adaptiverevisitfrontier.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="127" href="#127">127</a> <em>     *  Word chars, dash, period, comma, colon */</em>
<a name="128" href="#128">128</a>     <strong>protected</strong> <strong>final</strong> <strong>static</strong> String ACCEPTABLE_FORCE_QUEUE = <span class="string">"[-&#47;&#47;w&#47;&#47;.,:]*"</span>;
<a name="129" href="#129">129</a> 
<a name="130" href="#130">130</a>     <em>/**<em>* Should the queue assignment ignore www in hostnames, effectively </em></em>
<a name="131" href="#131">131</a> <em>     *  stripping them away. </em>
<a name="132" href="#132">132</a> <em>     */</em>
<a name="133" href="#133">133</a>     <strong>public</strong> <strong>final</strong> <strong>static</strong> String ATTR_QUEUE_IGNORE_WWW = <span class="string">"queue-ignore-www"</span>;
<a name="134" href="#134">134</a>     <strong>protected</strong> <strong>final</strong> <strong>static</strong> Boolean DEFAULT_QUEUE_IGNORE_WWW = <strong>new</strong> Boolean(false);
<a name="135" href="#135">135</a>     
<a name="136" href="#136">136</a>     <em>/**<em>* Should the Frontier use a seperate 'already included' datastructure</em></em>
<a name="137" href="#137">137</a> <em>     *  or rely on the queues'. </em>
<a name="138" href="#138">138</a> <em>     */</em>
<a name="139" href="#139">139</a>     <strong>public</strong> <strong>final</strong> <strong>static</strong> String ATTR_USE_URI_UNIQ_FILTER = <span class="string">"use-uri-uniq-filter"</span>;
<a name="140" href="#140">140</a>     <strong>protected</strong> <strong>final</strong> <strong>static</strong> Boolean DEFAULT_USE_URI_UNIQ_FILTER = <strong>new</strong> Boolean(false);
<a name="141" href="#141">141</a>     
<a name="142" href="#142">142</a>     <strong>private</strong> <a href="../../../../org/archive/crawler/framework/CrawlController.html">CrawlController</a> controller;
<a name="143" href="#143">143</a>     
<a name="144" href="#144">144</a>     <strong>private</strong> <a href="../../../../org/archive/crawler/frontier/AdaptiveRevisitQueueList.html">AdaptiveRevisitQueueList</a> hostQueues;
<a name="145" href="#145">145</a>     
<a name="146" href="#146">146</a>     <strong>private</strong> <a href="../../../../org/archive/crawler/datamodel/UriUniqFilter.html">UriUniqFilter</a> alreadyIncluded;
<a name="147" href="#147">147</a> 
<a name="148" href="#148">148</a>     <strong>private</strong> ThreadLocalQueue threadWaiting = <strong>new</strong> ThreadLocalQueue();
<a name="149" href="#149">149</a> 
<a name="150" href="#150">150</a>     <em>/**<em>* Policy for assigning CrawlURIs to named queues */</em></em>
<a name="151" href="#151">151</a>     <strong>private</strong> <a href="../../../../org/archive/crawler/frontier/QueueAssignmentPolicy.html">QueueAssignmentPolicy</a> queueAssignmentPolicy = <strong>null</strong>;
<a name="152" href="#152">152</a>     
<a name="153" href="#153">153</a>     <em class="comment">// top-level stats</em>
<a name="154" href="#154">154</a>     <strong>private</strong> <strong>long</strong> succeededFetchCount = 0;
<a name="155" href="#155">155</a>     <strong>private</strong> <strong>long</strong> failedFetchCount = 0;
<a name="156" href="#156">156</a>     <em class="comment">// URI's that are disregarded (for example because of robot.txt rules)</em>
<a name="157" href="#157">157</a>     <strong>private</strong> <strong>long</strong> disregardedUriCount = 0;
<a name="158" href="#158">158</a> 
<a name="159" href="#159">159</a>     <strong>private</strong> <strong>long</strong> totalProcessedBytes = 0;
<a name="160" href="#160">160</a>     
<a name="161" href="#161">161</a>     <em class="comment">// Flags indicating operator-specified crawl pause/end </em>
<a name="162" href="#162">162</a>     <strong>private</strong> <strong>boolean</strong> shouldPause = false;
<a name="163" href="#163">163</a>     <strong>private</strong> <strong>boolean</strong> shouldTerminate = false;
<a name="164" href="#164">164</a>     
<a name="165" href="#165">165</a> 
<a name="166" href="#166">166</a>     <strong>public</strong> <a href="../../../../org/archive/crawler/frontier/AdaptiveRevisitFrontier.html">AdaptiveRevisitFrontier</a>(String name) {
<a name="167" href="#167">167</a>         <strong>this</strong>(name, <span class="string">"AdaptiveRevisitFrontier. EXPERIMENTAL Frontier that "</span> +
<a name="168" href="#168">168</a>                 <span class="string">"will repeatedly visit all "</span> +
<a name="169" href="#169">169</a>                 <span class="string">"encountered URIs. Wait time between visits is configurable"</span> +
<a name="170" href="#170">170</a>                 <span class="string">" and is determined by seperate Processor(s). See "</span> +
<a name="171" href="#171">171</a>                 <span class="string">"WaitEvaluators "</span> +
<a name="172" href="#172">172</a>                 <span class="string">"See documentation for ARFrontier limitations."</span>);        
<a name="173" href="#173">173</a>     }
<a name="174" href="#174">174</a> 
<a name="175" href="#175">175</a>     <strong>public</strong> <a href="../../../../org/archive/crawler/frontier/AdaptiveRevisitFrontier.html">AdaptiveRevisitFrontier</a>(String name, String description) {
<a name="176" href="#176">176</a>         <strong>super</strong>(Frontier.ATTR_NAME, description);
<a name="177" href="#177">177</a>         addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_DELAY_FACTOR,
<a name="178" href="#178">178</a>                 <span class="string">"How many multiples of last fetch elapsed time to wait before "</span> +
<a name="179" href="#179">179</a>                 <span class="string">"recontacting same server"</span>, DEFAULT_DELAY_FACTOR));
<a name="180" href="#180">180</a>             addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_MAX_DELAY,
<a name="181" href="#181">181</a>                 <span class="string">"Never wait more than this long, regardless of multiple"</span>,
<a name="182" href="#182">182</a>                 DEFAULT_MAX_DELAY));
<a name="183" href="#183">183</a>             addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_MIN_DELAY,
<a name="184" href="#184">184</a>                 <span class="string">"Always wait this long after one completion before recontacting "</span> +
<a name="185" href="#185">185</a>                 <span class="string">"same server, regardless of multiple"</span>, DEFAULT_MIN_DELAY));
<a name="186" href="#186">186</a>              addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_MAX_RETRIES,
<a name="187" href="#187">187</a>                 <span class="string">"How often to retry fetching a URI that failed to be retrieved.\n"</span> +
<a name="188" href="#188">188</a>                 <span class="string">"If zero, the crawler will get the robots.txt only."</span>,
<a name="189" href="#189">189</a>                 DEFAULT_MAX_RETRIES));
<a name="190" href="#190">190</a>             addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_RETRY_DELAY,
<a name="191" href="#191">191</a>                     <span class="string">"How long to wait by default until we retry fetching a"</span> +
<a name="192" href="#192">192</a>                     <span class="string">" URI that failed to be retrieved (seconds). "</span>,
<a name="193" href="#193">193</a>                     DEFAULT_RETRY_DELAY));
<a name="194" href="#194">194</a>             addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_PREFERENCE_EMBED_HOPS,
<a name="195" href="#195">195</a>                     <span class="string">"Number of embedded (or redirected) hops up to which "</span> +
<a name="196" href="#196">196</a>                     <span class="string">"a URI has higher priority scheduling. For example, if set "</span> +
<a name="197" href="#197">197</a>                     <span class="string">"to 1 (the default), items such as inline images (1-hop "</span> +
<a name="198" href="#198">198</a>                     <span class="string">"embedded resources) will be scheduled ahead of all regular "</span> +
<a name="199" href="#199">199</a>                     <span class="string">"links (or many-hop resources, like nested frames). If set to "</span> +
<a name="200" href="#200">200</a>                     <span class="string">"zero, no preferencing will occur, and embeds/redirects are "</span> +
<a name="201" href="#201">201</a>                     <span class="string">"scheduled the same as regular links."</span>,
<a name="202" href="#202">202</a>                     DEFAULT_PREFERENCE_EMBED_HOPS));
<a name="203" href="#203">203</a>             <a href="../../../../org/archive/crawler/settings/Type.html">Type</a> t;
<a name="204" href="#204">204</a>             t = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_HOST_VALENCE,
<a name="205" href="#205">205</a>                     <span class="string">"Maximum number of simultaneous requests to a single"</span> +
<a name="206" href="#206">206</a>                     <span class="string">" host."</span>,
<a name="207" href="#207">207</a>                     DEFAULT_HOST_VALENCE));
<a name="208" href="#208">208</a>             t.setExpertSetting(<strong>true</strong>);
<a name="209" href="#209">209</a>             t = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_QUEUE_IGNORE_WWW,
<a name="210" href="#210">210</a>                     <span class="string">"If true then documents from x.com, www.x.com and any "</span> +
<a name="211" href="#211">211</a>                     <span class="string">"www[0-9]+.x.com will be assigned to the same queue."</span>,
<a name="212" href="#212">212</a>                     DEFAULT_QUEUE_IGNORE_WWW));
<a name="213" href="#213">213</a>             t.setExpertSetting(<strong>true</strong>);
<a name="214" href="#214">214</a>             t = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(
<a name="215" href="#215">215</a>                     ATTR_FORCE_QUEUE,
<a name="216" href="#216">216</a>                     <span class="string">"The queue name into which to force URIs. Should "</span>
<a name="217" href="#217">217</a>                     + <span class="string">"be left blank at global level.  Specify a "</span>
<a name="218" href="#218">218</a>                     + <span class="string">"per-domain/per-host override to force URIs into "</span>
<a name="219" href="#219">219</a>                     + <span class="string">"a particular named queue, regardless of the assignment "</span>
<a name="220" href="#220">220</a>                     + <span class="string">"policy in effect (domain or ip-based politeness). "</span>
<a name="221" href="#221">221</a>                     + <span class="string">"This could be used on domains known to all be from "</span>
<a name="222" href="#222">222</a>                     + <span class="string">"the same small set of IPs (eg blogspot, dailykos, etc.) "</span>
<a name="223" href="#223">223</a>                     + <span class="string">"to simulate IP-based politeness, or it could be used if "</span>
<a name="224" href="#224">224</a>                     + <span class="string">"you wanted to enforce politeness over a whole domain, even "</span>
<a name="225" href="#225">225</a>                     + <span class="string">"though the subdomains are split across many IPs."</span>,
<a name="226" href="#226">226</a>                     DEFAULT_FORCE_QUEUE));
<a name="227" href="#227">227</a>             t.setOverrideable(<strong>true</strong>);
<a name="228" href="#228">228</a>             t.setExpertSetting(<strong>true</strong>);
<a name="229" href="#229">229</a>             t.addConstraint(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/RegularExpressionConstraint.html">RegularExpressionConstraint</a>(ACCEPTABLE_FORCE_QUEUE,
<a name="230" href="#230">230</a>                     Level.WARNING, <span class="string">"This field must contain only alphanumeric "</span>
<a name="231" href="#231">231</a>                     + <span class="string">"characters plus period, dash, comma, colon, or underscore."</span>));
<a name="232" href="#232">232</a>             t = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_USE_URI_UNIQ_FILTER,
<a name="233" href="#233">233</a>                     <span class="string">"If true then the Frontier will use a seperate "</span> +
<a name="234" href="#234">234</a>                     <span class="string">"datastructure to detect and eliminate duplicates.\n"</span> +
<a name="235" href="#235">235</a>                     <span class="string">"This is required for Canonicalization rules to work."</span>,
<a name="236" href="#236">236</a>                     DEFAULT_USE_URI_UNIQ_FILTER));
<a name="237" href="#237">237</a>             t.setExpertSetting(<strong>true</strong>);
<a name="238" href="#238">238</a>             t.setOverrideable(false);
<a name="239" href="#239">239</a> 
<a name="240" href="#240">240</a>         <em class="comment">// Register persistent CrawlURI items </em>
<a name="241" href="#241">241</a>         CrawlURI.addAlistPersistentMember(A_CONTENT_STATE_KEY);
<a name="242" href="#242">242</a>         CrawlURI.addAlistPersistentMember(A_TIME_OF_NEXT_PROCESSING);
<a name="243" href="#243">243</a>     }
<a name="244" href="#244">244</a> 
<a name="245" href="#245">245</a>     <strong>public</strong> <strong>synchronized</strong> <strong>void</strong> initialize(<a href="../../../../org/archive/crawler/framework/CrawlController.html">CrawlController</a> c)
<a name="246" href="#246">246</a>             throws FatalConfigurationException, IOException {
<a name="247" href="#247">247</a>         controller = c;
<a name="248" href="#248">248</a>         controller.addCrawlStatusListener(<strong>this</strong>);
<a name="249" href="#249">249</a> 
<a name="250" href="#250">250</a>         queueAssignmentPolicy = <strong>new</strong> <a href="../../../../org/archive/crawler/frontier/HostnameQueueAssignmentPolicy.html">HostnameQueueAssignmentPolicy</a>();
<a name="251" href="#251">251</a>         
<a name="252" href="#252">252</a>         hostQueues = <strong>new</strong> <a href="../../../../org/archive/crawler/frontier/AdaptiveRevisitQueueList.html">AdaptiveRevisitQueueList</a>(c.getBdbEnvironment(),
<a name="253" href="#253">253</a>             c.getClassCatalog());
<a name="254" href="#254">254</a>         
<a name="255" href="#255">255</a>         <strong>if</strong>(((Boolean)getUncheckedAttribute(
<a name="256" href="#256">256</a>                 <strong>null</strong>,ATTR_USE_URI_UNIQ_FILTER)).booleanValue()){
<a name="257" href="#257">257</a>             alreadyIncluded = createAlreadyIncluded();
<a name="258" href="#258">258</a>         } <strong>else</strong> {
<a name="259" href="#259">259</a>             alreadyIncluded = <strong>null</strong>;
<a name="260" href="#260">260</a>         }
<a name="261" href="#261">261</a>