~ubuntu-branches/debian/jessie/sqlalchemy/jessie

    SQLAlchemy 0.8 Documentation

          VERSION:     '0.8.2',

    <link rel="top" title="SQLAlchemy 0.8 Documentation" href="index.html" />

    <h1>SQLAlchemy 0.8 Documentation</h1>

        Release: <span class="version-num">0.8.2</span> | Release Date: July 3, 2013

        <a href="index.html">SQLAlchemy 0.8 Documentation</a>

<dt id="term-annotations">annotations</dt>

<dd><p class="first">Annotations are a concept used internally by SQLAlchemy in order to store

additional information along with <a class="reference internal" href="core/expression_api.html#sqlalchemy.sql.expression.ClauseElement" title="sqlalchemy.sql.expression.ClauseElement"><tt class="xref py py-class docutils literal"><span class="pre">ClauseElement</span></tt></a> objects.  A Python

dictionary is associated with a copy of the object, which contains key/value

pairs significant to various internal systems, mostly within the ORM:</p>

<div class="highlight-python"><div class="highlight"><pre><span class="n">some_column</span> <span class="o">=</span> <span class="n">Column</span><span class="p">(</span><span class="s">&#39;some_column&#39;</span><span class="p">,</span> <span class="n">Integer</span><span class="p">)</span>

<span class="n">some_column_annotated</span> <span class="o">=</span> <span class="n">some_column</span><span class="o">.</span><span class="n">_annotate</span><span class="p">({</span><span class="s">&quot;entity&quot;</span><span class="p">:</span> <span class="n">User</span><span class="p">})</span></pre></div>

</div>

<p class="last">The annotation system differs from the public dictionary <a class="reference internal" href="core/schema.html#sqlalchemy.schema.Column.info" title="sqlalchemy.schema.Column.info"><tt class="xref py py-attr docutils literal"><span class="pre">Column.info</span></tt></a>

in that the above annotation operation creates a <em>copy</em> of the new <a class="reference internal" href="core/schema.html#sqlalchemy.schema.Column" title="sqlalchemy.schema.Column"><tt class="xref py py-class docutils literal"><span class="pre">Column</span></tt></a>,

rather than considering all annotation values to be part of a single

unit.  The ORM creates copies of expression objects in order to

apply annotations that are specific to their context, such as to differentiate

columns that should render themselves as relative to a joined-inheritance

entity versus those which should render relative to their immediate parent

table alone, as well as to differentiate columns within the &#8220;join condition&#8221;

of a relationship where the column in some cases needs to be expressed

in terms of one particular table alias or another, based on its position

within the join expression.</p>

</dd>

<dt id="term-columns-clause">columns clause</dt>

<dd><p class="first">The portion of the <tt class="docutils literal"><span class="pre">SELECT</span></tt> statement which enumerates the

SQL expressions to be returned in the result set.  The expressions

follow the <tt class="docutils literal"><span class="pre">SELECT</span></tt> keyword directly and are a comma-separated

list of individual expressions.</p>

<p>E.g.:</p>

<div class="highlight-sql"><div class="highlight"><pre><span class="k">SELECT</span> <span class="n">user_account</span><span class="p">.</span><span class="n">name</span><span class="p">,</span> <span class="n">user_account</span><span class="p">.</span><span class="n">email</span>

<span class="k">FROM</span> <span class="n">user_account</span> <span class="k">WHERE</span> <span class="n">user_account</span><span class="p">.</span><span class="n">name</span> <span class="o">=</span> <span class="s1">&#39;fred&#39;</span></pre></div>

</div>

<p class="last">Above, the list of columns <tt class="docutils literal"><span class="pre">user_acount.name</span></tt>,

<tt class="docutils literal"><span class="pre">user_account.email</span></tt> is the columns clause of the <tt class="docutils literal"><span class="pre">SELECT</span></tt>.</p>

</dd>

<dt id="term-correlates"><span id="term-correlated-subquery"></span><span id="term-correlated-subqueries"></span>correlates<br />correlated subquery<br />correlated subqueries</dt>

<dd><p class="first">A <a class="reference internal" href="#term-subquery"><em class="xref std std-term">subquery</em></a> is correlated if it depends on data in the

enclosing <tt class="docutils literal"><span class="pre">SELECT</span></tt>.</p>

<p>Below, a subquery selects the aggregate value <tt class="docutils literal"><span class="pre">MIN(a.id)</span></tt>

from the <tt class="docutils literal"><span class="pre">email_address</span></tt> table, such that

it will be invoked for each value of <tt class="docutils literal"><span class="pre">user_account.id</span></tt>, correlating

the value of this column against the <tt class="docutils literal"><span class="pre">email_address.user_account_id</span></tt>

column:</p>

<div class="highlight-sql"><div class="highlight"><pre><span class="k">SELECT</span> <span class="n">user_account</span><span class="p">.</span><span class="n">name</span><span class="p">,</span> <span class="n">email_address</span><span class="p">.</span><span class="n">email</span>

 <span class="k">FROM</span> <span class="n">user_account</span>

 <span class="k">JOIN</span> <span class="n">email_address</span> <span class="k">ON</span> <span class="n">user_account</span><span class="p">.</span><span class="n">id</span><span class="o">=</span><span class="n">email_address</span><span class="p">.</span><span class="n">user_account_id</span>

 <span class="k">WHERE</span> <span class="n">email_address</span><span class="p">.</span><span class="n">id</span> <span class="o">=</span> <span class="p">(</span>

    <span class="k">SELECT</span> <span class="k">MIN</span><span class="p">(</span><span class="n">a</span><span class="p">.</span><span class="n">id</span><span class="p">)</span> <span class="k">FROM</span> <span class="n">email_address</span> <span class="k">AS</span> <span class="n">a</span>

    <span class="k">WHERE</span> <span class="n">a</span><span class="p">.</span><span class="n">user_account_id</span><span class="o">=</span><span class="n">user_account</span><span class="p">.</span><span class="n">id</span>

 <span class="p">)</span></pre></div>

</div>

<p>The above subquery refers to the <tt class="docutils literal"><span class="pre">user_account</span></tt> table, which is not itself

in the <tt class="docutils literal"><span class="pre">FROM</span></tt> clause of this nested query.   Instead, the <tt class="docutils literal"><span class="pre">user_account</span></tt>

table is recieved from the enclosing query, where each row selected from

<tt class="docutils literal"><span class="pre">user_account</span></tt> results in a distinct execution of the subquery.</p>

<p>A correlated subquery is in most cases present in the <a class="reference internal" href="#term-where-clause"><em class="xref std std-term">WHERE clause</em></a>

or <a class="reference internal" href="#term-columns-clause"><em class="xref std std-term">columns clause</em></a> of the immediately enclosing <tt class="docutils literal"><span class="pre">SELECT</span></tt>

statement, as well as in the ORDER BY or HAVING clause.</p>

<p>In less common cases, a correlated subquery may be present in the

<a class="reference internal" href="#term-from-clause"><em class="xref std std-term">FROM clause</em></a> of an enclosing <tt class="docutils literal"><span class="pre">SELECT</span></tt>; in these cases the

correlation is typically due to the enclosing <tt class="docutils literal"><span class="pre">SELECT</span></tt> itself being

enclosed in the WHERE,

ORDER BY, columns or HAVING clause of another <tt class="docutils literal"><span class="pre">SELECT</span></tt>, such as:</p>

<div class="highlight-sql"><div class="highlight"><pre><span class="k">SELECT</span> <span class="n">parent</span><span class="p">.</span><span class="n">id</span> <span class="k">FROM</span> <span class="n">parent</span>

<span class="k">WHERE</span> <span class="k">EXISTS</span> <span class="p">(</span>

    <span class="k">SELECT</span> <span class="o">*</span> <span class="k">FROM</span> <span class="p">(</span>

        <span class="k">SELECT</span> <span class="n">child</span><span class="p">.</span><span class="n">id</span> <span class="k">AS</span> <span class="n">id</span><span class="p">,</span> <span class="n">child</span><span class="p">.</span><span class="n">parent_id</span> <span class="k">AS</span> <span class="n">parent_id</span><span class="p">,</span> <span class="n">child</span><span class="p">.</span><span class="n">pos</span> <span class="k">AS</span> <span class="n">pos</span>

        <span class="k">FROM</span> <span class="n">child</span>

        <span class="k">WHERE</span> <span class="n">child</span><span class="p">.</span><span class="n">parent_id</span> <span class="o">=</span> <span class="n">parent</span><span class="p">.</span><span class="n">id</span> <span class="k">ORDER</span> <span class="k">BY</span> <span class="n">child</span><span class="p">.</span><span class="n">pos</span>

    <span class="k">LIMIT</span> <span class="mi">3</span><span class="p">)</span>

<span class="k">WHERE</span> <span class="n">id</span> <span class="o">=</span> <span class="mi">7</span><span class="p">)</span></pre></div>

</div>

<p class="last">Correlation from one <tt class="docutils literal"><span class="pre">SELECT</span></tt> directly to one which encloses the correlated

query via its <tt class="docutils literal"><span class="pre">FROM</span></tt>

clause is not possible, because the correlation can only proceed once the

original source rows from the enclosing statement&#8217;s FROM clause are available.</p>

</dd>

<dt id="term-dbapi">DBAPI</dt>

<dd><p class="first">DBAPI is shorthand for the phrase &#8220;Python Database API

Specification&#8221;.  This is a widely used specification

within Python to define common usage patterns for all

database connection packages.   The DBAPI is a &#8220;low level&#8221;

API which is typically the lowest level system used

in a Python application to talk to a database.  SQLAlchemy&#8217;s

<em class="xref std std-term">dialect</em> system is constructed around the

operation of the DBAPI, providing individual dialect

classes which service a specific DBAPI on top of a

specific database engine; for example, the <a class="reference internal" href="core/engines.html#sqlalchemy.create_engine" title="sqlalchemy.create_engine"><tt class="xref py py-func docutils literal"><span class="pre">create_engine()</span></tt></a>

URL <tt class="docutils literal"><span class="pre">postgresql+psycopg2://&#64;localhost/test</span></tt>

refers to the <a class="reference internal" href="dialects/postgresql.html#module-sqlalchemy.dialects.postgresql.psycopg2" title="sqlalchemy.dialects.postgresql.psycopg2"><tt class="xref py py-mod docutils literal"><span class="pre">psycopg2</span></tt></a>

DBAPI/dialect combination, whereas the URL <tt class="docutils literal"><span class="pre">mysql+mysqldb://&#64;localhost/test</span></tt>

refers to the <a class="reference internal" href="dialects/mysql.html#module-sqlalchemy.dialects.mysql.mysqldb" title="sqlalchemy.dialects.mysql.mysqldb"><tt class="xref py py-mod docutils literal"><span class="pre">MySQL</span> <span class="pre">for</span> <span class="pre">Python</span></tt></a>

DBAPI DBAPI/dialect combination.</p>

<div class="admonition-see-also last admonition seealso">

<p class="first admonition-title">See also</p>

<p class="last"><a class="reference external" href="http://www.python.org/dev/peps/pep-0249/">PEP 249 - Python Database API Specification v2.0</a></p>

</div>

</dd>

<dt id="term-descriptor"><span id="term-descriptors"></span>descriptor<br />descriptors</dt>

<dd><p class="first">In Python, a descriptor is an object attribute with “binding behavior”, one whose attribute access has been overridden by methods in the <a class="reference external" href="http://docs.python.org/howto/descriptor.html">descriptor protocol</a>.

Those methods are __get__(), __set__(), and __delete__(). If any of those methods are defined

for an object, it is said to be a descriptor.</p>

<p>In SQLAlchemy, descriptors are used heavily in order to provide attribute behavior

on mapped classes.   When a class is mapped as such:</p>

<div class="highlight-python"><div class="highlight"><pre><span class="k">class</span> <span class="nc">MyClass</span><span class="p">(</span><span class="n">Base</span><span class="p">):</span>

    <span class="n">__tablename__</span> <span class="o">=</span> <span class="s">&#39;foo&#39;</span>

 

    <span class="nb">id</span> <span class="o">=</span> <span class="n">Column</span><span class="p">(</span><span class="n">Integer</span><span class="p">,</span> <span class="n">primary_key</span><span class="o">=</span><span class="bp">True</span><span class="p">)</span>

    <span class="n">data</span> <span class="o">=</span> <span class="n">Column</span><span class="p">(</span><span class="n">String</span><span class="p">)</span></pre></div>

</div>

<p>The <tt class="docutils literal"><span class="pre">MyClass</span></tt> class will be <a class="reference internal" href="#term-mapped"><em class="xref std std-term">mapped</em></a> when its definition

is complete, at which point the <tt class="docutils literal"><span class="pre">id</span></tt> and <tt class="docutils literal"><span class="pre">data</span></tt> attributes,

starting out as <a class="reference internal" href="core/schema.html#sqlalchemy.schema.Column" title="sqlalchemy.schema.Column"><tt class="xref py py-class docutils literal"><span class="pre">Column</span></tt></a> objects, will be replaced

by the <a class="reference internal" href="#term-instrumentation"><em class="xref std std-term">instrumentation</em></a> system with instances

of <a class="reference internal" href="orm/internals.html#sqlalchemy.orm.attributes.InstrumentedAttribute" title="sqlalchemy.orm.attributes.InstrumentedAttribute"><tt class="xref py py-class docutils literal"><span class="pre">InstrumentedAttribute</span></tt></a>, which are descriptors that

provide the above mentioned <tt class="docutils literal"><span class="pre">__get__()</span></tt>, <tt class="docutils literal"><span class="pre">__set__()</span></tt> and

<tt class="docutils literal"><span class="pre">__delete__()</span></tt> methods.   The <a class="reference internal" href="orm/internals.html#sqlalchemy.orm.attributes.InstrumentedAttribute" title="sqlalchemy.orm.attributes.InstrumentedAttribute"><tt class="xref py py-class docutils literal"><span class="pre">InstrumentedAttribute</span></tt></a>

will generate a SQL expression when used at the class level:</p>

<div class="highlight-python"><div class="highlight"><pre><span class="gp">&gt;&gt;&gt; </span><span class="k">print</span> <span class="n">MyClass</span><span class="o">.</span><span class="n">data</span> <span class="o">==</span> <span class="mi">5</span>

<span class="go">data = :data_1</span></pre></div>

</div>

<p>and at the instance level, keeps track of changes to values,

and also <a class="reference internal" href="#term-lazy-loads"><em class="xref std std-term">lazy loads</em></a> unloaded attributes

from the database:</p>

<div class="last highlight-python"><div class="highlight"><pre><span class="gp">&gt;&gt;&gt; </span><span class="n">m1</span> <span class="o">=</span> <span class="n">MyClass</span><span class="p">()</span>

<span class="gp">&gt;&gt;&gt; </span><span class="n">m1</span><span class="o">.</span><span class="n">id</span> <span class="o">=</span> <span class="mi">5</span>

<span class="gp">&gt;&gt;&gt; </span><span class="n">m1</span><span class="o">.</span><span class="n">data</span> <span class="o">=</span> <span class="s">&quot;some data&quot;</span>

 

<span class="gp">&gt;&gt;&gt; </span><span class="kn">from</span> <span class="nn">sqlalchemy</span> <span class="kn">import</span> <span class="n">inspect</span>

<span class="gp">&gt;&gt;&gt; </span><span class="n">inspect</span><span class="p">(</span><span class="n">m1</span><span class="p">)</span><span class="o">.</span><span class="n">attrs</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">history</span><span class="o">.</span><span class="n">added</span>

<span class="go">&quot;some data&quot;</span></pre></div>

</div>

</dd>

<dt id="term-discriminator">discriminator</dt>

<dd><p class="first">A result-set column which is used during <a class="reference internal" href="#term-polymorphic"><em class="xref std std-term">polymorphic</em></a> loading

to determine what kind of mapped class should be applied to a particular

incoming result row.   In SQLAlchemy, the classes are always part

of a hierarchy mapping using inheritance mapping.</p>

<div class="admonition-see-also last admonition seealso">

<p class="first admonition-title">See also</p>

<p class="last"><a class="reference internal" href="orm/inheritance.html"><em>Mapping Class Inheritance Hierarchies</em></a></p>

</div>

</dd>

<dt id="term-from-clause">FROM clause</dt>

<dd><p class="first">The portion of the <tt class="docutils literal"><span class="pre">SELECT</span></tt> statement which incicates the initial

source of rows.</p>

<p>A simple <tt class="docutils literal"><span class="pre">SELECT</span></tt> will feature one or more table names in its

FROM clause.  Multiple sources are separated by a comma:</p>

<div class="highlight-sql"><div class="highlight"><pre><span class="k">SELECT</span> <span class="k">user</span><span class="p">.</span><span class="n">name</span><span class="p">,</span> <span class="n">address</span><span class="p">.</span><span class="n">email_address</span>

<span class="k">FROM</span> <span class="k">user</span><span class="p">,</span> <span class="n">address</span>

<span class="k">WHERE</span> <span class="k">user</span><span class="p">.</span><span class="n">id</span><span class="o">=</span><span class="n">address</span><span class="p">.</span><span class="n">user_id</span></pre></div>

</div>

<p>The FROM clause is also where explicit joins are specified.  We can

rewrite the above <tt class="docutils literal"><span class="pre">SELECT</span></tt> using a single <tt class="docutils literal"><span class="pre">FROM</span></tt> element which consists

of a <tt class="docutils literal"><span class="pre">JOIN</span></tt> of the two tables:</p>

<div class="last highlight-sql"><div class="highlight"><pre><span class="k">SELECT</span> <span class="k">user</span><span class="p">.</span><span class="n">name</span><span class="p">,</span> <span class="n">address</span><span class="p">.</span><span class="n">email_address</span>

<span class="k">FROM</span> <span class="k">user</span> <span class="k">JOIN</span> <span class="n">address</span> <span class="k">ON</span> <span class="k">user</span><span class="p">.</span><span class="n">id</span><span class="o">=</span><span class="n">address</span><span class="p">.</span><span class="n">user_id</span></pre></div>

</div>

</dd>

<dt id="term-generative">generative</dt>

<dd>A term that SQLAlchemy uses to refer what&#8217;s normally known

as <a class="reference internal" href="#term-method-chaining"><em class="xref std std-term">method chaining</em></a>; see that term for details.</dd>

<dt id="term-instrumentation"><span id="term-instrumented"></span>instrumentation<br />instrumented</dt>

<dd>Instrumentation refers to the process of augmenting the functionality

and attribute set of a particular class.   Ideally, the

behavior of the class should remain close to a regular

class, except that additional behviors and features are

made available.  The SQLAlchemy <a class="reference internal" href="#term-mapping"><em class="xref std std-term">mapping</em></a> process,

among other things, adds database-enabled <a class="reference internal" href="#term-descriptors"><em class="xref std std-term">descriptors</em></a>

to a mapped

class which each represent a particular database column

or relationship to a related class.</dd>

<dt id="term-lazy-load"><span id="term-lazy-loads"></span>lazy load<br />lazy loads</dt>

<dd><p class="first">In object relational mapping, a &#8220;lazy load&#8221; refers to an

attribute that does not contain its database-side value

for some period of time, typically when the object is

first loaded.  Instead, the attribute receives a

<em>memoization</em> that causes it to go out to the database

and load its data when it&#8217;s first used.   Using this pattern,

the complexity and time spent within object fetches can

sometimes be reduced, in that

attributes for related tables don&#8217;t need to be addressed

immediately.</p>

<div class="admonition-see-also last admonition seealso">

<p class="first admonition-title">See also</p>

<p><a class="reference external" href="http://martinfowler.com/eaaCatalog/lazyLoad.html">Lazy Load (on Martin Fowler)</a></p>

<p><a class="reference internal" href="#term-n-plus-one-problem"><em class="xref std std-term">N plus one problem</em></a></p>

<p class="last"><a class="reference internal" href="orm/loading.html"><em>Relationship Loading Techniques</em></a></p>

</div>

</dd>

<dt id="term-mapping"><span id="term-mapped"></span>mapping<br />mapped</dt>

<dd>We say a class is &#8220;mapped&#8221; when it has been passed through the

<a class="reference internal" href="orm/mapper_config.html#sqlalchemy.orm.mapper" title="sqlalchemy.orm.mapper"><tt class="xref py py-func docutils literal"><span class="pre">orm.mapper()</span></tt></a> function.   This process associates the

class with a database table or other <em class="xref std std-term">selectable</em>

construct, so that instances of it can be persisted

using a <a class="reference internal" href="orm/session.html#sqlalchemy.orm.session.Session" title="sqlalchemy.orm.session.Session"><tt class="xref py py-class docutils literal"><span class="pre">Session</span></tt></a> as well as loaded using a

<a class="reference internal" href="orm/query.html#sqlalchemy.orm.query.Query" title="sqlalchemy.orm.query.Query"><tt class="xref py py-class docutils literal"><span class="pre">Query</span></tt></a>.</dd>

<dt id="term-method-chaining">method chaining</dt>

<dd><p class="first">An object-oriented technique whereby the state of an object

is constructed by calling methods on the object.   The

object features any number of methods, each of which return

a new object (or in some cases the same object) with

additional state added to the object.</p>

<p>The two SQLAlchemy objects that make the most use of

method chaining are the <a class="reference internal" href="core/expression_api.html#sqlalchemy.sql.expression.Select" title="sqlalchemy.sql.expression.Select"><tt class="xref py py-class docutils literal"><span class="pre">Select</span></tt></a>

object and the <a class="reference internal" href="orm/query.html#sqlalchemy.orm.query.Query" title="sqlalchemy.orm.query.Query"><tt class="xref py py-class docutils literal"><span class="pre">Query</span></tt></a> object.

For example, a <a class="reference internal" href="core/expression_api.html#sqlalchemy.sql.expression.Select" title="sqlalchemy.sql.expression.Select"><tt class="xref py py-class docutils literal"><span class="pre">Select</span></tt></a> object can

be assigned two expressions to its WHERE clause as well

as an ORDER BY clause by calling upon the <a class="reference internal" href="core/expression_api.html#sqlalchemy.sql.expression.Select.where" title="sqlalchemy.sql.expression.Select.where"><tt class="xref py py-meth docutils literal"><span class="pre">where()</span></tt></a>

and <a class="reference internal" href="core/expression_api.html#sqlalchemy.sql.expression.Select.order_by" title="sqlalchemy.sql.expression.Select.order_by"><tt class="xref py py-meth docutils literal"><span class="pre">order_by()</span></tt></a> methods:</p>

<div class="highlight-python"><pre>stmt = select([user.c.name]).\

            where(user.c.id &gt; 5).\

            where(user.c.name.like('e%').\

            order_by(user.c.name)</pre>

</div>

<p>Each method call above returns a copy of the original

<a class="reference internal" href="core/expression_api.html#sqlalchemy.sql.expression.Select" title="sqlalchemy.sql.expression.Select"><tt class="xref py py-class docutils literal"><span class="pre">Select</span></tt></a> object with additional qualifiers

added.</p>

<div class="admonition-see-also last admonition seealso">

<p class="first admonition-title">See also</p>

<p class="last"><a class="reference internal" href="#term-generative"><em class="xref std std-term">generative</em></a></p>

</div>

</dd>

<dt id="term-n-plus-one-problem">N plus one problem</dt>

<dd><p class="first">The N plus one problem is a common side effect of the

<a class="reference internal" href="#term-lazy-load"><em class="xref std std-term">lazy load</em></a> pattern, whereby an application wishes

to iterate through a related attribute or collection on

each member of a result set of objects, where that

attribute or collection is set to be loaded via the lazy

load pattern.   The net result is that a SELECT statement

is emitted to load the initial result set of parent objects;

then, as the application iterates through each member,

an additional SELECT statement is emitted for each member

in order to load the related attribute or collection for

that member.  The end result is that for a result set of

N parent objects, there will be N + 1 SELECT statements emitted.</p>

<p>The N plus one problem is alleviated using <em class="xref std std-term">eager loading</em>.</p>

<div class="admonition-see-also last admonition seealso">

<p class="first admonition-title">See also</p>

<p class="last"><a class="reference internal" href="orm/loading.html"><em>Relationship Loading Techniques</em></a></p>

</div>

</dd>

<dt id="term-polymorphic"><span id="term-polymorphically"></span>polymorphic<br />polymorphically</dt>

<dd><p class="first">Refers to a function that handles several types at once.  In SQLAlchemy,

the term is usually applied to the concept of an ORM mapped class

whereby a query operation will return different subclasses

based on information in the result set, typically by checking the

value of a particular column in the result known as the <a class="reference internal" href="#term-discriminator"><em class="xref std std-term">discriminator</em></a>.</p>

<p class="last">Polymorphic loading in SQLAlchemy implies that a one or a

combination of three different schemes are used to map a hierarchy

of classes; &#8220;joined&#8221;, &#8220;single&#8221;, and &#8220;concrete&#8221;.   The section

<a class="reference internal" href="orm/inheritance.html"><em>Mapping Class Inheritance Hierarchies</em></a> describes inheritance mapping fully.</p>

</dd>

<dd><p class="first">In the context of SQLAlchemy, the term &#8220;released&#8221;

refers to the process of ending the usage of a particular

database connection.    SQLAlchemy features the usage

of connection pools, which allows configurability as to

the lifespan of database connections.   When using a pooled

connection, the process of &#8220;closing&#8221; it, i.e. invoking

a statement like <tt class="docutils literal"><span class="pre">connection.close()</span></tt>, may have the effect

of the connection being returned to an existing pool,

or it may have the effect of actually shutting down the

underlying TCP/IP connection referred to by that connection -

which one takes place depends on configuration as well

as the current state of the pool.  So we used the term

<em>released</em> instead, to mean &#8220;do whatever it is you do

with connections when we&#8217;re done using them&#8221;.</p>

<p>The term will sometimes be used in the phrase, &#8220;release

transactional resources&#8221;, to indicate more explicitly that

what we are actually &#8220;releasing&#8221; is any transactional

state which as accumulated upon the connection.  In most

situations, the proces of selecting from tables, emitting

updates, etc. acquires <em class="xref std std-term">isolated</em> state upon

that connection as well as potential row or table locks.

This state is all local to a particular transaction

on the connection, and is released when we emit a rollback.

An important feature of the connection pool is that when

we return a connection to the pool, the <tt class="docutils literal"><span class="pre">connection.rollback()</span></tt>

method of the DBAPI is called as well, so that as the

connection is set up to be used again, it&#8217;s in a &#8220;clean&#8221;

state with no references held to the previous series

of operations.</p>

<dt id="term-subquery">subquery</dt>

<dd><p class="first">Refers to a <tt class="docutils literal"><span class="pre">SELECT</span></tt> statement that is embedded within an enclosing

<tt class="docutils literal"><span class="pre">SELECT</span></tt>.</p>

<p>A subquery comes in two general flavors, one known as a &#8220;scalar select&#8221;

which specifically must return exactly one row and one column, and the

other form which acts as a &#8220;derived table&#8221; and serves as a source of

rows for the FROM clause of another select.  A scalar select is eligble

to be placed in the <a class="reference internal" href="#term-where-clause"><em class="xref std std-term">WHERE clause</em></a>, <a class="reference internal" href="#term-columns-clause"><em class="xref std std-term">columns clause</em></a>,

ORDER BY clause or HAVING clause of the enclosing select, whereas the

derived table form is eligible to be placed in the FROM clause of the

enclosing <tt class="docutils literal"><span class="pre">SELECT</span></tt>.</p>

<p>Examples:</p>

<ol class="last arabic">

<li><p class="first">a scalar subquery placed in the <a class="reference internal" href="#term-columns-clause"><em class="xref std std-term">columns clause</em></a> of an enclosing

<tt class="docutils literal"><span class="pre">SELECT</span></tt>.  The subquery in this example is a <a class="reference internal" href="#term-correlated-subquery"><em class="xref std std-term">correlated subquery</em></a> because part

of the rows which it selects from are given via the enclosing statement.</p>

<div class="highlight-sql"><div class="highlight"><pre><span class="k">SELECT</span> <span class="n">id</span><span class="p">,</span> <span class="p">(</span><span class="k">SELECT</span> <span class="n">name</span> <span class="k">FROM</span> <span class="n">address</span> <span class="k">WHERE</span> <span class="n">address</span><span class="p">.</span><span class="n">user_id</span><span class="o">=</span><span class="k">user</span><span class="p">.</span><span class="n">id</span><span class="p">)</span>

<span class="k">FROM</span> <span class="k">user</span></pre></div>

</div>

</li>

<li><p class="first">a scalar subquery placed in the <a class="reference internal" href="#term-where-clause"><em class="xref std std-term">WHERE clause</em></a> of an enclosing

<tt class="docutils literal"><span class="pre">SELECT</span></tt>.  This subquery in this example is not correlated as it selects a fixed result.</p>

<div class="highlight-sql"><div class="highlight"><pre><span class="k">SELECT</span> <span class="n">id</span><span class="p">,</span> <span class="n">name</span> <span class="k">FROM</span> <span class="k">user</span>

<span class="k">WHERE</span> <span class="n">status</span><span class="o">=</span><span class="p">(</span><span class="k">SELECT</span> <span class="n">status_id</span> <span class="k">FROM</span> <span class="n">status_code</span> <span class="k">WHERE</span> <span class="n">code</span><span class="o">=</span><span class="s1">&#39;C&#39;</span><span class="p">)</span></pre></div>

</div>

</li>

<li><p class="first">a derived table subquery placed in the <a class="reference internal" href="#term-from-clause"><em class="xref std std-term">FROM clause</em></a> of an enclosing

<tt class="docutils literal"><span class="pre">SELECT</span></tt>.   Such a subquery is almost always given an alias name.</p>

<div class="highlight-sql"><div class="highlight"><pre><span class="k">SELECT</span> <span class="k">user</span><span class="p">.</span><span class="n">id</span><span class="p">,</span> <span class="k">user</span><span class="p">.</span><span class="n">name</span><span class="p">,</span> <span class="n">ad_subq</span><span class="p">.</span><span class="n">email_address</span>

<span class="k">FROM</span>

    <span class="k">user</span> <span class="k">JOIN</span>

    <span class="p">(</span><span class="k">select</span> <span class="n">user_id</span><span class="p">,</span> <span class="n">email_address</span> <span class="k">FROM</span> <span class="n">address</span> <span class="k">WHERE</span> <span class="n">address_type</span><span class="o">=</span><span class="s1">&#39;Q&#39;</span><span class="p">)</span> <span class="k">AS</span> <span class="n">ad_subq</span>

    <span class="k">ON</span> <span class="k">user</span><span class="p">.</span><span class="n">id</span> <span class="o">=</span> <span class="n">ad_subq</span><span class="p">.</span><span class="n">user_id</span></pre></div>

</div>

</li>

</ol>

</dd>

<dt id="term-unit-of-work">unit of work</dt>

<dd><p class="first">This pattern is where the system transparently keeps

track of changes to objects and periodically flushes all those

pending changes out to the database. SQLAlchemy&#8217;s Session

implements this pattern fully in a manner similar to that of

Hibernate.</p>

<div class="admonition-see-also last admonition seealso">

<p class="first admonition-title">See also</p>

<p><a class="reference external" href="http://martinfowler.com/eaaCatalog/unitOfWork.html">Unit of Work by Martin Fowler</a></p>

<p class="last"><a class="reference internal" href="orm/session.html"><em>Using the Session</em></a></p>

</div>

</dd>

<dt id="term-where-clause">WHERE clause</dt>

<dd><p class="first">The portion of the <tt class="docutils literal"><span class="pre">SELECT</span></tt> statement which indicates criteria

by which rows should be filtered.   It is a single SQL expression

which follows the keyword <tt class="docutils literal"><span class="pre">WHERE</span></tt>.</p>

<div class="highlight-sql"><div class="highlight"><pre><span class="k">SELECT</span> <span class="n">user_account</span><span class="p">.</span><span class="n">name</span><span class="p">,</span> <span class="n">user_account</span><span class="p">.</span><span class="n">email</span>

<span class="k">FROM</span> <span class="n">user_account</span>

<span class="k">WHERE</span> <span class="n">user_account</span><span class="p">.</span><span class="n">name</span> <span class="o">=</span> <span class="s1">&#39;fred&#39;</span> <span class="k">AND</span> <span class="n">user_account</span><span class="p">.</span><span class="n">status</span> <span class="o">=</span> <span class="s1">&#39;E&#39;</span></pre></div>

</div>

<p class="last">Above, the phrase <tt class="docutils literal"><span class="pre">WHERE</span> <span class="pre">user_account.name</span> <span class="pre">=</span> <span class="pre">'fred'</span> <span class="pre">AND</span> <span class="pre">user_account.status</span> <span class="pre">=</span> <span class="pre">'E'</span></tt>

comprises the WHERE clause of the <tt class="docutils literal"><span class="pre">SELECT</span></tt>.</p>

</dd>

        &copy; <a href="copyright.html">Copyright</a> 2007-2013, the SQLAlchemy authors and contributors.