Baoding-FenShuaJiang
/
Green_GoodERP18_FSJ_Standard
Observar 1
Favorito 0
Wiki 捐赠
gooderp18绿色标准版
Você não pode selecionar mais de 25 tópicos Os tópicos devem começar com uma letra ou um número, podem incluir traços ('-') e podem ter até 35 caracteres.
Branch: master
Branches Tags
${ item.name }
Criar branch ${ searchTerm }
de master
${ noResults }
Green_GoodERP18_FSJ_Standard/PostgreSQL/doc/postgresql/html/sql-expressions.html

995 linhas
64KB
Original Link permanente Anotar Histórico 

  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>4.2. Value Expressions</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-syntax-lexical.html" title="4.1. Lexical Structure" /><link rel="next" href="sql-syntax-calling-funcs.html" title="4.3. Calling Functions" /></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">4.2. Value Expressions</th></tr><tr><td width="10%" align="left"><a accesskey="p" href="sql-syntax-lexical.html" title="4.1. Lexical Structure">Prev</a> </td><td width="10%" align="left"><a accesskey="u" href="sql-syntax.html" title="Chapter 4. SQL Syntax">Up</a></td><th width="60%" align="center">Chapter 4. SQL Syntax</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-syntax-calling-funcs.html" title="4.3. Calling Functions">Next</a></td></tr></table><hr></hr></div><div class="sect1" id="SQL-EXPRESSIONS"><div class="titlepage"><div><div><h2 class="title" style="clear: both">4.2. Value Expressions</h2></div></div></div><div class="toc"><dl class="toc"><dt><span class="sect2"><a href="sql-expressions.html#SQL-EXPRESSIONS-COLUMN-REFS">4.2.1. Column References</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SQL-EXPRESSIONS-PARAMETERS-POSITIONAL">4.2.2. Positional Parameters</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SQL-EXPRESSIONS-SUBSCRIPTS">4.2.3. Subscripts</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#FIELD-SELECTION">4.2.4. Field Selection</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SQL-EXPRESSIONS-OPERATOR-CALLS">4.2.5. Operator Invocations</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SQL-EXPRESSIONS-FUNCTION-CALLS">4.2.6. Function Calls</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SYNTAX-AGGREGATES">4.2.7. Aggregate Expressions</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SYNTAX-WINDOW-FUNCTIONS">4.2.8. Window Function Calls</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SQL-SYNTAX-TYPE-CASTS">4.2.9. Type Casts</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SQL-SYNTAX-COLLATE-EXPRS">4.2.10. Collation Expressions</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SQL-SYNTAX-SCALAR-SUBQUERIES">4.2.11. Scalar Subqueries</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SQL-SYNTAX-ARRAY-CONSTRUCTORS">4.2.12. Array Constructors</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SQL-SYNTAX-ROW-CONSTRUCTORS">4.2.13. Row Constructors</a></span></dt><dt><span class="sect2"><a href="sql-expressions.html#SYNTAX-EXPRESS-EVAL">4.2.14. Expression Evaluation Rules</a></span></dt></dl></div><a id="id-1.5.3.6.2" class="indexterm"></a><a id="id-1.5.3.6.3" class="indexterm"></a><a id="id-1.5.3.6.4" class="indexterm"></a><p>

  3.    Value expressions are used in a variety of contexts, such

  4.    as in the target list of the <code class="command">SELECT</code> command, as

  5.    new column values in <code class="command">INSERT</code> or

  6.    <code class="command">UPDATE</code>, or in search conditions in a number of

  7.    commands.  The result of a value expression is sometimes called a

  8.    <em class="firstterm">scalar</em>, to distinguish it from the result of

  9.    a table expression (which is a table).  Value expressions are

  10.    therefore also called <em class="firstterm">scalar expressions</em> (or

  11.    even simply <em class="firstterm">expressions</em>).  The expression

  12.    syntax allows the calculation of values from primitive parts using

  13.    arithmetic, logical, set, and other operations.

  14.   </p><p>

  15.    A value expression is one of the following:

  16. 

  17.    </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>

  18.       A constant or literal value

  19.      </p></li><li class="listitem"><p>

  20.       A column reference

  21.      </p></li><li class="listitem"><p>

  22.       A positional parameter reference, in the body of a function definition

  23.       or prepared statement

  24.      </p></li><li class="listitem"><p>

  25.       A subscripted expression

  26.      </p></li><li class="listitem"><p>

  27.       A field selection expression

  28.      </p></li><li class="listitem"><p>

  29.       An operator invocation

  30.      </p></li><li class="listitem"><p>

  31.       A function call

  32.      </p></li><li class="listitem"><p>

  33.       An aggregate expression

  34.      </p></li><li class="listitem"><p>

  35.       A window function call

  36.      </p></li><li class="listitem"><p>

  37.       A type cast

  38.      </p></li><li class="listitem"><p>

  39.       A collation expression

  40.      </p></li><li class="listitem"><p>

  41.       A scalar subquery

  42.      </p></li><li class="listitem"><p>

  43.       An array constructor

  44.      </p></li><li class="listitem"><p>

  45.       A row constructor

  46.      </p></li><li class="listitem"><p>

  47.       Another value expression in parentheses (used to group

  48.       subexpressions and override

  49.       precedence<a id="id-1.5.3.6.6.1.15.1.1" class="indexterm"></a>)

  50.      </p></li></ul></div><p>

  51.   </p><p>

  52.    In addition to this list, there are a number of constructs that can

  53.    be classified as an expression but do not follow any general syntax

  54.    rules.  These generally have the semantics of a function or

  55.    operator and are explained in the appropriate location in <a class="xref" href="functions.html" title="Chapter 9. Functions and Operators">Chapter 9</a>.  An example is the <code class="literal">IS NULL</code>

  56.    clause.

  57.   </p><p>

  58.    We have already discussed constants in <a class="xref" href="sql-syntax-lexical.html#SQL-SYNTAX-CONSTANTS" title="4.1.2. Constants">Section 4.1.2</a>.  The following sections discuss

  59.    the remaining options.

  60.   </p><div class="sect2" id="SQL-EXPRESSIONS-COLUMN-REFS"><div class="titlepage"><div><div><h3 class="title">4.2.1. Column References</h3></div></div></div><a id="id-1.5.3.6.9.2" class="indexterm"></a><p>

  61.     A column can be referenced in the form:

  62. </p><pre class="synopsis">

  63. <em class="replaceable"><code>correlation</code></em>.<em class="replaceable"><code>columnname</code></em>

  64. </pre><p>

  65.    </p><p>

  66.     <em class="replaceable"><code>correlation</code></em> is the name of a

  67.     table (possibly qualified with a schema name), or an alias for a table

  68.     defined by means of a <code class="literal">FROM</code> clause.

  69.     The correlation name and separating dot can be omitted if the column name

  70.     is unique across all the tables being used in the current query.  (See also <a class="xref" href="queries.html" title="Chapter 7. Queries">Chapter 7</a>.)

  71.    </p></div><div class="sect2" id="SQL-EXPRESSIONS-PARAMETERS-POSITIONAL"><div class="titlepage"><div><div><h3 class="title">4.2.2. Positional Parameters</h3></div></div></div><a id="id-1.5.3.6.10.2" class="indexterm"></a><a id="id-1.5.3.6.10.3" class="indexterm"></a><p>

  72.     A positional parameter reference is used to indicate a value

  73.     that is supplied externally to an SQL statement.  Parameters are

  74.     used in SQL function definitions and in prepared queries.  Some

  75.     client libraries also support specifying data values separately

  76.     from the SQL command string, in which case parameters are used to

  77.     refer to the out-of-line data values.

  78.     The form of a parameter reference is:

  79. </p><pre class="synopsis">

  80. $<em class="replaceable"><code>number</code></em>

  81. </pre><p>

  82.    </p><p>

  83.     For example, consider the definition of a function,

  84.     <code class="function">dept</code>, as:

  85. 

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

  87. CREATE FUNCTION dept(text) RETURNS dept

  88.     AS $$ SELECT * FROM dept WHERE name = $1 $$

  89.     LANGUAGE SQL;

  90. </pre><p>

  91. 

  92.     Here the <code class="literal">$1</code> references the value of the first

  93.     function argument whenever the function is invoked.

  94.    </p></div><div class="sect2" id="SQL-EXPRESSIONS-SUBSCRIPTS"><div class="titlepage"><div><div><h3 class="title">4.2.3. Subscripts</h3></div></div></div><a id="id-1.5.3.6.11.2" class="indexterm"></a><p>

  95.     If an expression yields a value of an array type, then a specific

  96.     element of the array value can be extracted by writing

  97. </p><pre class="synopsis">

  98. <em class="replaceable"><code>expression</code></em>[<em class="replaceable"><code>subscript</code></em>]

  99. </pre><p>

  100.     or multiple adjacent elements (an <span class="quote">“<span class="quote">array slice</span>”</span>) can be extracted

  101.     by writing

  102. </p><pre class="synopsis">

  103. <em class="replaceable"><code>expression</code></em>[<em class="replaceable"><code>lower_subscript</code></em>:<em class="replaceable"><code>upper_subscript</code></em>]

  104. </pre><p>

  105.     (Here, the brackets <code class="literal">[ ]</code> are meant to appear literally.)

  106.     Each <em class="replaceable"><code>subscript</code></em> is itself an expression,

  107.     which must yield an integer value.

  108.    </p><p>

  109.     In general the array <em class="replaceable"><code>expression</code></em> must be

  110.     parenthesized, but the parentheses can be omitted when the expression

  111.     to be subscripted is just a column reference or positional parameter.

  112.     Also, multiple subscripts can be concatenated when the original array

  113.     is multidimensional.

  114.     For example:

  115. 

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

  117. mytable.arraycolumn[4]

  118. mytable.two_d_column[17][34]

  119. $1[10:42]

  120. (arrayfunction(a,b))[42]

  121. </pre><p>

  122. 

  123.     The parentheses in the last example are required.

  124.     See <a class="xref" href="arrays.html" title="8.15. Arrays">Section 8.15</a> for more about arrays.

  125.    </p></div><div class="sect2" id="FIELD-SELECTION"><div class="titlepage"><div><div><h3 class="title">4.2.4. Field Selection</h3></div></div></div><a id="id-1.5.3.6.12.2" class="indexterm"></a><p>

  126.     If an expression yields a value of a composite type (row type), then a

  127.     specific field of the row can be extracted by writing

  128. </p><pre class="synopsis">

  129. <em class="replaceable"><code>expression</code></em>.<em class="replaceable"><code>fieldname</code></em>

  130. </pre><p>

  131.    </p><p>

  132.     In general the row <em class="replaceable"><code>expression</code></em> must be

  133.     parenthesized, but the parentheses can be omitted when the expression

  134.     to be selected from is just a table reference or positional parameter.

  135.     For example:

  136. 

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

  138. mytable.mycolumn

  139. $1.somecolumn

  140. (rowfunction(a,b)).col3

  141. </pre><p>

  142. 

  143.     (Thus, a qualified column reference is actually just a special case

  144.     of the field selection syntax.)  An important special case is

  145.     extracting a field from a table column that is of a composite type:

  146. 

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

  148. (compositecol).somefield

  149. (mytable.compositecol).somefield

  150. </pre><p>

  151. 

  152.     The parentheses are required here to show that

  153.     <code class="structfield">compositecol</code> is a column name not a table name,

  154.     or that <code class="structname">mytable</code> is a table name not a schema name

  155.     in the second case.

  156.    </p><p>

  157.     You can ask for all fields of a composite value by

  158.     writing <code class="literal">.*</code>:

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

  160. (compositecol).*

  161. </pre><p>

  162.     This notation behaves differently depending on context;

  163.     see <a class="xref" href="rowtypes.html#ROWTYPES-USAGE" title="8.16.5. Using Composite Types in Queries">Section 8.16.5</a> for details.

  164.    </p></div><div class="sect2" id="SQL-EXPRESSIONS-OPERATOR-CALLS"><div class="titlepage"><div><div><h3 class="title">4.2.5. Operator Invocations</h3></div></div></div><a id="id-1.5.3.6.13.2" class="indexterm"></a><p>

  165.     There are three possible syntaxes for an operator invocation:

  166.     </p><table border="0" summary="Simple list" class="simplelist"><tr><td><em class="replaceable"><code>expression</code></em> <em class="replaceable"><code>operator</code></em> <em class="replaceable"><code>expression</code></em> (binary infix operator)</td></tr><tr><td><em class="replaceable"><code>operator</code></em> <em class="replaceable"><code>expression</code></em> (unary prefix operator)</td></tr><tr><td><em class="replaceable"><code>expression</code></em> <em class="replaceable"><code>operator</code></em> (unary postfix operator)</td></tr></table><p>

  167.     where the <em class="replaceable"><code>operator</code></em> token follows the syntax

  168.     rules of <a class="xref" href="sql-syntax-lexical.html#SQL-SYNTAX-OPERATORS" title="4.1.3. Operators">Section 4.1.3</a>, or is one of the

  169.     key words <code class="token">AND</code>, <code class="token">OR</code>, and

  170.     <code class="token">NOT</code>, or is a qualified operator name in the form:

  171. </p><pre class="synopsis">

  172. <code class="literal">OPERATOR(</code><em class="replaceable"><code>schema</code></em><code class="literal">.</code><em class="replaceable"><code>operatorname</code></em><code class="literal">)</code>

  173. </pre><p>

  174.     Which particular operators exist and whether

  175.     they are unary or binary depends on what operators have been

  176.     defined by the system or the user.  <a class="xref" href="functions.html" title="Chapter 9. Functions and Operators">Chapter 9</a>

  177.     describes the built-in operators.

  178.    </p></div><div class="sect2" id="SQL-EXPRESSIONS-FUNCTION-CALLS"><div class="titlepage"><div><div><h3 class="title">4.2.6. Function Calls</h3></div></div></div><a id="id-1.5.3.6.14.2" class="indexterm"></a><p>

  179.     The syntax for a function call is the name of a function

  180.     (possibly qualified with a schema name), followed by its argument list

  181.     enclosed in parentheses:

  182. 

  183. </p><pre class="synopsis">

  184. <em class="replaceable"><code>function_name</code></em> ([<span class="optional"><em class="replaceable"><code>expression</code></em> [<span class="optional">, <em class="replaceable"><code>expression</code></em> ... </span>]</span>] )

  185. </pre><p>

  186.    </p><p>

  187.     For example, the following computes the square root of 2:

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

  189. sqrt(2)

  190. </pre><p>

  191.    </p><p>

  192.     The list of built-in functions is in <a class="xref" href="functions.html" title="Chapter 9. Functions and Operators">Chapter 9</a>.

  193.     Other functions can be added by the user.

  194.    </p><p>

  195.     When issuing queries in a database where some users mistrust other users,

  196.     observe security precautions from <a class="xref" href="typeconv-func.html" title="10.3. Functions">Section 10.3</a> when

  197.     writing function calls.

  198.    </p><p>

  199.     The arguments can optionally have names attached.

  200.     See <a class="xref" href="sql-syntax-calling-funcs.html" title="4.3. Calling Functions">Section 4.3</a> for details.

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

  202.      A function that takes a single argument of composite type can

  203.      optionally be called using field-selection syntax, and conversely

  204.      field selection can be written in functional style.  That is, the

  205.      notations <code class="literal">col(table)</code> and <code class="literal">table.col</code> are

  206.      interchangeable.  This behavior is not SQL-standard but is provided

  207.      in <span class="productname">PostgreSQL</span> because it allows use of functions to

  208.      emulate <span class="quote">“<span class="quote">computed fields</span>”</span>.  For more information see

  209.      <a class="xref" href="rowtypes.html#ROWTYPES-USAGE" title="8.16.5. Using Composite Types in Queries">Section 8.16.5</a>.

  210.     </p></div></div><div class="sect2" id="SYNTAX-AGGREGATES"><div class="titlepage"><div><div><h3 class="title">4.2.7. Aggregate Expressions</h3></div></div></div><a id="id-1.5.3.6.15.2" class="indexterm"></a><a id="id-1.5.3.6.15.3" class="indexterm"></a><a id="id-1.5.3.6.15.4" class="indexterm"></a><a id="id-1.5.3.6.15.5" class="indexterm"></a><p>

  211.     An <em class="firstterm">aggregate expression</em> represents the

  212.     application of an aggregate function across the rows selected by a

  213.     query.  An aggregate function reduces multiple inputs to a single

  214.     output value, such as the sum or average of the inputs.  The

  215.     syntax of an aggregate expression is one of the following:

  216. 

  217. </p><pre class="synopsis">

  218. <em class="replaceable"><code>aggregate_name</code></em> (<em class="replaceable"><code>expression</code></em> [ , ... ] [ <em class="replaceable"><code>order_by_clause</code></em> ] ) [ FILTER ( WHERE <em class="replaceable"><code>filter_clause</code></em> ) ]

  219. <em class="replaceable"><code>aggregate_name</code></em> (ALL <em class="replaceable"><code>expression</code></em> [ , ... ] [ <em class="replaceable"><code>order_by_clause</code></em> ] ) [ FILTER ( WHERE <em class="replaceable"><code>filter_clause</code></em> ) ]

  220. <em class="replaceable"><code>aggregate_name</code></em> (DISTINCT <em class="replaceable"><code>expression</code></em> [ , ... ] [ <em class="replaceable"><code>order_by_clause</code></em> ] ) [ FILTER ( WHERE <em class="replaceable"><code>filter_clause</code></em> ) ]

  221. <em class="replaceable"><code>aggregate_name</code></em> ( * ) [ FILTER ( WHERE <em class="replaceable"><code>filter_clause</code></em> ) ]

  222. <em class="replaceable"><code>aggregate_name</code></em> ( [ <em class="replaceable"><code>expression</code></em> [ , ... ] ] ) WITHIN GROUP ( <em class="replaceable"><code>order_by_clause</code></em> ) [ FILTER ( WHERE <em class="replaceable"><code>filter_clause</code></em> ) ]

  223. </pre><p>

  224. 

  225.     where <em class="replaceable"><code>aggregate_name</code></em> is a previously

  226.     defined aggregate (possibly qualified with a schema name) and

  227.     <em class="replaceable"><code>expression</code></em> is

  228.     any value expression that does not itself contain an aggregate

  229.     expression or a window function call.  The optional

  230.     <em class="replaceable"><code>order_by_clause</code></em> and

  231.     <em class="replaceable"><code>filter_clause</code></em> are described below.

  232.    </p><p>

  233.     The first form of aggregate expression invokes the aggregate

  234.     once for each input row.

  235.     The second form is the same as the first, since

  236.     <code class="literal">ALL</code> is the default.

  237.     The third form invokes the aggregate once for each distinct value

  238.     of the expression (or distinct set of values, for multiple expressions)

  239.     found in the input rows.

  240.     The fourth form invokes the aggregate once for each input row; since no

  241.     particular input value is specified, it is generally only useful

  242.     for the <code class="function">count(*)</code> aggregate function.

  243.     The last form is used with <em class="firstterm">ordered-set</em> aggregate

  244.     functions, which are described below.

  245.    </p><p>

  246.     Most aggregate functions ignore null inputs, so that rows in which

  247.     one or more of the expression(s) yield null are discarded.  This

  248.     can be assumed to be true, unless otherwise specified, for all

  249.     built-in aggregates.

  250.    </p><p>

  251.     For example, <code class="literal">count(*)</code> yields the total number

  252.     of input rows; <code class="literal">count(f1)</code> yields the number of

  253.     input rows in which <code class="literal">f1</code> is non-null, since

  254.     <code class="function">count</code> ignores nulls; and

  255.     <code class="literal">count(distinct f1)</code> yields the number of

  256.     distinct non-null values of <code class="literal">f1</code>.

  257.    </p><p>

  258.     Ordinarily, the input rows are fed to the aggregate function in an

  259.     unspecified order.  In many cases this does not matter; for example,

  260.     <code class="function">min</code> produces the same result no matter what order it

  261.     receives the inputs in.  However, some aggregate functions

  262.     (such as <code class="function">array_agg</code> and <code class="function">string_agg</code>) produce

  263.     results that depend on the ordering of the input rows.  When using

  264.     such an aggregate, the optional <em class="replaceable"><code>order_by_clause</code></em> can be

  265.     used to specify the desired ordering.  The <em class="replaceable"><code>order_by_clause</code></em>

  266.     has the same syntax as for a query-level <code class="literal">ORDER BY</code> clause, as

  267.     described in <a class="xref" href="queries-order.html" title="7.5. Sorting Rows">Section 7.5</a>, except that its expressions

  268.     are always just expressions and cannot be output-column names or numbers.

  269.     For example:

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

  271. SELECT array_agg(a ORDER BY b DESC) FROM table;

  272. </pre><p>

  273.    </p><p>

  274.     When dealing with multiple-argument aggregate functions, note that the

  275.     <code class="literal">ORDER BY</code> clause goes after all the aggregate arguments.

  276.     For example, write this:

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

  278. SELECT string_agg(a, ',' ORDER BY a) FROM table;

  279. </pre><p>

  280.     not this:

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

  282. SELECT string_agg(a ORDER BY a, ',') FROM table;  -- incorrect

  283. </pre><p>

  284.     The latter is syntactically valid, but it represents a call of a

  285.     single-argument aggregate function with two <code class="literal">ORDER BY</code> keys

  286.     (the second one being rather useless since it's a constant).

  287.    </p><p>

  288.     If <code class="literal">DISTINCT</code> is specified in addition to an

  289.     <em class="replaceable"><code>order_by_clause</code></em>, then all the <code class="literal">ORDER BY</code>

  290.     expressions must match regular arguments of the aggregate; that is,

  291.     you cannot sort on an expression that is not included in the

  292.     <code class="literal">DISTINCT</code> list.

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

  294.      The ability to specify both <code class="literal">DISTINCT</code> and <code class="literal">ORDER BY</code>

  295.      in an aggregate function is a <span class="productname">PostgreSQL</span> extension.

  296.     </p></div><p>

  297.     Placing <code class="literal">ORDER BY</code> within the aggregate's regular argument

  298.     list, as described so far, is used when ordering the input rows for

  299.     general-purpose and statistical aggregates, for which ordering is

  300.     optional.  There is a

  301.     subclass of aggregate functions called <em class="firstterm">ordered-set

  302.     aggregates</em> for which an <em class="replaceable"><code>order_by_clause</code></em>

  303.     is <span class="emphasis"><em>required</em></span>, usually because the aggregate's computation is

  304.     only sensible in terms of a specific ordering of its input rows.

  305.     Typical examples of ordered-set aggregates include rank and percentile

  306.     calculations.  For an ordered-set aggregate,

  307.     the <em class="replaceable"><code>order_by_clause</code></em> is written

  308.     inside <code class="literal">WITHIN GROUP (...)</code>, as shown in the final syntax

  309.     alternative above.  The expressions in

  310.     the <em class="replaceable"><code>order_by_clause</code></em> are evaluated once per

  311.     input row just like regular aggregate arguments, sorted as per

  312.     the <em class="replaceable"><code>order_by_clause</code></em>'s requirements, and fed

  313.     to the aggregate function as input arguments.  (This is unlike the case

  314.     for a non-<code class="literal">WITHIN GROUP</code> <em class="replaceable"><code>order_by_clause</code></em>,

  315.     which is not treated as argument(s) to the aggregate function.)  The

  316.     argument expressions preceding <code class="literal">WITHIN GROUP</code>, if any, are

  317.     called <em class="firstterm">direct arguments</em> to distinguish them from

  318.     the <em class="firstterm">aggregated arguments</em> listed in

  319.     the <em class="replaceable"><code>order_by_clause</code></em>.  Unlike regular aggregate

  320.     arguments, direct arguments are evaluated only once per aggregate call,

  321.     not once per input row.  This means that they can contain variables only

  322.     if those variables are grouped by <code class="literal">GROUP BY</code>; this restriction

  323.     is the same as if the direct arguments were not inside an aggregate

  324.     expression at all.  Direct arguments are typically used for things like

  325.     percentile fractions, which only make sense as a single value per

  326.     aggregation calculation.  The direct argument list can be empty; in this

  327.     case, write just <code class="literal">()</code> not <code class="literal">(*)</code>.

  328.     (<span class="productname">PostgreSQL</span> will actually accept either spelling, but

  329.     only the first way conforms to the SQL standard.)

  330.    </p><p>

  331.     <a id="id-1.5.3.6.15.15.1" class="indexterm"></a>

  332.     An example of an ordered-set aggregate call is:

  333. 

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

  335. SELECT percentile_cont(0.5) WITHIN GROUP (ORDER BY income) FROM households;

  336.  percentile_cont

  337. -----------------

  338.            50489

  339. </pre><p>

  340. 

  341.    which obtains the 50th percentile, or median, value of

  342.    the <code class="structfield">income</code> column from table <code class="structname">households</code>.

  343.    Here, <code class="literal">0.5</code> is a direct argument; it would make no sense

  344.    for the percentile fraction to be a value varying across rows.

  345.    </p><p>

  346.     If <code class="literal">FILTER</code> is specified, then only the input

  347.     rows for which the <em class="replaceable"><code>filter_clause</code></em>

  348.     evaluates to true are fed to the aggregate function; other rows

  349.     are discarded.  For example:

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

  351. SELECT

  352.     count(*) AS unfiltered,

  353.     count(*) FILTER (WHERE i &lt; 5) AS filtered

  354. FROM generate_series(1,10) AS s(i);

  355.  unfiltered | filtered

  356. ------------+----------

  357.          10 |        4

  358. (1 row)

  359. </pre><p>

  360.    </p><p>

  361.     The predefined aggregate functions are described in <a class="xref" href="functions-aggregate.html" title="9.20. Aggregate Functions">Section 9.20</a>.  Other aggregate functions can be added

  362.     by the user.

  363.    </p><p>

  364.     An aggregate expression can only appear in the result list or

  365.     <code class="literal">HAVING</code> clause of a <code class="command">SELECT</code> command.

  366.     It is forbidden in other clauses, such as <code class="literal">WHERE</code>,

  367.     because those clauses are logically evaluated before the results

  368.     of aggregates are formed.

  369.    </p><p>

  370.     When an aggregate expression appears in a subquery (see

  371.     <a class="xref" href="sql-expressions.html#SQL-SYNTAX-SCALAR-SUBQUERIES" title="4.2.11. Scalar Subqueries">Section 4.2.11</a> and

  372.     <a class="xref" href="functions-subquery.html" title="9.22. Subquery Expressions">Section 9.22</a>), the aggregate is normally

  373.     evaluated over the rows of the subquery.  But an exception occurs

  374.     if the aggregate's arguments (and <em class="replaceable"><code>filter_clause</code></em>

  375.     if any) contain only outer-level variables:

  376.     the aggregate then belongs to the nearest such outer level, and is

  377.     evaluated over the rows of that query.  The aggregate expression

  378.     as a whole is then an outer reference for the subquery it appears in,

  379.     and acts as a constant over any one evaluation of that subquery.

  380.     The restriction about

  381.     appearing only in the result list or <code class="literal">HAVING</code> clause

  382.     applies with respect to the query level that the aggregate belongs to.

  383.    </p></div><div class="sect2" id="SYNTAX-WINDOW-FUNCTIONS"><div class="titlepage"><div><div><h3 class="title">4.2.8. Window Function Calls</h3></div></div></div><a id="id-1.5.3.6.16.2" class="indexterm"></a><a id="id-1.5.3.6.16.3" class="indexterm"></a><p>

  384.     A <em class="firstterm">window function call</em> represents the application

  385.     of an aggregate-like function over some portion of the rows selected

  386.     by a query.  Unlike non-window aggregate calls, this is not tied

  387.     to grouping of the selected rows into a single output row — each

  388.     row remains separate in the query output.  However the window function

  389.     has access to all the rows that would be part of the current row's

  390.     group according to the grouping specification (<code class="literal">PARTITION BY</code>

  391.     list) of the window function call.

  392.     The syntax of a window function call is one of the following:

  393. 

  394. </p><pre class="synopsis">

  395. <em class="replaceable"><code>function_name</code></em> ([<span class="optional"><em class="replaceable"><code>expression</code></em> [<span class="optional">, <em class="replaceable"><code>expression</code></em> ... </span>]</span>]) [ FILTER ( WHERE <em class="replaceable"><code>filter_clause</code></em> ) ] OVER <em class="replaceable"><code>window_name</code></em>

  396. <em class="replaceable"><code>function_name</code></em> ([<span class="optional"><em class="replaceable"><code>expression</code></em> [<span class="optional">, <em class="replaceable"><code>expression</code></em> ... </span>]</span>]) [ FILTER ( WHERE <em class="replaceable"><code>filter_clause</code></em> ) ] OVER ( <em class="replaceable"><code>window_definition</code></em> )

  397. <em class="replaceable"><code>function_name</code></em> ( * ) [ FILTER ( WHERE <em class="replaceable"><code>filter_clause</code></em> ) ] OVER <em class="replaceable"><code>window_name</code></em>

  398. <em class="replaceable"><code>function_name</code></em> ( * ) [ FILTER ( WHERE <em class="replaceable"><code>filter_clause</code></em> ) ] OVER ( <em class="replaceable"><code>window_definition</code></em> )

  399. </pre><p>

  400.     where <em class="replaceable"><code>window_definition</code></em>

  401.     has the syntax

  402. </p><pre class="synopsis">

  403. [ <em class="replaceable"><code>existing_window_name</code></em> ]

  404. [ PARTITION BY <em class="replaceable"><code>expression</code></em> [, ...] ]

  405. [ ORDER BY <em class="replaceable"><code>expression</code></em> [ ASC | DESC | USING <em class="replaceable"><code>operator</code></em> ] [ NULLS { FIRST | LAST } ] [, ...] ]

  406. [ <em class="replaceable"><code>frame_clause</code></em> ]

  407. </pre><p>

  408.     The optional <em class="replaceable"><code>frame_clause</code></em>

  409.     can be one of

  410. </p><pre class="synopsis">

  411. { RANGE | ROWS | GROUPS } <em class="replaceable"><code>frame_start</code></em> [ <em class="replaceable"><code>frame_exclusion</code></em> ]

  412. { RANGE | ROWS | GROUPS } BETWEEN <em class="replaceable"><code>frame_start</code></em> AND <em class="replaceable"><code>frame_end</code></em> [ <em class="replaceable"><code>frame_exclusion</code></em> ]

  413. </pre><p>

  414.     where <em class="replaceable"><code>frame_start</code></em>

  415.     and <em class="replaceable"><code>frame_end</code></em> can be one of

  416. </p><pre class="synopsis">

  417. UNBOUNDED PRECEDING

  418. <em class="replaceable"><code>offset</code></em> PRECEDING

  419. CURRENT ROW

  420. <em class="replaceable"><code>offset</code></em> FOLLOWING

  421. UNBOUNDED FOLLOWING

  422. </pre><p>

  423.     and <em class="replaceable"><code>frame_exclusion</code></em> can be one of

  424. </p><pre class="synopsis">

  425. EXCLUDE CURRENT ROW

  426. EXCLUDE GROUP

  427. EXCLUDE TIES

  428. EXCLUDE NO OTHERS

  429. </pre><p>

  430.    </p><p>

  431.     Here, <em class="replaceable"><code>expression</code></em> represents any value

  432.     expression that does not itself contain window function calls.

  433.    </p><p>

  434.     <em class="replaceable"><code>window_name</code></em> is a reference to a named window

  435.     specification defined in the query's <code class="literal">WINDOW</code> clause.

  436.     Alternatively, a full <em class="replaceable"><code>window_definition</code></em> can

  437.     be given within parentheses, using the same syntax as for defining a

  438.     named window in the <code class="literal">WINDOW</code> clause; see the

  439.     <a class="xref" href="sql-select.html" title="SELECT"><span class="refentrytitle">SELECT</span></a> reference page for details.  It's worth

  440.     pointing out that <code class="literal">OVER wname</code> is not exactly equivalent to

  441.     <code class="literal">OVER (wname ...)</code>; the latter implies copying and modifying the

  442.     window definition, and will be rejected if the referenced window

  443.     specification includes a frame clause.

  444.    </p><p>

  445.     The <code class="literal">PARTITION BY</code> clause groups the rows of the query into

  446.     <em class="firstterm">partitions</em>, which are processed separately by the window

  447.     function.  <code class="literal">PARTITION BY</code> works similarly to a query-level

  448.     <code class="literal">GROUP BY</code> clause, except that its expressions are always just

  449.     expressions and cannot be output-column names or numbers.

  450.     Without <code class="literal">PARTITION BY</code>, all rows produced by the query are

  451.     treated as a single partition.

  452.     The <code class="literal">ORDER BY</code> clause determines the order in which the rows

  453.     of a partition are processed by the window function.  It works similarly

  454.     to a query-level <code class="literal">ORDER BY</code> clause, but likewise cannot use

  455.     output-column names or numbers.  Without <code class="literal">ORDER BY</code>, rows are

  456.     processed in an unspecified order.

  457.    </p><p>

  458.     The <em class="replaceable"><code>frame_clause</code></em> specifies

  459.     the set of rows constituting the <em class="firstterm">window frame</em>, which is a

  460.     subset of the current partition, for those window functions that act on

  461.     the frame instead of the whole partition.  The set of rows in the frame

  462.     can vary depending on which row is the current row.  The frame can be

  463.     specified in <code class="literal">RANGE</code>, <code class="literal">ROWS</code>

  464.     or <code class="literal">GROUPS</code> mode; in each case, it runs from

  465.     the <em class="replaceable"><code>frame_start</code></em> to

  466.     the <em class="replaceable"><code>frame_end</code></em>.

  467.     If <em class="replaceable"><code>frame_end</code></em> is omitted, the end defaults

  468.     to <code class="literal">CURRENT ROW</code>.

  469.    </p><p>

  470.     A <em class="replaceable"><code>frame_start</code></em> of <code class="literal">UNBOUNDED PRECEDING</code> means

  471.     that the frame starts with the first row of the partition, and similarly

  472.     a <em class="replaceable"><code>frame_end</code></em> of <code class="literal">UNBOUNDED FOLLOWING</code> means

  473.     that the frame ends with the last row of the partition.

  474.    </p><p>

  475.     In <code class="literal">RANGE</code> or <code class="literal">GROUPS</code> mode,

  476.     a <em class="replaceable"><code>frame_start</code></em> of

  477.     <code class="literal">CURRENT ROW</code> means the frame starts with the current

  478.     row's first <em class="firstterm">peer</em> row (a row that the

  479.     window's <code class="literal">ORDER BY</code> clause sorts as equivalent to the

  480.     current row), while a <em class="replaceable"><code>frame_end</code></em> of

  481.     <code class="literal">CURRENT ROW</code> means the frame ends with the current

  482.     row's last peer row.

  483.     In <code class="literal">ROWS</code> mode, <code class="literal">CURRENT ROW</code> simply

  484.     means the current row.

  485.    </p><p>

  486.     In the <em class="replaceable"><code>offset</code></em> <code class="literal">PRECEDING</code>

  487.     and <em class="replaceable"><code>offset</code></em> <code class="literal">FOLLOWING</code> frame

  488.     options, the <em class="replaceable"><code>offset</code></em> must be an expression not

  489.     containing any variables, aggregate functions, or window functions.

  490.     The meaning of the <em class="replaceable"><code>offset</code></em> depends on the

  491.     frame mode:

  492.     </p><div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; "><li class="listitem"><p>

  493.        In <code class="literal">ROWS</code> mode,

  494.        the <em class="replaceable"><code>offset</code></em> must yield a non-null,

  495.        non-negative integer, and the option means that the frame starts or

  496.        ends the specified number of rows before or after the current row.

  497.       </p></li><li class="listitem"><p>

  498.        In <code class="literal">GROUPS</code> mode,

  499.        the <em class="replaceable"><code>offset</code></em> again must yield a non-null,

  500.        non-negative integer, and the option means that the frame starts or

  501.        ends the specified number of <em class="firstterm">peer groups</em>

  502.        before or after the current row's peer group, where a peer group is a

  503.        set of rows that are equivalent in the <code class="literal">ORDER BY</code>

  504.        ordering.  (There must be an <code class="literal">ORDER BY</code> clause

  505.        in the window definition to use <code class="literal">GROUPS</code> mode.)

  506.       </p></li><li class="listitem"><p>

  507.        In <code class="literal">RANGE</code> mode, these options require that

  508.        the <code class="literal">ORDER BY</code> clause specify exactly one column.

  509.        The <em class="replaceable"><code>offset</code></em> specifies the maximum

  510.        difference between the value of that column in the current row and

  511.        its value in preceding or following rows of the frame.  The data type

  512.        of the <em class="replaceable"><code>offset</code></em> expression varies depending

  513.        on the data type of the ordering column.  For numeric ordering

  514.        columns it is typically of the same type as the ordering column,

  515.        but for datetime ordering columns it is an <code class="type">interval</code>.

  516.        For example, if the ordering column is of type <code class="type">date</code>

  517.        or <code class="type">timestamp</code>, one could write <code class="literal">RANGE BETWEEN

  518.        '1 day' PRECEDING AND '10 days' FOLLOWING</code>.

  519.        The <em class="replaceable"><code>offset</code></em> is still required to be

  520.        non-null and non-negative, though the meaning

  521.        of <span class="quote">“<span class="quote">non-negative</span>”</span> depends on its data type.

  522.       </p></li></ul></div><p>

  523.     In any case, the distance to the end of the frame is limited by the

  524.     distance to the end of the partition, so that for rows near the partition

  525.     ends the frame might contain fewer rows than elsewhere.

  526.    </p><p>

  527.     Notice that in both <code class="literal">ROWS</code> and <code class="literal">GROUPS</code>

  528.     mode, <code class="literal">0 PRECEDING</code> and <code class="literal">0 FOLLOWING</code>

  529.     are equivalent to <code class="literal">CURRENT ROW</code>.  This normally holds

  530.     in <code class="literal">RANGE</code> mode as well, for an appropriate

  531.     data-type-specific meaning of <span class="quote">“<span class="quote">zero</span>”</span>.

  532.    </p><p>

  533.     The <em class="replaceable"><code>frame_exclusion</code></em> option allows rows around

  534.     the current row to be excluded from the frame, even if they would be

  535.     included according to the frame start and frame end options.

  536.     <code class="literal">EXCLUDE CURRENT ROW</code> excludes the current row from the

  537.     frame.

  538.     <code class="literal">EXCLUDE GROUP</code> excludes the current row and its

  539.     ordering peers from the frame.

  540.     <code class="literal">EXCLUDE TIES</code> excludes any peers of the current

  541.     row from the frame, but not the current row itself.

  542.     <code class="literal">EXCLUDE NO OTHERS</code> simply specifies explicitly the

  543.     default behavior of not excluding the current row or its peers.

  544.    </p><p>

  545.     The default framing option is <code class="literal">RANGE UNBOUNDED PRECEDING</code>,

  546.     which is the same as <code class="literal">RANGE BETWEEN UNBOUNDED PRECEDING AND

  547.     CURRENT ROW</code>.  With <code class="literal">ORDER BY</code>, this sets the frame to be

  548.     all rows from the partition start up through the current row's last

  549.     <code class="literal">ORDER BY</code> peer.  Without <code class="literal">ORDER BY</code>,

  550.     this means all rows of the partition are included in the window frame,

  551.     since all rows become peers of the current row.

  552.    </p><p>

  553.     Restrictions are that

  554.     <em class="replaceable"><code>frame_start</code></em> cannot be <code class="literal">UNBOUNDED FOLLOWING</code>,

  555.     <em class="replaceable"><code>frame_end</code></em> cannot be <code class="literal">UNBOUNDED PRECEDING</code>,

  556.     and the <em class="replaceable"><code>frame_end</code></em> choice cannot appear earlier in the

  557.     above list of <em class="replaceable"><code>frame_start</code></em>

  558.     and <em class="replaceable"><code>frame_end</code></em> options than

  559.     the <em class="replaceable"><code>frame_start</code></em> choice does — for example

  560.     <code class="literal">RANGE BETWEEN CURRENT ROW AND <em class="replaceable"><code>offset</code></em>

  561.     PRECEDING</code> is not allowed.

  562.     But, for example, <code class="literal">ROWS BETWEEN 7 PRECEDING AND 8

  563.     PRECEDING</code> is allowed, even though it would never select any

  564.     rows.

  565.    </p><p>

  566.     If <code class="literal">FILTER</code> is specified, then only the input

  567.     rows for which the <em class="replaceable"><code>filter_clause</code></em>

  568.     evaluates to true are fed to the window function; other rows

  569.     are discarded.  Only window functions that are aggregates accept

  570.     a <code class="literal">FILTER</code> clause.

  571.    </p><p>

  572.     The built-in window functions are described in <a class="xref" href="functions-window.html#FUNCTIONS-WINDOW-TABLE" title="Table 9.60. General-Purpose Window Functions">Table 9.60</a>.  Other window functions can be added by

  573.     the user.  Also, any built-in or user-defined general-purpose or

  574.     statistical aggregate can be used as a window function.  (Ordered-set

  575.     and hypothetical-set aggregates cannot presently be used as window functions.)

  576.    </p><p>

  577.     The syntaxes using <code class="literal">*</code> are used for calling parameter-less

  578.     aggregate functions as window functions, for example

  579.     <code class="literal">count(*) OVER (PARTITION BY x ORDER BY y)</code>.

  580.     The asterisk (<code class="literal">*</code>) is customarily not used for

  581.     window-specific functions.  Window-specific functions do not

  582.     allow <code class="literal">DISTINCT</code> or <code class="literal">ORDER BY</code> to be used within the

  583.     function argument list.

  584.    </p><p>

  585.     Window function calls are permitted only in the <code class="literal">SELECT</code>

  586.     list and the <code class="literal">ORDER BY</code> clause of the query.

  587.    </p><p>

  588.     More information about window functions can be found in

  589.     <a class="xref" href="tutorial-window.html" title="3.5. Window Functions">Section 3.5</a>,

  590.     <a class="xref" href="functions-window.html" title="9.21. Window Functions">Section 9.21</a>, and

  591.     <a class="xref" href="queries-table-expressions.html#QUERIES-WINDOW" title="7.2.5. Window Function Processing">Section 7.2.5</a>.

  592.    </p></div><div class="sect2" id="SQL-SYNTAX-TYPE-CASTS"><div class="titlepage"><div><div><h3 class="title">4.2.9. Type Casts</h3></div></div></div><a id="id-1.5.3.6.17.2" class="indexterm"></a><a id="id-1.5.3.6.17.3" class="indexterm"></a><a id="id-1.5.3.6.17.4" class="indexterm"></a><p>

  593.     A type cast specifies a conversion from one data type to another.

  594.     <span class="productname">PostgreSQL</span> accepts two equivalent syntaxes

  595.     for type casts:

  596. </p><pre class="synopsis">

  597. CAST ( <em class="replaceable"><code>expression</code></em> AS <em class="replaceable"><code>type</code></em> )

  598. <em class="replaceable"><code>expression</code></em>::<em class="replaceable"><code>type</code></em>

  599. </pre><p>

  600.     The <code class="literal">CAST</code> syntax conforms to SQL; the syntax with

  601.     <code class="literal">::</code> is historical <span class="productname">PostgreSQL</span>

  602.     usage.

  603.    </p><p>

  604.     When a cast is applied to a value expression of a known type, it

  605.     represents a run-time type conversion.  The cast will succeed only

  606.     if a suitable type conversion operation has been defined.  Notice that this

  607.     is subtly different from the use of casts with constants, as shown in

  608.     <a class="xref" href="sql-syntax-lexical.html#SQL-SYNTAX-CONSTANTS-GENERIC" title="4.1.2.7. Constants of Other Types">Section 4.1.2.7</a>.  A cast applied to an

  609.     unadorned string literal represents the initial assignment of a type

  610.     to a literal constant value, and so it will succeed for any type

  611.     (if the contents of the string literal are acceptable input syntax for the

  612.     data type).

  613.    </p><p>

  614.     An explicit type cast can usually be omitted if there is no ambiguity as

  615.     to the type that a value expression must produce (for example, when it is

  616.     assigned to a table column); the system will automatically apply a

  617.     type cast in such cases.  However, automatic casting is only done for

  618.     casts that are marked <span class="quote">“<span class="quote">OK to apply implicitly</span>”</span>

  619.     in the system catalogs.  Other casts must be invoked with

  620.     explicit casting syntax.  This restriction is intended to prevent

  621.     surprising conversions from being applied silently.

  622.    </p><p>

  623.     It is also possible to specify a type cast using a function-like

  624.     syntax:

  625. </p><pre class="synopsis">

  626. <em class="replaceable"><code>typename</code></em> ( <em class="replaceable"><code>expression</code></em> )

  627. </pre><p>

  628.     However, this only works for types whose names are also valid as

  629.     function names.  For example, <code class="literal">double precision</code>

  630.     cannot be used this way, but the equivalent <code class="literal">float8</code>

  631.     can.  Also, the names <code class="literal">interval</code>, <code class="literal">time</code>, and

  632.     <code class="literal">timestamp</code> can only be used in this fashion if they are

  633.     double-quoted, because of syntactic conflicts.  Therefore, the use of

  634.     the function-like cast syntax leads to inconsistencies and should

  635.     probably be avoided.

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

  637.      The function-like syntax is in fact just a function call.  When

  638.      one of the two standard cast syntaxes is used to do a run-time

  639.      conversion, it will internally invoke a registered function to

  640.      perform the conversion.  By convention, these conversion functions

  641.      have the same name as their output type, and thus the <span class="quote">“<span class="quote">function-like

  642.      syntax</span>”</span> is nothing more than a direct invocation of the underlying

  643.      conversion function.  Obviously, this is not something that a portable

  644.      application should rely on.  For further details see

  645.      <a class="xref" href="sql-createcast.html" title="CREATE CAST"><span class="refentrytitle">CREATE CAST</span></a>.

  646.     </p></div></div><div class="sect2" id="SQL-SYNTAX-COLLATE-EXPRS"><div class="titlepage"><div><div><h3 class="title">4.2.10. Collation Expressions</h3></div></div></div><a id="id-1.5.3.6.18.2" class="indexterm"></a><p>

  647.     The <code class="literal">COLLATE</code> clause overrides the collation of

  648.     an expression.  It is appended to the expression it applies to:

  649. </p><pre class="synopsis">

  650. <em class="replaceable"><code>expr</code></em> COLLATE <em class="replaceable"><code>collation</code></em>

  651. </pre><p>

  652.     where <em class="replaceable"><code>collation</code></em> is a possibly

  653.     schema-qualified identifier.  The <code class="literal">COLLATE</code>

  654.     clause binds tighter than operators; parentheses can be used when

  655.     necessary.

  656.    </p><p>

  657.     If no collation is explicitly specified, the database system

  658.     either derives a collation from the columns involved in the

  659.     expression, or it defaults to the default collation of the

  660.     database if no column is involved in the expression.

  661.    </p><p>

  662.     The two common uses of the <code class="literal">COLLATE</code> clause are

  663.     overriding the sort order in an <code class="literal">ORDER BY</code> clause, for

  664.     example:

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

  666. SELECT a, b, c FROM tbl WHERE ... ORDER BY a COLLATE "C";

  667. </pre><p>

  668.     and overriding the collation of a function or operator call that

  669.     has locale-sensitive results, for example:

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

  671. SELECT * FROM tbl WHERE a &gt; 'foo' COLLATE "C";

  672. </pre><p>

  673.     Note that in the latter case the <code class="literal">COLLATE</code> clause is

  674.     attached to an input argument of the operator we wish to affect.

  675.     It doesn't matter which argument of the operator or function call the

  676.     <code class="literal">COLLATE</code> clause is attached to, because the collation that is

  677.     applied by the operator or function is derived by considering all

  678.     arguments, and an explicit <code class="literal">COLLATE</code> clause will override the

  679.     collations of all other arguments.  (Attaching non-matching

  680.     <code class="literal">COLLATE</code> clauses to more than one argument, however, is an

  681.     error.  For more details see <a class="xref" href="collation.html" title="23.2. Collation Support">Section 23.2</a>.)

  682.     Thus, this gives the same result as the previous example:

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

  684. SELECT * FROM tbl WHERE a COLLATE "C" &gt; 'foo';

  685. </pre><p>

  686.     But this is an error:

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

  688. SELECT * FROM tbl WHERE (a &gt; 'foo') COLLATE "C";

  689. </pre><p>

  690.     because it attempts to apply a collation to the result of the

  691.     <code class="literal">&gt;</code> operator, which is of the non-collatable data type

  692.     <code class="type">boolean</code>.

  693.    </p></div><div class="sect2" id="SQL-SYNTAX-SCALAR-SUBQUERIES"><div class="titlepage"><div><div><h3 class="title">4.2.11. Scalar Subqueries</h3></div></div></div><a id="id-1.5.3.6.19.2" class="indexterm"></a><p>

  694.     A scalar subquery is an ordinary

  695.     <code class="command">SELECT</code> query in parentheses that returns exactly one

  696.     row with one column.  (See <a class="xref" href="queries.html" title="Chapter 7. Queries">Chapter 7</a> for information about writing queries.)

  697.     The <code class="command">SELECT</code> query is executed

  698.     and the single returned value is used in the surrounding value expression.

  699.     It is an error to use a query that

  700.     returns more than one row or more than one column as a scalar subquery.

  701.     (But if, during a particular execution, the subquery returns no rows,

  702.     there is no error; the scalar result is taken to be null.)

  703.     The subquery can refer to variables from the surrounding query,

  704.     which will act as constants during any one evaluation of the subquery.

  705.     See also <a class="xref" href="functions-subquery.html" title="9.22. Subquery Expressions">Section 9.22</a> for other expressions involving subqueries.

  706.    </p><p>

  707.     For example, the following finds the largest city population in each

  708.     state:

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

  710. SELECT name, (SELECT max(pop) FROM cities WHERE cities.state = states.name)

  711.     FROM states;

  712. </pre><p>

  713.    </p></div><div class="sect2" id="SQL-SYNTAX-ARRAY-CONSTRUCTORS"><div class="titlepage"><div><div><h3 class="title">4.2.12. Array Constructors</h3></div></div></div><a id="id-1.5.3.6.20.2" class="indexterm"></a><a id="id-1.5.3.6.20.3" class="indexterm"></a><p>

  714.     An array constructor is an expression that builds an

  715.     array value using values for its member elements.  A simple array

  716.     constructor

  717.     consists of the key word <code class="literal">ARRAY</code>, a left square bracket

  718.     <code class="literal">[</code>, a list of expressions (separated by commas) for the

  719.     array element values, and finally a right square bracket <code class="literal">]</code>.

  720.     For example:

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

  722. SELECT ARRAY[1,2,3+4];

  723.   array

  724. ---------

  725.  {1,2,7}

  726. (1 row)

  727. </pre><p>

  728.     By default,

  729.     the array element type is the common type of the member expressions,

  730.     determined using the same rules as for <code class="literal">UNION</code> or

  731.     <code class="literal">CASE</code> constructs (see <a class="xref" href="typeconv-union-case.html" title="10.5. UNION, CASE, and Related Constructs">Section 10.5</a>).

  732.     You can override this by explicitly casting the array constructor to the

  733.     desired type, for example:

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

  735. SELECT ARRAY[1,2,22.7]::integer[];

  736.   array

  737. ----------

  738.  {1,2,23}

  739. (1 row)

  740. </pre><p>

  741.     This has the same effect as casting each expression to the array

  742.     element type individually.

  743.     For more on casting, see <a class="xref" href="sql-expressions.html#SQL-SYNTAX-TYPE-CASTS" title="4.2.9. Type Casts">Section 4.2.9</a>.

  744.    </p><p>

  745.     Multidimensional array values can be built by nesting array

  746.     constructors.

  747.     In the inner constructors, the key word <code class="literal">ARRAY</code> can

  748.     be omitted.  For example, these produce the same result:

  749. 

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

  751. SELECT ARRAY[ARRAY[1,2], ARRAY[3,4]];

  752.      array

  753. ---------------

  754.  {{1,2},{3,4}}

  755. (1 row)

  756. 

  757. SELECT ARRAY[[1,2],[3,4]];

  758.      array

  759. ---------------

  760.  {{1,2},{3,4}}

  761. (1 row)

  762. </pre><p>

  763. 

  764.     Since multidimensional arrays must be rectangular, inner constructors

  765.     at the same level must produce sub-arrays of identical dimensions.

  766.     Any cast applied to the outer <code class="literal">ARRAY</code> constructor propagates

  767.     automatically to all the inner constructors.

  768.   </p><p>

  769.     Multidimensional array constructor elements can be anything yielding

  770.     an array of the proper kind, not only a sub-<code class="literal">ARRAY</code> construct.

  771.     For example:

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

  773. CREATE TABLE arr(f1 int[], f2 int[]);

  774. 

  775. INSERT INTO arr VALUES (ARRAY[[1,2],[3,4]], ARRAY[[5,6],[7,8]]);

  776. 

  777. SELECT ARRAY[f1, f2, '{{9,10},{11,12}}'::int[]] FROM arr;

  778.                      array

  779. ------------------------------------------------

  780.  {{{1,2},{3,4}},{{5,6},{7,8}},{{9,10},{11,12}}}

  781. (1 row)

  782. </pre><p>

  783.   </p><p>

  784.    You can construct an empty array, but since it's impossible to have an

  785.    array with no type, you must explicitly cast your empty array to the

  786.    desired type.  For example:

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

  788. SELECT ARRAY[]::integer[];

  789.  array

  790. -------

  791.  {}

  792. (1 row)

  793. </pre><p>

  794.   </p><p>

  795.    It is also possible to construct an array from the results of a

  796.    subquery.  In this form, the array constructor is written with the

  797.    key word <code class="literal">ARRAY</code> followed by a parenthesized (not

  798.    bracketed) subquery. For example:

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

  800. SELECT ARRAY(SELECT oid FROM pg_proc WHERE proname LIKE 'bytea%');

  801.                                  array

  802. -----------------------------------------------------------------------

  803.  {2011,1954,1948,1952,1951,1244,1950,2005,1949,1953,2006,31,2412,2413}

  804. (1 row)

  805. 

  806. SELECT ARRAY(SELECT ARRAY[i, i*2] FROM generate_series(1,5) AS a(i));

  807.               array

  808. ----------------------------------

  809.  {{1,2},{2,4},{3,6},{4,8},{5,10}}

  810. (1 row)

  811. </pre><p>

  812.    The subquery must return a single column.

  813.    If the subquery's output column is of a non-array type, the resulting

  814.    one-dimensional array will have an element for each row in the

  815.    subquery result, with an element type matching that of the

  816.    subquery's output column.

  817.    If the subquery's output column is of an array type, the result will be

  818.    an array of the same type but one higher dimension; in this case all

  819.    the subquery rows must yield arrays of identical dimensionality, else

  820.    the result would not be rectangular.

  821.   </p><p>

  822.    The subscripts of an array value built with <code class="literal">ARRAY</code>

  823.    always begin with one.  For more information about arrays, see

  824.    <a class="xref" href="arrays.html" title="8.15. Arrays">Section 8.15</a>.

  825.   </p></div><div class="sect2" id="SQL-SYNTAX-ROW-CONSTRUCTORS"><div class="titlepage"><div><div><h3 class="title">4.2.13. Row Constructors</h3></div></div></div><a id="id-1.5.3.6.21.2" class="indexterm"></a><a id="id-1.5.3.6.21.3" class="indexterm"></a><a id="id-1.5.3.6.21.4" class="indexterm"></a><p>

  826.     A row constructor is an expression that builds a row value (also

  827.     called a composite value) using values

  828.     for its member fields.  A row constructor consists of the key word

  829.     <code class="literal">ROW</code>, a left parenthesis, zero or more

  830.     expressions (separated by commas) for the row field values, and finally

  831.     a right parenthesis.  For example:

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

  833. SELECT ROW(1,2.5,'this is a test');

  834. </pre><p>

  835.     The key word <code class="literal">ROW</code> is optional when there is more than one

  836.     expression in the list.

  837.    </p><p>

  838.     A row constructor can include the syntax

  839.     <em class="replaceable"><code>rowvalue</code></em><code class="literal">.*</code>,

  840.     which will be expanded to a list of the elements of the row value,

  841.     just as occurs when the <code class="literal">.*</code> syntax is used at the top level

  842.     of a <code class="command">SELECT</code> list (see <a class="xref" href="rowtypes.html#ROWTYPES-USAGE" title="8.16.5. Using Composite Types in Queries">Section 8.16.5</a>).

  843.     For example, if table <code class="literal">t</code> has

  844.     columns <code class="literal">f1</code> and <code class="literal">f2</code>, these are the same:

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

  846. SELECT ROW(t.*, 42) FROM t;

  847. SELECT ROW(t.f1, t.f2, 42) FROM t;

  848. </pre><p>

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

  850.      Before <span class="productname">PostgreSQL</span> 8.2, the

  851.      <code class="literal">.*</code> syntax was not expanded in row constructors, so

  852.      that writing <code class="literal">ROW(t.*, 42)</code> created a two-field row whose first

  853.      field was another row value.  The new behavior is usually more useful.

  854.      If you need the old behavior of nested row values, write the inner

  855.      row value without <code class="literal">.*</code>, for instance

  856.      <code class="literal">ROW(t, 42)</code>.

  857.     </p></div><p>

  858.     By default, the value created by a <code class="literal">ROW</code> expression is of

  859.     an anonymous record type.  If necessary, it can be cast to a named

  860.     composite type — either the row type of a table, or a composite type

  861.     created with <code class="command">CREATE TYPE AS</code>.  An explicit cast might be needed

  862.     to avoid ambiguity.  For example:

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

  864. CREATE TABLE mytable(f1 int, f2 float, f3 text);

  865. 

  866. CREATE FUNCTION getf1(mytable) RETURNS int AS 'SELECT $1.f1' LANGUAGE SQL;

  867. 

  868. -- No cast needed since only one getf1() exists

  869. SELECT getf1(ROW(1,2.5,'this is a test'));

  870.  getf1

  871. -------

  872.      1

  873. (1 row)

  874. 

  875. CREATE TYPE myrowtype AS (f1 int, f2 text, f3 numeric);

  876. 

  877. CREATE FUNCTION getf1(myrowtype) RETURNS int AS 'SELECT $1.f1' LANGUAGE SQL;

  878. 

  879. -- Now we need a cast to indicate which function to call:

  880. SELECT getf1(ROW(1,2.5,'this is a test'));

  881. ERROR:  function getf1(record) is not unique

  882. 

  883. SELECT getf1(ROW(1,2.5,'this is a test')::mytable);

  884.  getf1

  885. -------

  886.      1

  887. (1 row)

  888. 

  889. SELECT getf1(CAST(ROW(11,'this is a test',2.5) AS myrowtype));

  890.  getf1

  891. -------

  892.     11

  893. (1 row)

  894. </pre><p>

  895.   </p><p>

  896.    Row constructors can be used to build composite values to be stored

  897.    in a composite-type table column, or to be passed to a function that

  898.    accepts a composite parameter.  Also,

  899.    it is possible to compare two row values or test a row with

  900.    <code class="literal">IS NULL</code> or <code class="literal">IS NOT NULL</code>, for example:

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

  902. SELECT ROW(1,2.5,'this is a test') = ROW(1, 3, 'not the same');

  903. 

  904. SELECT ROW(table.*) IS NULL FROM table;  -- detect all-null rows

  905. </pre><p>

  906.    For more detail see <a class="xref" href="functions-comparisons.html" title="9.23. Row and Array Comparisons">Section 9.23</a>.

  907.    Row constructors can also be used in connection with subqueries,

  908.    as discussed in <a class="xref" href="functions-subquery.html" title="9.22. Subquery Expressions">Section 9.22</a>.

  909.   </p></div><div class="sect2" id="SYNTAX-EXPRESS-EVAL"><div class="titlepage"><div><div><h3 class="title">4.2.14. Expression Evaluation Rules</h3></div></div></div><a id="id-1.5.3.6.22.2" class="indexterm"></a><p>

  910.     The order of evaluation of subexpressions is not defined.  In

  911.     particular, the inputs of an operator or function are not necessarily

  912.     evaluated left-to-right or in any other fixed order.

  913.    </p><p>

  914.     Furthermore, if the result of an expression can be determined by

  915.     evaluating only some parts of it, then other subexpressions

  916.     might not be evaluated at all.  For instance, if one wrote:

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

  918. SELECT true OR somefunc();

  919. </pre><p>

  920.     then <code class="literal">somefunc()</code> would (probably) not be called

  921.     at all. The same would be the case if one wrote:

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

  923. SELECT somefunc() OR true;

  924. </pre><p>

  925.     Note that this is not the same as the left-to-right

  926.     <span class="quote">“<span class="quote">short-circuiting</span>”</span> of Boolean operators that is found

  927.     in some programming languages.

  928.    </p><p>

  929.     As a consequence, it is unwise to use functions with side effects

  930.     as part of complex expressions.  It is particularly dangerous to

  931.     rely on side effects or evaluation order in <code class="literal">WHERE</code> and <code class="literal">HAVING</code> clauses,

  932.     since those clauses are extensively reprocessed as part of

  933.     developing an execution plan.  Boolean

  934.     expressions (<code class="literal">AND</code>/<code class="literal">OR</code>/<code class="literal">NOT</code> combinations) in those clauses can be reorganized

  935.     in any manner allowed by the laws of Boolean algebra.

  936.    </p><p>

  937.     When it is essential to force evaluation order, a <code class="literal">CASE</code>

  938.     construct (see <a class="xref" href="functions-conditional.html" title="9.17. Conditional Expressions">Section 9.17</a>) can be

  939.     used.  For example, this is an untrustworthy way of trying to

  940.     avoid division by zero in a <code class="literal">WHERE</code> clause:

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

  942. SELECT ... WHERE x &gt; 0 AND y/x &gt; 1.5;

  943. </pre><p>

  944.     But this is safe:

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

  946. SELECT ... WHERE CASE WHEN x &gt; 0 THEN y/x &gt; 1.5 ELSE false END;

  947. </pre><p>

  948.     A <code class="literal">CASE</code> construct used in this fashion will defeat optimization

  949.     attempts, so it should only be done when necessary.  (In this particular

  950.     example, it would be better to sidestep the problem by writing

  951.     <code class="literal">y &gt; 1.5*x</code> instead.)

  952.    </p><p>

  953.     <code class="literal">CASE</code> is not a cure-all for such issues, however.

  954.     One limitation of the technique illustrated above is that it does not

  955.     prevent early evaluation of constant subexpressions.

  956.     As described in <a class="xref" href="xfunc-volatility.html" title="37.7. Function Volatility Categories">Section 37.7</a>, functions and

  957.     operators marked <code class="literal">IMMUTABLE</code> can be evaluated when

  958.     the query is planned rather than when it is executed.  Thus for example

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

  960. SELECT CASE WHEN x &gt; 0 THEN x ELSE 1/0 END FROM tab;

  961. </pre><p>

  962.     is likely to result in a division-by-zero failure due to the planner

  963.     trying to simplify the constant subexpression,

  964.     even if every row in the table has <code class="literal">x &gt; 0</code> so that the

  965.     <code class="literal">ELSE</code> arm would never be entered at run time.

  966.    </p><p>

  967.     While that particular example might seem silly, related cases that don't

  968.     obviously involve constants can occur in queries executed within

  969.     functions, since the values of function arguments and local variables

  970.     can be inserted into queries as constants for planning purposes.

  971.     Within <span class="application">PL/pgSQL</span> functions, for example, using an

  972.     <code class="literal">IF</code>-<code class="literal">THEN</code>-<code class="literal">ELSE</code> statement to protect

  973.     a risky computation is much safer than just nesting it in a

  974.     <code class="literal">CASE</code> expression.

  975.    </p><p>

  976.     Another limitation of the same kind is that a <code class="literal">CASE</code> cannot

  977.     prevent evaluation of an aggregate expression contained within it,

  978.     because aggregate expressions are computed before other

  979.     expressions in a <code class="literal">SELECT</code> list or <code class="literal">HAVING</code> clause

  980.     are considered.  For example, the following query can cause a

  981.     division-by-zero error despite seemingly having protected against it:

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

  983. SELECT CASE WHEN min(employees) &gt; 0

  984.             THEN avg(expenses / employees)

  985.        END

  986.     FROM departments;

  987. </pre><p>

  988.     The <code class="function">min()</code> and <code class="function">avg()</code> aggregates are computed

  989.     concurrently over all the input rows, so if any row

  990.     has <code class="structfield">employees</code> equal to zero, the division-by-zero error

  991.     will occur before there is any opportunity to test the result of

  992.     <code class="function">min()</code>.  Instead, use a <code class="literal">WHERE</code>

  993.     or <code class="literal">FILTER</code> clause to prevent problematic input rows from

  994.     reaching an aggregate function in the first place.

  995.    </p></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="sql-syntax-lexical.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="sql-syntax.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="sql-syntax-calling-funcs.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">4.1. Lexical Structure </td><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td><td width="40%" align="right" valign="top"> 4.3. Calling Functions</td></tr></table></div></body></html>
上海开阖软件有限公司 沪ICP备12045867号-1