Documentation Source Text

Artifact Content

Artifact c0426fc7fbc2060620baf4ea29977282fdcc3ba8:

<!-- title>SQLite Requirements</title -->
<tcl>hd_keywords {requirements}</tcl>

<h1>SQLite Requirements</h1>

<h2>1.0 Introduction</h2>

<p>These requirements strive to describe the interfaces and operation of 
SQLite in sufficient detail that a compatible implementation of SQLite
can be written solely from these requirements and without reference to
the canonical SQLite source code.</p>

<p>Software development processes typically recognize a hierarchy of

<li>System requirements</li>
<li>High-level requirements</li>
<li>Derived high-level requirements</li>
<li>Low-level requirements</li>
<li>Derived low-level requirements</li>

<p>The usual distinction between high-level and low-level requirements is
that high-level requirements describe "what" the system does and the
low-level requirements describe "how" the system does it.  Since the
the requirements denoted here describe the
behavior of SQLite and not its implementation, they are best thought of
as high-level requirements.  Consistent with that view, most of
the requirements numbers begin with the letter "<b>H</b>" (for "high-level"). 
A few of the requirements presented here specify broad objectives that
SQLite strives to achieve.  These broad requirements can be thought of
as system requirements and are number with an initial letter "<b>S</b>".</p>

<p>These requirements are hierarchical in the sense that the more
specific requirements are derived from broader and more general
requirements.  When requirement B is derived from requirement A, we say
that A is the parent of B and that B is a child of A.  The parent/child
relationships of all requirements are tracked.  All requirements presented
here ultimately derive from a single very broad, very high-level, and
very general system requirement called "S10000".</p>

<p>Some behaviors of SQLite are undefined.  For example, the order in
which result rows are returned from a SELECT statement is undefined if
there is no ORDER BY clause.  As another example, many of the C interfaces
require a pointer to an open [database connection] as their first argument
and the behavior of those interfaces is undefined (and probably undesirable)
if a pointer to some other object is supplied instead.  Some, but not all
undefined behaviors are explicitly stated and number in this document
with numbers beginning with the "<b>U</b>" for "Undefined".  Applications
that use SQLite should never depend on undefined behavior.  If a behavior
is not explicitly defined by a requirement, then the behavior is undefined,
and so explicitly stating undefined behaviors in this document is
technically redundant.  Nevertheless, we find that explicitly stating
some undefined behaviors helps application developers to better understand
the boundaries of operation of SQLite and to generate safer and more
accurate programs that use SQLite.</p>

<h2><a href="sysreq.html">2.0 System Requirements</a></h2>

<h2><a href="hlr10000.html">3.0 Application C-Language Interfaces</a></h2>

<h2><a href="hlr20000.html">4.0 C-Language Interfaces For
Extending SQLite</a></h2>

<h2><a href="hlr30000.html">5.0 Database File Format</a></h2>

<h2><a href="hlr40000.html">6.0 SQL Language Specification</a></h2>