extractorimplieduri.html

用JAVA编写的,在做实验的时候留下来的,本来想删的,但是传上来,大家分享吧

<a name="85" href="#85">85</a>  <em>     * Constructor</em>
<a name="86" href="#86">86</a>  <em>     * </em>
<a name="87" href="#87">87</a>  <em>     * @param name</em>
<a name="88" href="#88">88</a>  <em>     */</em>
<a name="89" href="#89">89</a>      <strong>public</strong> <a href="../../../../org/archive/crawler/extractor/ExtractorImpliedURI.html">ExtractorImpliedURI</a>(String name) {
<a name="90" href="#90">90</a>          <strong>super</strong>(name, <span class="string">"Implied URI Extractor. Finds URIs implied by other "</span> +
<a name="91" href="#91">91</a>                  <span class="string">"URIs according to regex/replacement patterns. Should "</span> +
<a name="92" href="#92">92</a>                  <span class="string">"appear after most other extractors."</span>);
<a name="93" href="#93">93</a>  
<a name="94" href="#94">94</a>          addElementToDefinition(
<a name="95" href="#95">95</a>              <strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_TRIGGER_REGEXP, 
<a name="96" href="#96">96</a>                      <span class="string">"Triggering regular expression. When a discovered URI "</span> +
<a name="97" href="#97">97</a>                      <span class="string">"matches this pattern, the 'implied' URI will be "</span> +
<a name="98" href="#98">98</a>                      <span class="string">"built. The capturing groups of this expression are "</span> +
<a name="99" href="#99">99</a>                      <span class="string">"available for the build replacement pattern."</span>, <span class="string">""</span>));
<a name="100" href="#100">100</a>         addElementToDefinition(
<a name="101" href="#101">101</a>                 <strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_BUILD_PATTERN, 
<a name="102" href="#102">102</a>                     <span class="string">"Replacement pattern to build 'implied' URI, using "</span> +
<a name="103" href="#103">103</a>                     <span class="string">"captured groups of trigger expression."</span>, <span class="string">""</span>));
<a name="104" href="#104">104</a>     }
<a name="105" href="#105">105</a> 
<a name="106" href="#106">106</a>     <em>/**<em>*</em></em>
<a name="107" href="#107">107</a> <em>     * Perform usual extraction on a CrawlURI</em>
<a name="108" href="#108">108</a> <em>     * </em>
<a name="109" href="#109">109</a> <em>     * @param curi Crawl URI to process.</em>
<a name="110" href="#110">110</a> <em>     */</em>
<a name="111" href="#111">111</a>     <strong>public</strong> <strong>void</strong> extract(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="112" href="#112">112</a> 
<a name="113" href="#113">113</a>         <strong>this</strong>.numberOfCURIsHandled++;
<a name="114" href="#114">114</a>         <em class="comment">// use array copy because discoveriess will add to outlinks</em>
<a name="115" href="#115">115</a>         Object[] sourceLinks = curi.getOutLinks().toArray();
<a name="116" href="#116">116</a>         <strong>for</strong> (Object o : sourceLinks) {
<a name="117" href="#117">117</a>             <strong>if</strong>(o instanceof <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a>) {
<a name="118" href="#118">118</a>                 <strong>final</strong> <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a> wref = (Link)o;
<a name="119" href="#119">119</a>                 String implied = extractImplied(
<a name="120" href="#120">120</a>                         wref.getDestination(),
<a name="121" href="#121">121</a>                         (String)getUncheckedAttribute(curi,ATTR_TRIGGER_REGEXP),
<a name="122" href="#122">122</a>                         (String)getUncheckedAttribute(curi,ATTR_BUILD_PATTERN));
<a name="123" href="#123">123</a>                 <strong>if</strong> (implied!=<strong>null</strong>) {
<a name="124" href="#124">124</a>                     <strong>try</strong> {
<a name="125" href="#125">125</a>                         curi.createAndAddLink(
<a name="126" href="#126">126</a>                                 implied, 
<a name="127" href="#127">127</a>                                 Link.SPECULATIVE_MISC,
<a name="128" href="#128">128</a>                                 Link.SPECULATIVE_HOP);
<a name="129" href="#129">129</a>                         numberOfLinksExtracted++;
<a name="130" href="#130">130</a>                     } <strong>catch</strong> (URIException e) {
<a name="131" href="#131">131</a>                         LOGGER.log(Level.FINE, <span class="string">"bad URI"</span>, e);
<a name="132" href="#132">132</a>                     }
<a name="133" href="#133">133</a>                 }
<a name="134" href="#134">134</a>             } <strong>else</strong> {
<a name="135" href="#135">135</a>                 LOGGER.severe(<span class="string">"Unexpected type: "</span> + o);
<a name="136" href="#136">136</a>             }
<a name="137" href="#137">137</a>         }
<a name="138" href="#138">138</a>     }
<a name="139" href="#139">139</a>     
<a name="140" href="#140">140</a>     <em>/**<em>*</em></em>
<a name="141" href="#141">141</a> <em>     * Utility method for extracting 'implied' URI given a source uri, </em>
<a name="142" href="#142">142</a> <em>     * trigger pattern, and build pattern. </em>
<a name="143" href="#143">143</a> <em>     * </em>
<a name="144" href="#144">144</a> <em>     * @param uri source to check for implied URI</em>
<a name="145" href="#145">145</a> <em>     * @param trigger regex pattern which if matched implies another URI</em>
<a name="146" href="#146">146</a> <em>     * @param build replacement pattern to build the implied URI</em>
<a name="147" href="#147">147</a> <em>     * @return implied URI, or null if none</em>
<a name="148" href="#148">148</a> <em>     */</em>
<a name="149" href="#149">149</a>     <strong>protected</strong> <strong>static</strong> String extractImplied(CharSequence uri, String trigger, String build) {
<a name="150" href="#150">150</a>         <strong>if</strong>(trigger.length()==0) {
<a name="151" href="#151">151</a>             <em class="comment">// short-circuit empty-string trigger</em>
<a name="152" href="#152">152</a>             <strong>return</strong> <strong>null</strong>; 
<a name="153" href="#153">153</a>         }
<a name="154" href="#154">154</a>         Matcher m = TextUtils.getMatcher(trigger, uri);
<a name="155" href="#155">155</a>         <strong>if</strong>(m.matches()) {
<a name="156" href="#156">156</a>             String result = m.replaceFirst(build);
<a name="157" href="#157">157</a>             TextUtils.recycleMatcher(m);
<a name="158" href="#158">158</a>             <strong>return</strong> result; 
<a name="159" href="#159">159</a>         }
<a name="160" href="#160">160</a>         <strong>return</strong> <strong>null</strong>; 
<a name="161" href="#161">161</a>     }
<a name="162" href="#162">162</a> 
<a name="163" href="#163">163</a>     <strong>public</strong> String report() {
<a name="164" href="#164">164</a>         StringBuffer ret = <strong>new</strong> StringBuffer();
<a name="165" href="#165">165</a>         ret.append(<span class="string">"Processor: "</span>+ExtractorImpliedURI.<strong>class</strong>.getName()+<span class="string">"\n"</span>);
<a name="166" href="#166">166</a>         ret.append(<span class="string">"  Function:          Extracts links inside other URIs\n"</span>);
<a name="167" href="#167">167</a>         ret.append(<span class="string">"  CrawlURIs handled: "</span> + numberOfCURIsHandled + <span class="string">"\n"</span>);
<a name="168" href="#168">168</a>         ret.append(<span class="string">"  Links extracted:   "</span> + numberOfLinksExtracted + <span class="string">"\n\n"</span>);
<a name="169" href="#169">169</a> 
<a name="170" href="#170">170</a>         <strong>return</strong> ret.toString();
<a name="171" href="#171">171</a>     }
<a name="172" href="#172">172</a> }
</pre>
<hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/">Maven</a></div></body>
</html>