crawlserver.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="121" href="#121">121</a> <em>     * @throws IOException</em>
<a name="122" href="#122">122</a> <em>     */</em>
<a name="123" href="#123">123</a>     <strong>public</strong> <strong>void</strong> updateRobots(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="124" href="#124">124</a>         <a href="../../../../org/archive/crawler/datamodel/RobotsHonoringPolicy.html">RobotsHonoringPolicy</a> honoringPolicy =
<a name="125" href="#125">125</a>             settingsHandler.getOrder().getRobotsHonoringPolicy();
<a name="126" href="#126">126</a> 
<a name="127" href="#127">127</a>         robotsFetched = System.currentTimeMillis();
<a name="128" href="#128">128</a> 
<a name="129" href="#129">129</a>         <strong>boolean</strong> gotSomething = curi.getFetchStatus() > 0
<a name="130" href="#130">130</a>                 &amp;&amp; curi.isHttpTransaction();
<a name="131" href="#131">131</a>         <strong>if</strong> (!gotSomething &amp;&amp; curi.getFetchAttempts() &lt; MIN_ROBOTS_RETRIES) {
<a name="132" href="#132">132</a>             <em class="comment">// robots.txt lookup failed, no reason to consider IGNORE yet</em>
<a name="133" href="#133">133</a>             validRobots = false;
<a name="134" href="#134">134</a>             <strong>return</strong>;
<a name="135" href="#135">135</a>         }
<a name="136" href="#136">136</a>         
<a name="137" href="#137">137</a>         <a href="../../../../org/archive/crawler/settings/CrawlerSettings.html">CrawlerSettings</a> settings = getSettings(curi);
<a name="138" href="#138">138</a>         <strong>int</strong> type = honoringPolicy.getType(settings);
<a name="139" href="#139">139</a>         <strong>if</strong> (type == RobotsHonoringPolicy.IGNORE) {
<a name="140" href="#140">140</a>             <em class="comment">// IGNORE = ALLOWALL</em>
<a name="141" href="#141">141</a>             robots = RobotsExclusionPolicy.ALLOWALL;
<a name="142" href="#142">142</a>             validRobots = <strong>true</strong>;
<a name="143" href="#143">143</a>             <strong>return</strong>;
<a name="144" href="#144">144</a>         }
<a name="145" href="#145">145</a>         
<a name="146" href="#146">146</a>         <strong>if</strong>(!gotSomething) {
<a name="147" href="#147">147</a>             <em class="comment">// robots.txt lookup failed and policy not IGNORE</em>
<a name="148" href="#148">148</a>             validRobots = false;
<a name="149" href="#149">149</a>             <strong>return</strong>;
<a name="150" href="#150">150</a>         }
<a name="151" href="#151">151</a>         
<a name="152" href="#152">152</a>         <strong>if</strong> (!curi.is2XXSuccess()) {
<a name="153" href="#153">153</a>             <em class="comment">// Not found or anything but a status code in the 2xx range is</em>
<a name="154" href="#154">154</a>             <em class="comment">// treated as giving access to all of a sites' content.</em>
<a name="155" href="#155">155</a>             <em class="comment">// This is the prevailing practice of Google, since 4xx</em>
<a name="156" href="#156">156</a>             <em class="comment">// responses on robots.txt are usually indicative of a </em>
<a name="157" href="#157">157</a>             <em class="comment">// misconfiguration or blanket-block, not an intentional</em>
<a name="158" href="#158">158</a>             <em class="comment">// indicator of partial blocking. </em>
<a name="159" href="#159">159</a>             <em class="comment">// TODO: consider handling server errors, redirects differently</em>
<a name="160" href="#160">160</a>             robots = RobotsExclusionPolicy.ALLOWALL;
<a name="161" href="#161">161</a>             validRobots = <strong>true</strong>;
<a name="162" href="#162">162</a>             <strong>return</strong>;
<a name="163" href="#163">163</a>         }
<a name="164" href="#164">164</a> 
<a name="165" href="#165">165</a>         <a href="../../../../org/archive/io/ReplayInputStream.html">ReplayInputStream</a> contentBodyStream = <strong>null</strong>;
<a name="166" href="#166">166</a>         <strong>try</strong> {
<a name="167" href="#167">167</a>             <strong>try</strong> {
<a name="168" href="#168">168</a>                 BufferedReader reader;
<a name="169" href="#169">169</a>                 <strong>if</strong> (type == RobotsHonoringPolicy.CUSTOM) {
<a name="170" href="#170">170</a>                     reader = <strong>new</strong> BufferedReader(<strong>new</strong> StringReader(honoringPolicy
<a name="171" href="#171">171</a>                             .getCustomRobots(settings)));
<a name="172" href="#172">172</a>                 } <strong>else</strong> {
<a name="173" href="#173">173</a>                     contentBodyStream = curi.getHttpRecorder()
<a name="174" href="#174">174</a>                             .getRecordedInput().getContentReplayInputStream();
<a name="175" href="#175">175</a> 
<a name="176" href="#176">176</a>                     contentBodyStream.setToResponseBodyStart();
<a name="177" href="#177">177</a>                     reader = <strong>new</strong> BufferedReader(<strong>new</strong> InputStreamReader(
<a name="178" href="#178">178</a>                             contentBodyStream));
<a name="179" href="#179">179</a>                 }
<a name="180" href="#180">180</a>                 robots = RobotsExclusionPolicy.policyFor(settings,
<a name="181" href="#181">181</a>                         reader, honoringPolicy);
<a name="182" href="#182">182</a>                 validRobots = <strong>true</strong>;
<a name="183" href="#183">183</a>             } <strong>finally</strong> {
<a name="184" href="#184">184</a>                 <strong>if</strong> (contentBodyStream != <strong>null</strong>) {
<a name="185" href="#185">185</a>                     contentBodyStream.close();
<a name="186" href="#186">186</a>                 }
<a name="187" href="#187">187</a>             }
<a name="188" href="#188">188</a>         } <strong>catch</strong> (IOException e) {
<a name="189" href="#189">189</a>             robots = RobotsExclusionPolicy.ALLOWALL;
<a name="190" href="#190">190</a>             validRobots = <strong>true</strong>;
<a name="191" href="#191">191</a>             curi.addLocalizedError(getName(), e,
<a name="192" href="#192">192</a>                     <span class="string">"robots.txt parsing IOException"</span>);
<a name="193" href="#193">193</a>         }
<a name="194" href="#194">194</a>     }
<a name="195" href="#195">195</a> 
<a name="196" href="#196">196</a>     <em>/**<em>*</em></em>
<a name="197" href="#197">197</a> <em>     * @return Returns the time when robots.txt was fetched.</em>
<a name="198" href="#198">198</a> <em>     */</em>
<a name="199" href="#199">199</a>     <strong>public</strong> <strong>long</strong> getRobotsFetchedTime() {
<a name="200" href="#200">200</a>         <strong>return</strong> robotsFetched;
<a name="201" href="#201">201</a>     }
<a name="202" href="#202">202</a> 
<a name="203" href="#203">203</a>     <em>/**<em>*</em></em>
<a name="204" href="#204">204</a> <em>     * @return The server string which might include a port number.</em>
<a name="205" href="#205">205</a> <em>     */</em>
<a name="206" href="#206">206</a>     <strong>public</strong> String getName() {
<a name="207" href="#207">207</a>        <strong>return</strong> server;
<a name="208" href="#208">208</a>     }
<a name="209" href="#209">209</a> 
<a name="210" href="#210">210</a>     <em>/**<em>* Get the port number for this server.</em></em>
<a name="211" href="#211">211</a> <em>     *</em>
<a name="212" href="#212">212</a> <em>     * @return the port number or -1 if not known (uses default for protocol)</em>
<a name="213" href="#213">213</a> <em>     */</em>
<a name="214" href="#214">214</a>     <strong>public</strong> <strong>int</strong> getPort() {
<a name="215" href="#215">215</a>         <strong>return</strong> port;
<a name="216" href="#216">216</a>     }
<a name="217" href="#217">217</a> 
<a name="218" href="#218">218</a>     <em>/**<em>* </em></em>
<a name="219" href="#219">219</a> <em>     * Called when object is being deserialized.</em>
<a name="220" href="#220">220</a> <em>     * In addition to the default java deserialization, this method</em>
<a name="221" href="#221">221</a> <em>     * re-establishes the references to settings handler and robots honoring</em>
<a name="222" href="#222">222</a> <em>     * policy.</em>
<a name="223" href="#223">223</a> <em>     *</em>
<a name="224" href="#224">224</a> <em>     * @param stream the stream to deserialize from.</em>
<a name="225" href="#225">225</a> <em>     * @throws IOException if I/O errors occur</em>
<a name="226" href="#226">226</a> <em>     * @throws ClassNotFoundException If the class for an object being restored</em>
<a name="227" href="#227">227</a> <em>     *         cannot be found.</em>
<a name="228" href="#228">228</a> <em>     */</em>
<a name="229" href="#229">229</a>     <strong>private</strong> <strong>void</strong> readObject(ObjectInputStream stream)
<a name="230" href="#230">230</a>             throws IOException, ClassNotFoundException {
<a name="231" href="#231">231</a>         stream.defaultReadObject();
<a name="232" href="#232">232</a>         Thread t = Thread.currentThread();
<a name="233" href="#233">233</a>         <strong>if</strong> (t instanceof Checkpointer.CheckpointingThread) {
<a name="234" href="#234">234</a>             settingsHandler = ((Checkpointer.CheckpointingThread)t)
<a name="235" href="#235">235</a>         		.getController().getSettingsHandler();
<a name="236" href="#236">236</a>         } <strong>else</strong> <strong>if</strong> (t instanceof ToeThread) {
<a name="237" href="#237">237</a>             settingsHandler = ((ToeThread) Thread.currentThread())
<a name="238" href="#238">238</a>                 .getController().getSettingsHandler();
<a name="239" href="#239">239</a>         } <strong>else</strong> {
<a name="240" href="#240">240</a>             <em class="comment">// TODO: log differently? (if no throw here</em>
<a name="241" href="#241">241</a>             <em class="comment">// NPE is inevitable)</em>
<a name="242" href="#242">242</a>             <strong>throw</strong> <strong>new</strong> RuntimeException(<span class="string">"CrawlServer must deserialize "</span> +
<a name="243" href="#243">243</a>                     <span class="string">"in a ToeThread or CheckpointingThread"</span>);
<a name="244" href="#244">244</a>         }
<a name="245" href="#245">245</a>         postDeserialize();
<a name="246" href="#246">246</a>     }
<a name="247" href="#247">247</a>     
<a name="248" href="#248">248</a>     <strong>private</strong> <strong>void</strong> postDeserialize() {
<a name="249" href="#249">249</a>     	<strong>if</strong> (<strong>this</strong>.robots != <strong>null</strong>) {