Planet
navi homePPSaboutscreenshotsdownloaddevelopmentforum

source: downloads/boost_1_34_1/libs/numeric/interval/doc/interval.htm @ 29

Last change on this file since 29 was 29, checked in by landauf, 16 years ago

updated boost from 1_33_1 to 1_34_1

File size: 57.4 KB
Line 
1<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
2"http://www.w3.org/TR/html4/loose.dtd">
3
4<html>
5<head>
6  <meta http-equiv="Content-Language" content="en-us">
7  <meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
8  <link rel="stylesheet" type="text/css" href="../../../../boost.css">
9
10  <title>Boost Interval Arithmetic Library</title>
11</head>
12
13<body lang="en">
14  <h1><img src="../../../../boost.png" alt="boost.png (6897 bytes)" align=
15  "middle"> Interval Arithmetic Library</h1>
16
17  <center>
18    <table width="80%" summary="">
19      <tbody>
20        <tr>
21          <td><b>Contents of this page:</b><br>
22          <a href="#intro">Introduction</a><br>
23          <a href="#synopsis">Synopsis</a><br>
24          <a href="#interval">Template class <code>interval</code></a><br>
25          <a href="#opers">Operations and functions</a><br>
26          <a href="#interval_lib">Interval support library</a><br>
27          <!--<a href="#compil">Compilation notes</a><br>-->
28           <a href="#dangers">Common pitfalls and dangers</a><br>
29          <a href="#rationale">Rationale</a><br>
30          <a href="#acks">History and Acknowledgments</a></td>
31
32          <td><b>Other pages associated with this page:</b><br>
33          <a href="rounding.htm">Rounding policies</a><br>
34          <a href="checking.htm">Checking policies</a><br>
35          <a href="policies.htm">Policies manipulation</a><br>
36          <a href="comparisons.htm">Comparisons</a><br>
37          <a href="numbers.htm">Base number type requirements</a><br>
38          <a href="guide.htm">Choosing your own interval type</a><br>
39          <a href="examples.htm">Test and example programs</a><br>
40          <a href="includes.htm">Headers inclusion</a><br>
41          <a href="todo.htm">Some items on the todo list</a></td>
42        </tr>
43      </tbody>
44    </table>
45  </center>
46
47  <h2 id="intro">Introduction and Overview</h2>
48
49  <p>As implied by its name, this library is intended to help manipulating
50  mathematical intervals. It consists of a single header &lt;<a href=
51  "../../../../boost/numeric/interval.hpp">boost/numeric/interval.hpp</a>&gt;
52  and principally a type which can be used as <code>interval&lt;T&gt;</code>.
53  In fact, this interval template is declared as
54  <code>interval&lt;T,Policies&gt;</code> where <code>Policies</code> is a
55  policy class that controls the various behaviours of the interval class;
56  <code>interval&lt;T&gt;</code> just happens to pick the default policies
57  for the type <code>T</code>.</p>
58
59  <p><span style="color: #FF0000; font-weight: bold">Warning!</span>
60  Guaranteed interval arithmetic for native floating-point format is not
61  supported on every combination of processor, operating system, and
62  compiler. This is a list of systems known to work correctly when using
63  <code>interval&lt;float&gt;</code> and <code>interval&lt;double&gt;</code>
64  with basic arithmetic operators.</p>
65
66  <ul>
67    <li>x86-like hardware is supported by the library with GCC, Visual C++
68    &ge; 7.1, Intel compiler (&ge; 8 on Windows), CodeWarrior (&ge; 9), as
69    long as the traditional x87 floating-point unit is used for
70    floating-point computations (no <code>-mfpmath=sse2</code> support).</li>
71
72    <li>Sparc hardware is supported with GCC and Sun compiler.</li>
73
74    <li>PowerPC hardware is supported with GCC and CodeWarrior, when
75    floating-point computations are not done with the Altivec unit.</li>
76
77    <li>Alpha hardware is supported with GCC, except maybe for the square
78    root. The options <code>-mfp-rounding-mode=d -mieee</code> have to be
79    used.</li>
80  </ul>
81
82  <p>The previous list is not exhaustive. And even if a system does not
83  provide guaranteed computations for hardware floating-point types, the
84  interval library is still usable with user-defined types and for doing box
85  arithmetic.</p>
86
87  <h3>Interval Arithmetic</h3>
88
89  <p>An interval is a pair of numbers which represents all the numbers
90  between these two. (Intervals are considered close so the bounds are
91  included.) The purpose of this library is to extend the usual arithmetic
92  functions to intervals. These intervals will be written [<i>a</i>,<i>b</i>]
93  to represent all the numbers between <i>a</i> and <i>b</i> (included).
94  <i>a</i> and <i>b</i> can be infinite (but they can not be the same
95  infinite) and <i>a</i> &le; <i>b</i>.</p>
96
97  <p>The fundamental property of interval arithmetic is the
98  <em><strong>inclusion property</strong></em>:</p>
99
100  <dl>
101    <dd>``if <i>f</i> is a function on a set of numbers, <i>f</i> can be
102    extended to a new function defined on intervals. This new function
103    <i>f</i> takes one interval argument and returns an interval result such
104    as: &forall; <i>x</i> &isin; [<i>a</i>,<i>b</i>], <i>f</i>(<i>x</i>)
105    &isin; <i>f</i>([<i>a</i>,<i>b</i>]).''</dd>
106  </dl>
107
108  <p>Such a property is not limited to functions with only one argument.
109  Whenever possible, the interval result should be the smallest one able to
110  satisfy the property (it is not really useful if the new functions always
111  answer [-&infin;,+&infin;]).</p>
112
113  <p>There are at least two reasons a user would like to use this library.
114  The obvious one is when the user has to compute with intervals. One example
115  is when input data have some builtin imprecision: instead of a number, an
116  input variable can be passed as an interval. Another example application is
117  to solve equations, by bisecting an interval until the interval width is
118  small enough. A third example application is in computer graphics, where
119  computations with boxes, segments or rays can be reduced to computations
120  with points via intervals.</p>
121
122  <p>Another common reason to use interval arithmetic is when the computer
123  doesn't produce exact results: by using intervals, it is possible to
124  quantify the propagation of rounding errors. This approach is used often in
125  numerical computation. For example, let's assume the computer stores
126  numbers with ten decimal significant digits. To the question 1 + 1E-100 -
127  1, the computer will answer 0 although the correct answer would be 1E-100.
128  With the help of interval arithmetic, the computer will answer [0,1E-9].
129  This is quite a huge interval for such a little result, but the precision
130  is now known, without having to compute error propagation.</p>
131
132  <h3>Numbers, rounding, and exceptional behavior</h3>
133
134  <p>The <em><strong>base number type</strong></em> is the type that holds
135  the bounds of the interval. In order to successfully use interval
136  arithmetic, the base number type must present some <a href=
137  "rounding.htm">characteristics</a>. Firstly, due to the definition of an
138  interval, the base numbers have to be totally ordered so, for instance,
139  <code>complex&lt;T&gt;</code> is not usable as base number type for
140  intervals. The mathematical functions for the base number type should also
141  be compatible with the total order (for instance if x&gt;y and z&gt;t, then
142  it should also hold that x+z &gt; y+t), so modulo types are not usable
143  either.</p>
144
145  <p>Secondly, the computations must be exact or provide some rounding
146  methods (for instance, toward minus or plus infinity) if we want to
147  guarantee the inclusion property. Note that we also may explicitely specify
148  no rounding, for instance if the base number type is exact, i.e. the result
149  of a mathematic operation is always computed and represented without loss
150  of precision. If the number type is not exact, we may still explicitely
151  specify no rounding, with the obvious consequence that the inclusion
152  property is no longuer guaranteed.</p>
153
154  <p>Finally, because heavy loss of precision is always possible, some
155  numbers have to represent infinities or an exceptional behavior must be
156  defined. The same situation also occurs for NaN (<i>Not a Number</i>).</p>
157
158  <p>Given all this, one may want to limit the template argument T of the
159  class template <code>interval</code> to the floating point types
160  <code>float</code>, <code>double</code>, and <code>long double</code>, as
161  defined by the IEEE-754 Standard. Indeed, if the interval arithmetic is
162  intended to replace the arithmetic provided by the floating point unit of a
163  processor, these types are the best choice. Unlike
164  <code>std::complex</code>, however, we don't want to limit <code>T</code>
165  to these types. This is why we allow the rounding and exceptional behaviors
166  to be given by the two policies (rounding and checking). We do nevertheless
167  provide highly optimized rounding and checking class specializations for
168  the above-mentioned floating point types.</p>
169
170  <h3>Operations and functions</h3>
171
172  <p>It is straightforward to define the elementary arithmetic operations on
173  intervals, being guided by the inclusion property. For instance, if [a,b]
174  and [c,d] are intervals, [a,b]+[c,d] can be computed by taking the smallest
175  interval that contains all the numbers x+y for x in [a,b] and y in [c,d];
176  in this case, rounding a+c down and b+d up will suffice. Other operators
177  and functions are similarly defined (see their definitions below).</p>
178
179  <h3>Comparisons</h3>
180
181  <p>It is also possible to define some comparison operators. Given two
182  intervals, the result is a tri-state boolean type
183  {<i>false</i>,<i>true,indeterminate</i>}. The answers <i>false</i> and
184  <i>true</i> are easy to manipulate since they can directly be mapped on the
185  boolean <i>true</i> and <i>false</i>. But it is not the case for the answer
186  <em>indeterminate</em> since comparison operators are supposed to be
187  boolean functions. So, what to do in order to obtain boolean answers?</p>
188
189  <p>One solution consists of deciding to adopt an exceptional behavior, such
190  as a failed assertion or raising an exception. In this case, the
191  exceptional behavior will be triggered when the result is
192  indeterminate.</p>
193
194  <p>Another solution is to map <em>indeterminate</em> always to
195  <i>false,</i> or always to <i>true</i>. If <i>false</i> is chosen, the
196  comparison will be called "<i>certain</i>;" indeed, the result of
197  [<i>a</i>,<i>b</i>] &lt; [<i>c</i>,<i>d</i>] will be <i>true</i> if and
198  only if: &forall; <i>x</i> &isin; [<i>a</i>,<i>b</i>] &forall; <i>y</i>
199  &isin; [<i>c</i>,<i>d</i>], <i>x</i> &lt; <i>y</i>. If <i>true</i> is
200  chosen, the comparison will be called "<i>possible</i>;" indeed, the result
201  of [<i>a</i>,<i>b</i>] &lt; [<i>c</i>,<i>d</i>] will be <i>true</i> if and
202  only if: &exist; <i>x</i> &isin; [<i>a</i>,<i>b</i>] &exist; <i>y</i>
203  &isin; [<i>c</i>,<i>d</i>], <i>x</i> &lt; <i>y</i>.</p>
204
205  <p>Since any of these solution has a clearly defined semantics, it is not
206  clear that we should enforce either of them. For this reason, the default
207  behavior consists to mimic the real comparisons by throwing an exception in
208  the indeterminate case. Other behaviors can be selected bu using specific
209  comparison namespace. There is also a bunch of explicitely named comparison
210  functions. See <a href="comparisons.htm">comparisons</a> pages for further
211  details.</p>
212
213  <h3>Overview of the library, and usage</h3>
214
215  <p>This library provides two quite distinct levels of usage. One is to use
216  the basic class template <code>interval&lt;T&gt;</code> without specifying
217  the policy. This only requires to know and understand the concepts
218  developed above and the content of the namespace boost. In addition to the
219  class <code>interval&lt;T&gt;</code>, this level of usage provides
220  arithmetic operators (<code>+</code>, <code>-</code>, <code>*</code>,
221  <code>/</code>), algebraic and piecewise-algebraic functions
222  (<code>abs</code>, <code>square</code>, <code>sqrt</code>,
223  <code>pow</code>), transcendental and trigonometric functions
224  (<code>exp</code>, <code>log</code>, <code>sin</code>, <code>cos</code>,
225  <code>tan</code>, <code>asin</code>, <code>acos</code>, <code>atan</code>,
226  <code>sinh</code>, <code>cosh</code>, <code>tanh</code>,
227  <code>asinh</code>, <code>acosh</code>, <code>atanh</code>), and the
228  standard comparison operators (<code>&lt;</code>, <code>&lt;=</code>,
229  <code>&gt;</code>, <code>&gt;=</code>, <code>==</code>, <code>!=</code>),
230  as well as several interval-specific functions (<code>min</code>,
231  <code>max</code>, which have a different meaning than <code>std::min</code>
232  and <code>std::max</code>; <code>lower</code>, <code>upper</code>,
233  <code>width</code>, <code>median</code>, <code>empty</code>,
234  <code>singleton</code>, <code>equal</code>, <code>in</code>,
235  <code>zero_in</code>, <code>subset</code>, <code>proper_subset</code>,
236  <code>overlap</code>, <code>intersection</code>, <code>hull</code>,
237  <code>bisect</code>).</p>
238
239  <p>For some functions which take several parameters of type
240  <code>interval&lt;T&gt;</code>, all combinations of argument types
241  <code>T</code> and <code>interval&lt;T&gt;</code> which contain at least
242  one <code>interval&lt;T&gt;</code>, are considered in order to avoid a
243  conversion from the arguments of type <code>T</code> to a singleton of type
244  <code>interval&lt;T&gt;</code>. This is done for efficiency reasons (the
245  fact that an argument is a singleton sometimes renders some tests
246  unnecessary).</p>
247
248  <p>A somewhat more advanced usage of this library is to hand-pick the
249  policies <code>Rounding</code> and <code>Checking</code> and pass them to
250  <code>interval&lt;T, Policies&gt;</code> through the use of <code>Policies
251  := boost::numeric::interval_lib::policies&lt;Rounding,Checking&gt;</code>.
252  Appropriate policies can be fabricated by using the various classes
253  provided in the namespace <code>boost::numeric::interval_lib</code> as
254  detailed in section <a href="#interval_lib">Interval Support Library</a>.
255  It is also possible to choose the comparison scheme by overloading
256  operators through namespaces.</p>
257
258  <h2><a name="synopsis" id="synopsis"></a>Synopsis</h2>
259  <pre>
260namespace boost {
261namespace numeric {
262
263namespace interval_lib {
264
265/* this declaration is necessary for the declaration of interval */
266template &lt;class T&gt; struct default_policies;
267
268/* ... ; the full synopsis of namespace interval_lib can be found <a href=
269"#interval_lib">here</a> */
270
271} // namespace interval_lib
272
273
274/* template interval_policies; class definition can be found <a href=
275"policies.htm">here</a> */
276template&lt;class Rounding, class Checking&gt;
277struct interval_policies;
278
279/* template class interval; class definition can be found <a href=
280"#interval">here</a> */
281template&lt;class T, class Policies = typename interval_lib::default_policies&lt;T&gt;::type &gt; class interval;
282
283/* arithmetic operators involving intervals */
284template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator+(const interval&lt;T, Policies&gt;&amp; x);
285template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator-(const interval&lt;T, Policies&gt;&amp; x);
286
287template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator+(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
288template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator+(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
289template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator+(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
290
291template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator-(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
292template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator-(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
293template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator-(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
294
295template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator*(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
296template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator*(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
297template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator*(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
298
299template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator/(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
300template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator/(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
301template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; operator/(const T&amp; r, const interval&lt;T, Policies&gt;&amp; x);
302
303/* algebraic functions: sqrt, abs, square, pow, root */
304template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; abs(const interval&lt;T, Policies&gt;&amp; x);
305template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; sqrt(const interval&lt;T, Policies&gt;&amp; x);
306template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; square(const interval&lt;T, Policies&gt;&amp; x);
307template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; pow(const interval&lt;T, Policies&gt;&amp; x, int y);
308template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; root(const interval&lt;T, Policies&gt;&amp; x, int y);
309
310/* transcendental functions: exp, log */
311template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; exp(const interval&lt;T, Policies&gt;&amp; x);
312template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; log(const interval&lt;T, Policies&gt;&amp; x);
313
314/* fmod, for trigonometric function argument reduction (see below) */
315template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; fmod(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
316template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; fmod(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
317template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; fmod(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
318
319/* trigonometric functions */
320template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; sin(const interval&lt;T, Policies&gt;&amp; x);
321template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; cos(const interval&lt;T, Policies&gt;&amp; x);
322template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; tan(const interval&lt;T, Policies&gt;&amp; x);
323template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; asin(const interval&lt;T, Policies&gt;&amp; x);
324template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; acos(const interval&lt;T, Policies&gt;&amp; x);
325template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; atan(const interval&lt;T, Policies&gt;&amp; x);
326
327/* hyperbolic trigonometric functions */
328template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; sinh(const interval&lt;T, Policies&gt;&amp; x);
329template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; cosh(const interval&lt;T, Policies&gt;&amp; x);
330template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; tanh(const interval&lt;T, Policies&gt;&amp; x);
331template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; asinh(const interval&lt;T, Policies&gt;&amp; x);
332template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; acosh(const interval&lt;T, Policies&gt;&amp; x);
333template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; atanh(const interval&lt;T, Policies&gt;&amp; x);
334
335/* min, max external functions (NOT std::min/max, see below) */
336template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; max(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
337template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; max(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
338template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; max(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
339template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; min(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
340template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; min(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
341template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; min(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
342
343/* bounds-related interval functions */
344template &lt;class T, class Policies&gt;  T lower(const interval&lt;T, Policies&gt;&amp; x);
345template &lt;class T, class Policies&gt;  T upper(const interval&lt;T, Policies&gt;&amp; x);
346template &lt;class T, class Policies&gt;  T width(const interval&lt;T, Policies&gt;&amp; x);
347template &lt;class T, class Policies&gt;  T median(const interval&lt;T, Policies&gt;&amp; x);
348template &lt;class T, class Policies&gt;  T norm(const interval&lt;T, Policies&gt;&amp; x);
349
350/* bounds-related interval functions */
351template &lt;class T, class Policies&gt;  bool empty(const interval&lt;T, Policies&gt;&amp; b);
352template &lt;class T, class Policies&gt;  bool singleton(const interval&lt;T, Policies&gt;&amp; x);
353template &lt;class T, class Policies&gt;  bool equal(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
354template &lt;class T, class Policies&gt;  bool in(const T&amp; r, const interval&lt;T, Policies&gt;&amp; b);
355template &lt;class T, class Policies&gt;  bool zero_in(const interval&lt;T, Policies&gt;&amp; b);
356template &lt;class T, class Policies&gt;  bool subset(const interval&lt;T, Policies&gt;&amp; a, const interval&lt;T, Policies&gt;&amp; b);
357template &lt;class T, class Policies&gt;  bool proper_subset(const interval&lt;T, Policies&gt;&amp; a, const interval&lt;T, Policies&gt;&amp; b);
358template &lt;class T, class Policies&gt;  bool overlap(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
359
360/* set manipulation interval functions */
361template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; intersection(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
362template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; hull(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
363template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; hull(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
364template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; hull(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
365template &lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; hull(const T&amp; x, const T&amp; y);
366template &lt;class T, class Policies&gt;  std::pair&lt;interval&lt;T, Policies&gt;, interval&lt;T, Policies&gt; &gt; bisect(const interval&lt;T, Policies&gt;&amp; x);
367
368/* interval comparison operators */
369template&lt;class T, class Policies&gt;  bool operator&lt;(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
370template&lt;class T, class Policies&gt;  bool operator&lt;(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
371template&lt;class T, class Policies&gt;  bool operator&lt;(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
372
373template&lt;class T, class Policies&gt;  bool operator&lt;=(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
374template&lt;class T, class Policies&gt;  bool operator&lt;=(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
375template&lt;class T, class Policies&gt;  bool operator&lt;=(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
376
377template&lt;class T, class Policies&gt;  bool operator&gt;(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
378template&lt;class T, class Policies&gt;  bool operator&gt;(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
379template&lt;class T, class Policies&gt;  bool operator&gt;(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
380
381template&lt;class T, class Policies&gt;  bool operator&gt;=(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
382template&lt;class T, class Policies&gt;  bool operator&gt;=(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
383template&lt;class T, class Policies&gt;  bool operator&gt;=(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
384</pre>
385  <pre>
386template&lt;class T, class Policies&gt;  bool operator==(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
387template&lt;class T, class Policies&gt;  bool operator==(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
388template&lt;class T, class Policies&gt;  bool operator==(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
389
390template&lt;class T, class Policies&gt;  bool operator!=(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
391template&lt;class T, class Policies&gt;  bool operator!=(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
392template&lt;class T, class Policies&gt;  bool operator!=(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
393
394namespace interval_lib {
395
396template&lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; division_part1(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&amp; y, bool&amp; b);
397template&lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; division_part2(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&amp; y, bool b = true);
398template&lt;class T, class Policies&gt;  interval&lt;T, Policies&gt; multiplicative_inverse(const interval&lt;T, Policies&gt;&amp; x);
399
400template&lt;class I&gt;  I add(const typename I::base_type&amp; x, const typename I::base_type&amp; y);
401template&lt;class I&gt;  I sub(const typename I::base_type&amp; x, const typename I::base_type&amp; y);
402template&lt;class I&gt;  I mul(const typename I::base_type&amp; x, const typename I::base_type&amp; y);
403template&lt;class I&gt;  I div(const typename I::base_type&amp; x, const typename I::base_type&amp; y);
404
405} // namespace interval_lib
406
407} // namespace numeric
408} // namespace boost
409</pre>
410
411  <h2><a name="interval" id="interval"></a>Template class
412  <code>interval</code></h2>The public interface of the template class
413  interval itself is kept at a simplest minimum:
414  <pre>
415template &lt;class T, class Policies = typename interval_lib::default_policies&lt;T&gt;::type&gt;
416class interval
417{
418  public:
419  typedef T base_type;
420  typedef Policies traits_type;
421
422  interval();
423  interval(T const &amp;v);
424  template&lt;class T1&gt; interval(T1 const &amp;v);
425  interval(T const &amp;l, T const &amp;u);
426  template&lt;class T1, class T2&gt; interval(T1 const &amp;l, T2 const &amp;u);
427  interval(interval&lt;T, Policies&gt; const &amp;r);
428  template&lt;class Policies1&gt; interval(interval&lt;T, Policies1&gt; const &amp;r);
429  template&lt;class T1, class Policies1&gt; interval(interval&lt;T1, Policies1&gt; const &amp;r);
430
431  interval &amp;operator=(T const &amp;v);
432  template&lt;class T1&gt; interval &amp;operator=(T1 const &amp;v);
433  interval &amp;operator=(interval&lt;T, Policies&gt; const &amp;r);
434  template&lt;class Policies1&gt; interval &amp;operator=(interval&lt;T, Policies1&gt; const &amp;r);
435  template&lt;class T1, class Policies1&gt; interval &amp;operator=(interval&lt;T1, Policies1&gt; const &amp;r);
436
437  void assign(T const &amp;l, T const &amp;u);
438
439  T const &amp;lower() const;
440  T const &amp;upper() const;
441
442  static interval empty();
443  static interval whole();
444  static interval hull(T const &amp;x, T const &amp;y);
445
446  interval&amp; operator+= (T const &amp;r);
447  interval&amp; operator-= (T const &amp;r);
448  interval&amp; operator*= (T const &amp;r);
449  interval&amp; operator/= (T const &amp;r);
450  interval&amp; operator+= (interval const &amp;r);
451  interval&amp; operator-= (interval const &amp;r);
452  interval&amp; operator*= (interval const &amp;r);
453  interval&amp; operator/= (interval const &amp;r);
454};
455</pre>
456
457  <p>The constructors create an interval enclosing their arguments. If there
458  are two arguments, the first one is assumed to be the left bound and the
459  second one is the right bound. Consequently, the arguments need to be
460  ordered. If the property !(l &lt;= u) is not respected, the checking policy
461  will be used to create an empty interval. If no argument is given, the
462  created interval is the singleton zero.</p>
463
464  <p>If the type of the arguments is the same as the base number type, the
465  values are directly used for the bounds. If it is not the same type, the
466  library will use the rounding policy in order to convert the arguments
467  (<code>conv_down</code> and <code>conv_up</code>) and create an enclosing
468  interval. When the argument is an interval with a different policy, the
469  input interval is checked in order to correctly propagate its emptiness (if
470  empty).</p>
471
472  <p>The assignment operators behave similarly, except they obviously take
473  one argument only. There is also an <code>assign</code> function in order
474  to directly change the bounds of an interval. It behaves like the
475  two-arguments constructors if the bounds are not ordered. There is no
476  assign function that directly takes an interval or only one number as a
477  parameter; just use the assignment operators in such a case.</p>
478
479  <p>The type of the bounds and the policies of the interval type define the
480  set of values the intervals contain. E.g. with the default policies,
481  intervals are subsets of the set of real numbers. The static functions
482  <code>empty</code> and <code>whole</code> produce the intervals/subsets
483  that are repectively the empty subset and the whole set. They are static
484  member functions rather than global functions because they cannot guess
485  their return types. Likewise for <code>hull</code>. <code>empty</code> and
486  <code>whole</code> involve the checking policy in order to get the bounds
487  of the resulting intervals.</p>
488
489  <h2><a name="opers" id="opers"></a>Operations and Functions</h2>
490
491  <p>Some of the following functions expect <code>min</code> and
492  <code>max</code> to be defined for the base type. Those are the only
493  requirements for the <code>interval</code> class (but the policies can have
494  other requirements).</p>
495
496  <h4>Operators <code>+</code> <code>-</code> <code>*</code> <code>/</code>
497  <code>+=</code> <code>-=</code> <code>*=</code> <code>/=</code></h4>
498
499  <p>The basic operations are the unary minus and the binary <code>+</code>
500  <code>-</code> <code>*</code> <code>/</code>. The unary minus takes an
501  interval and returns an interval. The binary operations take two intervals,
502  or one interval and a number, and return an interval. If an argument is a
503  number instead of an interval, you can expect the result to be the same as
504  if the number was first converted to an interval. This property will be
505  true for all the following functions and operators.</p>
506
507  <p>There are also some assignment operators <code>+=</code> <code>-=</code>
508  <code>*=</code> <code>/=</code>. There is not much to say: <code>x op=
509  y</code> is equivalent to <code>x = x op y</code>. If an exception is
510  thrown during the computations, the l-value is not modified (but it may be
511  corrupt if an exception is thrown by the base type during an
512  assignment).</p>
513
514  <p>The operators <code>/</code> and <code>/=</code> will try to produce an
515  empty interval if the denominator is exactly zero. If the denominator
516  contains zero (but not only zero), the result will be the smallest interval
517  containing the set of division results; so one of its bound will be
518  infinite, but it may not be the whole interval.</p>
519
520  <h4><code>lower</code> <code>upper</code> <code>median</code>
521  <code>width</code> <code>norm</code></h4>
522
523  <p><code>lower</code>, <code>upper</code>, <code>median</code> respectively
524  compute the lower bound, the upper bound, and the median number of an
525  interval (<code>(lower+upper)/2</code> rounded to nearest).
526  <code>width</code> computes the width of an interval
527  (<code>upper-lower</code> rounded toward plus infinity). <code>norm</code>
528  computes an upper bound of the interval in absolute value; it is a
529  mathematical norm (hence the name) similar to the absolute value for real
530  numbers.</p>
531
532  <h4><code>min</code> <code>max</code> <code>abs</code> <code>square</code>
533  <code>pow</code> <code>root</code> <code>division_part?</code>
534  <code>multiplicative_inverse</code></h4>
535
536  <p>The functions <code>min</code>, <code>max</code> and <code>abs</code>
537  are also defined. Please do not mistake them for the functions defined in
538  the standard library (aka <code>a&lt;b?a:b</code>, <code>a&gt;b?a:b</code>,
539  <code>a&lt;0?-a:a</code>). These functions are compatible with the
540  elementary property of interval arithmetic. For example,
541  max([<i>a</i>,<i>b</i>], [<i>c</i>,<i>d</i>]) = {max(<i>x</i>,<i>y</i>)
542  such that <i>x</i> in [<i>a</i>,<i>b</i>] and <i>y</i> in
543  [<i>c</i>,<i>d</i>]}. They are not defined in the <code>std</code>
544  namespace but in the boost namespace in order to avoid conflict with the
545  other definitions.</p>
546
547  <p>The <code>square</code> function is quite particular. As you can expect
548  from its name, it computes the square of its argument. The reason this
549  function is provided is: <code>square(x)</code> is not <code>x*x</code> but
550  only a subset when <code>x</code> contains zero. For example, [-2,2]*[-2,2]
551  = [-4,4] but [-2,2]&sup2; = [0,4]; the result is a smaller interval.
552  Consequently, <code>square(x)</code> should be used instead of
553  <code>x*x</code> because of its better accuracy and a small performance
554  improvement.</p>
555
556  <p>As for <code>square</code>, <code>pow</code> provides an efficient and
557  more accurate way to compute the integer power of an interval. Please note:
558  when the power is 0 and the interval is not empty, the result is 1, even if
559  the input interval contains 0. <code>root</code> computes the integer root
560  of an interval (<code>root(pow(x,k),k)</code> encloses <code>x</code> or
561  <code>abs(x)</code> depending on whether <code>k</code> is odd or even).
562  The behavior of <code>root</code> is not defined if the integer argument is
563  not positive. <code>multiplicative_inverse</code> computes
564  <code>1/x</code>.</p>
565
566  <p>The functions <code>division_part1</code> and
567  <code>division_part2</code> are useful when the user expects the division
568  to return disjoint intervals if necessary. For example, the narrowest
569  closed set containg [2,3] / [-2,1] is not ]-&infin;,&infin;[ but the union
570  of ]-&infin;,-1] and [2,&infin;[. When the result of the division is
571  representable by only one interval, <code>division_part1</code> returns
572  this interval and sets the boolean reference to <code>false</code>.
573  However, if the result needs two intervals, <code>division_part1</code>
574  returns the negative part and sets the boolean reference to
575  <code>true</code>; a call to <code>division_part2</code> is now needed to
576  get the positive part. This second function can take the boolean returned
577  by the first function as last argument. If this bool is not given, its
578  value is assumed to be true and the behavior of the function is then
579  undetermined if the division does not produce a second interval.</p>
580
581  <h4><code>intersect</code> <code>hull</code> <code>overlap</code>
582  <code>in</code> <code>zero_in</code> <code>subset</code>
583  <code>proper_subset</code> <code>empty</code> <code>singleton</code>
584  <code>equal</code></h4>
585
586  <p><code>intersect</code> computes the set intersection of two closed sets,
587  <code>hull</code> computes the smallest interval which contains the two
588  parameters; those parameters can be numbers or intervals. If one of the
589  arguments is an invalid number or an empty interval, the function will only
590  use the other argument to compute the resulting interval (if allowed by the
591  checking policy).</p>
592
593  <p>There is no union function since the union of two intervals is not an
594  interval if they do not overlap. If they overlap, the <code>hull</code>
595  function computes the union.</p>
596
597  <p>The function <code>overlap</code> tests if two intervals have some
598  common subset. <code>in</code> tests if a number is in an interval;
599  <code>zero_in</code> is a variant which tests if zero is in the interval.
600  <code>subset</code> tests if the first interval is a subset of the second;
601  and <code>proper_subset</code> tests if it is a proper subset.
602  <code>empty</code> and <code>singleton</code> test if an interval is empty
603  or is a singleton. Finally, <code>equal</code> tests if two intervals are
604  equal.</p>
605
606  <h4><code>sqrt</code> <code>log</code> <code>exp</code> <code>sin</code>
607  <code>cos</code> <code>tan</code> <code>asin</code> <code>acos</code>
608  <code>atan</code> <code>sinh</code> <code>cosh</code> <code>tanh</code>
609  <code>asinh</code> <code>acosh</code> <code>atanh</code>
610  <code>fmod</code></h4>
611
612  <p>The functions <code>sqrt</code>, <code>log</code>, <code>exp</code>,
613  <code>sin</code>, <code>cos</code>, <code>tan</code>, <code>asin</code>,
614  <code>acos</code>, <code>atan</code>, <code>sinh</code>, <code>cosh</code>,
615  <code>tanh</code>, <code>asinh</code>, <code>acosh</code>,
616  <code>atanh</code> are also defined. There is not much to say; these
617  functions extend the traditional functions to the intervals and respect the
618  basic property of interval arithmetic. They use the <a href=
619  "checking.htm">checking</a> policy to produce empty intervals when the
620  input interval is strictly outside of the domain of the function.</p>
621
622  <p>The function <code>fmod(interval x, interval y)</code> expects the lower
623  bound of <code>y</code> to be strictly positive and returns an interval
624  <code>z</code> such as <code>0 &lt;= z.lower() &lt; y.upper()</code> and
625  such as <code>z</code> is a superset of <code>x-n*y</code> (with
626  <code>n</code> being an integer). So, if the two arguments are positive
627  singletons, this function <code>fmod(interval, interval)</code> will behave
628  like the traditional function <code>fmod(double, double)</code>.</p>
629
630  <p>Please note that <code>fmod</code> does not respect the inclusion
631  property of arithmetic interval. For example, the result of
632  <code>fmod</code>([13,17],[7,8]) should be [0,8] (since it must contain
633  [0,3] and [5,8]). But this answer is not really useful when the purpose is
634  to restrict an interval in order to compute a periodic function. It is the
635  reason why <code>fmod</code> will answer [5,10].</p>
636
637  <h4><code>add</code> <code>sub</code> <code>mul</code>
638  <code>div</code></h4>
639
640  <p>These four functions take two numbers and return the enclosing interval
641  for the operations. It avoids converting a number to an interval before an
642  operation, it can result in a better code with poor optimizers.</p>
643
644  <h3>Constants</h3>
645
646  <p>Some constants are hidden in the
647  <code>boost::numeric::interval_lib</code> namespace. They need to be
648  explicitely templated by the interval type. The functions are
649  <code>pi&lt;I&gt;()</code>, <code>pi_half&lt;I&gt;()</code> and
650  <code>pi_twice&lt;I&gt;()</code>, and they return an object of interval
651  type <code>I</code>. Their respective values are &pi;, &pi;/2 and
652  2&pi;.</p>
653
654  <h3>Exception throwing</h3>
655
656  <p>The interval class and all the functions defined around this class never
657  throw any exceptions by themselves. However, it does not mean that an
658  operation will never throw an exception. For example, let's consider the
659  copy constructor. As explained before, it is the default copy constructor
660  generated by the compiler. So it will not throw an exception if the copy
661  constructor of the base type does not throw an exception.</p>
662
663  <p>The same situation applies to all the functions: exceptions will only be
664  thrown if the base type or one of the two policies throws an exception.</p>
665
666  <h2 id="interval_lib">Interval Support Library</h2>
667
668  <p>The interval support library consists of a collection of classes that
669  can be used and combined to fabricate almost various commonly-needed
670  interval policies. In contrast to the basic classes and functions which are
671  used in conjunction with <code>interval&lt;T&gt;</code> (and the default
672  policies as the implicit second template parameter in this type), which
673  belong simply to the namespace <code>boost</code>, these components belong
674  to the namespace <code>boost::numeric::interval_lib</code>.</p>
675
676  <p>We merely give the synopsis here and defer each section to a separate
677  web page since it is only intended for the advanced user. This allows to
678  expand on each topic with examples, without unduly stretching the limits of
679  this document.</p>
680
681  <h4>Synopsis</h4>
682  <pre>
683namespace boost {
684namespace numeric {
685namespace interval_lib {
686
687<span style=
688"color: #FF0000">/* built-in rounding policy and its specializations */</span>
689template &lt;class T&gt;  struct rounded_math;
690template &lt;&gt;         struct rounded_math&lt;float&gt;;
691template &lt;&gt;         struct rounded_math&lt;double&gt;;
692template &lt;&gt;         struct rounded_math&lt;long double&gt;;
693
694<span style=
695"color: #FF0000">/* built-in rounding construction blocks */</span>
696template &lt;class T&gt;  struct rounding_control;
697
698template &lt;class T, class Rounding = rounding_control&lt;T&gt; &gt;  struct rounded_arith_exact;
699template &lt;class T, class Rounding = rounding_control&lt;T&gt; &gt;  struct rounded_arith_std;
700template &lt;class T, class Rounding = rounding_control&lt;T&gt; &gt;  struct rounded_arith_opp;
701
702template &lt;class T, class Rounding&gt;  struct rounded_transc_dummy;
703template &lt;class T, class Rounding = rounded_arith_exact&lt;T&gt; &gt;  struct rounded_transc_exact;
704template &lt;class T, class Rounding = rounded_arith_std  &lt;T&gt; &gt;  struct rounded_transc_std;
705template &lt;class T, class Rounding = rounded_arith_opp  &lt;T&gt; &gt;  struct rounded_transc_opp;
706
707template &lt;class Rounding&gt; struct save_state;
708template &lt;class Rounding&gt; struct save_state_nothing;
709
710<span style="color: #FF0000">/* built-in checking policies */</span>
711template &lt;class T&gt; struct checking_base;
712template &lt;class T, class Checking = checking_base&lt;T&gt;, class Exception = exception_create_empty&gt;   struct checking_no_empty;
713template &lt;class T, class Checking = checking_base&lt;T&gt; &gt;                                            struct checking_no_nan;
714template &lt;class T, class Checking = checking_base&lt;T&gt;, class Exception = exception_invalid_number&gt; struct checking_catch_nan;
715template &lt;class T&gt; struct checking_strict;
716
717<span style=
718"color: #FF0000">/* some metaprogramming to manipulate interval policies */</span>
719template &lt;class Rounding, class Checking&gt; struct policies;
720template &lt;class OldInterval, class NewRounding&gt; struct change_rounding;
721template &lt;class OldInterval, class NewChecking&gt; struct change_checking;
722template &lt;class OldInterval&gt; struct unprotect;
723
724<span style=
725"color: #FF0000">/* constants, need to be explicitly templated */</span>
726template&lt;class I&gt; I pi();
727template&lt;class I&gt; I pi_half();
728template&lt;class I&gt; I pi_twice();
729
730<span style="color: #FF0000">/* interval explicit comparison functions:
731 * the mode can be cer=certainly or pos=possibly,
732 * the function lt=less_than, gt=greater_than, le=less_than_or_equal_to, ge=greater_than_or_equal_to
733 *   eq=equal_to, ne= not_equal_to */</span>
734template &lt;class T, class Policies&gt;  bool cerlt(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
735template &lt;class T, class Policies&gt;  bool cerlt(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
736template &lt;class T, class Policies&gt;  bool cerlt(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
737
738template &lt;class T, class Policies&gt;  bool cerle(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
739template &lt;class T, class Policies&gt;  bool cerle(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
740template &lt;class T, class Policies&gt;  bool cerle(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
741
742template &lt;class T, class Policies&gt;  bool cergt(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
743template &lt;class T, class Policies&gt;  bool cergt(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
744template &lt;class T, class Policies&gt;  bool cergt(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
745
746template &lt;class T, class Policies&gt;  bool cerge(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
747template &lt;class T, class Policies&gt;  bool cerge(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
748template &lt;class T, class Policies&gt;  bool cerge(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
749
750template &lt;class T, class Policies&gt;  bool cereq(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
751template &lt;class T, class Policies&gt;  bool cereq(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
752template &lt;class T, class Policies&gt;  bool cereq(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
753
754template &lt;class T, class Policies&gt;  bool cerne(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
755template &lt;class T, class Policies&gt;  bool cerne(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
756template &lt;class T, class Policies&gt;  bool cerne(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
757
758template &lt;class T, class Policies&gt;  bool poslt(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
759template &lt;class T, class Policies&gt;  bool poslt(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
760template &lt;class T, class Policies&gt;  bool poslt(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
761
762template &lt;class T, class Policies&gt;  bool posle(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
763template &lt;class T, class Policies&gt;  bool posle(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
764template &lt;class T, class Policies&gt;  bool posle(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
765
766template &lt;class T, class Policies&gt;  bool posgt(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
767template &lt;class T, class Policies&gt;  bool posgt(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
768template &lt;class T, class Policies&gt;  bool posgt(const T&amp; x, const interval&lt;T, Policies&gt; &amp; y);
769
770template &lt;class T, class Policies&gt;  bool posge(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
771template &lt;class T, class Policies&gt;  bool posge(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
772template &lt;class T, class Policies&gt;  bool posge(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
773
774template &lt;class T, class Policies&gt;  bool poseq(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
775template &lt;class T, class Policies&gt;  bool poseq(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
776template &lt;class T, class Policies&gt;  bool poseq(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
777
778template &lt;class T, class Policies&gt;  bool posne(const interval&lt;T, Policies&gt;&amp; x, const interval&lt;T, Policies&gt;&amp; y);
779template &lt;class T, class Policies&gt;  bool posne(const interval&lt;T, Policies&gt;&amp; x, const T&amp; y);
780template &lt;class T, class Policies&gt;  bool posne(const T&amp; x, const interval&lt;T, Policies&gt;&amp; y);
781
782<span style="color: #FF0000">/* comparison namespaces */</span>
783namespace compare {
784  namespace certain;
785  namespace possible;
786  namespace lexicographic;
787  namespace set;
788  namespace tribool;
789} // namespace compare
790
791} // namespace interval_lib
792} // namespace numeric
793} // namespace boost
794</pre>
795
796  <p>Each component of the interval support library is detailed in its own
797  page.</p>
798
799  <ul>
800    <li><a href="comparisons.htm">Comparisons</a></li>
801
802    <li><a href="rounding.htm">Rounding</a></li>
803
804    <li><a href="checking.htm">Checking</a></li>
805  </ul>
806
807  <h2 id="dangers">Common Pitfalls and Dangers</h2>
808
809  <h4>Comparisons</h4>
810
811  <p>One of the biggest problems is problably the correct use of the
812  comparison functions and operators. First, functions and operators do not
813  try to know if two intervals are the same mathematical object. So, if the
814  comparison used is "certain", then <code>x != x</code> is always true
815  unless <code>x</code> is a singleton interval; and the same problem arises
816  with <code>cereq</code> and <code>cerne</code>.</p>
817
818  <p>Another misleading interpretation of the comparison is: you cannot
819  always expect [a,b] &lt; [c,d] to be !([a,b] &gt;= [c,d]) since the
820  comparison is not necessarily total. Equality and less comparison should be
821  seen as two distincts relational operators. However the default comparison
822  operators do respect this property since they throw an exception whenever
823  [a,b] and [c,d] overlap.</p>
824
825  <h4>Interval values and references</h4>
826
827  <p>This problem is a corollary of the previous problem with <code>x !=
828  x</code>. All the functions of the library only consider the value of an
829  interval and not the reference of an interval. In particular, you should
830  not expect (unless <code>x</code> is a singleton) the following values to
831  be equal: <code>x/x</code> and 1, <code>x*x</code> and
832  <code>square(x)</code>, <code>x-x</code> and 0, etc. So the main cause of
833  wide intervals is that interval arithmetic does not identify different
834  occurences of the same variable. So, whenever possible, the user has to
835  rewrite the formulas to eliminate multiple occurences of the same variable.
836  For example, <code>square(x)-2*x</code> is far less precise than
837  <code>square(x-1)-1</code>.</p>
838
839  <h4>Unprotected rounding</h4>
840
841  <p>As explained in <a href="rounding.htm#perf">this section</a>, a good way
842  to speed up computations when the base type is a basic floating-point type
843  is to unprotect the intervals at the hot spots of the algorithm. This
844  method is safe and really an improvement for interval computations. But
845  please remember that any basic floating-point operation executed inside the
846  unprotection blocks will probably have an undefined behavior (but only for
847  the current thread). And do not forget to create a rounding object as
848  explained in the <a href="rounding.htm#perfexp">example</a>.</p>
849
850  <h2 id="rationale">Rationale</h2>
851
852  <p>The purpose of this library is to provide an efficient and generalized
853  way to deal with interval arithmetic through the use of a templatized class
854  <code>boost::interval</code>. The big contention for which we provide a
855  rationale is the format of this class template.</p>
856
857  <p>It would have been easier to provide a class interval whose base type is
858  double. Or to follow <code>std::complex</code> and allow only
859  specializations for <code>float</code>, <code>double</code>, and <code>long
860  double</code>. We decided not to do this to allow intervals on custom
861  types, e.g. fixed-precision bigfloat library types (MPFR, etc), rational
862  numbers, and so on.</p>
863
864  <p><strong>Policy design.</strong> Although it was tempting to make it a
865  class template with only one template argument, the diversity of uses for
866  an interval arithmetic practically forced us to use policies. The behavior
867  of this class can be fixed by two policies. These policies are packaged
868  into a single policy class, rather than making <code>interval</code> with
869  three template parameters. This is both for ease of use (the policy class
870  can be picked by default) and for readability.</p>
871
872  <p>The first policy provides all the mathematical functions on the base
873  type needed to define the functions on the interval type. The second one
874  sets the way exceptional cases encountered during computations are
875  handled.</p>
876
877  <p>We could foresee situations where any combination of these policies
878  would be appropriate. Moreover, we wanted to enable the user of the library
879  to reuse the <code>interval</code> class template while at the same time
880  choosing his own behavior. See this <a href="guide.htm">page</a> for some
881  examples.</p>
882
883  <p><strong>Rounding policy.</strong> The library provides specialized
884  implementations of the rounding policy for the primitive types float and
885  double. In order for these implementations to be correct and fast, the
886  library needs to work a lot with rounding modes. Some processors are
887  directly dealt with and some mecanisms are provided in order to speed up
888  the computations. It seems to be heavy and hazardous optimizations for a
889  gain of only a few computer cycles; but in reality, the speed-up factor can
890  easily go past 2 or 3 depending on the computer. Moreover, these
891  optimizations do not impact the interface in any major way (with the design
892  we have chosen, everything can be added by specialization or by passing
893  different template parameters).</p>
894
895  <p><strong>Pred/succ.</strong> In a previous version, two functions
896  <code>pred</code> and <code>succ</code>, with various corollaries like
897  <code>widen</code>, were supplied. The intent was to enlarge the interval
898  by one ulp (as little as possible), e.g. to ensure the inclusion property.
899  Since making interval a template of T, we could not define <i>ulp</i> for a
900  random parameter. In turn, rounding policies let us eliminate entirely the
901  use of ulp while making the intervals tighter (if a result is a
902  representable singleton, there is no use to widen the interval). We decided
903  to drop those functions.</p>
904
905  <p><strong>Specialization of <code>std::less</code>.</strong> Since the
906  operator <code>&lt;</code> depends on the comparison namespace locally
907  chosen by the user, it is not possible to correctly specialize
908  <code>std::less</code>. So you have to explicitely provide such a class to
909  all the algorithms and templates that could require it (for example,
910  <code>std::map</code>).</p>
911
912  <p><strong>Input/output.</strong> The interval library does not include I/O
913  operators. Printing an interval value allows a lot of customization: some
914  people may want to output the bounds, others may want to display the median
915  and the width of intervals, and so on. The example file io.cpp shows some
916  possibilities and may serve as a foundation in order for the user to define
917  her own operators.</p>
918
919  <p><strong>Mixed operations with integers.</strong> When using and reusing
920  template codes, it is common there are operations like <code>2*x</code>.
921  However, the library does not provide them by default because the
922  conversion from <code>int</code> to the base number type is not always
923  correct (think about the conversion from a 32bit integer to a single
924  precision floating-point number). So the functions have been put in a
925  separate header and the user needs to include them explicitely if she wants
926  to benefit from these mixed operators. Another point, there is no mixed
927  comparison operators due to the technical way they are defined.</p>
928
929  <p><strong>Interval-aware functions.</strong> All the functions defined by
930  the library are obviously aware they manipulate intervals and they do it
931  accordingly to general interval arithmetic principles. Consequently they
932  may have a different behavior than the one commonly encountered with
933  functions not interval-aware. For example, <code>max</code> is defined by
934  canonical set extension and the result is not always one of the two
935  arguments (if the intervals do not overlap, then the result is one of the
936  two intervals).</p>
937
938  <p>This behavior is different from <code>std::max</code> which returns a
939  reference on one of its arguments. So if the user expects a reference to be
940  returned, she should use <code>std::max</code> since it is exactly what
941  this function does. Please note that <code>std::max</code> will throw an
942  exception when the intervals overlap. This behavior does not predate the
943  one described by the C++ standard since the arguments are not "equivalent"
944  and it allows to have an equivalence between <code>a &lt;= b</code> and
945  <code>&amp;b == &amp;std::max(a,b)</code>(some particular cases may be
946  implementation-defined). However it is different from the one described by
947  SGI since it does not return the first argument even if "neither is greater
948  than the other".</p>
949
950  <h2 id="acks">History and Acknowledgments</h2>
951
952  <p>This library was mostly inspired by previous work from Jens Maurer. Some
953  discussions about his work are reproduced <a href=
954  "http://www.mscs.mu.edu/%7Egeorgec/IFAQ/maurer1.html">here</a>. Jeremy Siek
955  and Maarten Keijzer provided some rounding control for MSVC and Sparc
956  platforms.</p>
957
958  <p>Guillaume Melquiond, Herv&eacute; Br&ouml;nnimann and Sylvain Pion
959  started from the library left by Jens and added the policy design.
960  Guillaume and Sylvain worked hard on the code, especially the porting and
961  mostly tuning of the rounding modes to the different architectures.
962  Guillaume did most of the coding, while Sylvain and Herv&eacute; have
963  provided some useful comments in order for this library to be written.
964  Herv&eacute; reorganized and wrote chapters of the documentation based on
965  Guillaume's great starting point.</p>
966
967  <p>This material is partly based upon work supported by the National
968  Science Foundation under NSF CAREER Grant CCR-0133599. Any opinions,
969  findings and conclusions or recommendations expressed in this material are
970  those of the author(s) and do not necessarily reflect the views of the
971  National Science Foundation (NSF).</p>
972  <hr>
973
974  <p><a href="http://validator.w3.org/check?uri=referer"><img border="0" src=
975  "http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01 Transitional"
976  height="31" width="88"></a></p>
977
978  <p>Revised
979  <!--webbot bot="Timestamp" s-type="EDITED" s-format="%Y-%m-%d" startspan -->2006-12-25<!--webbot bot="Timestamp" endspan i-checksum="12174" --></p>
980
981  <p><i>Copyright &copy; 2002 Guillaume Melquiond, Sylvain Pion, Herv&eacute;
982  Br&ouml;nnimann, Polytechnic University<br>
983  Copyright &copy; 2003-2006 Guillaume Melquiond, ENS Lyon</i></p>
984
985  <p><i>Distributed under the Boost Software License, Version 1.0. (See
986  accompanying file <a href="../../../../LICENSE_1_0.txt">LICENSE_1_0.txt</a>
987  or copy at <a href=
988  "http://www.boost.org/LICENSE_1_0.txt">http://www.boost.org/LICENSE_1_0.txt</a>)</i></p>
989</body>
990</html>
Note: See TracBrowser for help on using the repository browser.