in degrees, whereas <code class="function">sin()</code>
        measures in radians.  These functions go to some lengths to
        deliver exact results for values where an exact result can be
        expected, for instance <code class="literal">sind(30) = 0.5</code>.
       </p></li><li class="listitem"><p>
        Ensure that trigonometric functions handle <code class="literal">infinity</code>
        and <code class="literal">NaN</code> inputs per the <acronym class="acronym">POSIX</acronym> standard
        (Dean Rasheed)
       </p><p>
        The <acronym class="acronym">POSIX</acronym> standard says that these functions should
        return <code class="literal">NaN</code> for <code class="literal">NaN</code> input, and should throw
        an error for out-of-range inputs including <code class="literal">infinity</code>.
        Previously our behavior varied across platforms.
       </p></li><li class="listitem"><p>
        Make <a class="link" href="functions-datetime.html#FUNCTIONS-DATETIME-TABLE" title="Table 9.30. Date/Time Functions"><code class="function">to_timestamp(float8)</code></a>
        convert float <code class="literal">infinity</code> to
        timestamp <code class="literal">infinity</code> (Vitaly Burovoy)
       </p><p>
        Formerly it just failed on an infinite input.
       </p></li><li class="listitem"><p>
        Add new functions for <code class="type">tsvector</code> data (Stas Kelvich)
       </p><p>
        The new functions are <a class="link" href="functions-textsearch.html#TEXTSEARCH-FUNCTIONS-TABLE" title="Table 9.41. Text Search Functions"><code class="function">ts_delete()</code></a>,
        <code class="function">ts_filter()</code>, <code class="function">unnest()</code>,
        <code class="function">tsvector_to_array()</code>, <code class="function">array_to_tsvector()</code>,
        and a variant of <code class="function">setweight()</code> that sets the weight
        only for specified lexeme(s).
       </p></li><li class="listitem"><p>
        Allow <a class="link" href="textsearch-features.html#TEXTSEARCH-STATISTICS" title="12.4.4. Gathering Document Statistics"><code class="function">ts_stat()</code></a>
        and <a class="link" href="textsearch-features.html#TEXTSEARCH-UPDATE-TRIGGERS" title="12.4.3. Triggers for Automatic Updates"><code class="function">tsvector_update_trigger()</code></a>
        to operate on values that are of types binary-compatible with the
        expected argument type, not just exactly that type; for example
        allow <code class="type">citext</code> where <code class="type">text</code> is expected (Teodor
        Sigaev)
       </p></li><li class="listitem"><p>
        Add variadic functions <a class="link" href="functions-comparison.html#FUNCTIONS-COMPARISON-FUNC-TABLE" title="Table 9.3. Comparison Functions"><code class="function">num_nulls()</code></a>
        and <code class="function">num_nonnulls()</code> that count the number of their
        arguments that are null or non-null (Marko Tiikkaja)
       </p><p>
        An example usage is <code class="literal">CHECK(num_nonnulls(a,b,c) = 1)</code>
        which asserts that exactly one of a,b,c is not <code class="literal">NULL</code>.
        These functions can also be used to count the number of null or
        nonnull elements in an array.
       </p></li><li class="listitem"><p>
        Add function <a class="link" href="functions-string.html#FUNCTIONS-STRING-OTHER" title="Table 9.9. Other String Functions"><code class="function">parse_ident()</code></a>
        to split a qualified, possibly quoted <acronym class="acronym">SQL</acronym> identifier
        into its parts (Pavel Stehule)
       </p></li><li class="listitem"><p>
        In <a class="link" href="functions-formatting.html#FUNCTIONS-FORMATTING-TABLE" title="Table 9.23. Formatting Functions"><code class="function">to_number()</code></a>,
        interpret a <code class="literal">V</code> format code as dividing by 10 to the
        power of the number of digits following <code class="literal">V</code> (Bruce
        Momjian)
       </p><p>
        This makes it operate in an inverse fashion to
        <code class="function">to_char()</code>.
       </p></li><li class="listitem"><p>
        Make the <a class="link" href="functions-info.html#FUNCTIONS-INFO-CATALOG-TABLE" title="Table 9.63. System Catalog Information Functions"><code class="function">to_reg*()</code></a>
        functions accept type <code class="type">text</code> not <code class="type">cstring</code>
        (Petr Korobeinikov)
       </p><p>
        This avoids the need to write an explicit cast in most cases
        where the argument is not a simple literal constant.
       </p></li><li class="listitem"><p>
        Add <a class="link" href="functions-admin.html#FUNCTIONS-ADMIN-DBSIZE" title="Table 9.84. Database Object Size Functions"><code class="function">pg_size_bytes()</code></a>
        function to convert human-readable size strings to numbers (Pavel
        Stehule, Vitaly Burovoy, Dean Rasheed)
       </p><p>
        This function converts strings like those produced by
        <code class="function">pg_size_pretty()</code> into bytes.  An example
        usage is <code class="literal">SELECT oid::regclass FROM pg_class WHERE
        pg_total_relation_size(oid) &gt; pg_size_bytes('10 GB')</code>.
       </p></li><li class="listitem"><p>
        In <a class="link" href="functions-admin.html#FUNCTIONS-ADMIN-DBSIZE" title="Table 9.84. Database Object Size Functions"><code class="function">pg_size_pretty()</code></a>,
        format negative numbers similarly to positive ones (Adrian
        Vondendriesch)
       </p><p>
        Previously, negative numbers were never abbreviated, just printed
        in bytes.
       </p></li><li class="listitem"><p>
        Add an optional <em class="replaceable"><code>missing_ok</code></em> argument to the <a class="link" href="functions-admin.html#FUNCTIONS-ADMIN-SET-TABLE" title="Table 9.77. Configuration Settings Functions"><code class="function">current_setting()</code></a>
        function (David Christensen)
       </p><p>
        This allows avoiding an error for an unrecognized parameter
        name, instead returning a <code class="literal">NULL</code>.
       </p></li><li class="listitem"><p>
        Change various catalog-inspection functions to return
        <code class="literal">NULL</code> for invalid input (Michael Paquier)
       </p><p>
        <a class="link" href="functions-info.html#FUNCTIONS-INFO-CATALOG-TABLE" title="Table 9.63. System Catalog Information Functions"><code class="function">pg_get_viewdef()</code></a>
        now returns <code class="literal">NULL</code> if given an invalid view <code class="type">OID</code>,
        and several similar functions likewise return <code class="literal">NULL</code> for
        bad input.  Previously, such cases usually led to <span class="quote">“<span class="quote">cache
        lookup failed</span>”</span> errors, which are not meant to occur in
        user-facing cases.
       </p></li><li class="listitem"><p>
        Fix <a class="link" href="functions-admin.html#PG-REPLICATION-ORIGIN-XACT-RESET"><code class="function">pg_replication_origin_xact_reset()</code></a>
        to not have any arguments (Fujii Masao)
       </p><p>
        The documentation said that it has no arguments, and the C code did
        not expect any arguments, but the entry in <code class="structname">pg_proc</code>
        mistakenly specified two arguments.
       </p></li></ul></div></div><div class="sect3" id="id-1.11.6.25.5.10"><div class="titlepage"><div><div><h4 class="title">E.21.3.8. Server-Side Languages</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        In <a class="link" href="plpgsql.html" title="Chapter 43. PL/pgSQL - SQL Procedural Language">PL/pgSQL</a>, detect mismatched
        <code class="command">CONTINUE</code> and <code class="command">EXIT</code> statements while
        compiling a function, rather than at execution time
        (Jim Nasby)
       </p></li><li class="listitem"><p>
        Extend <span class="application">PL/Python</span>'s error-reporting and
        message-reporting functions to allow specifying additional message
        fields besides the primary error message (Pavel Stehule)
       </p></li><li class="listitem"><p>
        Allow PL/Python functions to call themselves recursively
        via <span class="application">SPI</span>, and fix the behavior when multiple
        set-returning PL/Python functions are called within one query
        (Alexey Grishchenko, Tom Lane)
       </p></li><li class="listitem"><p>
        Fix session-lifespan memory leaks in PL/Python (Heikki Linnakangas,
        Haribabu Kommi, Tom Lane)
       </p></li><li class="listitem"><p>
        Modernize <span class="application">PL/Tcl</span> to use Tcl's <span class="quote">“<span class="quote">object</span>”</span>
        <acronym class="acronym">API</acronym>s instead of simple strings (Jim Nasby, Karl
        Lehenbauer)
       </p><p>
        This can improve performance substantially in some cases.
        Note that <span class="application">PL/Tcl</span> now requires Tcl 8.4 or later.
       </p></li><li class="listitem"><p>
        In <span class="application">PL/Tcl</span>, make database-reported errors return
        additional information in Tcl's <code class="varname">errorCode</code> global
        variable (Jim Nasby, Tom Lane)
       </p><p>
        This feature follows the Tcl convention for returning auxiliary
        data about an error.
       </p></li><li class="listitem"><p>
        Fix <span class="application">PL/Tcl</span> to perform encoding conversion between
        the database encoding and <code class="literal">UTF-8</code>, which is what Tcl
        expects (Tom Lane)
       </p><p>
        Previously, strings were passed through without conversion,
        leading to misbehavior with non-<code class="literal">ASCII</code> characters when
        the database encoding was not <code class="literal">UTF-8</code>.
       </p></li></ul></div></div><div class="sect3" id="id-1.11.6.25.5.11"><div class="titlepage"><div><div><h4 class="title">E.21.3.9. Client Interfaces</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add a nonlocalized version of
        the <a class="link" href="protocol-error-fields.html" title="53.8. Error and Notice Message Fields">severity field</a> in
        error and notice messages (Tom Lane)
       </p><p>
        This change allows client code to determine severity of an error or
        notice without having to worry about localized variants of the
        severity strings.
       </p></li><li class="listitem"><p>
        Introduce a feature in <span class="application">libpq</span> whereby the
        <code class="literal">CONTEXT</code> field of messages can be suppressed, either
        always or only for non-error messages (Pavel Stehule)
       </p><p>
        The default behavior of <a class="link" href="libpq-status.html#LIBPQ-PQERRORMESSAGE"><code class="function">PQerrorMessage()</code></a>
        is now to print <code class="literal">CONTEXT</code>
        only for errors.  The new function <a class="link" href="libpq-control.html#LIBPQ-PQSETERRORCONTEXTVISIBILITY"><code class="function">PQsetErrorContextVisibility()</code></a>
        can be used to adjust this.
       </p></li><li class="listitem"><p>
        Add support in <span class="application">libpq</span> for regenerating an error
        message with a different verbosity level (Alex Shulgin)
       </p><p>
        This is done with the new function <a class="link" href="libpq-exec.html#LIBPQ-PQRESULTVERBOSEERRORMESSAGE"><code class="function">PQresultVerboseErrorMessage()</code></a>.
        This supports <span class="application">psql</span>'s new <code class="literal">\errverbose</code>
        feature, and may be useful for other clients as well.
       </p></li><li class="listitem"><p>
        Improve <span class="application">libpq</span>'s <a class="link" href="libpq-status.html#LIBPQ-PQHOST"><code class="function">PQhost()</code></a> function to return
        useful data for default Unix-socket connections (Tom Lane)
       </p><p>
        Previously it would return <code class="literal">NULL</code> if no explicit host
        specification had been given; now it returns the default socket
        directory path.
       </p></li><li class="listitem"><p>
        Fix <span class="application">ecpg</span>'s lexer to handle line breaks within
        comments starting on preprocessor directive lines (Michael Meskes)
       </p></li></ul></div></div><div class="sect3" id="id-1.11.6.25.5.12"><div class="titlepage"><div><div><h4 class="title">E.21.3.10. Client Applications</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add a <code class="option">--strict-names</code> option
        to <a class="link" href="app-pgdump.html" title="pg_dump"><span class="application">pg_dump</span></a>
        and <a class="link" href="app-pgrestore.html" title="pg_restore"><span class="application">pg_restore</span></a>
        (Pavel Stehule)
       </p><p>
        This option causes the program to complain if there is no match
        for a <code class="option">-t</code> or <code class="option">-n</code> option, rather
        than silently doing nothing.
       </p></li><li class="listitem"><p>
        In <span class="application">pg_dump</span>, dump locally-made changes of privilege
        assignments for system objects (Stephen Frost)
       </p><p>
        While it has always been possible for a superuser to change
        the privilege assignments for built-in or extension-created
        objects, such changes were formerly lost in a dump and reload.
        Now, <span class="application">pg_dump</span> recognizes and dumps such changes.
        (This works only when dumping from a 9.6 or later server, however.)
       </p></li><li class="listitem"><p>
        Allow <span class="application">pg_dump</span> to dump non-extension-owned objects
        that are within an extension-owned schema
        (Martín Marqués)
       </p><p>
        Previously such objects were ignored because they were mistakenly
        assumed to belong to the extension owning their schema.
       </p></li><li class="listitem"><p>
        In <span class="application">pg_dump</span> output, include the table name in object
        tags for object types that are only uniquely named per-table
        (for example, triggers) (Peter Eisentraut)
       </p></li></ul></div><div class="sect4" id="id-1.11.6.25.5.12.3"><div class="titlepage"><div><div><h5 class="title">E.21.3.10.1. <a class="xref" href="app-psql.html" title="psql"><span class="refentrytitle"><span class="application">psql</span></span></a></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Support multiple <code class="option">-c</code> and <code class="option">-f</code>
        command-line options (Pavel Stehule, Catalin Iacob)
       </p><p>
        The specified operations are carried out in the order in which the
        options are given, and then <span class="application">psql</span> terminates.
       </p></li><li class="listitem"><p>
        Add a <code class="command">\crosstabview</code> command that prints the results of
        a query in a cross-tabulated display (Daniel Vérité)
       </p><p>
        In the crosstab display, data values from one query result column
        are placed in a grid whose column and row headers come from other
        query result columns.
       </p></li><li class="listitem"><p>
        Add an <code class="literal">\errverbose</code> command that shows the last server
        error at full verbosity (Alex Shulgin)
       </p><p>
        This is useful after getting an unexpected error — you
        no longer need to adjust the <code class="varname">VERBOSITY</code> variable and
        recreate the failure in order to see error fields that are not
        shown by default.
       </p></li><li class="listitem"><p>
        Add <code class="literal">\ev</code> and <code class="literal">\sv</code> commands for editing and
        showing view definitions (Petr Korobeinikov)
       </p><p>
        These are parallel to the existing <code class="literal">\ef</code> and
        <code class="literal">\sf</code> commands for functions.
       </p></li><li class="listitem"><p>
        Add a <code class="command">\gexec</code> command that executes a query and
        re-submits the result(s) as new queries (Corey Huinker)
       </p></li><li class="listitem"><p>
        Allow <code class="literal">\pset C <em class="replaceable"><code>string</code></em></code>
        to set the table title, for consistency with <code class="literal">\C
        <em class="replaceable"><code>string</code></em></code> (Bruce Momjian)
       </p></li><li class="listitem"><p>
        In <code class="literal">\pset expanded auto</code> mode, do not use expanded
        format for query results with only one column (Andreas Karlsson,
        Robert Haas)
       </p></li><li class="listitem"><p>
        Improve the headers output by the <code class="command">\watch</code> command
        (Michael Paquier, Tom Lane)
       </p><p>
        Include the <code class="command">\pset title</code> string if one has
        been set, and shorten the prefabricated part of the
        header to be <code class="literal"><em class="replaceable"><code>timestamp</code></em> (every
        <em class="replaceable"><code>N</code></em>s)</code>.  Also, the timestamp format now
        obeys <span class="application">psql</span>'s locale environment.
       </p></li><li class="listitem"><p>
        Improve tab-completion logic to consider the entire input query,
        not only the current line (Tom Lane)
       </p><p>
        Previously, breaking a command into multiple lines defeated any
        tab completion rules that needed to see words on earlier lines.
       </p></li><li class="listitem"><p>
        Numerous minor improvements in tab-completion behavior (Peter
        Eisentraut, Vik Fearing, Kevin Grittner, Kyotaro Horiguchi, Jeff
        Janes, Andreas Karlsson, Fujii Masao, Thomas Munro, Masahiko
        Sawada, Pavel Stehule)
       </p></li><li class="listitem"><p>
        Add a <code class="literal">PROMPT</code> option <code class="literal">%p</code> to insert the
        process ID of the connected backend (Julien Rouhaud)
       </p></li><li class="listitem"><p>
        Introduce a feature whereby the <code class="literal">CONTEXT</code> field of
        messages can be suppressed, either always or only for non-error
        messages (Pavel Stehule)
       </p><p>
        Printing <code class="literal">CONTEXT</code> only for errors is now the default
        behavior.  This can be changed by setting the special variable
        <code class="varname">SHOW_CONTEXT</code>.
       </p></li><li class="listitem"><p>
        Make <code class="command">\df+</code> show function access privileges and
        parallel-safety attributes (Michael Paquier)
       </p></li></ul></div></div><div class="sect4" id="id-1.11.6.25.5.12.4"><div class="titlepage"><div><div><h5 class="title">E.21.3.10.2. <a class="xref" href="pgbench.html" title="pgbench"><span class="refentrytitle"><span class="application">pgbench</span></span></a></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        SQL commands in <span class="application">pgbench</span> scripts are now ended by
        semicolons, not newlines (Kyotaro Horiguchi, Tom Lane)
       </p><p>
        This change allows SQL commands in scripts to span multiple lines.
        Existing custom scripts will need to be modified to add a semicolon
        at the end of each line that does not have one already.  (Doing so
        does not break the script for use with older versions
        of <span class="application">pgbench</span>.)
       </p></li><li class="listitem"><p>
        Support floating-point arithmetic, as well as some <a class="link" href="pgbench.html#PGBENCH-BUILTIN-FUNCTIONS" title="Built-In Functions">built-in functions</a>, in
        expressions in backslash commands (Fabien Coelho)
       </p></li><li class="listitem"><p>
        Replace <code class="command">\setrandom</code> with built-in functions (Fabien
        Coelho)
       </p><p>
        The new built-in functions include <a class="link" href="pgbench.html#PGBENCH-FUNCTIONS" title="Table 243. pgbench Functions"><code class="function">random()</code></a>,
        <code class="function">random_exponential()</code>, and
        <code class="function">random_gaussian()</code>, which perform the same work as
        <code class="command">\setrandom</code>, but are easier to use since they can be
        embedded in larger expressions.  Since these additions have made
        <code class="command">\setrandom</code> obsolete, remove it.
       </p></li><li class="listitem"><p>
        Allow invocation of multiple copies of the built-in scripts,
        not only custom scripts (Fabien Coelho)
       </p><p>
        This is done with the new <code class="option">-b</code> switch, which works
        similarly to <code class="option">-f</code> for custom scripts.
       </p></li><li class="listitem"><p>
        Allow changing the selection probabilities (weights) for scripts
        (Fabien Coelho)
       </p><p>
        When multiple scripts are specified, each <span class="application">pgbench</span>
        transaction randomly chooses one to execute.  Formerly this was
        always done with uniform probability, but now different selection
        probabilities can be specified for different scripts.
       </p></li><li class="listitem"><p>
        Collect statistics for each script in a multi-script run (Fabien
        Coelho)
       </p><p>
        This feature adds an intermediate level of detail to existing
        global and per-command statistics printouts.
       </p></li><li class="listitem"><p>
        Add a <code class="option">--progress-timestamp</code> option to report progress
        with Unix epoch timestamps, instead of time since the run started
        (Fabien Coelho)
       </p></li><li class="listitem"><p>
        Allow the number of client connections (<code class="option">-c</code>) to not
        be an exact multiple of the number of threads (<code class="option">-j</code>)
        (Fabien Coelho)
       </p></li><li class="listitem"><p>
        When the <code class="option">-T</code> option is used, stop promptly at the end
        of the specified time (Fabien Coelho)
       </p><p>
        Previously, specifying a low transaction rate could cause
        <span class="application">pgbench</span> to wait significantly longer than
        specified.
       </p></li></ul></div></div></div><div class="sect3" id="id-1.11.6.25.5.13"><div class="titlepage"><div><div><h4 class="title">E.21.3.11. Server Applications</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Improve error reporting during <span class="application">initdb</span>'s
        post-bootstrap phase (Tom Lane)
       </p><p>
        Previously, an error here led to reporting the entire input
        file as the <span class="quote">“<span class="quote">failing query</span>”</span>; now just the current
        query is reported.  To get the desired behavior, queries in
        <span class="application">initdb</span>'s input files must be separated by blank
        lines.
       </p></li><li class="listitem"><p>
        Speed up <span class="application">initdb</span> by using just one
        standalone-backend session for all the post-bootstrap steps
        (Tom Lane)
       </p></li><li class="listitem"><p>
        Improve <a class="link" href="app-pgrewind.html" title="pg_rewind"><span class="application">pg_rewind</span></a>
        so that it can work when the target timeline changes (Alexander
        Korotkov)
       </p><p>
        This allows, for example, rewinding a promoted standby back to
        some state of the old master's timeline.
       </p></li></ul></div></div><div class="sect3" id="id-1.11.6.25.5.14"><div class="titlepage"><div><div><h4 class="title">E.21.3.12. Source Code</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Remove obsolete
        <code class="function">heap_formtuple</code>/<code class="function">heap_modifytuple</code>/<code class="function">heap_deformtuple</code>
        functions (Peter Geoghegan)
       </p></li><li class="listitem"><p>
        Add macros to make <code class="function">AllocSetContextCreate()</code> calls simpler
        and safer (Tom Lane)
       </p><p>
        Writing out the individual sizing parameters for a memory context
        is now deprecated in favor of using one of the new
        macros <code class="symbol">ALLOCSET_DEFAULT_SIZES</code>,
        <code class="symbol">ALLOCSET_SMALL_SIZES</code>,
        or <code class="symbol">ALLOCSET_START_SMALL_SIZES</code>.
        Existing code continues to work, however.
       </p></li><li class="listitem"><p>
        Unconditionally use <code class="literal">static inline</code> functions in header
        files (Andres Freund)
       </p><p>
        This may result in warnings and/or wasted code space with very
        old compilers, but the notational improvement seems worth it.
       </p></li><li class="listitem"><p>
        Improve <span class="application">TAP</span> testing infrastructure (Michael
        Paquier, Craig Ringer, Álvaro Herrera, Stephen Frost)
       </p><p>
        Notably, it is now possible to test recovery scenarios using
        this infrastructure.
       </p></li><li class="listitem"><p>
        Make <code class="varname">trace_lwlocks</code> identify individual locks by name
        (Robert Haas)
       </p></li><li class="listitem"><p>
        Improve <span class="application">psql</span>'s tab-completion code infrastructure
        (Thomas Munro, Michael Paquier)
       </p><p>
        Tab-completion rules are now considerably easier to write, and
        more compact.
       </p></li><li class="listitem"><p>
        Nail the <code class="structname">pg_shseclabel</code> system catalog into cache,
        so that it is available for access during connection authentication
        (Adam Brightwell)
       </p><p>
        The core code does not use this catalog for authentication,
        but extensions might wish to consult it.
       </p></li><li class="listitem"><p>
        Restructure <a class="link" href="indexam.html" title="Chapter 61. Index Access Method Interface Definition">index access
        method <acronym class="acronym">API</acronym></a> to hide most of it at
        the <span class="application">C</span> level (Alexander Korotkov, Andrew Gierth)
       </p><p>
        This change modernizes the index <acronym class="acronym">AM API</acronym> to look more
        like the designs we have adopted for foreign data wrappers and
        tablesample handlers.  This simplifies the <span class="application">C</span> code
        and makes it much more practical to define index access methods in
        installable extensions.  A consequence is that most of the columns
        of the <code class="structname">pg_am</code> system catalog have disappeared.
        New <a class="link" href="functions-info.html#FUNCTIONS-INFO-CATALOG-TABLE" title="Table 9.63. System Catalog Information Functions">inspection
        functions</a> have been added to allow SQL queries to determine
        index AM properties that used to be discoverable
        from <code class="structname">pg_am</code>.
       </p></li><li class="listitem"><p>
        Add <a class="link" href="catalog-pg-init-privs.html" title="52.28. pg_init_privs"><code class="structname">pg_init_privs</code></a>
        system catalog to hold original privileges
        of <span class="application">initdb</span>-created and extension-created objects
        (Stephen Frost)
       </p><p>
        This infrastructure allows <span class="application">pg_dump</span> to dump changes
        that an installation may have made in privileges attached to
        system objects.  Formerly, such changes would be lost in a dump
        and reload, but now they are preserved.
       </p></li><li class="listitem"><p>
        Change the way that extensions allocate custom <code class="literal">LWLocks</code>
        (Amit Kapila, Robert Haas)
       </p><p>
        The <code class="function">RequestAddinLWLocks()</code> function is removed,
        and replaced by <code class="function">RequestNamedLWLockTranche()</code>.
        This allows better identification of custom <code class="literal">LWLocks</code>,
        and is less error-prone.
       </p></li><li class="listitem"><p>
        Improve the isolation tester to allow multiple sessions to wait
        concurrently, allowing testing of deadlock scenarios (Robert Haas)
       </p></li><li class="listitem"><p>
        Introduce extensible node types (KaiGai Kohei)
       </p><p>
        This change allows <span class="application">FDW</span>s or custom scan providers
        to store data in a plan tree in a more convenient format than
        was previously possible.
       </p></li><li class="listitem"><p>
        Make the planner deal with post-scan/join query steps by generating
        and comparing <code class="literal">Paths</code>, replacing a lot of ad-hoc logic
        (Tom Lane)
       </p><p>
        This change provides only marginal user-visible improvements today,
        but it enables future work on a lot of upper-planner improvements
        that were impractical to tackle using the old code structure.
       </p></li><li class="listitem"><p>
        Support partial aggregation (David Rowley, Simon Riggs)
       </p><p>
        This change allows the computation of an aggregate function to be
        split into separate parts, for example so that parallel worker
        processes can cooperate on computing an aggregate.  In future
        it might allow aggregation across local and remote data to occur
        partially on the remote end.
       </p></li><li class="listitem"><p>
        Add a generic command progress reporting facility (Vinayak Pokale,
        Rahila Syed, Amit Langote, Robert Haas)
       </p></li><li class="listitem"><p>
        Separate out <span class="application">psql</span>'s <span class="application">flex</span> lexer to
        make it usable by other client programs (Tom Lane, Kyotaro
        Horiguchi)
       </p><p>
        This eliminates code duplication for programs that need to be able
        to parse SQL commands well enough to identify command boundaries.
        Doing that in full generality is more painful than one could
        wish, and up to now only <span class="application">psql</span> has really gotten
        it right among our supported client programs.
       </p><p>
        A new source-code subdirectory <code class="filename">src/fe_utils/</code> has
        been created to hold this and other code that is shared across
        our client programs.  Formerly such sharing was accomplished by
        symbolic linking or copying source files at build time, which
        was ugly and required duplicate compilation.
       </p></li><li class="listitem"><p>
        Introduce <code class="literal">WaitEventSet</code> <acronym class="acronym">API</acronym> to allow
        efficient waiting for event sets that usually do not change from
        one wait to the next (Andres Freund, Amit Kapila)
       </p></li><li class="listitem"><p>
        Add a generic interface for writing <acronym class="acronym">WAL</acronym> records
        (Alexander Korotkov, Petr Jelínek, Markus Nullmeier)
       </p><p>
        This change allows extensions to write <acronym class="acronym">WAL</acronym> records for
        changes to pages using a standard layout.  The problem of needing to
        replay <acronym class="acronym">WAL</acronym> without access to the extension is solved by
        having generic replay code.  This allows extensions to implement,
        for example, index access methods and have <acronym class="acronym">WAL</acronym>
        support for them.
       </p></li><li class="listitem"><p>
        Support generic <acronym class="acronym">WAL</acronym> messages for logical decoding
        (Petr Jelínek, Andres Freund)
       </p><p>
        This feature allows extensions to insert data into the
        <acronym class="acronym">WAL</acronym> stream that can be read by logical-decoding
        plugins, but is not connected to physical data restoration.
       </p></li><li class="listitem"><p>
        Allow SP-GiST operator classes to store an arbitrary
        <span class="quote">“<span class="quote">traversal value</span>”</span> while descending the index (Alexander
        Lebedev, Teodor Sigaev)
       </p><p>
        This is somewhat like the <span class="quote">“<span class="quote">reconstructed value</span>”</span>, but it
        could be any arbitrary chunk of data, not necessarily of the same
        data type as the indexed column.
       </p></li><li class="listitem"><p>
        Introduce a <code class="literal">LOG_SERVER_ONLY</code> message level for
        <code class="function">ereport()</code> (David Steele)
       </p><p>
        This level acts like <code class="literal">LOG</code> except that the message is
        never sent to the client.  It is meant for use in auditing and
        similar applications.
       </p></li><li class="listitem"><p>
        Provide a <code class="filename">Makefile</code> target to build all generated
        headers (Michael Paquier, Tom Lane)
       </p><p>
        <code class="literal">submake-generated-headers</code> can now be invoked to ensure
        that generated backend header files are up-to-date.  This is
        useful in subdirectories that might be built <span class="quote">“<span class="quote">standalone</span>”</span>.
       </p></li><li class="listitem"><p>
        Support OpenSSL 1.1.0 (Andreas Karlsson, Heikki Linnakangas)
       </p></li></ul></div></div><div class="sect3" id="id-1.11.6.25.5.15"><div class="titlepage"><div><div><h4 class="title">E.21.3.13. Additional Modules</h4></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Add configuration parameter <code class="literal">auto_explain.sample_rate</code> to
        allow <a class="link" href="auto-explain.html" title="F.4. auto_explain"><code class="filename">contrib/auto_explain</code></a>
        to capture just a configurable fraction of all queries (Craig
        Ringer, Julien Rouhaud)
       </p><p>
        This allows reduction of overhead for heavy query traffic, while
        still getting useful information on average.
       </p></li><li class="listitem"><p>
        Add <a class="link" href="bloom.html" title="F.5. bloom"><code class="filename">contrib/bloom</code></a> module that
        implements an index access method based on Bloom filtering (Teodor
        Sigaev, Alexander Korotkov)
       </p><p>
        This is primarily a proof-of-concept for non-core index access
        methods, but it could be useful in its own right for queries that
        search many columns.
       </p></li><li class="listitem"><p>
        In <a class="link" href="cube.html" title="F.9. cube"><code class="filename">contrib/cube</code></a>, introduce
        distance operators for cubes, and support kNN-style searches in
        GiST indexes on cube columns (Stas Kelvich)
       </p></li><li class="listitem"><p>
        Make <code class="filename">contrib/hstore</code>'s <a class="link" href="hstore.html#HSTORE-FUNC-TABLE" title="Table F.8. hstore Functions"><code class="function">hstore_to_jsonb_loose()</code></a>
        and <code class="function">hstore_to_json_loose()</code> functions agree on what
        is a number (Tom Lane)
       </p><p>
        Previously, <code class="function">hstore_to_jsonb_loose()</code> would convert
        numeric-looking strings to <acronym class="acronym">JSON</acronym> numbers, rather than
        strings, even if they did not exactly match the <acronym class="acronym">JSON</acronym>
        syntax specification for numbers.  This was inconsistent with
        <code class="function">hstore_to_json_loose()</code>, so tighten the test to match
        the <acronym class="acronym">JSON</acronym> syntax.
       </p></li><li class="listitem"><p>
        Add selectivity estimation functions for
        <a class="link" href="intarray.html" title="F.18. intarray"><code class="filename">contrib/intarray</code></a> operators
        to improve plans for queries using those operators (Yury Zhuravlev,
        Alexander Korotkov)
       </p></li><li class="listitem"><p>
        Make <a class="link" href="pageinspect.html" title="F.22. pageinspect"><code class="filename">contrib/pageinspect</code></a>'s
        <code class="function">heap_page_items()</code> function show the raw data in each
        tuple, and add new functions <code class="function">tuple_data_split()</code> and
        <code class="function">heap_page_item_attrs()</code> for inspection of individual
        tuple fields (Nikolay Shaplov)
       </p></li><li class="listitem"><p>
        Add an optional <code class="literal">S2K</code> iteration count parameter to
        <a class="link" href="pgcrypto.html" title="F.25. pgcrypto"><code class="filename">contrib/pgcrypto</code></a>'s
        <code class="function">pgp_sym_encrypt()</code> function (Jeff Janes)
       </p></li><li class="listitem"><p>
        Add support for <span class="quote">“<span class="quote">word similarity</span>”</span> to
        <a class="link" href="pgtrgm.html" title="F.31. pg_trgm"><code class="filename">contrib/pg_trgm</code></a>
        (Alexander Korotkov, Artur Zakirov)
       </p><p>
        These functions and operators measure the similarity between one
        string and the most similar single word of another string.
       </p></li><li class="listitem"><p>
        Add configuration parameter
        <code class="varname">pg_trgm.similarity_threshold</code> for
        <code class="filename">contrib/pg_trgm</code>'s similarity threshold (Artur Zakirov)
       </p><p>
        This threshold has always been configurable, but formerly it was
        controlled by special-purpose functions <code class="function">set_limit()</code>
        and <code class="function">show_limit()</code>.  Those are now deprecated.
       </p></li><li class="listitem"><p>
        Improve <code class="filename">contrib/pg_trgm</code>'s GIN operator class to
        speed up index searches in which both common and rare keys appear
        (Jeff Janes)
       </p></li><li class="listitem"><p>
        Improve performance of similarity searches in
        <code class="filename">contrib/pg_trgm</code> GIN indexes (Christophe Fornaroli)
       </p></li><li class="listitem"><p>
        Add <a class="link" href="pgvisibility.html" title="F.32. pg_visibility"><code class="filename">contrib/pg_visibility</code></a> module
        to allow examining table visibility maps (Robert Haas)
       </p></li><li class="listitem"><p>
        Add <a class="link" href="sslinfo.html" title="F.37. sslinfo"><code class="function">ssl_extension_info()</code></a>
        function to <code class="filename">contrib/sslinfo</code>, to print information
        about <acronym class="acronym">SSL</acronym> extensions present in the <code class="literal">X509</code>
        certificate used for the current connection (Dmitry Voronin)
       </p></li></ul></div><div class="sect4" id="id-1.11.6.25.5.15.3"><div class="titlepage"><div><div><h5 class="title">E.21.3.13.1. <a class="link" href="postgres-fdw.html" title="F.33. postgres_fdw"><code class="filename">postgres_fdw</code></a></h5></div></div></div><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>
        Allow extension-provided operators and functions to be sent for
        remote execution, if the extension is whitelisted in the foreign
        server's options (Paul Ramsey)
       </p><p>
        Users can enable this feature when the extension is known to exist
        in a compatible version in the remote database.  It allows more
        efficient execution of queries involving extension operators.
       </p></li><li class="listitem"><p>
        Consider performing sorts on the remote server (Ashutosh Bapat)
       </p></li><li class="listitem"><p>
        Consider performing joins on the remote server (Shigeru Hanada,
        Ashutosh Bapat)
       </p></li><li class="listitem"><p>
        When feasible, perform <code class="command">UPDATE</code> or <code class="command">DELETE</code>
        entirely on the remote server (Etsuro Fujita)
       </p><p>
        Formerly, remote updates involved sending a <code class="command">SELECT FOR UPDATE</code>
        command and then updating or deleting the selected rows one-by-one.
        While that is still necessary if the operation requires any local
        processing, it can now be done remotely if all elements of the
        query are safe to send to the remote server.
       </p></li><li class="listitem"><p>
        Allow the fetch size to be set as a server or table option
        (Corey Huinker)
       </p><p>
        Formerly, <code class="filename">postgres_fdw</code> always fetched 100 rows at
        a time from remote queries; now that behavior is configurable.
       </p></li><li class="listitem"><p>
        Use a single foreign-server connection for local user IDs that
        all map to the same remote user (Ashutosh Bapat)
       </p></li><li class="listitem"><p>
        Transmit query cancellation requests to the remote server
        (Michael Paquier, Etsuro Fujita)
       </p><p>
        Previously, a local query cancellation request did not cause an
        already-sent remote query to terminate early.
       </p></li></ul></div></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-9-6-1.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-9-5-15.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">E.20. Release 9.6.1 </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> E.22. Release 9.5.15</td></tr></table></div></body></html>