extractoruri.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="95" href="#95">95</a>          Object[] sourceLinks = curi.getOutLinks().toArray();
<a name="96" href="#96">96</a>          <strong>for</strong> (Object o : sourceLinks) {
<a name="97" href="#97">97</a>              <strong>if</strong>(o instanceof <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a>) {
<a name="98" href="#98">98</a>                  <strong>final</strong> <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a> wref = (Link)o;
<a name="99" href="#99">99</a>                  extractLink(curi,wref);
<a name="100" href="#100">100</a>             } <strong>else</strong> {
<a name="101" href="#101">101</a>                 LOGGER.severe(<span class="string">"Unexpected type: "</span> + o);
<a name="102" href="#102">102</a>             }
<a name="103" href="#103">103</a>         }
<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>     * Consider a single Link for internal URIs</em>
<a name="108" href="#108">108</a> <em>     * </em>
<a name="109" href="#109">109</a> <em>     * @param curi CrawlURI to add discoveries to </em>
<a name="110" href="#110">110</a> <em>     * @param wref Link to examine for internal URIs</em>
<a name="111" href="#111">111</a> <em>     */</em>
<a name="112" href="#112">112</a>     <strong>protected</strong> <strong>void</strong> extractLink(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi, <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a> wref) {
<a name="113" href="#113">113</a>         <a href="../../../../org/archive/net/UURI.html">UURI</a> source = UURI.from(wref.getDestination());
<a name="114" href="#114">114</a>         <strong>if</strong>(source == <strong>null</strong>) {
<a name="115" href="#115">115</a>             <em class="comment">// shouldn't happen</em>
<a name="116" href="#116">116</a>             <strong>return</strong>; 
<a name="117" href="#117">117</a>         }
<a name="118" href="#118">118</a>         List&lt;String> found = extractQueryStringLinks(source);
<a name="119" href="#119">119</a>         <strong>for</strong> (String uri : found) {
<a name="120" href="#120">120</a>             <strong>try</strong> {
<a name="121" href="#121">121</a>                 curi.createAndAddLink(
<a name="122" href="#122">122</a>                         uri, 
<a name="123" href="#123">123</a>                         Link.SPECULATIVE_MISC,
<a name="124" href="#124">124</a>                         Link.SPECULATIVE_HOP);
<a name="125" href="#125">125</a>                 numberOfLinksExtracted++;
<a name="126" href="#126">126</a>             } <strong>catch</strong> (URIException e) {
<a name="127" href="#127">127</a>                 LOGGER.log(Level.FINE, <span class="string">"bad URI"</span>, e);
<a name="128" href="#128">128</a>             }
<a name="129" href="#129">129</a>         }
<a name="130" href="#130">130</a>         <em class="comment">// TODO: consider path URIs too</em>
<a name="131" href="#131">131</a>         
<a name="132" href="#132">132</a>     }
<a name="133" href="#133">133</a> 
<a name="134" href="#134">134</a>     <em>/**<em>*</em></em>
<a name="135" href="#135">135</a> <em>     * Look for URIs inside the supplied UURI.</em>
<a name="136" href="#136">136</a> <em>     * </em>
<a name="137" href="#137">137</a> <em>     * Static for ease of testing or outside use. </em>
<a name="138" href="#138">138</a> <em>     * </em>
<a name="139" href="#139">139</a> <em>     * @param source UURI to example</em>
<a name="140" href="#140">140</a> <em>     * @return List of discovered String URIs.</em>
<a name="141" href="#141">141</a> <em>     */</em>
<a name="142" href="#142">142</a>     <strong>protected</strong> <strong>static</strong> List&lt;String> extractQueryStringLinks(<a href="../../../../org/archive/net/UURI.html">UURI</a> source) {
<a name="143" href="#143">143</a>         List&lt;String> results = <strong>new</strong> ArrayList&lt;String>(); 
<a name="144" href="#144">144</a>         String decodedQuery;
<a name="145" href="#145">145</a>         <strong>try</strong> {
<a name="146" href="#146">146</a>             decodedQuery = source.getQuery();
<a name="147" href="#147">147</a>         } <strong>catch</strong> (URIException e1) {
<a name="148" href="#148">148</a>             <em class="comment">// shouldn't happen</em>
<a name="149" href="#149">149</a>             <strong>return</strong> results;
<a name="150" href="#150">150</a>         }
<a name="151" href="#151">151</a>         <strong>if</strong>(decodedQuery==<strong>null</strong>) {
<a name="152" href="#152">152</a>             <strong>return</strong> results;
<a name="153" href="#153">153</a>         }
<a name="154" href="#154">154</a>         <em class="comment">// check if full query-string appears to be http(s) URI</em>
<a name="155" href="#155">155</a>         Matcher m = TextUtils.getMatcher(ABS_HTTP_URI_PATTERN,decodedQuery);
<a name="156" href="#156">156</a>         <strong>if</strong>(m.matches()) {
<a name="157" href="#157">157</a>             TextUtils.recycleMatcher(m);
<a name="158" href="#158">158</a>             results.add(decodedQuery);
<a name="159" href="#159">159</a>         }
<a name="160" href="#160">160</a>         <em class="comment">// split into params, see if any param value is http(s) URI</em>
<a name="161" href="#161">161</a>         String rawQuery = <strong>new</strong> String(source.getRawQuery());
<a name="162" href="#162">162</a>         String[] params = rawQuery.split(<span class="string">"&amp;"</span>);
<a name="163" href="#163">163</a>         <strong>for</strong> (String param : params) {
<a name="164" href="#164">164</a>             String[] keyVal = param.split(<span class="string">"="</span>);
<a name="165" href="#165">165</a>             <strong>if</strong>(keyVal.length==2) {
<a name="166" href="#166">166</a>                 String candidate;
<a name="167" href="#167">167</a>                 <strong>try</strong> {
<a name="168" href="#168">168</a>                     candidate = LaxURLCodec.DEFAULT.decode(keyVal[1]);
<a name="169" href="#169">169</a>                 } <strong>catch</strong> (DecoderException e) {
<a name="170" href="#170">170</a>                     <strong>continue</strong>;
<a name="171" href="#171">171</a>                 }
<a name="172" href="#172">172</a>                 <em class="comment">// TODO: use other non-UTF8 codecs when appropriate</em>
<a name="173" href="#173">173</a>                 m.reset(candidate);
<a name="174" href="#174">174</a>                 <strong>if</strong>(m.matches()) {
<a name="175" href="#175">175</a>                     results.add(candidate);
<a name="176" href="#176">176</a>                 }
<a name="177" href="#177">177</a>             }
<a name="178" href="#178">178</a>         }
<a name="179" href="#179">179</a>         <strong>return</strong> results;
<a name="180" href="#180">180</a>     }
<a name="181" href="#181">181</a> 
<a name="182" href="#182">182</a>     <strong>public</strong> String report() {
<a name="183" href="#183">183</a>         StringBuffer ret = <strong>new</strong> StringBuffer();
<a name="184" href="#184">184</a>         ret.append(<span class="string">"Processor: "</span>+ExtractorURI.<strong>class</strong>.getName()+<span class="string">"\n"</span>);
<a name="185" href="#185">185</a>         ret.append(<span class="string">"  Function:          Extracts links inside other URIs\n"</span>);
<a name="186" href="#186">186</a>         ret.append(<span class="string">"  CrawlURIs handled: "</span> + numberOfCURIsHandled + <span class="string">"\n"</span>);
<a name="187" href="#187">187</a>         ret.append(<span class="string">"  Links extracted:   "</span> + numberOfLinksExtracted + <span class="string">"\n\n"</span>);
<a name="188" href="#188">188</a> 
<a name="189" href="#189">189</a>         <strong>return</strong> ret.toString();
<a name="190" href="#190">190</a>     }
<a name="191" href="#191">191</a> }
</pre>
<hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/">Maven</a></div></body>
</html>