ceph/src/boost/libs/spirit/classic/doc/faq.html

   1 <html>
   2 <head>
   3 <title>FAQ</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>FAQ</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="techniques.html"><img src="theme/l_arr.gif" border="0"></a></td>
  23     <td width="30"><a href="rationale.html"><img src="theme/r_arr.gif" border="0"></a></td>
  24   </tr>
  25 </table>
  26 <ul>
  27   <li><a href="#scanner_business">The Scanner Business</a></li>
  28   <li><a href="#left_recursion">Eliminating Left Recursion</a> </li>
  29   <li><a href="#right_associativity">Implementing Right Associativity</a></li>
  30   <li><a href="#lexeme_and_rules">The lexeme_d directive and rules</a></li>
  31   <li><a href="#kleene_star">Kleene Star infinite loop</a></li>
  32   <li><a href="#CVS">Boost CVS and Spirit CVS</a></li>
  33   <li><a href="#compilation_times">How to reduce compilation times with complex
  34     Spirit grammars</a></li>
  35   <li><strong><a href="#frame_assertion">Closure frame assertion</a></strong></li>
  36   <li><strong><a href="#greedy_rd">Greedy RD</a></strong></li>
  37   <li><strong><a href="#referencing_a_rule_at_construction">Referencing a rule
  38     at construction time</a></strong></li>
  39   <li><strong><a href="#storing_rules">Storing Rules</a></strong></li>
  40   <li><strong><a href="#parsing_ints_and_reals">Parsing ints and reals</a> </strong></li>
  41   <li><strong><a href="#output_operator">BOOST_SPIRIT_DEBUG and missing <tt>operator&lt;&lt;</tt></a></strong></li>
  42   <li><strong><a href="#repository">Applications that used to be part of spirit</a></strong></li>
  43 </ul>
  44 <p><b> <a name="scanner_business" id="scanner_business"></a> The Scanner Business</b></p>
  45 <p><font color="#FF0000">Question:</font> Why doesn't this compile?</p>
  46 <pre><code><font color="#000000"><span class=special>    </span><span class=identifier>rule</span><span class=special>&lt;&gt; </span><span class=identifier>r </span><span class=special>= /*...*/;
  47 </span>    <span class=identifier>parse</span><span class=special>(</span><span class=string>"hello world"</span><span class=special>, </span><span class=identifier>r</span><span class=special>, </span><span class=identifier>space_p</span><span class=special>); </span><span class=comment>// BAD [attempts phrase level parsing]</span></font></code></pre>
  48 <p>But if I <font color="#000000">remove the skip-parser, everything goes back
  49   to normal again:<code></code></font></p>
  50 <pre><code><font color="#000000">    <span class=identifier>rule</span><span class=special>&lt;&gt; </span><span class=identifier>r </span><span class=special>= *</span><span class=identifier>anychar_p</span><span class=special>;
  51     </span><span class=identifier>parse</span><span class=special>(</span><span class=string>"hello world"</span><span class=special>, </span><span class=identifier>r</span><span class=special>); </span><span class=comment>// OK [character level parsing]</span></font></code></pre>
  52 <p>Sometimes you'll want to pass in a rule to one of the functions parse functions
  53   that Spirit provides. The problem is that the rule is a template class that
  54   is parameterized by the scanner type. This is rather awkward but unavoidable:
  55   <strong>the rule is tied to a scanner</strong>. What's not obvious is that this
  56   scanner must be compatible with the scanner that is ultimately passed to the
  57   rule's parse member function. Otherwise, the compiler will complain. </p>
  58 <p>Why does the first call to parse not compile? Because of scanner incompatibility.
  59   Behind the scenes, the free parse function creates a scanner from the iterators
  60   passed in. In the first call to parse, the scanner created is a plain vanilla
  61   <tt>scanner&lt;&gt;</tt>. This is compatible with the default scanner type of
  62   <tt>rule&lt;&gt;</tt> [see default template parameters of <a href="rule.html">the
  63   rule</a>]. The second call creates a scanner of type <tt><a href="scanner.html#phrase_scanner_t">phrase_scanner_t</a></tt>.
  64   Thus, in order for the second call to succeed, the rule must be parameterized
  65   as <tt>rule&lt;phrase_scanner_t&gt;</tt>:</p>
  66 <pre><code><font color="#000000"><span class=comment>    </span><span class=identifier>rule</span><span class=special>&lt;</span><span class=identifier>phrase_scanner_t</span><span class=special>&gt; </span><span class=identifier>r </span><span class=special>= </span><span class=special>*</span><span class=identifier>anychar_p</span><span class=special>;
  67     </span><span class=identifier>parse</span><span class=special>(</span><span class=string>"hello world"</span><span class=special>, </span><span class=identifier>r</span><span class=special>, </span><span class=identifier>space_p</span><span class=special>);       </span><span class=comment>//  OK [phrase level parsing]</span></font></code></pre>
  68 <p>Take note however that <tt>phrase_scanner_t</tt> is compatible only when you
  69   are using <tt>char const*</tt> iterators and <tt>space_p</tt> as the skip parser.
  70   Other than that, you'll have to find the right type of scanner. This is tedious
  71   to do correctly. In light of this issue, <strong>it is best to avoid rules as
  72   arguments to the parse functions</strong>. Keep in mind that this happens only
  73   with rules. The rule is the only parser that has to be tied to a particular
  74   scanner type. For instance:</p>
  75 <pre><span class=comment>    </span><span class=identifier>parse</span><span class=special>(</span><span class=string>"hello world"</span><span class=special>, *</span><span class=identifier>anychar_p</span><span class=special>);           </span><span class=comment><code><font color="#000000"><span class=comment>//  OK  [character level parsing]</span></font></code>
  76     </span><span class=identifier>parse</span><span class=special>(</span><span class=string>"hello world"</span><span class=special>, *</span><span class=identifier>anychar_p</span><span class=special>, </span><span class=identifier>space_p</span><span class=special>);  </span><span class="comment">//  OK  [phrase level parsing]</span></pre>
  77 <table width="80%" border="0" align="center">
  78   <tr>
  79     <td class="note_box"> <strong><img src="theme/note.gif" width="16" height="16">
  80       Multiple Scanner Support</strong><br>
  81       <br>
  82       As of v1.8.0, rules can use one or more scanner types. There are cases,
  83       for instance, where we need a rule that can work on the phrase and character
  84       levels. Rule/scanner mismatch has been a source of confusion and is the
  85       no. 1 <a href="faq.html#scanner_business">FAQ</a>. To address this issue,
  86       we now have <a href="rule.html#multiple_scanner_support">multiple scanner
  87       support</a>. <br>
  88       <br>
  89       <img src="theme/bulb.gif" width="13" height="18"> See the techniques section
  90       for an <a href="techniques.html#multiple_scanner_support">example</a> of
  91       a <a href="grammar.html">grammar</a> using a multiple scanner enabled rule,
  92       <a href="scanner.html#lexeme_scanner">lexeme_scanner</a> and <a href="scanner.html#as_lower_scanner">as_lower_scanner.</a></td>
  93   </tr>
  94 </table>
  95 <p><b> <a name="left_recursion"></a> Eliminating Left Recursion </b></p>
  96 <p><font color="#FF0000">Question:</font> I ported a grammar from YACC. It's &quot;kinda&quot;
  97   working - the parser itself compiles with no errors. But when I try to parse,
  98   it gives me an &quot;invalid page fault&quot;. I tracked down the problem to
  99   this grammar snippet:</p>
 100 <pre>    <span class=identifier>or_expr </span><span class=special>= </span><span class=identifier>xor_expr </span><span class=special>| (</span><span class=identifier>or_expr </span><span class=special>&gt;&gt; </span><span class=identifier>VBAR </span><span class=special>&gt;&gt; </span><span class=identifier>xor_expr</span><span class=special>);</span></pre>
 101 <p>What you should do is to eliminate direct and indirect left-recursion. This
 102   causes the invalid page fault because the program enters an infinite loop. The
 103   code above is good for bottom up parsers such as YACC but not for LL parsers
 104   such as Spirit.</p>
 105 <p>This is similar to a rule in Hartmut Kaiser's C
 106   parser (this should be available for download from <a href="http://spirit.sf.net">Spirit's site</a> as soon as you read this).</p>
 107 <pre>
 108     <span class=identifier>inclusive_or_expression
 109     </span><span class=special>= </span><span class=identifier>exclusive_or_expression
 110     </span><span class=special>| </span><span class=identifier>inclusive_or_expression </span><span class=special>&gt;&gt; </span><span class=identifier>OR </span><span class=special>&gt;&gt; </span><span class=identifier>exclusive_or_expression
 111     </span><span class=special>;</span></pre>
 112 <p><span class=special></span>Transforming left recursion to right recursion,
 113   we have:</p>
 114 <pre>    <span class=identifier>inclusive_or_expression
 115     </span><span class=special>= </span><span class=identifier>exclusive_or_expression </span><span class=special>&gt;&gt; </span><span class=identifier>inclusive_or_expression_helper
 116     </span><span class=special>;
 117
 118     </span><span class=identifier>inclusive_or_expression_helper
 119     </span><span class=special>= </span><span class=identifier>OR </span><span class=special>&gt;&gt; </span><span class=identifier>exclusive_or_expression </span><span class=special>&gt;&gt; </span><span class=identifier>inclusive_or_expression_helper
 120     </span><span class=special>| </span><span class=identifier>epsilon_p
 121     </span><span class=special>;</span></pre>
 122 <p><span class=special></span>I'd go further. Since:</p>
 123 <pre>    <span class=identifier>r </span><span class=special>= </span><span class=identifier>a </span><span class=special>| </span><span class=identifier>epsilon_p</span><span class=special>;</span></pre>
 124 <p><span class=special></span>is equivalent to:<span class=special><br>
 125   </span></p>
 126 <pre>    <span class=identifier>r </span><span class=special>= !</span><span class=identifier>a</span><span class=special>;</span></pre>
 127 <p>we can simplify <tt>inclusive_or_expression_helper</tt> thus:</p>
 128 <pre>    <span class=identifier>inclusive_or_expression_helper
 129     </span><span class=special>= !(</span><span class=identifier>OR </span><span class=special>&gt;&gt; </span><span class=identifier>exclusive_or_expression </span><span class=special>&gt;&gt; </span><span class=identifier>inclusive_or_expression_helper</span><span class=special>)
 130     ;</span></pre>
 131 <p><span class=special></span>Now, since:</p>
 132 <pre>    <span class=identifier>r </span><span class=special>= !(</span><span class=identifier>a </span><span class=special>&gt;&gt; </span><span class=identifier>r</span><span class=special>);</span></pre>
 133 <p><span class=special></span>is equivalent to:</p>
 134 <pre>    <span class=identifier>r </span><span class=special>= *</span><span class=identifier>a</span><span class=special>;</span></pre>
 135 <p><span class=special></span>we have:</p>
 136 <pre>    <span class=identifier>inclusive_or_expression_helper
 137     </span><span class=special>= *(</span><span class=identifier>OR </span><span class=special>&gt;&gt; </span><span class=identifier>exclusive_or_expression</span><span class=special>)
 138     ;</span></pre>
 139 <p><span class=special></span>Now simplifying <tt>inclusive_or_expression</tt>
 140   fully, we have:</p>
 141 <pre>    <span class=identifier>inclusive_or_expression
 142     </span><span class=special>= </span><span class=identifier>exclusive_or_expression </span><span class=special>&gt;&gt; *(</span><span class=identifier>OR </span><span class=special>&gt;&gt; </span><span class=identifier>exclusive_or_expression</span><span class=special>)
 143     ;</span></pre>
 144 <p><span class=special></span>Reminds me of the calculators. So in short:</p>
 145 <pre>    <span class=identifier>a </span><span class=special>= </span><span class=identifier>b </span><span class=special>| </span><span class=identifier>a </span><span class=special>&gt;&gt; </span><span class=identifier>op </span><span class=special>&gt;&gt; </span><span class=identifier>b</span><span class=special>;</span></pre>
 146 <p><span class=special></span><span class=identifier>in </span><span class=identifier>pseudo-YACC
 147   </span><span class=identifier>is</span><span class=special>:</span></p>
 148 <pre>    <span class=identifier>a </span><span class=special>= </span><span class=identifier>b </span><span class=special>&gt;&gt; *(</span><span class=identifier>op </span><span class=special>&gt;&gt; </span><span class=identifier>b</span><span class=special>);</span></pre>
 149 <p><span class=special></span>in Spirit. What could be simpler? Look Ma, no recursion,
 150   just iteration.</p>
 151 <p><b> <a name="right_associativity" id="right_associativity"></a> Implementing Right Associativity </b></p>
 152 <p> <font color="#FF0000">Question:</font> I tried adding <tt>'^'</tt> as an operator to compute the power to a calculator grammer. The following code
 153 </p>
 154 <pre>    <span class=identifier>pow_expression
 155     </span><span class=special>= </span><span class=identifier>pow_operand </span><span class=special>&gt;&gt; </span><span class=special>*( </span><span class=literal>'^' </span><span class=special>&gt;&gt; </span><span class=identifier>pow_operand </span><span class=special>[ </span><span class=special>&amp; </span><span class=identifier>do_pow </span><span class=special>]
 156                       </span><span class=special>)
 157     </span><span class=special>;</span>
 158 </pre>
 159 <p>parses the input correctly, but I want the operator to be evalutated from right to left. In other words, the expression <tt>2^3^4</tt> is supposed to have the same semantics as <tt>2^(3^4)</tt> instead of <tt>(2^3)^4</tt>. How do I do it?
 160 </p>
 161 <p> The "textbook recipe" for Right Associativity is Right Recursion. In BNF that means:
 162 <pre>    &lt;pow_expression&gt; ::=  &lt;pow_operand&gt; '^' &lt;pow_expression&gt; | &lt;pow_operand&gt;
 163 </pre>
 164 <p>But we better don't take the theory too literally here, because if the first alternative fails, the semantic actions within <tt>pow_operand</tt> might have been executed already and will then be executed again when trying the second alternative. So let's apply Left Factorization to factor out <tt>pow_operand</tt>:
 165 <pre>    &lt;pow_expression&gt; ::=  &lt;pow_operand&gt; &lt;pow_expression_helper&gt;
 166     &lt;pow_expression_helper&gt; ::= '^' &lt;pow_expression&gt; | <i>&#949;</i>
 167 </pre>
 168 <p>The production <tt>pow_expression_helper</tt> matches the empty string <i>&#949;</i>, so we can replace the alternative with the optional operator in Spirit code.
 169 </p>
 170 <pre>    <span class=identifier>pow_expression
 171     </span><span class=special>= </span><span class=identifier>pow_operand </span><span class=special>&gt;&gt; </span><span class=special>!( </span><span class=literal>'^' </span><span class=special>&gt;&gt; </span><span class=identifier>pow_expression </span><span class=special>[ </span><span class=special>&amp; </span><span class=identifier>do_pow </span><span class=special>]
 172                       </span><span class=special>)
 173     </span><span class=special>;</span>
 174 </pre>
 175 <p>Now any semantic actions within <tt>pow_operand</tt> can safely be executed. For stack-based evaluation that means that each match of <tt>pow_operand</tt> will leave one value on the stack and the recursion makes sure there are (at least) two values on the stack when <tt>do_pow</tt> is fired to reduce these two values to their power.
 176 </p>
 177 <p>In cases where this technique isn't applicable, such as C-style assignment
 178 <pre>    <span class=identifier>assignment
 179     </span><span class=special>= </span><span class=identifier>lvalue </span><span class=special>&gt;&gt; </span><span class=literal>'=' </span><span class=special>&gt;&gt; </span><span class=identifier>assignment
 180     </span><span class=special>| </span><span class=identifier>ternary_conditional
 181     </span><span class=special>;</span>
 182 </pre>
 183 <p>you can append <tt>| epsilon_p [ <i>action</i> ] &gt;&gt; nothing_p</tt> to a parser to correct the semantic context when backtracking occurs (in the example case that would be dropping the address pushed by <tt>lvalue</tt> off the evaluation stack):
 184 </p>
 185 <pre>    <span class=identifier>assignment
 186     </span><span class=special>= </span><span class=identifier>lvalue </span><span class=special>&gt;&gt; </span><span class=special>( </span><span class=literal>'=' </span><span class=special>&gt;&gt; </span><span class=identifier>assignment    </span></span><span class=special>[ </span><span class=special>&amp; </span><span class=identifier>do_store </span><span class=special>]
 187                 </span><span class=special>| </span><span class=identifier>epsilon_p            </span><span class=special>[ </span><span class=special>&amp; </span><span class=identifier>do_drop  </span><span class=special>]
 188                   </span><span class=special>&gt;&gt; </span><span class=identifier>nothing_p
 189                 </span><span class=special>)
 190     </span><span class=special>| </span><span class=identifier>ternary_conditional
 191     </span><span class=special>;</span>
 192 </pre>
 193 <p>However, this trick compromises the clear separation of syntax and semantics, so you also might want to consider using an <a href="trees.html">AST</a> instead of semantic actions so you can just go with the first definition of <tt>assignment</tt>.
 194 </p>
 195 <p><b> <a name="lexeme_and_rules" id="lexeme_and_rules"></a> The lexeme_d directive
 196   and rules</b></p>
 197 <p> <font color="#FF0000">Question:</font> Does lexeme_d not support expressions
 198   which include rules? In the example below, the definition of atomicRule compiles,
 199 </p>
 200 <pre>    <span class=identifier></span><span class=identifier>rule</span><span class=special>&lt;</span><span class=identifier>phrase_scanner_t</span><span class=special>&gt; </span><span class=identifier>atomicRule</span>
 201         <span class=special>= </span><span class=identifier>lexeme_d</span><span class=special>[(</span><span class=identifier>alpha_p </span><span class=special>| </span><span class=literal>'_'</span><span class=special>) &gt;&gt; *(</span><span class=identifier>alnum_p </span><span class=special>| </span><span class=literal>'.' </span><span class=special>| </span><span class=literal>'-' </span><span class=special>| </span><span class=literal>'_'</span><span class=special>)];</span></pre>
 202 <p>but if I move <tt>alnum_p | '.' | '-' | '_'</tt> into its own rule, the compiler
 203   complains about conversion from <tt>const scanner&lt;...&gt;</tt> to <tt>const
 204   phrase_scaner_t&amp;</tt>. </p>
 205 <pre>    <span class=identifier>rule</span><span class=special>&lt;</span><span class=identifier>phrase_scanner_t</span><span class=special>&gt; </span><span class=identifier>ch </span><span class=special>
 206         = </span><span class=identifier>alnum_p </span><span class=special>| </span><span class=literal>'.' </span><span class=special>| </span><span class=literal>'-' </span><span class=special>| </span><span class=literal>'_'</span><span class=special>;</span>
 207
 208 <span class=identifier>    rule</span><span class=special>&lt;</span><span class=identifier>phrase_scanner_t</span><span class=special>&gt; </span><span class=identifier>compositeRule</span>
 209         <span class=special>= </span><span class=identifier>lexeme_d</span><span class=special>[(</span><span class=identifier>alpha_p </span><span class=special>| </span><span class=literal>'_'</span><span class=special>) &gt;&gt; *(</span><span class=identifier>ch</span><span class=special>)]; </span><span class="comment">// &lt;- error source</span></pre>
 210 <p>You might get the impression that the <tt>lexeme_d</tt> directive and rules
 211   do not mix. Actually, this problem is related to the first FAQ entry: The Scanner
 212   Business. More precisely, the <tt>lexeme_d</tt> directive and rules with incompatible
 213   scanner types do not mix. This problem is more subtle. What's causing the scanner
 214   incompatibility is the directive itself. The <tt>lexeme_d</tt> directive transforms
 215   the scanner it receives into something that disables the skip parser. This non-skipping
 216   scanner, unfortunately, is incompatible with the original scanner before transformation
 217   took place.</p>
 218 <p>The simplest solution is not to use rules in the <tt>lexeme_d</tt>. Instead,
 219   you can definitely apply <tt>lexeme_d</tt> to subrules and grammars if you really
 220   need more complex parsers inside the <tt>lexeme_d</tt>. If you really must use
 221   a rule, you need to know the exact scanner used by the directive. The <tt>lexeme_scanner</tt>
 222   metafunction is your friend here. The example above will work as expected once
 223   we give the <tt>ch</tt> rule a correct scanner type:</p>
 224 <pre>    <span class=identifier>rule</span><span class=special>&lt;</span><span class=identifier>lexeme_scanner</span><span class="special">&lt;</span><span class=identifier>phrase_scanner_t</span><span class=special>&gt;::</span><span class="identifier">type</span><span class=special>&gt; </span><span class=identifier>ch </span><span class=special>
 225         = </span><span class=identifier>alnum_p </span><span class=special>| </span><span class=literal>'.' </span><span class=special>| </span><span class=literal>'-' </span><span class=special>| </span><span class=literal>'_'</span><span class=special>;</span></pre>
 226 <p>Note: make sure to add &quot;<tt>typename</tt>&quot; before <tt>lexeme_scanner</tt>
 227   when this is used inside a template class or function.</p>
 228 <p>The same thing happens when rules are used inside the <tt>as_lower_d</tt> directive.
 229   In such cases, you can use the <tt>as_lower_scanner</tt>. See the <span class=identifier><tt><a href="scanner.html#lexeme_scanner">lexeme_scanner</a></tt></span>
 230   and <tt><a href="scanner.html#as_lower_scanner">as_lower_scanner</a></tt>.</p>
 231 <table width="80%" border="0" align="center">
 232   <tr>
 233     <td class="note_box"><img src="theme/bulb.gif" width="13" height="18"> See
 234       the techniques section for an <a href="techniques.html#multiple_scanner_support">example</a>
 235       of a <a href="grammar.html">grammar</a> using a <a href="rule.html#multiple_scanner_support">multiple
 236       scanner enabled rule,</a> <a href="scanner.html#lexeme_scanner">lexeme_scanner</a>
 237       and <a href="scanner.html#as_lower_scanner">as_lower_scanner.</a></td>
 238   </tr>
 239 </table>
 240 <p><strong><a name="kleene_star"></a>Kleene Star infinite loop</strong></p>
 241 <p><font color="#FF0000">Question</font>: Why Does This Loop Forever?</p>
 242 <pre>    <span class=identifier>rule</span><span class=special>&lt;&gt; </span><span class=identifier>optional </span><span class=special>= !(</span>str_p<span class="special">(</span><span class="string">&quot;optional&quot;</span><span class="special">));
 243     </span><span class=identifier>rule</span><span class=special>&lt;&gt; </span><span class="identifier">list_of_optional </span><span class=special>= *</span><span class=identifier>optional</span><span class="special">;</span></pre>
 244 <p>The problem with this is that the kleene star will continue looping until it
 245   gets a no-match from it's enclosed parser. Because the <tt>optional</tt> rule
 246   is optional, it will always return a match. Even if the input doesn't match
 247   &quot;optional&quot; it will return a zero length match. <tt>list_of_optional</tt>
 248   will keep calling optional forever since optional will never return a no-match.
 249   So in general, any rule that can be &quot;nullable&quot; (meaning it can return
 250   a zero length match) must not be put inside a kleene star.</p>
 251 <p><strong><a name="CVS"></a>Boost CVS and Spirit CVS</strong></p>
 252 <p><font color="#FF0000">Question:</font> There is Boost CVS and Spirit CVS. Which
 253   is used for further development of Spirit?</p>
 254 <p> Generally, development takes place in Spirit's CVS. However, from time to
 255   time a new version of Spirit will be integrated in Boost. When this happens
 256   development takes place in the Boost CVS. There will be announcements on the
 257   Spirit mailing lists whenever the status of the Spirit CVS changes.<br>
 258 </p>
 259 <table width="80%" border="0" align="center">
 260   <tr>
 261     <td class="note_box"><img src="theme/alert.gif" width="16" height="16">
 262     During development of Spirit v1.8.1 (released as part of boost-1.32.0) and
 263     v1.6.2, Spirit's developers decided to stop maintaining Spirit CVS for
 264     BRANCH_1_8 and BRANCH_1_6. This was necessary to reduce the added work of
 265     maintaining and synch'ing two repositories. The maintenance of these branches
 266     will take place on Boost CVS. At this time, new developments towards Spirit
 267     v2 and other experimental developments are expected to happen in Spirit
 268     CVS.</td>
 269   </tr>
 270 </table>
 271 <p><strong><a name="compilation_times"></a>How to reduce compilation times with
 272   complex Spirit grammars </strong></p>
 273 <p><font color="#FF0000">Question:</font> Are there any techniques to minimize
 274   compile times using spirit? For simple parsers compile time doesn't seem to
 275   be a big issue, but recently I created a parser with about 78 rules
 276   and it took about 2 hours to compile. I would like to break the grammar up into
 277   smaller chunks, but it is not as easy as I thought it would be because rules
 278   in two grammar capsules are defined in terms of each other. Any thoughts?</p>
 279 <p> The only way to reduce compile times is </p>
 280 <ul>
 281   <li> to split up your grammars into smaller chunks</li>
 282   <li> prevent the compiler from seeing all grammar definitions at the same time
 283     (in the same compilation unit)</li>
 284 </ul>
 285 <p>The first task is merely logistical, the second is rather a technical one. </p>
 286 <p>A good example of solving the first task is given in the Spirit cpp_lexer example
 287   written by JCAB (you may find it on the <a href="http://spirit.sourceforge.net/repository/applications/show_contents.php">applications' repository</a>).
 288 </p>
 289 <p>The cross referencing problems may be solved by some kind of forward declaration,
 290   or, if this does not work, by introducing some dummy template argument to the
 291   non-templated grammars. Thus allows the instantiation time to be defered until the
 292   compiler has seen all the defintions:</p>
 293 <pre>    <span class="keyword">template</span> &lt;<span class="keyword">typename</span> T = <span class="keyword">int</span>&gt;<br>    grammar2;</p>
 294
 295     <span class="keyword">template</span> &lt;<span class="keyword">typename</span> T = <span class="keyword">int</span>&gt;<br>    <span class="keyword">struct</span> grammar1 : <span class="keyword">public</span> grammar&lt;grammar1&gt;<br>    {
 296     <span class="comment">// refers to grammar2&lt;&gt;</span>
 297     };
 298
 299     <span class="keyword">template</span> &lt;typename T&gt;
 300     <span class="keyword">struct</span> grammar2 : <span class="keyword">public</span> grammar&lt;grammar2&gt;
 301     {
 302     <span class="comment">// refers to grammar1&lt;&gt;</span>
 303     };
 304
 305     //...
 306     grammar1&lt;&gt; g; <span class="comment">// both grammars instantiated here</span>
 307 </pre>
 308 <p>The second task is slightly more complex. You must ensure that in the first
 309   compilation unit the compiler sees only some function/template <strong>declaration</strong>
 310   and in the second compilation unit the function/template <strong>definition</strong>.
 311   Still no problem, if no templates are involved. If templates are involved,
 312   you need to manually (explicitly) instantiate these templates with the correct
 313   template parameters inside a separate compilation unit. This way the compilation
 314   time is split between several compilation units, reducing the overall
 315   required time drastically too. </p>
 316 <p>For a sample, showing how to achieve this, you may want to look at the <tt>Wave</tt>
 317   preprocessor library, where this technique is used extensively.  (this should be available for download from <a href="http://spirit.sf.net">Spirit's site</a> as soon as you read this).</p>
 318 <p><strong><a name="frame_assertion" id="frame_assertion"></a>Closure frame assertion</strong></p>
 319 <p><font color="#FF0000">Question:</font> When I run the parser I get an assertion
 320   <span class="string">&quot;frame.get() != 0 in file closures.hpp&quot;</span>.
 321   What am I doing wrong?</p>
 322 <p>Basically, the assertion fires when you are accessing a closure variable that
 323   is not constructed yet. Here's an example. We have three rules <tt>a</tt>, <tt>b</tt>
 324   and <tt>c</tt>. Consider that the rule <tt>a</tt> has a closure member <tt>m</tt>.
 325   Now:</p>
 326 <pre>    <span class="identifier">a</span> <span class="special">=</span> <span class="identifier">b</span><span class="special">;</span>
 327     <span class="identifier">b</span> <span class="special">=</span> <span class="identifier">int_p</span><span class="special">[</span><span class="identifier">a</span><span class="special">.</span><span class="identifier">m</span> <span class="special">=</span> 123<span class="special">];</span>
 328     <span class="identifier">c</span> <span class="special">=</span> <span class="identifier">b</span><span class="special">;</span></pre>
 329 <p>When the rule <tt>a</tt> is invoked, its frame is set, along with its member
 330   <tt>m</tt>. So, when <tt>b</tt> is called from <tt>a</tt>, the semantic action
 331   <tt>[a.m = 123]</tt>will store <tt>123</tt> into <tt>a</tt>'s closure member
 332   <tt>m</tt>. On the other hand, when <tt>c</tt> is invoked, and <tt>c</tt> attempts
 333   to call <tt>b</tt>, no frame for <tt>a</tt> is set. Thus, when <tt>b</tt> is
 334   called from <tt>c</tt>, the semantic action <tt>[a.m = 123]</tt>will fire the
 335   <span class="string">&quot;frame.get() != 0 in file closures.hpp&quot;</span>
 336   assertion.</p>
 337 <p><strong><a name="greedy_rd" id="greedy_rd"></a>Greedy RD</strong></p>
 338 <p><font color="#FF0000">Question:</font> I'm wondering why the this won't work
 339   when parsed:</p>
 340 <pre>
 341 <span class="identifier">    a</span> <span class="special">= +</span><span class="identifier">anychar_p</span><span class="special">;</span>
 342     <span class="identifier">b</span> = <span class="string">'('</span> <span class="special">&gt;&gt;</span> <span class="identifier">a</span> <span class="special">&gt;&gt;</span> <span class="string">')'</span><span class="special">;</span></pre>
 343 <p>Try this:</p>
 344 <pre>
 345 <span class="identifier">    a</span> <span class="special">= +(</span><span class="identifier">anychar_p - </span><span class="string">')'</span><span class="special">);</span>
 346     <span class="identifier">b</span> <span class="special">=</span> <span class="string">'('</span> <span class="special">&gt;&gt;</span> <span class="identifier">a</span> <span class="special">&gt;&gt;</span> <span class="string">')'</span><span class="special">;</span></pre>
 347 <p>David Held writes: That's because it's like the langoliers--it eats everything
 348   up. You usually want to say what it shouldn't eat up by subtracting the terminating
 349   character from the parser. The moral being: Using <tt>*anychar_p</tt> or <tt>+anychar_p</tt>
 350   all by itself is usually a <em>Bad Thing</em>&#8482;.</p>
 351 <p>In other words: Recursive Descent is inherently greedy (however, see <a href="rationale.html#exhaustive_rd">Exhaustive
 352   backtracking and greedy RD</a>).</p>
 353 <p><span class="special"></span><strong><a name="referencing_a_rule_at_construction" id="referencing_a_rule_at_construction"></a>Referencing
 354   a rule at construction time</strong></p>
 355 <p><font color="#FF0000">Question:</font> The code below terminates with a segmentation
 356   fault, but I'm (obviously) confused about what I'm doing wrong.</p>
 357 <pre>    rule<span class="special">&lt;</span>ScannerT<span class="special">,</span> clos<span class="special">::</span>context_t<span class="special">&gt;</span> id <span class="special">=</span> int_p<span class="special">[</span>id<span class="special">.</span>i <span class="special">=</span> arg1<span class="special">];</span></pre>
 358 <p>You have a rule <tt>id</tt> being constructed. Before it is constructed, you
 359   reference <tt>id.i</tt> in the RHS of the constructor. It's a chicken and egg
 360   thing. The closure member <tt>id.i</tt> is not yet constructed at that point.
 361   Using assignment will solve the problem. Try this instead:</p>
 362 <pre>    rule<span class="special">&lt;</span>ScannerT<span class="special">,</span> clos<span class="special">::</span>context_t<span class="special">&gt;</span> id<span class="special">;</span>
 363     id <span class="special">=</span> int_p<span class="special">[</span>id<span class="special">.</span>i <span class="special">=</span> arg1<span class="special">];</span></pre>
 364 <p><span class="special"></span><strong><a name="storing_rules" id="storing_rules"></a>Storing
 365   Rules </strong></p>
 366 <p><font color="#FF0000">Question:</font> Why can't I store rules in STL containers
 367   for later use and why can't I pass and return rules to and from functions by
 368   value? </p>
 369 <p>EBNF is primarily declarative. Like in functional programming, It's a static
 370   recipe and there's no notion of do this then that. However, in Spirit, we managed
 371   to coax imperative C++ to take in declarative EBNF. Hah! Fun!... We did that
 372   by masquerading the C++ assignment operator to mimic EBNF's <tt>::=</tt>, among
 373   other things (e.g. <tt>&gt;&gt;</tt>, <tt>|</tt>, <tt>&amp;</tt> etc.). We used
 374   the rule class to let us do that by giving its assignment operator (and copy
 375   constructor) a different meaning and semantics. Doing so made the rule unlike
 376   any other C++ object. You can't copy it. You can't assign it. You can't place
 377   it in a container (vector, stack, etc).Heck, you can't even return it from a
 378   function *by value*.</p>
 379 <table width="80%" border="0" align="center">
 380   <tr>
 381     <td class="note_box"><img src="theme/alert.gif" width="16" height="16"> The
 382       rule is a weird object, unlike any other C++ object. It does not have the
 383       proper copy and assignment semantics and cannot be stored and passed around
 384       by value.</td>
 385   </tr>
 386 </table>
 387 <p>However nice declarative EBNF is, the dynamic nature of C++ can be an advantage.
 388   We've seen this in action here and there. There are indeed some interesting
 389   applications of dynamic parsers using Spirit. Yet, we haven't fully utilized
 390   the power of dynamic parsing, unless(!), we have a rule that's not so alien
 391   to C++ (i.e. behaves as a good C++ object). With such a beast, we can write
 392   parsers that's defined at run time, as opposed to at compile time.</p>
 393 <p>Now that I started focusing on rules (hey, check out the hunky new rule features),
 394   it might be a good time to implement the rule-holder. It is basically just a
 395   rule, but with C++ object semantics. Yet it's not as simple. Without true garbage
 396   collection, the implementation will be a bit tricky. We can't simply use reference
 397   counting because a rule-holder (hey, anyone here has a better name?) *is-a*
 398   rule, and rules are typically recursive and thus cyclic. The problem is which
 399   will own which.</p>
 400 <p>Ok... this will do for now. You'll definitely see more of the rule-holder in
 401   the coming days.</p>
 402 <p><strong><a name="parsing_ints_and_reals"></a>Parsing Ints and Reals</strong></p>
 403 <p> <font color="#FF0000">Question:</font>  I was trying to parse an int or float value with the <tt>longest_d</tt> directive and put some actors on the alternatives to visualize the results. When I parse &quot;123.456&quot;, the output reports:</p>
 404 <ol>
 405   <li>(int) has been matched: full match = false</li>
 406   <li> (double) has been matched: full match = true</li>
 407 </ol>
 408 <p>That is not what I expected. What am I missing? </p>
 409 <p> Actually, the problem is that both semantic actions of the int and real branch will be triggered because both branches will be tried. This doesn't buy us much. What actually wins in the end is what you expected. But there's no easy way to know which one wins. The problem stems from the ambiguity. </p>
 410 <blockquote>
 411   <p>Case1: Consider this input: &quot;2&quot;. Is it an int or a real? They are both (strictly following the grammar of a real). </p>
 412   <p>Case2 : Now how about &quot;1.0&quot;? Is it an int or a real? They are both, albeit the int part gets a partial match: &quot;1&quot;. That is why you are getting a (partial) match for your <em>int</em> rule (full match = false). </p>
 413 </blockquote>
 414 <p>  Instead of using the <tt>longest_d</tt> to parse ints and reals, what I suggest is to remove the ambiguity and use the plain short-circuiting alternatives. The first step is to use <tt><a href="numerics.html#strict_reals">strict_real_p</a> </tt>to make the first case unambiguous. Unlike
 415
 416
 417   <tt>real_p</tt>, <tt>strict_real_p</tt> requires a dot to be present for a number to be considered a successful match.
 418
 419 Your grammar can be written unambiguously as:</p>
 420 <pre>    strict_real_p<span class="special"> | </span>int_p</pre>
 421 <p> Note that because ambiguity is resolved, attaching actions to both branches is safe. Only one will be triggered:</p>
 422 <pre>    strict_real_p<span class="special">[</span>R<span class="special">] | </span>int_p<span class="special">[</span>I<span class="special">]</span></pre>
 423 <blockquote>
 424   <p> &quot;1.0&quot; ---&gt; triggers R<br>
 425 &quot;2&quot; ---&gt; triggers I</p>
 426 </blockquote>
 427 <p> Again, as a rule of thumb, it is always best to resolve as much ambiguity as possible. The best grammars are those which involve no backtracking at all: an LL(1) grammar. Backtracking and semantic actions do not mix well.</p>
 428 <p><b><a name="output_operator" id="output_operator"></a>BOOST_SPIRIT_DEBUG and missing <tt>operator&lt;&lt;</tt></b></p>
 429 <p><font color="#FF0000">Question:</font> My code compiles fine in release mode but when I try to define <tt>BOOST_SPIRIT_DEBUG</tt> the compiler complains about a missing <tt><span class="keyword">operator</span><span class="special">&lt;&lt;</span></tt>.</p>
 430 <p>When <tt>BOOST_SPIRIT_DEBUG</tt> is defined debug output is generated for
 431   spirit parsers. To this end it is expected that each closure member has the
 432   default output operator defined.</p>
 433 <p>You may provide the operator overload either in the namespace where the
 434   class is declared (will be found through Argument Dependent Lookup) or make it visible where it is
 435   used, that is <tt><span class="keyword">namespace</span> <span
 436   class="identifier">boost</span><span class="special">::</span><span
 437   class="identifier">spirit</span></tt>. Here's an example for <tt><span
 438   class="identifier">std</span><span class="special">::</span><span
 439   class="identifier">pair</span></tt>:</p>
 440 <pre><code>
 441     <span class="preprocessor">#include</span> <span class="string">&lt;iosfwd&gt;</span>
 442     <span class="preprocessor">#include</span> <span class="string">&lt;utility&gt;</span>
 443
 444     <span class="keyword">namespace</span> <span class="identifier">std</span> <span class="special">{</span>
 445
 446         <span class="keyword">template</span> <span class="special">&lt;</span>
 447             <span class="keyword">typename</span> <span class="identifier">C</span><span class="special">,</span>
 448             <span class="keyword">typename</span> <span class="identifier">E</span><span class="special">,</span>
 449             <span class="keyword">typename</span> <span class="identifier">T1</span><span class="special">,</span>
 450             <span class="keyword">typename</span> <span class="identifier">T2</span>
 451         <span class="special">&gt;</span>
 452         <span class="identifier">basic_ostream</span><span class="special">&lt;</span><span class="identifier">C</span><span class="special">,</span> <span class="identifier">E</span><span class="special">&gt;</span> <span class="special">&amp;</span> <span class="keyword">operator</span><span class="special">&lt;&lt;(</span>
 453             <span class="identifier">basic_ostream</span><span class="special">&lt;</span><span class="identifier">C</span><span class="special">,</span> <span class="identifier">E</span><span class="special">&gt;</span> <span class="special">&amp;</span> <span class="identifier">out</span><span class="special">,</span>
 454             <span class="identifier">pair</span><span class="special">&lt;</span><span class="identifier">T1</span><span class="special">,</span> <span class="identifier">T2</span><span class="special">&gt;</span> <span class="keyword">const</span> <span class="special">&amp;</span> <span class="identifier">what</span><span class="special">)</span>
 455         <span class="special">{</span>
 456             <span class="keyword">return</span> <span class="identifier">out</span> <span class="special">&lt;&lt;</span> <span class="string">'('</span> <span class="special">&lt;&lt;</span> <span class="identifier">what</span><span class="special">.</span><span class="identifier">first</span> <span class="special">&lt;&lt;</span> <span class="string">", "</span>
 457                 <span class="special">&lt;&lt;</span> <span class="identifier">what</span><span class="special">.</span><span class="identifier">second</span> <span class="special">&lt;&lt;</span> <span class="string">')'</span><span class="special">;</span>
 458         <span class="special">}</span>
 459
 460     <span class="special">}</span>
 461
 462 </code></pre>
 463 <p><b><a name="repository" id="repository"></a>Applications that used to be part of spirit</b></p>
 464 <p><font color="#FF0000">Question:</font> Where can I find <i>&lt;insert great application&gt;</i>, that used to be part of the Spirit distribution?</p>
 465 <p>Old versions of Spirit used to include applications built with it.
 466   In order to streamline the distribution they were moved to a separate
 467   <a href="http://spirit.sourceforge.net/repository/applications/show_contents.php">applications repository</a>.
 468   In that page you'll find links to full applications that use the Spirit
 469   parser framework. We encourage you to send in your own applications for
 470   inclusion (see the page for instructions).</p>
 471   <p>You may also check out the <a href="http://spirit.sourceforge.net/repository/grammars/show_contents.php">grammars' repository</a>.</p>
 472 <table width="80%" border="0" align="center">
 473   <tr>
 474     <td class="note_box">
 475       <img src="theme/note.gif" width="16" height="16"> You'll still find the
 476       example applications that complement (actually are part of) the
 477       documentation in the usual place: <code>libs/spirit/example</code>.<br>
 478       <br>
 479       <img src="theme/alert.gif" width="16" height="16"> The applications and
 480       grammars listed in the repositories are works of the respective authors.
 481       It is the author's responsibility to provide support and maintenance.
 482       Should you have any questions, please send the author an email.
 483     </td>
 484   </tr>
 485 </table>
 486 <br>
 487 <table border="0">
 488   <tr>
 489     <td width="10"></td>
 490     <td width="30"><a href="../index.html"><img src="theme/u_arr.gif" border="0"></a></td>
 491     <td width="30"><a href="techniques.html"><img src="theme/l_arr.gif" border="0"></a></td>
 492     <td width="30"><a href="rationale.html"><img src="theme/r_arr.gif" border="0"></a></td>
 493   </tr>
 494 </table>
 495 <br>
 496 <hr size="1">
 497 <p class="copyright">Copyright &copy; 1998-2003 Joel de Guzman<br>
 498 <span class="copyright">Copyright &copy; 2002-2003 Hartmut Kaiser </span><br>
 499 <span class="copyright">Copyright &copy; 2006-2007 Tobias Schwinger </span><br>
 500   <br>
 501   <font size="2">Use, modification and distribution is subject to the Boost Software
 502       License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at
 503 http://www.boost.org/LICENSE_1_0.txt)</font></p>
 504 <p class="copyright">&nbsp;</p>
 505 </body>
 506 </html>