domainsensitivefrontier.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="134" href="#134">134</a>                     counterMode.equalsIgnoreCase(COUNT_HOST))
<a name="135" href="#135">135</a>                 countPerOverride = false;
<a name="136" href="#136">136</a>             <strong>else</strong>
<a name="137" href="#137">137</a>                 countPerOverride = <strong>true</strong>;
<a name="138" href="#138">138</a>         } <strong>catch</strong> (AttributeNotFoundException e) {
<a name="139" href="#139">139</a>             e.printStackTrace();
<a name="140" href="#140">140</a>         } <strong>catch</strong> (MBeanException e) {
<a name="141" href="#141">141</a>             e.printStackTrace();
<a name="142" href="#142">142</a>         } <strong>catch</strong> (ReflectionException e) {
<a name="143" href="#143">143</a>             e.printStackTrace();
<a name="144" href="#144">144</a>         }
<a name="145" href="#145">145</a>     }
<a name="146" href="#146">146</a>     
<a name="147" href="#147">147</a>     <em>/**<em>*</em></em>
<a name="148" href="#148">148</a> <em>     * Check if the max document download limit for this host or domain has</em>
<a name="149" href="#149">149</a> <em>     * been reached.</em>
<a name="150" href="#150">150</a> <em>     * </em>
<a name="151" href="#151">151</a> <em>     * If so, delete the rest of the URIs for this host or domain waiting in</em>
<a name="152" href="#152">152</a> <em>     * the queue. Then add an URIRegExpFilter for this host or domain, so</em>
<a name="153" href="#153">153</a> <em>     * we won't get any more URIs from this one later on.</em>
<a name="154" href="#154">154</a> <em>     * @param curi CrawlURI.</em>
<a name="155" href="#155">155</a> <em>     * @return True if discarded queue.</em>
<a name="156" href="#156">156</a> <em>     */</em>
<a name="157" href="#157">157</a>     <strong>private</strong> <strong>synchronized</strong> <strong>boolean</strong> checkDownloadLimits(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="158" href="#158">158</a>         <strong>long</strong> thisMaxDocs = 0;
<a name="159" href="#159">159</a>         <strong>long</strong> thisCounter = 0;
<a name="160" href="#160">160</a>         <strong>boolean</strong> discarded = false;
<a name="161" href="#161">161</a>         <strong>boolean</strong> retVal = false;
<a name="162" href="#162">162</a>         <strong>if</strong> (curi.getUURI().getScheme().equals(<span class="string">"dns"</span>)) {
<a name="163" href="#163">163</a>             <strong>return</strong> false;
<a name="164" href="#164">164</a>         }
<a name="165" href="#165">165</a>         <strong>try</strong> {
<a name="166" href="#166">166</a>             String host = curi.getUURI().getHost();
<a name="167" href="#167">167</a>             <a href="../../../../org/archive/crawler/settings/CrawlerSettings.html">CrawlerSettings</a> cs = controller.getSettingsHandler().
<a name="168" href="#168">168</a>                 getSettings(host);
<a name="169" href="#169">169</a>             <strong>do</strong> {
<a name="170" href="#170">170</a>                 String scope;
<a name="171" href="#171">171</a>                 <strong>if</strong>(counterMode.equalsIgnoreCase(COUNT_OVERRIDE))
<a name="172" href="#172">172</a>                     scope = cs.getScope() != <strong>null</strong> ? cs.getScope() : <span class="string">"root"</span>;
<a name="173" href="#173">173</a>                 <strong>else</strong> <strong>if</strong>(counterMode.equalsIgnoreCase(COUNT_HOST))    
<a name="174" href="#174">174</a>                     scope = host;
<a name="175" href="#175">175</a>                 <strong>else</strong>{ <em class="comment">//Get domain part of host</em>
<a name="176" href="#176">176</a>                     <strong>int</strong> i = host.lastIndexOf(<span class="string">"."</span>);
<a name="177" href="#177">177</a>                     i = host.lastIndexOf(<span class="string">"."</span>, i-1);
<a name="178" href="#178">178</a>                     scope = host.substring(i+1, host.length());
<a name="179" href="#179">179</a>                 }
<a name="180" href="#180">180</a>                 thisMaxDocs =
<a name="181" href="#181">181</a>                     ((Long) getAttribute(cs, ATTR_MAX_DOCS)).longValue();
<a name="182" href="#182">182</a>                 <strong>this</strong>Counter = <strong>this</strong>.hostCounters.get(scope) != <strong>null</strong> ?
<a name="183" href="#183">183</a>                     ((Long) <strong>this</strong>.hostCounters.get(scope)).longValue(): 0;
<a name="184" href="#184">184</a>                 <em class="comment">// Have we hit the max document download limit for this host</em>
<a name="185" href="#185">185</a>                 <em class="comment">// or domain?</em>
<a name="186" href="#186">186</a>                 <strong>if</strong> ((thisMaxDocs > 0 &amp;&amp; thisCounter >= thisMaxDocs)) {
<a name="187" href="#187">187</a>                     logger.fine(<span class="string">"Discarding Queue: "</span> + host + <span class="string">" "</span>);
<a name="188" href="#188">188</a>                     curi.addAnnotation(<span class="string">"dsfLimit"</span>);
<a name="189" href="#189">189</a>                    <strong>if</strong> (!discarded) {
<a name="190" href="#190">190</a>                         <strong>long</strong> count = 0;
<a name="191" href="#191">191</a>                         <a href="../../../../org/archive/crawler/frontier/WorkQueue.html">WorkQueue</a> wq = getQueueFor(curi);
<a name="192" href="#192">192</a>                         wq.unpeek();
<a name="193" href="#193">193</a>                         count += wq.deleteMatching(<strong>this</strong>, <span class="string">".*"</span>);
<a name="194" href="#194">194</a>                         decrementQueuedCount(count);
<a name="195" href="#195">195</a>                         discarded = <strong>true</strong>;
<a name="196" href="#196">196</a>                         <em class="comment">// I tried adding annotation but we're past log time</em>
<a name="197" href="#197">197</a>                         <em class="comment">// for Curi so it doesn't work.</em>
<a name="198" href="#198">198</a>                         <em class="comment">// curi.addAnnotation("maxDocsForHost");</em>
<a name="199" href="#199">199</a>                     }
<a name="200" href="#200">200</a>                     <em class="comment">// Adding an exclude filter for this host or domain</em>
<a name="201" href="#201">201</a>                     <a href="../../../../org/archive/crawler/filter/OrFilter.html">OrFilter</a> or = (OrFilter) <strong>this</strong>.controller.getScope()
<a name="202" href="#202">202</a>                             .getAttribute(ClassicScope.ATTR_EXCLUDE_FILTER);
<a name="203" href="#203">203</a>                     <em class="comment">// If we have hit max for root, block everything. Else</em>
<a name="204" href="#204">204</a>                     <em class="comment">// just the scope.</em>
<a name="205" href="#205">205</a>                     String filter = scope.equalsIgnoreCase(<span class="string">"root"</span>) ?
<a name="206" href="#206">206</a>                         <span class="string">".*"</span> : <span class="string">"^((https?://)?[a-zA-Z0-9&#47;&#47;.]*)"</span> + scope +
<a name="207" href="#207">207</a>                             <span class="string">"($|/.*)"</span>;
<a name="208" href="#208">208</a>                     logger.fine(<span class="string">"Adding filter: ["</span> + filter + <span class="string">"]."</span>);
<a name="209" href="#209">209</a>                     <a href="../../../../org/archive/crawler/filter/URIRegExpFilter.html">URIRegExpFilter</a> urf =
<a name="210" href="#210">210</a>                         <strong>new</strong> <a href="../../../../org/archive/crawler/filter/URIRegExpFilter.html">URIRegExpFilter</a>(curi.toString(), filter);
<a name="211" href="#211">211</a>                     or.addFilter(<strong>this</strong>.controller.getSettingsHandler().
<a name="212" href="#212">212</a>                         getSettings(<strong>null</strong>), urf);
<a name="213" href="#213">213</a>                     thisMaxDocs = 0;
<a name="214" href="#214">214</a>                     thisCounter = 0;
<a name="215" href="#215">215</a>                     retVal = <strong>true</strong>;
<a name="216" href="#216">216</a>                 }
<a name="217" href="#217">217</a>             } <strong>while</strong> ((cs = cs.getParent()) != <strong>null</strong> &amp;&amp; countPerOverride);
<a name="218" href="#218">218</a>         } <strong>catch</strong> (Exception e) {
<a name="219" href="#219">219</a>             logger.severe(<span class="string">"ERROR: checkDownloadLimits(), "</span>
<a name="220" href="#220">220</a>                     + <span class="string">"while processing {"</span> + curi.toString() + <span class="string">"}"</span>
<a name="221" href="#221">221</a>                     + e.getClass()
<a name="222" href="#222">222</a>                     + <span class="string">"message: "</span> + e.getMessage() + <span class="string">".  Stack trace:"</span>);
<a name="223" href="#223">223</a>             e.printStackTrace();
<a name="224" href="#224">224</a>         }
<a name="225" href="#225">225</a>         <strong>return</strong> retVal;
<a name="226" href="#226">226</a>     }
<a name="227" href="#227">227</a>     
<a name="228" href="#228">228</a>     <strong>protected</strong> <strong>synchronized</strong> <strong>void</strong> incrementHostCounters(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="229" href="#229">229</a>         <strong>if</strong> (!curi.getUURI().toString().startsWith(<span class="string">"dns:"</span>)) {
<a name="230" href="#230">230</a>             <strong>try</strong> {
<a name="231" href="#231">231</a>                 String host = curi.getUURI().getHost();
<a name="232" href="#232">232</a>                 <a href="../../../../org/archive/crawler/settings/CrawlerSettings.html">CrawlerSettings</a> cs =
<a name="233" href="#233">233</a>                     controller.getSettingsHandler().getSettings(host);
<a name="234" href="#234">234</a>                 <strong>do</strong> {
<a name="235" href="#235">235</a>                     String scope;
<a name="236" href="#236">236</a>                     <strong>if</strong>(counterMode.equalsIgnoreCase(COUNT_OVERRIDE))
<a name="237" href="#237">237</a>                         scope = cs.getScope() != <strong>null</strong>? cs.getScope() : <span class="string">"root"</span>;
<a name="238" href="#238">238</a>                     <strong>else</strong> <strong>if</strong>(counterMode.equalsIgnoreCase(COUNT_HOST))    
<a name="239" href="#239">239</a>                         scope = host;
<a name="240" href="#240">240</a>                     <strong>else</strong>{ <em class="comment">//Get only domain part of host</em>
<a name="241" href="#241">241</a>                         <strong>int</strong> i = host.lastIndexOf(<span class="string">"."</span>);
<a name="242" href="#242">242</a>                         i = host.lastIndexOf(<span class="string">"."</span>, i-1);
<a name="243" href="#243">243</a>                         scope = host.substring(i+1, host.length());
<a name="244" href="#244">244</a>                     }
<a name="245" href="#245">245</a>                     <strong>long</strong> counter = <strong>this</strong>.hostCounters.get(scope) != <strong>null</strong> ?
<a name="246" href="#246">246</a>                         ((Long)<strong>this</strong>.hostCounters.get(scope)).longValue(): 0;
<a name="247" href="#247">247</a>                     <strong>this</strong>.hostCounters.put(scope, <strong>new</strong> Long(++counter));
<a name="248" href="#248">248</a>                 } <strong>while</strong> ((cs = cs.getParent()) != <strong>null</strong> &amp;&amp; countPerOverride);
<a name="249" href="#249">249</a>             } <strong>catch</strong> (Exception e) {
<a name="250" href="#250">250</a>                 logger.severe(<span class="string">"ERROR: incrementHostCounters() "</span> +
<a name="251" href="#251">251</a>                     e.getMessage());
<a name="252" href="#252">252</a>             }
<a name="253" href="#253">253</a>         }
<a name="254" href="#254">254</a>     }
<a name="255" href="#255">255</a>     
<a name="256" href="#256">256</a>     <strong>public</strong> <strong>void</strong> crawledURISuccessful(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="257" href="#257">257</a>         incrementHostCounters(curi);
<a name="258" href="#258">258</a>         checkDownloadLimits(curi);
<a name="259" href="#259">259</a>     }
<a name="260" href="#260">260</a> 
<a name="261" href="#261">261</a>     <strong>public</strong> <strong>void</strong> crawledURINeedRetry(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="262" href="#262">262</a>     }
<a name="263" href="#263">263</a> 
<a name="264" href="#264">264</a>     <strong>public</strong> <strong>void</strong> crawledURIDisregard(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="265" href="#265">265</a>     }
<a name="266" href="#266">266</a> 
<a name="267" href="#267">267</a>     <strong>public</strong> <strong>void</strong> crawledURIFailure(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="268" href="#268">268</a>     }
<a name="269" href="#269">269</a> }
</pre>
<hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/">Maven</a></div></body>
</html>