extractorjs.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="96" href="#96">96</a>              (!curi.toString().toLowerCase().endsWith(<span class="string">".js"</span>)) &amp;&amp;
<a name="97" href="#97">97</a>              (curi.getViaContext() == <strong>null</strong> || !curi.getViaContext().
<a name="98" href="#98">98</a>                  toString().toLowerCase().startsWith(<span class="string">"script"</span>))) {
<a name="99" href="#99">99</a>              <strong>return</strong>;
<a name="100" href="#100">100</a>         }
<a name="101" href="#101">101</a> 
<a name="102" href="#102">102</a>         <strong>this</strong>.numberOfCURIsHandled++;
<a name="103" href="#103">103</a> 
<a name="104" href="#104">104</a>         <a href="../../../../org/archive/io/ReplayCharSequence.html">ReplayCharSequence</a> cs = <strong>null</strong>;
<a name="105" href="#105">105</a>         <strong>try</strong> {
<a name="106" href="#106">106</a>             cs = curi.getHttpRecorder().getReplayCharSequence();
<a name="107" href="#107">107</a>         } <strong>catch</strong> (IOException e) {
<a name="108" href="#108">108</a>             curi.addLocalizedError(<strong>this</strong>.getName(), e,
<a name="109" href="#109">109</a>             	<span class="string">"Failed get of replay char sequence."</span>);
<a name="110" href="#110">110</a>         }
<a name="111" href="#111">111</a>         <strong>if</strong> (cs == <strong>null</strong>) {
<a name="112" href="#112">112</a>             LOGGER.warning(<span class="string">"Failed getting ReplayCharSequence: "</span> +
<a name="113" href="#113">113</a>                 curi.toString());
<a name="114" href="#114">114</a>             <strong>return</strong>;
<a name="115" href="#115">115</a>         }
<a name="116" href="#116">116</a> 
<a name="117" href="#117">117</a>         <strong>try</strong> {
<a name="118" href="#118">118</a>             <strong>try</strong> {
<a name="119" href="#119">119</a>                 numberOfLinksExtracted += considerStrings(curi, cs,
<a name="120" href="#120">120</a>                         getController(), <strong>true</strong>);
<a name="121" href="#121">121</a>             } <strong>catch</strong> (StackOverflowError e) {
<a name="122" href="#122">122</a>                 DevUtils.warnHandle(e, <span class="string">"ExtractorJS StackOverflowError"</span>);
<a name="123" href="#123">123</a>             }
<a name="124" href="#124">124</a>             <em class="comment">// Set flag to indicate that link extraction is completed.</em>
<a name="125" href="#125">125</a>             curi.linkExtractorFinished();
<a name="126" href="#126">126</a>         } <strong>finally</strong> {
<a name="127" href="#127">127</a>             <em class="comment">// Done w/ the ReplayCharSequence. Close it.</em>
<a name="128" href="#128">128</a>             <strong>if</strong> (cs != <strong>null</strong>) {
<a name="129" href="#129">129</a>                 <strong>try</strong> {
<a name="130" href="#130">130</a>                     cs.close();
<a name="131" href="#131">131</a>                 } <strong>catch</strong> (IOException ioe) {
<a name="132" href="#132">132</a>                     LOGGER.warning(TextUtils.exceptionToString(
<a name="133" href="#133">133</a>                         <span class="string">"Failed close of ReplayCharSequence."</span>, ioe));
<a name="134" href="#134">134</a>                 }
<a name="135" href="#135">135</a>             }
<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>     <strong>public</strong> <strong>static</strong> <strong>long</strong> considerStrings(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi, CharSequence cs,
<a name="140" href="#140">140</a>             <a href="../../../../org/archive/crawler/framework/CrawlController.html">CrawlController</a> controller, <strong>boolean</strong> handlingJSFile) {
<a name="141" href="#141">141</a>         <strong>long</strong> foundLinks = 0;
<a name="142" href="#142">142</a>         Matcher strings =
<a name="143" href="#143">143</a>             TextUtils.getMatcher(JAVASCRIPT_STRING_EXTRACTOR, cs);
<a name="144" href="#144">144</a>         <strong>while</strong>(strings.find()) {
<a name="145" href="#145">145</a>             CharSequence subsequence =
<a name="146" href="#146">146</a>                 cs.subSequence(strings.start(2), strings.end(2));
<a name="147" href="#147">147</a>             Matcher uri =
<a name="148" href="#148">148</a>                 TextUtils.getMatcher(STRING_URI_DETECTOR, subsequence);
<a name="149" href="#149">149</a>             <strong>if</strong>(uri.matches()) {
<a name="150" href="#150">150</a>                 String string = uri.group();
<a name="151" href="#151">151</a>                 string = TextUtils.replaceAll(ESCAPED_AMP, string, AMP);
<a name="152" href="#152">152</a>                 foundLinks++;
<a name="153" href="#153">153</a>                 <strong>try</strong> {
<a name="154" href="#154">154</a>                     <strong>if</strong> (handlingJSFile) {
<a name="155" href="#155">155</a>                         curi.createAndAddLinkRelativeToVia(string,
<a name="156" href="#156">156</a>                             Link.JS_MISC, Link.SPECULATIVE_HOP);
<a name="157" href="#157">157</a>                     } <strong>else</strong> {
<a name="158" href="#158">158</a>                         curi.createAndAddLinkRelativeToBase(string,
<a name="159" href="#159">159</a>                             Link.JS_MISC, Link.SPECULATIVE_HOP);
<a name="160" href="#160">160</a>                     }
<a name="161" href="#161">161</a>                 } <strong>catch</strong> (URIException e) {
<a name="162" href="#162">162</a>                     <em class="comment">// There may not be a controller (e.g. If we're being run</em>
<a name="163" href="#163">163</a>                     <em class="comment">// by the extractor tool).</em>
<a name="164" href="#164">164</a>                     <strong>if</strong> (controller != <strong>null</strong>) {
<a name="165" href="#165">165</a>                         controller.logUriError(e, curi.getUURI(), string);
<a name="166" href="#166">166</a>                     } <strong>else</strong> {
<a name="167" href="#167">167</a>                         LOGGER.info(curi + <span class="string">", "</span> + string + <span class="string">": "</span> +
<a name="168" href="#168">168</a>                             e.getMessage());
<a name="169" href="#169">169</a>                     }
<a name="170" href="#170">170</a>                 }
<a name="171" href="#171">171</a>             } <strong>else</strong> {
<a name="172" href="#172">172</a>                foundLinks += considerStrings(curi, subsequence,
<a name="173" href="#173">173</a>                    controller, handlingJSFile);
<a name="174" href="#174">174</a>             }
<a name="175" href="#175">175</a>             TextUtils.recycleMatcher(uri);
<a name="176" href="#176">176</a>         }
<a name="177" href="#177">177</a>         TextUtils.recycleMatcher(strings);
<a name="178" href="#178">178</a>         <strong>return</strong> foundLinks;
<a name="179" href="#179">179</a>     }
<a name="180" href="#180">180</a> 
<a name="181" href="#181">181</a>     <em class="comment">/*<em class="comment"> (non-Javadoc)</em></em>
<a name="182" href="#182">182</a> <em class="comment">     * @see org.archive.crawler.framework.Processor#report()</em>
<a name="183" href="#183">183</a> <em class="comment">     */</em>
<a name="184" href="#184">184</a>     <strong>public</strong> String report() {
<a name="185" href="#185">185</a>         StringBuffer ret = <strong>new</strong> StringBuffer();
<a name="186" href="#186">186</a>         ret.append(<span class="string">"Processor: org.archive.crawler.extractor.ExtractorJS\n"</span>);
<a name="187" href="#187">187</a>         ret.append(<span class="string">"  Function:          Link extraction on JavaScript code\n"</span>);
<a name="188" href="#188">188</a>         ret.append(<span class="string">"  CrawlURIs handled: "</span> + numberOfCURIsHandled + <span class="string">"\n"</span>);
<a name="189" href="#189">189</a>         ret.append(<span class="string">"  Links extracted:   "</span> + numberOfLinksExtracted + <span class="string">"\n\n"</span>);
<a name="190" href="#190">190</a> 
<a name="191" href="#191">191</a>         <strong>return</strong> ret.toString();
<a name="192" href="#192">192</a>     }
<a name="193" href="#193">193</a> }
</pre>
<hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/">Maven</a></div></body>
</html>