Baoding-FenShuaJiang
/
Green_GoodERP18_FSJ_Standard
Suivre 1
Ajouter aux favoris 0
Wiki 捐赠
gooderp18绿色标准版
Vous ne pouvez pas sélectionner plus de 25 sujets Les noms de sujets doivent commencer par une lettre ou un nombre, peuvent contenir des tirets ('-') et peuvent comporter jusqu'à 35 caractères.
Branche: master
Branches Tags
${ item.name }
Créer la branche ${ searchTerm }
de 'master'
${ noResults }
Green_GoodERP18_FSJ_Standard/PostgreSQL/doc/postgresql/html/sql-createindex.html

505 lignes
36KB
Brut Lien permanent Annotations Historique 

  1. <?xml version="1.0" encoding="UTF-8" standalone="no"?>

  2. <!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>CREATE INDEX</title><link rel="stylesheet" type="text/css" href="stylesheet.css" /><link rev="made" href="pgsql-docs@lists.postgresql.org" /><meta name="generator" content="DocBook XSL Stylesheets V1.79.1" /><link rel="prev" href="sql-creategroup.html" title="CREATE GROUP" /><link rel="next" href="sql-createlanguage.html" title="CREATE LANGUAGE" /></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">CREATE INDEX</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="sql-creategroup.html" title="CREATE GROUP">Prev</a> </td><td width="10%" align="left"><a accesskey="u" href="sql-commands.html" title="SQL Commands">Up</a></td><th width="60%" align="center">SQL Commands</th><td width="10%" align="right"><a accesskey="h" href="index.html" title="PostgreSQL 12.4 Documentation">Home</a></td><td width="10%" align="right"> <a accesskey="n" href="sql-createlanguage.html" title="CREATE LANGUAGE">Next</a></td></tr></table><hr></hr></div><div class="refentry" id="SQL-CREATEINDEX"><div class="titlepage"></div><a id="id-1.9.3.69.1" class="indexterm"></a><div class="refnamediv"><h2><span class="refentrytitle">CREATE INDEX</span></h2><p>CREATE INDEX — define a new index</p></div><div class="refsynopsisdiv"><h2>Synopsis</h2><pre class="synopsis">

  3. CREATE [ UNIQUE ] INDEX [ CONCURRENTLY ] [ [ IF NOT EXISTS ] <em class="replaceable"><code>name</code></em> ] ON [ ONLY ] <em class="replaceable"><code>table_name</code></em> [ USING <em class="replaceable"><code>method</code></em> ]

  4.     ( { <em class="replaceable"><code>column_name</code></em> | ( <em class="replaceable"><code>expression</code></em> ) } [ COLLATE <em class="replaceable"><code>collation</code></em> ] [ <em class="replaceable"><code>opclass</code></em> ] [ ASC | DESC ] [ NULLS { FIRST | LAST } ] [, ...] )

  5.     [ INCLUDE ( <em class="replaceable"><code>column_name</code></em> [, ...] ) ]

  6.     [ WITH ( <em class="replaceable"><code>storage_parameter</code></em> [= <em class="replaceable"><code>value</code></em>] [, ... ] ) ]

  7.     [ TABLESPACE <em class="replaceable"><code>tablespace_name</code></em> ]

  8.     [ WHERE <em class="replaceable"><code>predicate</code></em> ]

  9. </pre></div><div class="refsect1" id="id-1.9.3.69.5"><h2>Description</h2><p>

  10.    <code class="command">CREATE INDEX</code> constructs an index on the specified column(s)

  11.    of the specified relation, which can be a table or a materialized view.

  12.    Indexes are primarily used to enhance database performance (though

  13.    inappropriate use can result in slower performance).

  14.   </p><p>

  15.    The key field(s) for the index are specified as column names,

  16.    or alternatively as expressions written in parentheses.

  17.    Multiple fields can be specified if the index method supports

  18.    multicolumn indexes.

  19.   </p><p>

  20.    An index field can be an expression computed from the values of

  21.    one or more columns of the table row.  This feature can be used

  22.    to obtain fast access to data based on some transformation of

  23.    the basic data. For example, an index computed on

  24.    <code class="literal">upper(col)</code> would allow the clause

  25.    <code class="literal">WHERE upper(col) = 'JIM'</code> to use an index.

  26.   </p><p>

  27.    <span class="productname">PostgreSQL</span> provides the index methods

  28.    B-tree, hash, GiST, SP-GiST, GIN, and BRIN.  Users can also define their own

  29.    index methods, but that is fairly complicated.

  30.   </p><p>

  31.     When the <code class="literal">WHERE</code> clause is present, a

  32.     <em class="firstterm">partial index</em> is created.

  33.     A partial index is an index that contains entries for only a portion of

  34.     a table, usually a portion that is more useful for indexing than the

  35.     rest of the table. For example, if you have a table that contains both

  36.     billed and unbilled orders where the unbilled orders take up a small

  37.     fraction of the total table and yet that is an often used section, you

  38.     can improve performance by creating an index on just that portion.

  39.     Another possible application is to use <code class="literal">WHERE</code> with

  40.     <code class="literal">UNIQUE</code> to enforce uniqueness over a subset of a

  41.     table.  See <a class="xref" href="indexes-partial.html" title="11.8. Partial Indexes">Section 11.8</a> for more discussion.

  42.   </p><p>

  43.     The expression used in the <code class="literal">WHERE</code> clause can refer

  44.     only to columns of the underlying table, but it can use all columns,

  45.     not just the ones being indexed.  Presently, subqueries and

  46.     aggregate expressions are also forbidden in <code class="literal">WHERE</code>.

  47.     The same restrictions apply to index fields that are expressions.

  48.   </p><p>

  49.    All functions and operators used in an index definition must be

  50.    <span class="quote">“<span class="quote">immutable</span>”</span>, that is, their results must depend only on

  51.    their arguments and never on any outside influence (such as

  52.    the contents of another table or the current time).  This restriction

  53.    ensures that the behavior of the index is well-defined.  To use a

  54.    user-defined function in an index expression or <code class="literal">WHERE</code>

  55.    clause, remember to mark the function immutable when you create it.

  56.   </p></div><div class="refsect1" id="id-1.9.3.69.6"><h2>Parameters</h2><div class="variablelist"><dl class="variablelist"><dt><span class="term"><code class="literal">UNIQUE</code></span></dt><dd><p>

  57.         Causes the system to check for

  58.         duplicate values in the table when the index is created (if data

  59.         already exist) and each time data is added. Attempts to

  60.         insert or update data which would result in duplicate entries

  61.         will generate an error.

  62.        </p><p>

  63.         Additional restrictions apply when unique indexes are applied to

  64.         partitioned tables; see <a class="xref" href="sql-createtable.html" title="CREATE TABLE"><span class="refentrytitle">CREATE TABLE</span></a>.

  65.        </p></dd><dt><span class="term"><code class="literal">CONCURRENTLY</code></span></dt><dd><p>

  66.         When this option is used, <span class="productname">PostgreSQL</span> will build the

  67.         index without taking any locks that prevent concurrent inserts,

  68.         updates, or deletes on the table; whereas a standard index build

  69.         locks out writes (but not reads) on the table until it's done.

  70.         There are several caveats to be aware of when using this option

  71.         — see <a class="xref" href="sql-createindex.html#SQL-CREATEINDEX-CONCURRENTLY" title="Building Indexes Concurrently">Building Indexes Concurrently</a>.

  72.        </p><p>

  73.         For temporary tables, <code class="command">CREATE INDEX</code> is always

  74.         non-concurrent, as no other session can access them, and

  75.         non-concurrent index creation is cheaper.

  76.        </p></dd><dt><span class="term"><code class="literal">IF NOT EXISTS</code></span></dt><dd><p>

  77.         Do not throw an error if a relation with the same name already exists.

  78.         A notice is issued in this case. Note that there is no guarantee that

  79.         the existing index is anything like the one that would have been created.

  80.         Index name is required when <code class="literal">IF NOT EXISTS</code> is specified.

  81.        </p></dd><dt><span class="term"><code class="literal">INCLUDE</code></span></dt><dd><p>

  82.         The optional <code class="literal">INCLUDE</code> clause specifies a

  83.         list of columns which will be included in the index

  84.         as <em class="firstterm">non-key</em> columns.  A non-key column cannot

  85.         be used in an index scan search qualification, and it is disregarded

  86.         for purposes of any uniqueness or exclusion constraint enforced by

  87.         the index.  However, an index-only scan can return the contents of

  88.         non-key columns without having to visit the index's table, since

  89.         they are available directly from the index entry.  Thus, addition of

  90.         non-key columns allows index-only scans to be used for queries that

  91.         otherwise could not use them.

  92.        </p><p>

  93.         It's wise to be conservative about adding non-key columns to an

  94.         index, especially wide columns.  If an index tuple exceeds the

  95.         maximum size allowed for the index type, data insertion will fail.

  96.         In any case, non-key columns duplicate data from the index's table

  97.         and bloat the size of the index, thus potentially slowing searches.

  98.        </p><p>

  99.         Columns listed in the <code class="literal">INCLUDE</code> clause don't need

  100.         appropriate operator classes; the clause can include

  101.         columns whose data types don't have operator classes defined for

  102.         a given access method.

  103.        </p><p>

  104.         Expressions are not supported as included columns since they cannot be

  105.         used in index-only scans.

  106.        </p><p>

  107.         Currently, the B-tree and the GiST index access methods support this

  108.         feature.  In B-tree and the GiST indexes, the values of columns listed

  109.         in the <code class="literal">INCLUDE</code> clause are included in leaf tuples

  110.         which correspond to heap tuples, but are not included in upper-level

  111.         index entries used for tree navigation.

  112.        </p></dd><dt><span class="term"><em class="replaceable"><code>name</code></em></span></dt><dd><p>

  113.         The name of the index to be created.  No schema name can be included

  114.         here; the index is always created in the same schema as its parent

  115.         table.  If the name is omitted, <span class="productname">PostgreSQL</span> chooses a

  116.         suitable name based on the parent table's name and the indexed column

  117.         name(s).

  118.        </p></dd><dt><span class="term"><code class="literal">ONLY</code></span></dt><dd><p>

  119.         Indicates not to recurse creating indexes on partitions, if the

  120.         table is partitioned.  The default is to recurse.

  121.        </p></dd><dt><span class="term"><em class="replaceable"><code>table_name</code></em></span></dt><dd><p>

  122.         The name (possibly schema-qualified) of the table to be indexed.

  123.        </p></dd><dt><span class="term"><em class="replaceable"><code>method</code></em></span></dt><dd><p>

  124.         The name of the index method to be used.  Choices are

  125.         <code class="literal">btree</code>, <code class="literal">hash</code>,

  126.         <code class="literal">gist</code>, <code class="literal">spgist</code>, <code class="literal">gin</code>, and

  127.         <code class="literal">brin</code>.

  128.         The default method is <code class="literal">btree</code>.

  129.        </p></dd><dt><span class="term"><em class="replaceable"><code>column_name</code></em></span></dt><dd><p>

  130.         The name of a column of the table.

  131.        </p></dd><dt><span class="term"><em class="replaceable"><code>expression</code></em></span></dt><dd><p>

  132.         An expression based on one or more columns of the table.  The

  133.         expression usually must be written with surrounding parentheses,

  134.         as shown in the syntax.  However, the parentheses can be omitted

  135.         if the expression has the form of a function call.

  136.        </p></dd><dt><span class="term"><em class="replaceable"><code>collation</code></em></span></dt><dd><p>

  137.         The name of the collation to use for the index.  By default,

  138.         the index uses the collation declared for the column to be

  139.         indexed or the result collation of the expression to be

  140.         indexed.  Indexes with non-default collations can be useful for

  141.         queries that involve expressions using non-default collations.

  142.        </p></dd><dt><span class="term"><em class="replaceable"><code>opclass</code></em></span></dt><dd><p>

  143.         The name of an operator class. See below for details.

  144.        </p></dd><dt><span class="term"><code class="literal">ASC</code></span></dt><dd><p>

  145.         Specifies ascending sort order (which is the default).

  146.        </p></dd><dt><span class="term"><code class="literal">DESC</code></span></dt><dd><p>

  147.         Specifies descending sort order.

  148.        </p></dd><dt><span class="term"><code class="literal">NULLS FIRST</code></span></dt><dd><p>

  149.         Specifies that nulls sort before non-nulls.  This is the default

  150.         when <code class="literal">DESC</code> is specified.

  151.        </p></dd><dt><span class="term"><code class="literal">NULLS LAST</code></span></dt><dd><p>

  152.         Specifies that nulls sort after non-nulls.  This is the default

  153.         when <code class="literal">DESC</code> is not specified.

  154.        </p></dd><dt><span class="term"><em class="replaceable"><code>storage_parameter</code></em></span></dt><dd><p>

  155.         The name of an index-method-specific storage parameter.  See

  156.         <a class="xref" href="sql-createindex.html#SQL-CREATEINDEX-STORAGE-PARAMETERS" title="Index Storage Parameters">Index Storage Parameters</a>

  157.         for details.

  158.        </p></dd><dt><span class="term"><em class="replaceable"><code>tablespace_name</code></em></span></dt><dd><p>

  159.         The tablespace in which to create the index.  If not specified,

  160.         <a class="xref" href="runtime-config-client.html#GUC-DEFAULT-TABLESPACE">default_tablespace</a> is consulted, or

  161.         <a class="xref" href="runtime-config-client.html#GUC-TEMP-TABLESPACES">temp_tablespaces</a> for indexes on temporary

  162.         tables.

  163.        </p></dd><dt><span class="term"><em class="replaceable"><code>predicate</code></em></span></dt><dd><p>

  164.         The constraint expression for a partial index.

  165.        </p></dd></dl></div><div class="refsect2" id="SQL-CREATEINDEX-STORAGE-PARAMETERS"><h3>Index Storage Parameters</h3><p>

  166.     The optional <code class="literal">WITH</code> clause specifies <em class="firstterm">storage

  167.     parameters</em> for the index.  Each index method has its own set of allowed

  168.     storage parameters.  The B-tree, hash, GiST and SP-GiST index methods all

  169.     accept this parameter:

  170.    </p><div class="variablelist"><dl class="variablelist"><dt id="INDEX-RELOPTION-FILLFACTOR"><span class="term"><code class="literal">fillfactor</code>

  171.      <a id="id-1.9.3.69.6.3.3.1.1.2" class="indexterm"></a>

  172.     </span></dt><dd><p>

  173.       The fillfactor for an index is a percentage that determines how full

  174.       the index method will try to pack index pages.  For B-trees, leaf pages

  175.       are filled to this percentage during initial index build, and also

  176.       when extending the index at the right (adding new largest key values).

  177.       If pages

  178.       subsequently become completely full, they will be split, leading to

  179.       gradual degradation in the index's efficiency.  B-trees use a default

  180.       fillfactor of 90, but any integer value from 10 to 100 can be selected.

  181.       If the table is static then fillfactor 100 is best to minimize the

  182.       index's physical size, but for heavily updated tables a smaller

  183.       fillfactor is better to minimize the need for page splits.  The

  184.       other index methods use fillfactor in different but roughly analogous

  185.       ways; the default fillfactor varies between methods.

  186.      </p></dd></dl></div><p>

  187.     B-tree indexes additionally accept this parameter:

  188.    </p><div class="variablelist"><dl class="variablelist"><dt id="INDEX-RELOPTION-VACUUM-CLEANUP-INDEX-SCALE-FACTOR"><span class="term"><code class="literal">vacuum_cleanup_index_scale_factor</code>

  189.      <a id="id-1.9.3.69.6.3.5.1.1.2" class="indexterm"></a>

  190.     </span></dt><dd><p>

  191.       Per-index value for <a class="xref" href="runtime-config-client.html#GUC-VACUUM-CLEANUP-INDEX-SCALE-FACTOR">vacuum_cleanup_index_scale_factor</a>.

  192.     </p></dd></dl></div><p>

  193.     GiST indexes additionally accept this parameter:

  194.    </p><div class="variablelist"><dl class="variablelist"><dt id="INDEX-RELOPTION-BUFFERING"><span class="term"><code class="literal">buffering</code>

  195.      <a id="id-1.9.3.69.6.3.7.1.1.2" class="indexterm"></a>

  196.     </span></dt><dd><p>

  197.      Determines whether the buffering build technique described in

  198.      <a class="xref" href="gist-implementation.html#GIST-BUFFERING-BUILD" title="64.4.1. GiST Buffering Build">Section 64.4.1</a> is used to build the index. With

  199.      <code class="literal">OFF</code> it is disabled, with <code class="literal">ON</code> it is enabled, and

  200.      with <code class="literal">AUTO</code> it is initially disabled, but turned on

  201.      on-the-fly once the index size reaches <a class="xref" href="runtime-config-query.html#GUC-EFFECTIVE-CACHE-SIZE">effective_cache_size</a>. The default is <code class="literal">AUTO</code>.

  202.     </p></dd></dl></div><p>

  203.     GIN indexes accept different parameters:

  204.    </p><div class="variablelist"><dl class="variablelist"><dt id="INDEX-RELOPTION-FASTUPDATE"><span class="term"><code class="literal">fastupdate</code>

  205.      <a id="id-1.9.3.69.6.3.9.1.1.2" class="indexterm"></a>

  206.     </span></dt><dd><p>

  207.      This setting controls usage of the fast update technique described in

  208.      <a class="xref" href="gin-implementation.html#GIN-FAST-UPDATE" title="66.4.1. GIN Fast Update Technique">Section 66.4.1</a>.  It is a Boolean parameter:

  209.      <code class="literal">ON</code> enables fast update, <code class="literal">OFF</code> disables it.

  210.      (Alternative spellings of <code class="literal">ON</code> and <code class="literal">OFF</code> are

  211.      allowed as described in <a class="xref" href="config-setting.html" title="19.1. Setting Parameters">Section 19.1</a>.)  The

  212.      default is <code class="literal">ON</code>.

  213.     </p><div class="note"><h3 class="title">Note</h3><p>

  214.       Turning <code class="literal">fastupdate</code> off via <code class="command">ALTER INDEX</code> prevents

  215.       future insertions from going into the list of pending index entries,

  216.       but does not in itself flush previous entries.  You might want to

  217.       <code class="command">VACUUM</code> the table or call <code class="function">gin_clean_pending_list</code>

  218.       function afterward to ensure the pending list is emptied.

  219.      </p></div></dd></dl></div><div class="variablelist"><dl class="variablelist"><dt id="INDEX-RELOPTION-GIN-PENDING-LIST-LIMIT"><span class="term"><code class="literal">gin_pending_list_limit</code>

  220.      <a id="id-1.9.3.69.6.3.10.1.1.2" class="indexterm"></a>

  221.     </span></dt><dd><p>

  222.      Custom <a class="xref" href="runtime-config-client.html#GUC-GIN-PENDING-LIST-LIMIT">gin_pending_list_limit</a> parameter.

  223.      This value is specified in kilobytes.

  224.     </p></dd></dl></div><p>

  225.     <acronym class="acronym">BRIN</acronym> indexes accept different parameters:

  226.    </p><div class="variablelist"><dl class="variablelist"><dt id="INDEX-RELOPTION-PAGES-PER-RANGE"><span class="term"><code class="literal">pages_per_range</code>

  227.      <a id="id-1.9.3.69.6.3.12.1.1.2" class="indexterm"></a>

  228.     </span></dt><dd><p>

  229.      Defines the number of table blocks that make up one block range for

  230.      each entry of a <acronym class="acronym">BRIN</acronym> index (see <a class="xref" href="brin-intro.html" title="67.1. Introduction">Section 67.1</a>

  231.      for more details).  The default is <code class="literal">128</code>.

  232.     </p></dd><dt id="INDEX-RELOPTION-AUTOSUMMARIZE"><span class="term"><code class="literal">autosummarize</code>

  233.      <a id="id-1.9.3.69.6.3.12.2.1.2" class="indexterm"></a>

  234.     </span></dt><dd><p>

  235.      Defines whether a summarization run is invoked for the previous page

  236.      range whenever an insertion is detected on the next one.

  237.     </p></dd></dl></div></div><div class="refsect2" id="SQL-CREATEINDEX-CONCURRENTLY"><h3>Building Indexes Concurrently</h3><a id="id-1.9.3.69.6.4.2" class="indexterm"></a><p>

  238.     Creating an index can interfere with regular operation of a database.

  239.     Normally <span class="productname">PostgreSQL</span> locks the table to be indexed against

  240.     writes and performs the entire index build with a single scan of the

  241.     table. Other transactions can still read the table, but if they try to

  242.     insert, update, or delete rows in the table they will block until the

  243.     index build is finished. This could have a severe effect if the system is

  244.     a live production database.  Very large tables can take many hours to be

  245.     indexed, and even for smaller tables, an index build can lock out writers

  246.     for periods that are unacceptably long for a production system.

  247.    </p><p>

  248.     <span class="productname">PostgreSQL</span> supports building indexes without locking

  249.     out writes.  This method is invoked by specifying the

  250.     <code class="literal">CONCURRENTLY</code> option of <code class="command">CREATE INDEX</code>.

  251.     When this option is used,

  252.     <span class="productname">PostgreSQL</span> must perform two scans of the table, and in

  253.     addition it must wait for all existing transactions that could potentially

  254.     modify or use the index to terminate.  Thus

  255.     this method requires more total work than a standard index build and takes

  256.     significantly longer to complete.  However, since it allows normal

  257.     operations to continue while the index is built, this method is useful for

  258.     adding new indexes in a production environment.  Of course, the extra CPU

  259.     and I/O load imposed by the index creation might slow other operations.

  260.    </p><p>

  261.     In a concurrent index build, the index is actually entered into

  262.     the system catalogs in one transaction, then two table scans occur in

  263.     two more transactions.  Before each table scan, the index build must

  264.     wait for existing transactions that have modified the table to terminate.

  265.     After the second scan, the index build must wait for any transactions

  266.     that have a snapshot (see <a class="xref" href="mvcc.html" title="Chapter 13. Concurrency Control">Chapter 13</a>) predating the second

  267.     scan to terminate.  Then finally the index can be marked ready for use,

  268.     and the <code class="command">CREATE INDEX</code> command terminates.

  269.     Even then, however, the index may not be immediately usable for queries:

  270.     in the worst case, it cannot be used as long as transactions exist that

  271.     predate the start of the index build.

  272.    </p><p>

  273.     If a problem arises while scanning the table, such as a deadlock or a

  274.     uniqueness violation in a unique index, the <code class="command">CREATE INDEX</code>

  275.     command will fail but leave behind an <span class="quote">“<span class="quote">invalid</span>”</span> index. This index

  276.     will be ignored for querying purposes because it might be incomplete;

  277.     however it will still consume update overhead. The <span class="application">psql</span>

  278.     <code class="command">\d</code> command will report such an index as <code class="literal">INVALID</code>:

  279. 

  280. </p><pre class="programlisting">

  281. postgres=# \d tab

  282.        Table "public.tab"

  283.  Column |  Type   | Collation | Nullable | Default 

  284. --------+---------+-----------+----------+---------

  285.  col    | integer |           |          | 

  286. Indexes:

  287.     "idx" btree (col) INVALID

  288. </pre><p>

  289. 

  290.     The recommended recovery

  291.     method in such cases is to drop the index and try again to perform

  292.     <code class="command">CREATE INDEX CONCURRENTLY</code>.  (Another possibility is

  293.     to rebuild the index with <code class="command">REINDEX INDEX CONCURRENTLY</code>).

  294.    </p><p>

  295.     Another caveat when building a unique index concurrently is that the

  296.     uniqueness constraint is already being enforced against other transactions

  297.     when the second table scan begins.  This means that constraint violations

  298.     could be reported in other queries prior to the index becoming available

  299.     for use, or even in cases where the index build eventually fails.  Also,

  300.     if a failure does occur in the second scan, the <span class="quote">“<span class="quote">invalid</span>”</span> index

  301.     continues to enforce its uniqueness constraint afterwards.

  302.    </p><p>

  303.     Concurrent builds of expression indexes and partial indexes are supported.

  304.     Errors occurring in the evaluation of these expressions could cause

  305.     behavior similar to that described above for unique constraint violations.

  306.    </p><p>

  307.     Regular index builds permit other regular index builds on the

  308.     same table to occur simultaneously, but only one concurrent index build

  309.     can occur on a table at a time.  In either case, schema modification of the

  310.     table is not allowed while the index is being built.  Another difference is

  311.     that a regular <code class="command">CREATE INDEX</code> command can be performed

  312.     within a transaction block, but <code class="command">CREATE INDEX CONCURRENTLY</code>

  313.     cannot.

  314.    </p><p>

  315.     Concurrent builds for indexes on partitioned tables are currently not

  316.     supported.  However, you may concurrently build the index on each

  317.     partition individually and then finally create the partitioned index

  318.     non-concurrently in order to reduce the time where writes to the

  319.     partitioned table will be locked out.  In this case, building the

  320.     partitioned index is a metadata only operation.

  321.    </p></div></div><div class="refsect1" id="id-1.9.3.69.7"><h2>Notes</h2><p>

  322.    See <a class="xref" href="indexes.html" title="Chapter 11. Indexes">Chapter 11</a> for information about when indexes can

  323.    be used, when they are not used, and in which particular situations

  324.    they can be useful.

  325.   </p><p>

  326.    Currently, only the B-tree, GiST, GIN, and BRIN index methods support

  327.    multicolumn indexes. Up to 32 fields can be specified by default.

  328.    (This limit can be altered when building

  329.    <span class="productname">PostgreSQL</span>.)  Only B-tree currently

  330.    supports unique indexes.

  331.   </p><p>

  332.    An <em class="firstterm">operator class</em> can be specified for each

  333.    column of an index. The operator class identifies the operators to be

  334.    used by the index for that column. For example, a B-tree index on

  335.    four-byte integers would use the <code class="literal">int4_ops</code> class;

  336.    this operator class includes comparison functions for four-byte

  337.    integers. In practice the default operator class for the column's data

  338.    type is usually sufficient. The main point of having operator classes

  339.    is that for some data types, there could be more than one meaningful

  340.    ordering. For example, we might want to sort a complex-number data

  341.    type either by absolute value or by real part. We could do this by

  342.    defining two operator classes for the data type and then selecting

  343.    the proper class when creating an index.  More information about

  344.    operator classes is in <a class="xref" href="indexes-opclass.html" title="11.10. Operator Classes and Operator Families">Section 11.10</a> and in <a class="xref" href="xindex.html" title="37.16. Interfacing Extensions to Indexes">Section 37.16</a>.

  345.   </p><p>

  346.    When <code class="literal">CREATE INDEX</code> is invoked on a partitioned

  347.    table, the default behavior is to recurse to all partitions to ensure

  348.    they all have matching indexes.

  349.    Each partition is first checked to determine whether an equivalent

  350.    index already exists, and if so, that index will become attached as a

  351.    partition index to the index being created, which will become its

  352.    parent index.

  353.    If no matching index exists, a new index will be created and

  354.    automatically attached; the name of the new index in each partition

  355.    will be determined as if no index name had been specified in the

  356.    command.

  357.    If the <code class="literal">ONLY</code> option is specified, no recursion

  358.    is done, and the index is marked invalid.

  359.    (<code class="command">ALTER INDEX ... ATTACH PARTITION</code> marks the index

  360.    valid, once all partitions acquire matching indexes.)  Note, however,

  361.    that any partition that is created in the future using

  362.    <code class="command">CREATE TABLE ... PARTITION OF</code> will automatically

  363.    have a matching index, regardless of whether <code class="literal">ONLY</code> is

  364.    specified.

  365.   </p><p>

  366.    For index methods that support ordered scans (currently, only B-tree),

  367.    the optional clauses <code class="literal">ASC</code>, <code class="literal">DESC</code>, <code class="literal">NULLS

  368.    FIRST</code>, and/or <code class="literal">NULLS LAST</code> can be specified to modify

  369.    the sort ordering of the index.  Since an ordered index can be

  370.    scanned either forward or backward, it is not normally useful to create a

  371.    single-column <code class="literal">DESC</code> index — that sort ordering is already

  372.    available with a regular index.  The value of these options is that

  373.    multicolumn indexes can be created that match the sort ordering requested

  374.    by a mixed-ordering query, such as <code class="literal">SELECT ... ORDER BY x ASC, y

  375.    DESC</code>.  The <code class="literal">NULLS</code> options are useful if you need to support

  376.    <span class="quote">“<span class="quote">nulls sort low</span>”</span> behavior, rather than the default <span class="quote">“<span class="quote">nulls

  377.    sort high</span>”</span>, in queries that depend on indexes to avoid sorting steps.

  378.   </p><p>

  379.    For most index methods, the speed of creating an index is

  380.    dependent on the setting of <a class="xref" href="runtime-config-resource.html#GUC-MAINTENANCE-WORK-MEM">maintenance_work_mem</a>.

  381.    Larger values will reduce the time needed for index creation, so long

  382.    as you don't make it larger than the amount of memory really available,

  383.    which would drive the machine into swapping.

  384.   </p><p>

  385.    <span class="productname">PostgreSQL</span> can build indexes while

  386.    leveraging multiple CPUs in order to process the table rows faster.

  387.    This feature is known as <em class="firstterm">parallel index

  388.    build</em>.  For index methods that support building indexes

  389.    in parallel (currently, only B-tree),

  390.    <code class="varname">maintenance_work_mem</code> specifies the maximum

  391.    amount of memory that can be used by each index build operation as

  392.    a whole, regardless of how many worker processes were started.

  393.    Generally, a cost model automatically determines how many worker

  394.    processes should be requested, if any.

  395.   </p><p>

  396.    Parallel index builds may benefit from increasing

  397.    <code class="varname">maintenance_work_mem</code> where an equivalent serial

  398.    index build will see little or no benefit.  Note that

  399.    <code class="varname">maintenance_work_mem</code> may influence the number of

  400.    worker processes requested, since parallel workers must have at

  401.    least a <code class="literal">32MB</code> share of the total

  402.    <code class="varname">maintenance_work_mem</code> budget.  There must also be

  403.    a remaining <code class="literal">32MB</code> share for the leader process.

  404.    Increasing <a class="xref" href="runtime-config-resource.html#GUC-MAX-PARALLEL-WORKERS-MAINTENANCE">max_parallel_maintenance_workers</a>

  405.    may allow more workers to be used, which will reduce the time

  406.    needed for index creation, so long as the index build is not

  407.    already I/O bound.  Of course, there should also be sufficient

  408.    CPU capacity that would otherwise lie idle.

  409.   </p><p>

  410.    Setting a value for <code class="literal">parallel_workers</code> via <a class="xref" href="sql-altertable.html" title="ALTER TABLE"><span class="refentrytitle">ALTER TABLE</span></a> directly controls how many parallel

  411.    worker processes will be requested by a <code class="command">CREATE

  412.    INDEX</code> against the table.  This bypasses the cost model

  413.    completely, and prevents <code class="varname">maintenance_work_mem</code>

  414.    from affecting how many parallel workers are requested.  Setting

  415.    <code class="literal">parallel_workers</code> to 0 via <code class="command">ALTER

  416.    TABLE</code> will disable parallel index builds on the table in

  417.    all cases.

  418.   </p><div class="tip"><h3 class="title">Tip</h3><p>

  419.     You might want to reset <code class="literal">parallel_workers</code> after

  420.     setting it as part of tuning an index build.  This avoids

  421.     inadvertent changes to query plans, since

  422.     <code class="literal">parallel_workers</code> affects

  423.     <span class="emphasis"><em>all</em></span> parallel table scans.

  424.    </p></div><p>

  425.    While <code class="command">CREATE INDEX</code> with the

  426.    <code class="literal">CONCURRENTLY</code> option supports parallel builds

  427.    without special restrictions, only the first table scan is actually

  428.    performed in parallel.

  429.   </p><p>

  430.    Use <a class="xref" href="sql-dropindex.html" title="DROP INDEX"><span class="refentrytitle">DROP INDEX</span></a>

  431.    to remove an index.

  432.   </p><p>

  433.    Prior releases of <span class="productname">PostgreSQL</span> also had an

  434.    R-tree index method.  This method has been removed because

  435.    it had no significant advantages over the GiST method.

  436.    If <code class="literal">USING rtree</code> is specified, <code class="command">CREATE INDEX</code>

  437.    will interpret it as <code class="literal">USING gist</code>, to simplify conversion

  438.    of old databases to GiST.

  439.   </p></div><div class="refsect1" id="id-1.9.3.69.8"><h2>Examples</h2><p>

  440.    To create a unique B-tree index on the column <code class="literal">title</code> in

  441.    the table <code class="literal">films</code>:

  442. </p><pre class="programlisting">

  443. CREATE UNIQUE INDEX title_idx ON films (title);

  444. </pre><p>

  445.   </p><p>

  446.    To create a unique B-tree index on the column <code class="literal">title</code>

  447.    with included columns <code class="literal">director</code>

  448.    and <code class="literal">rating</code> in the table <code class="literal">films</code>:

  449. </p><pre class="programlisting">

  450. CREATE UNIQUE INDEX title_idx ON films (title) INCLUDE (director, rating);

  451. </pre><p>

  452.   </p><p>

  453.    To create an index on the expression <code class="literal">lower(title)</code>,

  454.    allowing efficient case-insensitive searches:

  455. </p><pre class="programlisting">

  456. CREATE INDEX ON films ((lower(title)));

  457. </pre><p>

  458.    (In this example we have chosen to omit the index name, so the system

  459.    will choose a name, typically <code class="literal">films_lower_idx</code>.)

  460.   </p><p>

  461.    To create an index with non-default collation:

  462. </p><pre class="programlisting">

  463. CREATE INDEX title_idx_german ON films (title COLLATE "de_DE");

  464. </pre><p>

  465.   </p><p>

  466.    To create an index with non-default sort ordering of nulls:

  467. </p><pre class="programlisting">

  468. CREATE INDEX title_idx_nulls_low ON films (title NULLS FIRST);

  469. </pre><p>

  470.   </p><p>

  471.    To create an index with non-default fill factor:

  472. </p><pre class="programlisting">

  473. CREATE UNIQUE INDEX title_idx ON films (title) WITH (fillfactor = 70);

  474. </pre><p>

  475.   </p><p>

  476.    To create a <acronym class="acronym">GIN</acronym> index with fast updates disabled:

  477. </p><pre class="programlisting">

  478. CREATE INDEX gin_idx ON documents_table USING GIN (locations) WITH (fastupdate = off);

  479. </pre><p>

  480.   </p><p>

  481.    To create an index on the column <code class="literal">code</code> in the table

  482.    <code class="literal">films</code> and have the index reside in the tablespace

  483.    <code class="literal">indexspace</code>:

  484. </p><pre class="programlisting">

  485. CREATE INDEX code_idx ON films (code) TABLESPACE indexspace;

  486. </pre><p>

  487.   </p><p>

  488.    To create a GiST index on a point attribute so that we

  489.    can efficiently use box operators on the result of the

  490.    conversion function:

  491. </p><pre class="programlisting">

  492. CREATE INDEX pointloc

  493.     ON points USING gist (box(location,location));

  494. SELECT * FROM points

  495.     WHERE box(location,location) &amp;&amp; '(0,0),(1,1)'::box;

  496. </pre><p>

  497.   </p><p>

  498.    To create an index without locking out writes to the table:

  499. </p><pre class="programlisting">

  500. CREATE INDEX CONCURRENTLY sales_quantity_index ON sales_table (quantity);

  501. </pre></div><div class="refsect1" id="id-1.9.3.69.9"><h2>Compatibility</h2><p>

  502.    <code class="command">CREATE INDEX</code> is a

  503.    <span class="productname">PostgreSQL</span> language extension.  There

  504.    are no provisions for indexes in the SQL standard.

  505.   </p></div><div class="refsect1" id="id-1.9.3.69.10"><h2>See Also</h2><span class="simplelist"><a class="xref" href="sql-alterindex.html" title="ALTER INDEX"><span class="refentrytitle">ALTER INDEX</span></a>, <a class="xref" href="sql-dropindex.html" title="DROP INDEX"><span class="refentrytitle">DROP INDEX</span></a>, <a class="xref" href="sql-reindex.html" title="REINDEX"><span class="refentrytitle">REINDEX</span></a></span></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="sql-creategroup.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="sql-commands.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="sql-createlanguage.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">CREATE GROUP </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> CREATE LANGUAGE</td></tr></table></div></body></html>
上海开阖软件有限公司 沪ICP备12045867号-1