extractorimplieduri.html

来自「网络爬虫开源代码」· HTML 代码 · 共 204 行 · 第 1/2 页

<a name="94" href="#94">94</a>                      <span class="string">"built. The capturing groups of this expression are "</span> +
<a name="95" href="#95">95</a>                      <span class="string">"available for the build replacement pattern."</span>, <span class="string">""</span>));
<a name="96" href="#96">96</a>          addElementToDefinition(
<a name="97" href="#97">97</a>                  <strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_BUILD_PATTERN, 
<a name="98" href="#98">98</a>                      <span class="string">"Replacement pattern to build 'implied' URI, using "</span> +
<a name="99" href="#99">99</a>                      <span class="string">"captured groups of trigger expression."</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_REMOVE_TRIGGER_URIS, 
<a name="102" href="#102">102</a>                     <span class="string">"If true, all URIs that match trigger regular expression "</span> +
<a name="103" href="#103">103</a>                     <span class="string">"are removed from the list of extracted URIs. "</span> +
<a name="104" href="#104">104</a>                     <span class="string">"Default is false."</span>, Boolean.FALSE));
<a name="105" href="#105">105</a>     }
<a name="106" href="#106">106</a> 
<a name="107" href="#107">107</a>     <em>/**<em>*</em></em>
<a name="108" href="#108">108</a> <em>     * Perform usual extraction on a CrawlURI</em>
<a name="109" href="#109">109</a> <em>     * </em>
<a name="110" href="#110">110</a> <em>     * @param curi Crawl URI to process.</em>
<a name="111" href="#111">111</a> <em>     */</em>
<a name="112" href="#112">112</a>     <strong>public</strong> <strong>void</strong> extract(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="113" href="#113">113</a> 
<a name="114" href="#114">114</a>         <strong>this</strong>.numberOfCURIsHandled++;
<a name="115" href="#115">115</a>         <em class="comment">// use array copy because discoveriess will add to outlinks</em>
<a name="116" href="#116">116</a>         Collection&lt;Link> links = curi.getOutLinks();
<a name="117" href="#117">117</a>         <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a>[] sourceLinks = links.toArray(<strong>new</strong> <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a>[links.size()]);
<a name="118" href="#118">118</a>         <strong>for</strong> (Link wref: sourceLinks) {
<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>                 	
<a name="130" href="#130">130</a>                     numberOfLinksExtracted++;
<a name="131" href="#131">131</a>                 	
<a name="132" href="#132">132</a>                     <strong>final</strong> <strong>boolean</strong> removeTriggerURI = 
<a name="133" href="#133">133</a>                     	((Boolean)getUncheckedAttribute(curi,
<a name="134" href="#134">134</a>                     			ATTR_REMOVE_TRIGGER_URIS)).booleanValue();
<a name="135" href="#135">135</a> 
<a name="136" href="#136">136</a>                     <em class="comment">// remove trigger URI from the outlinks if configured so.</em>
<a name="137" href="#137">137</a>                     <strong>if</strong> (removeTriggerURI) {
<a name="138" href="#138">138</a>                     	<strong>if</strong> (curi.getOutLinks().remove(wref)) {
<a name="139" href="#139">139</a>                     		LOGGER.log(Level.FINE, wref.getDestination() + 
<a name="140" href="#140">140</a>                     				<span class="string">" has been removed from "</span> + 
<a name="141" href="#141">141</a>                     				wref.getSource() + <span class="string">" outlinks list."</span>);                    	
<a name="142" href="#142">142</a>                     		numberOfLinksExtracted--;
<a name="143" href="#143">143</a> 
<a name="144" href="#144">144</a>                     	} <strong>else</strong> {
<a name="145" href="#145">145</a>                         	LOGGER.log(Level.FINE, <span class="string">"Failed to remove "</span> + 
<a name="146" href="#146">146</a>                         			wref.getDestination() + <span class="string">" from "</span> + 
<a name="147" href="#147">147</a>                         			wref.getSource()+ <span class="string">" outlinks list."</span>);             		
<a name="148" href="#148">148</a>                     	}
<a name="149" href="#149">149</a>                     }
<a name="150" href="#150">150</a>                     
<a name="151" href="#151">151</a>                 } <strong>catch</strong> (URIException e) {
<a name="152" href="#152">152</a>                     LOGGER.log(Level.FINE, <span class="string">"bad URI"</span>, e);
<a name="153" href="#153">153</a>                 }
<a name="154" href="#154">154</a>             }
<a name="155" href="#155">155</a>         }
<a name="156" href="#156">156</a>     }
<a name="157" href="#157">157</a>     
<a name="158" href="#158">158</a>     <em>/**<em>*</em></em>
<a name="159" href="#159">159</a> <em>     * Utility method for extracting 'implied' URI given a source uri, </em>
<a name="160" href="#160">160</a> <em>     * trigger pattern, and build pattern. </em>
<a name="161" href="#161">161</a> <em>     * </em>
<a name="162" href="#162">162</a> <em>     * @param uri source to check for implied URI</em>
<a name="163" href="#163">163</a> <em>     * @param trigger regex pattern which if matched implies another URI</em>
<a name="164" href="#164">164</a> <em>     * @param build replacement pattern to build the implied URI</em>
<a name="165" href="#165">165</a> <em>     * @return implied URI, or null if none</em>
<a name="166" href="#166">166</a> <em>     */</em>
<a name="167" href="#167">167</a>     <strong>protected</strong> <strong>static</strong> String extractImplied(CharSequence uri, String trigger, String build) {
<a name="168" href="#168">168</a>         <strong>if</strong>(trigger.length()==0) {
<a name="169" href="#169">169</a>             <em class="comment">// short-circuit empty-string trigger</em>
<a name="170" href="#170">170</a>             <strong>return</strong> <strong>null</strong>; 
<a name="171" href="#171">171</a>         }
<a name="172" href="#172">172</a>         Matcher m = TextUtils.getMatcher(trigger, uri);
<a name="173" href="#173">173</a>         <strong>if</strong>(m.matches()) {
<a name="174" href="#174">174</a>             String result = m.replaceFirst(build);
<a name="175" href="#175">175</a>             TextUtils.recycleMatcher(m);
<a name="176" href="#176">176</a>             <strong>return</strong> result; 
<a name="177" href="#177">177</a>         }
<a name="178" href="#178">178</a>         <strong>return</strong> <strong>null</strong>; 
<a name="179" href="#179">179</a>     }
<a name="180" href="#180">180</a> 
<a name="181" href="#181">181</a>     <strong>public</strong> String report() {
<a name="182" href="#182">182</a>         StringBuffer ret = <strong>new</strong> StringBuffer();
<a name="183" href="#183">183</a>         ret.append(<span class="string">"Processor: "</span>+ExtractorImpliedURI.<strong>class</strong>.getName()+<span class="string">"\n"</span>);
<a name="184" href="#184">184</a>         ret.append(<span class="string">"  Function:          Extracts links inside other URIs\n"</span>);
<a name="185" href="#185">185</a>         ret.append(<span class="string">"  CrawlURIs handled: "</span> + numberOfCURIsHandled + <span class="string">"\n"</span>);
<a name="186" href="#186">186</a>         ret.append(<span class="string">"  Links extracted:   "</span> + numberOfLinksExtracted + <span class="string">"\n\n"</span>);
<a name="187" href="#187">187</a> 
<a name="188" href="#188">188</a>         <strong>return</strong> ret.toString();
<a name="189" href="#189">189</a>     }
<a name="190" href="#190">190</a> }
</pre>
<hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/">Maven</a></div></body>
</html>