Planet
navi homePPSaboutscreenshotsdownloaddevelopmentforum

source: downloads/boost_1_34_1/libs/spirit/doc/rationale.html @ 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: 9.9 KB
Line 
1<html>
2<head>
3<title>Rationale</title>
4<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
5<link rel="stylesheet" href="theme/style.css" type="text/css">
6</head>
7
8<body>
9<table width="100%" border="0" background="theme/bkd2.gif" cellspacing="2">
10  <tr> 
11    <td width="10"> 
12    </td>
13    <td width="85%"> <font size="6" face="Verdana, Arial, Helvetica, sans-serif"><b>Rationale</b></font></td>
14    <td width="112"><a href="http://spirit.sf.net"><img src="theme/spirit.gif" width="112" height="48" align="right" border="0"></a></td>
15  </tr>
16</table>
17<br>
18<table border="0">
19  <tr>
20    <td width="10"></td>
21    <td width="30"><a href="../index.html"><img src="theme/u_arr.gif" border="0"></a></td>
22    <td width="30"><a href="faq.html"><img src="theme/l_arr.gif" border="0"></a></td>
23    <td width="30"><a href="acknowledgments.html"><img src="theme/r_arr.gif" border="0"></a></td>
24   </tr>
25</table>
26<p><img src="theme/lens.gif" width="15" height="16"> <strong>Virtual functions:
27  From static to dynamic C++</strong></p>
28<p>Rules straddle the border between static and dynamic C++. In effect, a rule
29  transforms compile-time polymorphism (using templates) into run-time polymorphism
30  (using virtual functions). This is necessary due to C++'s inability to automatically
31  declare a variable of a type deduced from an arbitrarily complex expression
32  in the right-hand side (rhs) of an assignment. Basically, we want to do something
33  like:</p>
34<pre><code><font color="#000000">    <span class=identifier>T </span><span class=identifier>rule </span><span class=special>= </span><span class=identifier>an_arbitrarily_complex_expression</span><span class=special>;</span></font></code></pre>
35<p>without having to know or care about the resulting type of the right-hand side
36  (rhs) of the assignment expression. Apart from this, we also need a facility
37  to forward declare an unknown type:</p>
38<pre><code><font color="#000000"><span class=special>    </span><span class=identifier>T </span><span class=identifier>rule</span><span class=special>;
39    </span><span class=special>...
40    </span><span class=identifier>rule </span><span class=special>= </span><span class=identifier>a </span><span class=special>| </span><span class=identifier>b</span><span class=special>;</span></font></code></pre>
41<p>These limitations lead us to this implementation of rules. This comes at the
42  expense of the overhead of a virtual function call, once through each invocation
43  of a rule.</p>
44<p><img src="theme/lens.gif" width="15" height="16"> <strong>Multiple declaration
45  </strong> </p>
46<p>Some BNF variants allow multiple declarations of a <tt>rule</tt>. The declarations
47  are taken as alternatives. Example:</p>
48<pre>
49   <span class=identifier><code>r </code></span><code><span class=special>= </span><span class=identifier>a</span><span class=special>;    </span><span class=identifier>
50   r </span><span class=special>= </span><span class=identifier>b</span><span class=special>;</span></code></pre>
51<p> is equivalent to: </p>
52<pre>
53   <span class=identifier><code>r </code></span><code><span class=special>= </span><span class=identifier>a </span><span class=special>| </span><span class=identifier>b</span><span class=special>;</span></code></pre>
54<p>Spirit v1.3 allowed this behavior. However, the current version of Spirit <b>no
55  longer</b> allows this because experience shows that this behavior leads to
56  unwanted gotchas (for instance, it does not allow rules to be held in containers).
57  In the current release of Spirit, a second assignment to a rule will simply
58  redefine it. The old definition is destructed. This follows more closely C++
59  semantics and is more in line with what the user expects the rule to behave.</p>
60<p><img src="theme/lens.gif" width="15" height="16"> <b>Sequencing Syntax</b> 
61  <br>
62  <br>
63  The comma operator as in a, b seems to be a better candidate, syntax-wise. But
64  then the problem is with its precedence. It has the lowest precedence in C/C++,
65  which makes it virtually useless. <br>
66  <br>
67  Bjarne Stroustrup, in his article <a href="references.html#generalized_overloading">&quot;Generalizing
68  Overloading for C++2000&quot;</a> talks about overloading whitespace. Such a
69  feature would allow juxtapositioning of parser objects exactly as we do in (E)BNF
70  (e.g. a b | c instead of a &gt;&gt; b | c). Unfortunately, the article was dated
71  April 1, 1998. Oh well.</p>
72<p><img src="theme/lens.gif" width="15" height="16"> <b>Forward iterators</b><br>
73  <br>
74  In general, the scanner expects at least a standard conforming forward iterator.
75  Forward iterators are needed for backtracking where the iterator needs to be
76  saved and restored later. Generally speaking, Spirit is a backtracking parser.
77  The implication of this is that at some point, the iterator position will have
78  to be saved to allow the parser to backtrack to a previous point. Thus, for
79  backtracking to work, the framework requires at least a forward iterator.<br>
80  <br>
81  Some parsers might require more specialized iterators (bi-directional or even
82  random access). Perhaps in the future, deterministic parsers when added to the
83  framework, will perform no backtracking and will need just a single token lookahead,
84  hence will require input iterators only.</p>
85<p><img src="theme/lens.gif" width="15" height="16"><b> Why are subrules important?</b><br>
86  <br>
87  Subrules open up the oportunity to do aggressive meta programming as well because
88  they do not rely on virtual functions. The virtual function is the meta-programmer's
89  hell. Not only does it slow down the program due to the virtual function indirect
90  call, it is also an opaque wall where no metaprogram can get past. It kills
91  all meta-information beyond the virtual function call. Worse, the virtual function
92  cannot be templated. Which means that its arguments have to be tied to a actual
93  types. Many problems stem from this limitation. <br>
94  <br>
95  While Spirit is a currently classified as a non-deterministic recursive-descent
96  parser, Doug Gregor first noted that other parsing techniques apart from top-down
97  recursive descent may be applied. For instance, apart from non-deterministic
98  recursive descent, deterministic LL(1) and LR(1) can theoretically be implemented
99  using the same expression template front end. Spirit rules use virtual functions
100  to encode the RHS parser expression in an opaque abstract parser type. While
101  it serves its purpose well, the rule's virtual functions are the stumbling blocks
102  to more advanced metaprogramming. Subrules are free from virtual functions.</p>
103<p><img src="theme/lens.gif" width="15" height="16"><b> <a name="exhaustive_rd"></a>Exhaustive
104  backtracking and greedy RD</b></p>
105<p>Spirit doesn't do exhaustive backtracking like regular expressions are expected
106  to. For example:</p>
107<pre>    <span class="special">*</span>chlit_p<span class="special">(</span><span class="quotes">'a'</span><span class="special">) &gt;&gt;</span> chlit_p<span class="special">(</span><span class="quotes">'a'</span><span class="special">);</span></pre>
108<p>will always fail to match because Spirit's Kleene star does not back off when
109  the rest of the rule fails to match. </p>
110<p>Actually, there's a solution to this greedy RD problem. Such a scheme is discussed
111  in section 6.6.2 of <a 
112href="http://www.cs.vu.nl/%7Edick/PTAPG.html">Parsing Techniques: A Practical
113  Guide</a>. The trick involves passing a <em>tail</em> parser (in addition to
114  the scanner) to each parser. The start parser will then simply be: <tt>start
115  &gt;&gt; end_p;</tt> (end_p is the start's tail). </p>
116<p>Spirit is greedy --using straight forward, naive RD. It is certainly possible
117  to implement the fully backtracking scheme presented above, but there will be
118  also certainly be a performance hit. The scheme will always try to match all
119  possible parser paths (full parser hierarchy traversal) until it reaches a point
120  of certainty, that the whole thing matches or fails to match. </p>
121<table border="0" width="80%" align="center">
122  <tr> 
123    <td class="note_box"><p><img src="theme/note.gif" width="16" height="16"><b>Backtracking
124        and Greedy RD </b><br>
125        <br>
126        Spirit is quite consistent and intuitive about when it backtracks and
127        to where, although it may not be obvious to those coming from different
128        backgrounds. In general, any (sub)parser will, given the same input, always
129        match the same portion of the input (or fail to match the input at all).
130        This means that Spirit is inherently greedy. Spirit will only backtrack
131        when a (sub)parser fails to match the input, and it will always backtrack
132        to the next choice point upward (not backward) in the parser structure.
133        In other words abb|ab will match &quot;ab&quot;, as will a(bb|b), but
134        (ab|a)b won't because the (ab|a) subparser will always match the 'b' after
135        the 'a' if it is available.</p>
136      <p>--Rainer Deyke</p>
137      </td>
138  </tr>
139</table>
140<p>There's a strong preference on &quot;simplicity with all the knobs when you
141  need them&quot; approach, right now. On the other hand, the flexibility of Spirit
142  makes it possible to have different optional schemes available. It might be
143  possible to implement an exhaustive backtracking RD scheme as an optional feature
144  in the future. </p>
145<table border="0">
146  <tr> 
147    <td width="10"></td>
148    <td width="30"><a href="../index.html"><img src="theme/u_arr.gif" border="0"></a></td>
149    <td width="30"><a href="faq.html"><img src="theme/l_arr.gif" border="0"></a></td>
150    <td width="30"><a href="acknowledgments.html"><img src="theme/r_arr.gif" border="0"></a></td>
151  </tr>
152</table>
153<br>
154<hr size="1">
155<p class="copyright">Copyright &copy; 1998-2003 Joel de Guzman<br>
156  <br>
157  <font size="2">Use, modification and distribution is subject to the Boost Software
158    License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at
159    http://www.boost.org/LICENSE_1_0.txt)</font></p>
160<p class="copyright">&nbsp;</p>
161</body>
162</html>
Note: See TracBrowser for help on using the repository browser.