1<html> 2<head> 3<title>Techniques</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>Techniques</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="style_guide.html"><img src="theme/l_arr.gif" border="0"></a></td> 23 <td width="30"><a href="faq.html"><img src="theme/r_arr.gif" border="0"></a></td> 24 </tr> 25</table> 26<ul> 27 <li><a href="#templatized_functors">Templatized Functors</a></li> 28 <li><a href="#multiple_scanner_support">Rule With Multiple Scanners</a></li> 29 <li><a href="#no_rules">Look Ma' No Rules!</a></li> 30 <li><a href="#typeof">typeof</a></li> 31 <li><a href="#nabialek_trick">Nabialek trick</a></li> 32</ul> 33<h3><a name="templatized_functors"></a> Templatized Functors</h3> 34<p>For the sake of genericity, it is often better to make the functor's member 35 <tt>operator()</tt> a template. That way, we do not have to concern ourselves 36 with the type of the argument to expect as long as the behavior is appropriate. 37 For instance, rather than hard-coding <tt>char const*</tt> as the argument of 38 a generic semantic action, it is better to make it a template member function. 39 That way, it can accept any type of iterator:</p> 40<pre><code><font color="#000000"><span class=special> </span><span class=keyword>struct </span><span class=identifier>my_functor 41 </span><span class=special>{ 42 </span><span class=keyword>template </span><span class=special><</span><span class=keyword>typename </span><span class=identifier>IteratorT</span><span class=special>> 43 </span><span class=keyword>void </span><span class=keyword>operator</span><span class=special>()(</span><span class=identifier>IteratorT </span><span class=identifier>first</span><span class=special>, </span><span class=identifier>IteratorT </span><span class=identifier>last</span><span class=special>) </span><span class=keyword>const</span><span class=special>; 44 </span><span class=special>};</span></font></code></pre> 45<p>Take note that this is only possible with functors. It is not possible to pass 46 in template functions as semantic actions unless you cast it to the correct 47 function signature; in which case, you <em>monomorphize</em> the function. This 48 clearly shows that functors are superior to plain functions.</p> 49<h3><b><a name="multiple_scanner_support" id="multiple_scanner_support"></a> Rule 50 With Multiple Scanners</b></h3> 51<p>As of v1.8.0, rules can use one or more scanner types. There are cases, for 52 instance, where we need a rule that can work on the phrase and character levels. 53 Rule/scanner mismatch has been a source of confusion and is the no. 1 <a href="faq.html#scanner_business">FAQ</a>. 54 To address this issue, we now have <a href="rule.html#multiple_scanner_support">multiple 55 scanner support</a>. </p> 56<p>Here is an example of a grammar with a rule <tt>r</tt> that can be called with 57 3 types of scanners (phrase-level, lexeme, and lower-case). See the <a href="rule.html">rule</a>, 58 <a href="grammar.html">grammar</a>, <a href="scanner.html#lexeme_scanner">lexeme_scanner</a> 59 and <a href="scanner.html#as_lower_scanner">as_lower_scanner </a>for more information. 60</p> 61<p>Here's the grammar (see <a href="../example/techniques/multiple_scanners.cpp">multiple_scanners.cpp</a>): 62</p> 63<pre><span class=special> </span><span class=keyword>struct </span><span class=identifier>my_grammar </span><span class=special>: </span><span class=identifier>grammar</span><span class=special><</span><span class=identifier>my_grammar</span><span class=special>> 64 </span><span class=special>{ 65 </span><span class=keyword>template </span><span class=special><</span><span class=keyword>typename </span><span class=identifier>ScannerT</span><span class=special>> 66 </span><span class=keyword>struct </span><span class=identifier>definition 67 </span><span class=special>{ 68 </span><span class=identifier>definition</span><span class=special>(</span><span class=identifier>my_grammar </span><span class=keyword>const</span><span class=special>& </span><span class=identifier>self</span><span class=special>) 69 </span><span class=special>{ 70 </span><span class=identifier>r </span><span class=special>= </span><span class=identifier>lower_p</span><span class=special>; 71 </span><span class=identifier>rr </span><span class=special>= </span><span class=special>+(</span><span class=identifier>lexeme_d</span><span class=special>[</span><span class=identifier>r</span><span class=special>] </span><span class=special>>> </span><span class=identifier>as_lower_d</span><span class=special>[</span><span class=identifier>r</span><span class=special>] </span><span class=special>>> </span><span class=identifier>r</span><span class=special>); 72 </span><span class=special>} 73 74 </span><span class=keyword>typedef </span><span class=identifier>scanner_list</span><span class=special>< 75 </span><span class=identifier>ScannerT 76 </span><span class=special>, </span><span class=keyword>typename </span><span class=identifier>lexeme_scanner</span><span class=special><</span><span class=identifier>ScannerT</span><span class=special>>::</span><span class=identifier>type 77 </span><span class=special>, </span><span class=keyword>typename </span><span class=identifier>as_lower_scanner</span><span class=special><</span><span class=identifier>ScannerT</span><span class=special>>::</span><span class=identifier>type 78 </span><span class=special>> </span><span class=identifier>scanners</span><span class=special>; 79 80 </span><span class=identifier>rule</span><span class=special><</span><span class=identifier>scanners</span><span class=special>> </span><span class=identifier>r</span><span class=special>; 81 </span><span class=identifier>rule</span><span class=special><</span><span class=identifier>ScannerT</span><span class=special>> </span><span class=identifier>rr</span><span class=special>; 82 </span><span class=identifier>rule</span><span class=special><</span><span class=identifier>ScannerT</span><span class=special>> </span><span class=keyword>const</span><span class=special>& </span><span class=identifier>start</span><span class=special>() </span><span class=keyword>const </span><span class=special>{ </span><span class=keyword>return </span><span class=identifier>rr</span><span class=special>; </span><span class=special>} 83 </span><span class=special>}; 84 </span><span class=special>};</span></pre> 85<p>By default support for multiple scanners is disabled. The macro 86 <tt>BOOST_SPIRIT_RULE_SCANNERTYPE_LIMIT</tt> must be defined to the 87 maximum number of scanners allowed in a scanner_list. The value must 88 be greater than 1 to enable multiple scanners. Given the 89 example above, to define a limit of three scanners for the list, the 90 following line must be inserted into the source file before the 91 inclusion of Spirit headers: 92</p> 93<pre><span class=special> </span><span class=preprocessor>#define </span><span class=identifier>BOOST_SPIRIT_RULE_SCANNERTYPE_LIMIT</span> <span class=literal>3</span></pre> 94<h3><span class=special></span><b> <a name="no_rules" id="no_rules"></a> Look 95 Ma' No Rules</b></h3> 96<p>You use grammars and you use lots of 'em? Want a fly-weight, no-cholesterol, 97 super-optimized grammar? Read on...</p> 98<p>I have a love-hate relationship with rules. I guess you know the reasons why. 99 A lot of problems stem from the limitation of rules. Dynamic polymorphism and 100 static polymorphism in C++ do not mix well. There is no notion of virtual template 101 functions in C++; at least not just yet. Thus, the <strong>rule is tied to a 102 specific scanner type</strong>. This results in problems such as the <a href="faq.html#scanner_business">scanner 103 business</a>, our no. 1 FAQ. Apart from that, the virtual functions in rules 104 slow down parsing, kill all meta-information, and kills inlining, hence bloating 105 the generated code, especially for very tiny rules such as:</p> 106<pre> r <span class="special">=</span> ch_p<span class="special">(</span><span class="quotes">'x'</span><span class="special">) >></span> uint_p<span class="special">;</span></pre> 107<p> The rule's limitation is the main reason why the grammar is designed the way 108 it is now, with a nested template definition class. The rule's limitation is 109 also the reason why subrules exists. But do we really need rules? Of course! 110 Before C++ adopts some sort of auto-type deduction, such as that proposed by 111 David Abrahams in clc++m:</p> 112<pre> 113 <code><span class=keyword>auto </span><span class=identifier>r </span><span class=special>= ...</span><span class=identifier>definition </span><span class=special>...</span></code></pre> 114<p> we are tied to the rule as RHS placeholders. However.... in some occasions 115 we can get by without rules! For instance, rather than writing:</p> 116<pre> 117 <code><span class=identifier>rule</span><span class=special><> </span><span class=identifier>x </span><span class=special>= </span><span class=identifier>ch_p</span><span class=special>(</span><span class=literal>'x'</span><span class=special>);</span></code></pre> 118<p> It's better to write:</p> 119<pre> 120 <code><span class=identifier>chlit</span><span class=special><> </span><span class=identifier>x </span><span class=special>= </span><span class=identifier>ch_p</span><span class=special>(</span><span class=literal>'x'</span><span class=special>);</span></code></pre> 121<p> That's trivial. But what if the rule is rather complicated? Ok, let's proceed 122 stepwise... I'll investigate a simple skip_parser based on the C grammar from 123 Hartmut Kaiser. Basically, the grammar is written as (see <a href="../example/techniques/no_rules/no_rule1.cpp">no_rule1.cpp</a>):</p> 124<pre><code> <span class=keyword>struct </span><span class=identifier>skip_grammar </span><span class=special>: </span><span class=identifier>grammar</span><span class=special><</span><span class=identifier>skip_grammar</span><span class=special>> 125 { 126 </span><span class=keyword>template </span><span class=special><</span><span class=keyword>typename </span><span class=identifier>ScannerT</span><span class=special>> 127 </span><span class=keyword>struct </span><span class=identifier>definition 128 </span><span class=special>{ 129 </span><span class=identifier>definition</span><span class=special>(</span><span class=identifier>skip_grammar </span><span class=keyword>const</span><span class=special>& /*</span><span class=identifier>self</span><span class=special>*/) 130 { 131 </span><span class=identifier>skip 132 </span><span class=special>= </span><span class=identifier>space_p 133 </span><span class=special>| </span><span class=string>"//" </span><span class=special>>> *(</span><span class=identifier>anychar_p </span><span class=special>- </span><span class=literal>'\n'</span><span class=special>) >> </span><span class=literal>'\n' 134 </span><span class=special>| </span><span class=string>"/*" </span><span class=special>>> *(</span><span class=identifier>anychar_p </span><span class=special>- </span><span class=string>"*/"</span><span class=special>) >> </span><span class=string>"*/" 135 </span><span class=special>; 136 } 137 138 </span><span class=identifier>rule</span><span class=special><</span><span class=identifier>ScannerT</span><span class=special>> </span><span class=identifier>skip</span><span class=special>; 139 140 </span><span class=identifier>rule</span><span class=special><</span><span class=identifier>ScannerT</span><span class=special>> </span><span class=keyword>const</span><span class=special>& 141 </span><span class=identifier>start</span><span class=special>() </span><span class=keyword>const </span><span class=special>{ </span><span class=keyword>return </span><span class=identifier>skip</span><span class=special>; } 142 }; 143 };</span></code></pre> 144<p> Ok, so far so good. Can we do better? Well... since there are no recursive 145 rules there (in fact there's only one rule), you can expand the type of rule's 146 RHS as the rule type (see <a href="../example/techniques/no_rules/no_rule2.cpp">no_rule2.cpp</a>):</p> 147<pre><code><span class=special> </span><span class=keyword>struct </span><span class=identifier>skip_grammar </span><span class=special>: </span><span class=identifier>grammar</span><span class=special><</span><span class=identifier>skip_grammar</span><span class=special>> 148 { 149 </span><span class=keyword>template </span><span class=special><</span><span class=keyword>typename </span><span class=identifier>ScannerT</span><span class=special>> 150 </span><span class=keyword>struct </span><span class=identifier>definition 151 </span><span class=special>{ 152</span> <span class=identifier>definition</span><span class=special>(</span><span class=identifier>skip_grammar </span><span class=keyword>const</span><span class=special>& /*</span><span class=identifier>self</span><span class=special>*/) 153 : </span><span class=identifier>skip</span><span class=special> 154 ( </span><span class=identifier>space_p 155 </span><span class=special>| </span><span class=string>"//" </span><span class=special>>> *(</span><span class=identifier>anychar_p </span><span class=special>- </span><span class=literal>'\n'</span><span class=special>) >> </span><span class=literal>'\n' 156 </span><span class=special>| </span><span class=string>"/*" </span><span class=special>>> *(</span><span class=identifier>anychar_p </span><span class=special>- </span><span class=string>"*/"</span><span class=special>) >> </span><span class=string>"*/" 157 </span><span class=special>) 158 { 159 } 160 161 </span><span class=keyword>typedef 162 </span><span class=identifier>alternative</span><span class=special><</span><span class=identifier>alternative</span><span class=special><</span><span class=identifier>space_parser</span><span class=special>, </span><span class=identifier>sequence</span><span class=special><</span><span class=identifier>sequence</span><span class=special>< 163 </span><span class=identifier>strlit</span><span class=special><</span><span class=keyword>const </span><span class=keyword>char</span><span class=special>*>, </span><span class=identifier>kleene_star</span><span class=special><</span><span class=identifier>difference</span><span class=special><</span><span class=identifier>anychar_parser</span><span class=special>, 164 </span><span class=identifier>chlit</span><span class=special><</span><span class=keyword>char</span><span class=special>> > > >, </span><span class=identifier>chlit</span><span class=special><</span><span class=keyword>char</span><span class=special>> > >, </span><span class=identifier>sequence</span><span class=special><</span><span class=identifier>sequence</span><span class=special>< 165 </span><span class=identifier>strlit</span><span class=special><</span><span class=keyword>const </span><span class=keyword>char</span><span class=special>*>, </span><span class=identifier>kleene_star</span><span class=special><</span><span class=identifier>difference</span><span class=special><</span><span class=identifier>anychar_parser</span><span class=special>, 166 </span><span class=identifier>strlit</span><span class=special><</span><span class=keyword>const </span><span class=keyword>char</span><span class=special>*> > > >, </span><span class=identifier>strlit</span><span class=special><</span><span class=keyword>const </span><span class=keyword>char</span><span class=special>*> > > 167 </span><span class=identifier>skip_t</span><span class=special>; 168</span><span class=special> </span><span class=identifier>skip_t </span><span class=identifier>skip</span><span class=special>; 169 170 </span><span class=identifier>skip_t </span><span class=keyword>const</span><span class=special>& 171 </span><span class=identifier>start</span><span class=special>() </span><span class=keyword>const </span><span class=special>{ </span><span class=keyword>return </span><span class=identifier>skip</span><span class=special>; } 172 }; 173 };</span></code></pre> 174<p> Ughhh! How did I do that? How was I able to get at the complex typedef? Am 175 I insane? Well, not really... there's a trick! What you do is define the typedef 176 <tt>skip_t</tt> first as int:</p> 177<pre> 178 <code><span class=keyword>typedef </span><span class=keyword>int </span><span class=identifier>skip_t</span><span class=special>;</span></code></pre> 179<p> Try to compile. Then, the compiler will generate an obnoxious error message 180 such as:</p> 181<pre> 182 <code><span class=string>"cannot convert boost::spirit::alternative<... blah blah...to int"</span><span class=special>.</span></code></pre> 183<p> <strong>THERE YOU GO!</strong> You got it's type! I just copy and paste the 184 correct type (removing explicit qualifications, if preferred).</p> 185<p> Can we still go further? Yes. Remember that the grammar was designed for rules. 186 The nested template definition class is needed to get around the rule's limitations. 187 Without rules, I propose a new class called <tt>sub_grammar</tt>, the grammar's 188 low-fat counterpart:</p> 189<pre><code><span class=special> </span><span class=keyword>namespace </span><span class=identifier>boost </span><span class=special>{ </span><span class=keyword>namespace </span><span class=identifier>spirit 190 </span><span class=special>{ 191 </span><span class=keyword>template </span><span class=special><</span><span class=keyword>typename </span><span class=identifier>DerivedT</span><span class=special>> 192 </span><span class=keyword>struct </span><span class=identifier>sub_grammar </span><span class=special>: </span><span class=identifier>parser</span><span class=special><</span><span class=identifier>DerivedT</span><span class=special>> 193 { 194 </span><span class=keyword>typedef </span><span class=identifier>sub_grammar </span><span class=identifier>self_t</span><span class=special>; 195 </span><span class=keyword>typedef </span><span class=identifier>DerivedT </span><span class=keyword>const</span><span class=special>& </span><span class=identifier>embed_t</span><span class=special>; 196 197 </span><span class=keyword>template </span><span class=special><</span><span class=keyword>typename </span><span class=identifier>ScannerT</span><span class=special>> 198 </span><span class=keyword>struct </span><span class=identifier>result 199 </span><span class=special>{ 200 </span><span class=keyword>typedef </span><span class=keyword>typename </span><span class=identifier>parser_result</span><span class=special>< 201 </span><span class=keyword>typename </span><span class=identifier>DerivedT</span><span class=special>::</span><span class=identifier>start_t</span><span class=special>, </span><span class=identifier>ScannerT</span><span class=special>>::</span><span class=identifier>type 202 </span><span class=identifier>type</span><span class=special>; 203 }; 204 205 </span><span class=identifier>DerivedT </span><span class=keyword>const</span><span class=special>& </span><span class=identifier>derived</span><span class=special>() </span><span class=keyword>const 206 </span><span class=special>{ </span><span class=keyword>return </span><span class=special>*</span><span class=keyword>static_cast</span><span class=special><</span><span class=identifier>DerivedT </span><span class=keyword>const</span><span class=special>*>(</span><span class=keyword>this</span><span class=special>); } 207 208 </span><span class=keyword>template </span><span class=special><</span><span class=keyword>typename </span><span class=identifier>ScannerT</span><span class=special>> 209 </span><span class=keyword>typename </span><span class=identifier>parser_result</span><span class=special><</span><span class=identifier>self_t</span><span class=special>, </span><span class=identifier>ScannerT</span><span class=special>>::</span><span class=identifier>type 210 </span><span class=identifier>parse</span><span class=special>(</span><span class=identifier>ScannerT </span><span class=keyword>const</span><span class=special>& </span><span class=identifier>scan</span><span class=special>) </span><span class=keyword>const 211 </span><span class=special>{ 212 </span><span class=keyword>return </span><span class=identifier>derived</span><span class=special>().</span><span class=identifier>start</span><span class=special>.</span><span class=identifier>parse</span><span class=special>(</span><span class=identifier>scan</span><span class=special>); 213 } 214 }; 215 }}</span></code></pre> 216<p>With the <tt>sub_grammar</tt> class, we can define our skipper grammar this 217 way (see <a href="../example/techniques/no_rules/no_rule3.cpp">no_rule3.cpp</a>):</p> 218<pre><code><span class=special> </span><span class=keyword>struct </span><span class=identifier>skip_grammar </span><span class=special>: </span><span class=identifier>sub_grammar</span><span class=special><</span><span class=identifier>skip_grammar</span><span class=special>> 219 { 220 </span><span class=keyword>typedef 221 </span><span class=identifier>alternative</span><span class=special><</span><span class=identifier>alternative</span><span class=special><</span><span class=identifier>space_parser</span><span class=special>, </span><span class=identifier>sequence</span><span class=special><</span><span class=identifier>sequence</span><span class=special>< 222 </span><span class=identifier>strlit</span><span class=special><</span><span class=keyword>const </span><span class=keyword>char</span><span class=special>*>, </span><span class=identifier>kleene_star</span><span class=special><</span><span class=identifier>difference</span><span class=special><</span><span class=identifier>anychar_parser</span><span class=special>, 223 </span><span class=identifier>chlit</span><span class=special><</span><span class=keyword>char</span><span class=special>> > > >, </span><span class=identifier>chlit</span><span class=special><</span><span class=keyword>char</span><span class=special>> > >, </span><span class=identifier>sequence</span><span class=special><</span><span class=identifier>sequence</span><span class=special>< 224 </span><span class=identifier>strlit</span><span class=special><</span><span class=keyword>const </span><span class=keyword>char</span><span class=special>*>, </span><span class=identifier>kleene_star</span><span class=special><</span><span class=identifier>difference</span><span class=special><</span><span class=identifier>anychar_parser</span><span class=special>, 225 </span><span class=identifier>strlit</span><span class=special><</span><span class=keyword>const </span><span class=keyword>char</span><span class=special>*> > > >, </span><span class=identifier>strlit</span><span class=special><</span><span class=keyword>const </span><span class=keyword>char</span><span class=special>*> > > 226 </span><span class=identifier>start_t</span><span class=special>; 227 228 </span><span class=identifier>skip_grammar</span><span class=special>() 229 : </span><span class=identifier>start 230 </span><span class=special>( 231 </span><span class=identifier>space_p 232 </span><span class=special>| </span><span class=string>"//" </span><span class=special>>> *(</span><span class=identifier>anychar_p </span><span class=special>- </span><span class=literal>'\n'</span><span class=special>) >> </span><span class=literal>'\n' 233 </span><span class=special>| </span><span class=string>"/*" </span><span class=special>>> *(</span><span class=identifier>anychar_p </span><span class=special>- </span><span class=string>"*/"</span><span class=special>) >> </span><span class=string>"*/" 234 </span><span class=special>) 235 {} 236 237 </span><span class=identifier>start_t </span><span class=identifier>start</span><span class=special>; 238 };</span></code></pre> 239<p>But what for, you ask? You can simply use the <tt>start_t</tt> type above as-is. 240 It's already a parser! We can just type:</p> 241<pre> 242 <code><span class=identifier>skipper_t </span><span class=identifier>skipper </span><span class=special>= 243 </span><span class=identifier>space_p 244 </span><span class=special>| </span><span class=string>"//" </span><span class=special>>> *(</span><span class=identifier>anychar_p </span><span class=special>- </span><span class=literal>'\n'</span><span class=special>) >> </span><span class=literal>'\n' </span><br> <span class=special>| </span><span class=string>"/*" </span><span class=special>>> *(</span><span class=identifier>anychar_p </span><span class=special>- </span><span class=string>"*/"</span><span class=special>) >> </span><span class=string>"*/"</span> 245 <span class=special> ;</span></code></pre> 246<p> and use <tt>skipper</tt> just as we would any parser? Well, a subtle difference 247 is that <tt>skipper</tt>, used this way will be embedded <strong>by value </strong>when<strong> 248 </strong>you compose more complex parsers using it. That is, if we use <tt>skipper</tt> 249 inside another production, the whole thing will be stored in the composite. 250 Heavy!</p> 251<p> The proposed <tt>sub_grammar</tt> OTOH will be held by reference. Note:</p> 252<pre><code> <span class=keyword>typedef </span><span class=identifier>DerivedT </span><span class=keyword>const</span><span class=special>& </span><span class=identifier>embed_t</span><span class=special>;</span></code></pre> 253<p>The proposed <tt>sub_grammar</tt> does not have the inherent limitations of 254 rules, is very lighweight, and should be blazingly fast (can be fully inlined 255 and does not use virtual functions). Perhaps this class will be part of a future 256 spirit release. </p> 257<table width="80%" border="0" align="center"> 258 <tr> 259 <td class="note_box"><img src="theme/note.gif" width="16" height="16"> <strong>The 260 no-rules result</strong><br> <br> 261 So, how much did we save? On MSVCV7.1, the original code: <a href="../example/techniques/no_rules/no_rule1.cpp">no_rule1.cpp</a> 262 compiles to <strong>28k</strong>. Eliding rules, <a href="../example/techniques/no_rules/no_rule2.cpp">no_rule2.cpp</a>, 263 we got <strong>24k</strong>. Not bad, we shaved off 4k amounting to a 14% 264 reduction. But you'll be in for a surprise. The last version, using the 265 sub-grammar: <a href="../example/techniques/no_rules/no_rule3.cpp">no_rule3.cpp</a>, 266 compiles to <strong>5.5k</strong>! That's a whopping 80% reduction.<br> 267 <br> 268 <table width="100%" border="1"> 269 <tr> 270 <td><a href="../example/techniques/no_rules/no_rule1.cpp">no_rule1.cpp</a></td> 271 <td><strong>28k</strong></td> 272 <td>standard rule and grammar</td> 273 </tr> 274 <tr> 275 <td><a href="../example/techniques/no_rules/no_rule2.cpp">no_rule2.cpp</a></td> 276 <td><strong>24k</strong></td> 277 <td>standard grammar, no rule</td> 278 </tr> 279 <tr> 280 <td><a href="../example/techniques/no_rules/no_rule3.cpp">no_rule3.cpp</a></td> 281 <td><strong>5.5k</strong></td> 282 <td>sub_grammar, no rule, no grammar</td> 283 </tr> 284 </table> </td> 285 </tr> 286</table> 287<h3><b> <a name="typeof" id="typeof"></a> typeof</b></h3> 288<p>Some compilers already support the <tt>typeof</tt> keyword. Examples are g++ 289 and Metrowerks CodeWarrior. Someday, <tt>typeof</tt> will become commonplace. 290 It is worth noting that we can use <tt>typeof</tt> to define non-recursive rules 291 without using the rule class. To give an example, we'll use the skipper example 292 above; this time using <tt>typeof</tt>. First, to avoid redundancy, we'll introduce 293 a macro <tt>RULE</tt>: </p> 294<pre><code> <span class=preprocessor>#define </span><span class=identifier>RULE</span><span class=special>(</span><span class=identifier>name</span><span class=special>, </span><span class=identifier>definition</span><span class=special>) </span><span class="keyword">typeof</span><span class=special>(</span><span class=identifier>definition</span><span class=special>) </span><span class=identifier>name </span><span class=special>= </span><span class=identifier>definition</span></code></pre> 295<p>Then, simply:</p> 296<pre><code><span class=identifier> </span><span class=identifier>RULE</span><span class=special>( 297 </span><span class=identifier>skipper</span><span class=special>, 298 ( </span><span class=identifier>space_p 299 </span><span class=special>| </span><span class=string>"//" </span><span class=special>>> *(</span><span class=identifier>anychar_p </span><span class=special>- </span><span class=literal>'\n'</span><span class=special>) >> </span><span class=literal>'\n' 300 </span><span class=special>| </span><span class=string>"/*" </span><span class=special>>> *(</span><span class=identifier>anychar_p </span><span class=special>- </span><span class=string>"*/"</span><span class=special>) >> </span><span class=string>"*/" 301 </span><span class=special>) 302 );</span></code></pre> 303<p>(see <a href="../example/techniques/typeof.cpp">typeof.cpp</a>)</p> 304<p>That's it! Now you can use skipper just as you would any parser. Be reminded, 305 however, that <tt>skipper</tt> above will be embedded by value when<strong> 306 </strong>you compose more complex parsers using it (see <tt>sub_grammar</tt> rationale above). You can use the <tt>sub_grammar</tt> class to avoid this problem.</p> 307<h3><a name="nabialek_trick"></a> Nabialek trick</h3> 308<p>This technique, I'll call the <strong><em>"Nabialek trick" </em></strong>(from the name of its inventor, Sam Nabialek), can improve the rule dispatch from linear non-deterministic to deterministic. The trick applies to grammars where a keyword (operator, etc), precedes a production. There are lots of grammars similar to this:</p> 309<pre> <span class=identifier>r </span><span class=special>= 310 </span><span class=identifier>keyword1 </span><span class=special>>> </span><span class=identifier>production1 311 </span><span class=special>| </span><span class=identifier>keyword2 </span><span class=special>>> </span><span class=identifier>production2 312 </span><span class=special>| </span><span class=identifier>keyword3 </span><span class=special>>> </span><span class=identifier>production3 313 </span><span class=special>| </span><span class=identifier>keyword4 </span><span class=special>>> </span><span class=identifier>production4 314 </span><span class=special>| </span><span class=identifier>keyword5 </span><span class=special>>> </span><span class=identifier>production5 315 </span><span class=comment>/*** etc ***/ 316 </span><span class=special>;</span></pre> 317<p>The cascaded alternatives are tried one at a time through trial and error until something matches. The Nabialek trick takes advantage of the <a href="symbols.html">symbol table</a>'s search properties to optimize the dispatching of the alternatives. For an example, see <a href="../example/techniques/nabialek.cpp">nabialek.cpp</a>. The grammar works as follows. There are two rules (<tt>one</tt> and <tt>two</tt>). When "one" is recognized, rule <tt>one</tt> is invoked. When "two" is recognized, rule <tt>two</tt> is invoked. Here's the grammar:</p> 318<pre><span class=special> </span><span class=identifier>one </span><span class=special>= </span><span class=identifier>name</span><span class=special>; 319 </span><span class=identifier>two </span><span class=special>= </span><span class=identifier>name </span><span class=special>>> </span><span class=literal>',' </span><span class=special>>> </span><span class=identifier>name</span><span class=special>; 320 321 </span><span class=identifier>continuations</span><span class=special>.</span><span class=identifier>add 322 </span><span class=special>(</span><span class=string>"one"</span><span class=special>, &</span><span class=identifier>one</span><span class=special>) 323 </span><span class=special>(</span><span class=string>"two"</span><span class=special>, &</span><span class=identifier>two</span><span class=special>) 324 </span><span class=special>; 325 326 </span><span class=identifier>line </span><span class=special>= </span><span class=identifier>continuations</span><span class=special>[</span><span class=identifier>set_rest</span><span class=special><</span><span class=identifier>rule_t</span><span class=special>>(</span><span class=identifier>rest</span><span class=special>)] </span><span class=special>>> </span><span class=identifier>rest</span><span class=special>;</span></pre> 327<p>where continuations is a <a href="symbols.html">symbol table</a> with pointer to rule_t slots. one, two, name, line and rest are rules:</p> 328<pre><span class=special> </span><span class=identifier>rule_t </span><span class=identifier>name</span><span class=special>; 329 </span><span class=identifier>rule_t </span><span class=identifier>line</span><span class=special>; 330 </span><span class=identifier>rule_t </span><span class=identifier>rest</span><span class=special>; 331 </span><span class=identifier>rule_t </span><span class=identifier>one</span><span class=special>; 332 </span><span class=identifier>rule_t </span><span class=identifier>two</span><span class=special>; 333 334 </span><span class=identifier>symbols</span><span class=special><</span><span class=identifier>rule_t</span><span class=special>*> </span><span class=identifier>continuations</span><span class=special>;</span></pre> 335<p>set_rest, the semantic action attached to continuations is:</p> 336<pre><span class=special> </span><span class=keyword>template </span><span class=special><</span><span class=keyword>typename </span><span class=identifier>Rule</span><span class=special>> 337 </span><span class=keyword>struct </span><span class=identifier>set_rest 338 </span><span class=special>{ 339 </span><span class=identifier>set_rest</span><span class=special>(</span><span class=identifier>Rule</span><span class=special>& </span><span class=identifier>the_rule</span><span class=special>) 340 </span><span class=special>: </span><span class=identifier>the_rule</span><span class=special>(</span><span class=identifier>the_rule</span><span class=special>) </span><span class=special>{} 341 342 </span><span class=keyword>void </span><span class=keyword>operator</span><span class=special>()(</span><span class=identifier>Rule</span><span class=special>* </span><span class=identifier>newRule</span><span class=special>) </span><span class=keyword>const 343 </span><span class=special>{ </span><span class=identifier>m_theRule </span><span class=special>= </span><span class=special>*</span><span class=identifier>newRule</span><span class=special>; </span><span class=special>} 344 345 </span><span class=identifier>Rule</span><span class=special>& </span><span class=identifier>the_rule</span><span class=special>; 346 </span><span class=special>};</span></pre> 347<p>Notice how the rest <tt>rule</tt> gets set dynamically when the set_rule action is called. The dynamic grammar parses inputs such as:</p> 348<p> "one only"<br> 349"one again"<br> 350"two first, second"</p> 351<p>The cool part is that the <tt>rest</tt> rule is set (by the <tt>set_rest</tt> action) depending on what the symbol table got. If it got a <em>"one"</em> then rest = one. If it got <em>"two"</em>, then rest = two. Very nifty! This technique should be very fast, especially when there are lots of keywords. It would be nice to add special facilities to make this easy to use. I imagine:</p> 352<pre><span class=special> </span><span class=identifier>r </span><span class=special>= </span><span class=identifier>keywords </span><span class=special>>> </span><span class=identifier>rest</span><span class=special>;</span></pre> 353<p>where <tt>keywords</tt> is a special parser (based on the symbol table) that automatically sets its RHS (rest) depending on the acquired symbol. This, I think, is mighty cool! Someday perhaps... </p> 354<p><img src="theme/note.gif" width="16" height="16"> Also, see the <a href="switch_parser.html">switch parser</a> for another deterministic parsing trick for character/token prefixes. </p> 355<span class=special></span> 356<table border="0"> 357 <tr> 358 <td width="10"></td> 359 <td width="30"><a href="../index.html"><img src="theme/u_arr.gif" border="0"></a></td> 360 <td width="30"><a href="style_guide.html"><img src="theme/l_arr.gif" border="0"></a></td> 361 <td width="30"><a href="faq.html"><img src="theme/r_arr.gif" border="0"></a></td> 362 </tr> 363</table> 364<br> 365<hr size="1"> 366<p class="copyright">Copyright © 1998-2003 Joel de Guzman<br> 367 <br> 368<font size="2">Use, modification and distribution is subject to the Boost Software 369 License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at 370 http://www.boost.org/LICENSE_1_0.txt)</font></p> 371<p class="copyright"> </p> 372</body> 373</html> 374
