statisticstracker.html

来自「网络爬虫开源代码」· HTML 代码 · 共 738 行 · 第 1/5 页

<a name="140" href="#140">140</a>     <strong>protected</strong> <strong>float</strong> congestionRatio = 0; 
<a name="141" href="#141">141</a>     <strong>protected</strong> <strong>long</strong> deepestUri;
<a name="142" href="#142">142</a>     <strong>protected</strong> <strong>long</strong> averageDepth;
<a name="143" href="#143">143</a>     
<a name="144" href="#144">144</a>     <em class="comment">/*</em>
<a name="145" href="#145">145</a> <em class="comment">     * Cumulative data</em>
<a name="146" href="#146">146</a> <em class="comment">     */</em>
<a name="147" href="#147">147</a>     <em>/**<em>* tally sizes novel, verified (same hash), vouched (not-modified) */</em> </em>
<a name="148" href="#148">148</a>     <strong>protected</strong> <a href="../../../../org/archive/crawler/util/CrawledBytesHistotable.html">CrawledBytesHistotable</a> crawledBytes = <strong>new</strong> <a href="../../../../org/archive/crawler/util/CrawledBytesHistotable.html">CrawledBytesHistotable</a>();
<a name="149" href="#149">149</a>     
<a name="150" href="#150">150</a>     <em>/**<em>* Keep track of the file types we see (mime type -> count) */</em></em>
<a name="151" href="#151">151</a>     <strong>protected</strong> Hashtable&lt;String,LongWrapper> mimeTypeDistribution
<a name="152" href="#152">152</a>      = <strong>new</strong> Hashtable&lt;String,LongWrapper>();
<a name="153" href="#153">153</a>     <strong>protected</strong> Hashtable&lt;String,LongWrapper> mimeTypeBytes
<a name="154" href="#154">154</a>      = <strong>new</strong> Hashtable&lt;String,LongWrapper>();
<a name="155" href="#155">155</a>     
<a name="156" href="#156">156</a>     <em>/**<em>* Keep track of fetch status codes */</em></em>
<a name="157" href="#157">157</a>     <strong>protected</strong> Hashtable&lt;String,LongWrapper> statusCodeDistribution
<a name="158" href="#158">158</a>      = <strong>new</strong> Hashtable&lt;String,LongWrapper>();
<a name="159" href="#159">159</a>     
<a name="160" href="#160">160</a>     <em>/**<em>* Keep track of hosts. </em></em>
<a name="161" href="#161">161</a> <em>     * </em>
<a name="162" href="#162">162</a> <em>     * Each of these Maps are individually unsynchronized, and cannot </em>
<a name="163" href="#163">163</a> <em>     * be trivially synchronized with the Collections wrapper. Thus</em>
<a name="164" href="#164">164</a> <em>     * their synchronized access is enforced by this class.</em>
<a name="165" href="#165">165</a> <em>     * </em>
<a name="166" href="#166">166</a> <em>     * &lt;p>They're transient because usually bigmaps that get reconstituted</em>
<a name="167" href="#167">167</a> <em>     * on recover from checkpoint.</em>
<a name="168" href="#168">168</a> <em>     */</em>
<a name="169" href="#169">169</a>     <strong>protected</strong> <strong>transient</strong> Map&lt;String,LongWrapper> hostsDistribution = <strong>null</strong>;
<a name="170" href="#170">170</a>     <strong>protected</strong> <strong>transient</strong> Map&lt;String,LongWrapper> hostsBytes = <strong>null</strong>;
<a name="171" href="#171">171</a>     <strong>protected</strong> <strong>transient</strong> Map&lt;String,Long> hostsLastFinished = <strong>null</strong>;
<a name="172" href="#172">172</a> 
<a name="173" href="#173">173</a>     <em>/**<em>* Keep track of URL counts per host per seed */</em></em>
<a name="174" href="#174">174</a>     <strong>protected</strong> <strong>transient</strong> 
<a name="175" href="#175">175</a>     Map&lt;String,HashMap&lt;String,LongWrapper>> sourceHostDistribution = <strong>null</strong>;
<a name="176" href="#176">176</a> 
<a name="177" href="#177">177</a>     <em>/**<em>*</em></em>
<a name="178" href="#178">178</a> <em>     * Record of seeds' latest actions.</em>
<a name="179" href="#179">179</a> <em>     */</em>
<a name="180" href="#180">180</a>     <strong>protected</strong> <strong>transient</strong> Map&lt;String,SeedRecord> processedSeedsRecords;
<a name="181" href="#181">181</a> 
<a name="182" href="#182">182</a>     <em class="comment">// seeds tallies: ONLY UPDATED WHEN SEED REPORT WRITTEN</em>
<a name="183" href="#183">183</a>     <strong>private</strong> <strong>int</strong> seedsCrawled;
<a name="184" href="#184">184</a>     <strong>private</strong> <strong>int</strong> seedsNotCrawled;
<a name="185" href="#185">185</a>     <em class="comment">// sExitMessage: only set at crawl-end</em>
<a name="186" href="#186">186</a>     <strong>private</strong> String sExitMessage = <span class="string">"Before crawl end"</span>;
<a name="187" href="#187">187</a> 
<a name="188" href="#188">188</a> 
<a name="189" href="#189">189</a>     <strong>public</strong> <a href="../../../../org/archive/crawler/admin/StatisticsTracker.html">StatisticsTracker</a>(String name) {
<a name="190" href="#190">190</a>         <strong>super</strong>( name, <span class="string">"A statistics tracker thats integrated into "</span> +
<a name="191" href="#191">191</a>             <span class="string">"the web UI and that creates the progress-statistics log."</span>);
<a name="192" href="#192">192</a>     }
<a name="193" href="#193">193</a> 
<a name="194" href="#194">194</a>     <strong>public</strong> <strong>void</strong> initialize(<a href="../../../../org/archive/crawler/framework/CrawlController.html">CrawlController</a> c)
<a name="195" href="#195">195</a>     throws <a href="../../../../org/archive/crawler/framework/exceptions/FatalConfigurationException.html">FatalConfigurationException</a> {
<a name="196" href="#196">196</a>         <strong>super</strong>.initialize(c);
<a name="197" href="#197">197</a>         <strong>try</strong> {
<a name="198" href="#198">198</a>             <strong>this</strong>.sourceHostDistribution = c.getBigMap(<span class="string">"sourceHostDistribution"</span>,
<a name="199" href="#199">199</a>             	    String.<strong>class</strong>, HashMap.<strong>class</strong>);
<a name="200" href="#200">200</a>             <strong>this</strong>.hostsDistribution = c.getBigMap(<span class="string">"hostsDistribution"</span>,
<a name="201" href="#201">201</a>                 String.<strong>class</strong>, LongWrapper.<strong>class</strong>);
<a name="202" href="#202">202</a>             <strong>this</strong>.hostsBytes = c.getBigMap(<span class="string">"hostsBytes"</span>, String.<strong>class</strong>,
<a name="203" href="#203">203</a>                 LongWrapper.<strong>class</strong>);
<a name="204" href="#204">204</a>             <strong>this</strong>.hostsLastFinished = c.getBigMap(<span class="string">"hostsLastFinished"</span>,
<a name="205" href="#205">205</a>                 String.<strong>class</strong>, Long.<strong>class</strong>);
<a name="206" href="#206">206</a>             <strong>this</strong>.processedSeedsRecords = c.getBigMap(<span class="string">"processedSeedsRecords"</span>,
<a name="207" href="#207">207</a>                     String.<strong>class</strong>, SeedRecord.<strong>class</strong>);
<a name="208" href="#208">208</a>         } <strong>catch</strong> (Exception e) {
<a name="209" href="#209">209</a>             <strong>throw</strong> <strong>new</strong> <a href="../../../../org/archive/crawler/framework/exceptions/FatalConfigurationException.html">FatalConfigurationException</a>(<span class="string">"Failed setup of"</span> +
<a name="210" href="#210">210</a>                 <span class="string">" StatisticsTracker: "</span> + e);
<a name="211" href="#211">211</a>         }
<a name="212" href="#212">212</a>         controller.addCrawlURIDispositionListener(<strong>this</strong>);
<a name="213" href="#213">213</a>     }
<a name="214" href="#214">214</a> 
<a name="215" href="#215">215</a>     <strong>protected</strong> <strong>void</strong> finalCleanup() {
<a name="216" href="#216">216</a>         <strong>super</strong>.finalCleanup();
<a name="217" href="#217">217</a>         <strong>if</strong> (<strong>this</strong>.hostsBytes != <strong>null</strong>) {
<a name="218" href="#218">218</a>             <strong>this</strong>.hostsBytes.clear();
<a name="219" href="#219">219</a>             <strong>this</strong>.hostsBytes = <strong>null</strong>;
<a name="220" href="#220">220</a>         }
<a name="221" href="#221">221</a>         <strong>if</strong> (<strong>this</strong>.hostsDistribution != <strong>null</strong>) {
<a name="222" href="#222">222</a>             <strong>this</strong>.hostsDistribution.clear();
<a name="223" href="#223">223</a>             <strong>this</strong>.hostsDistribution = <strong>null</strong>;
<a name="224" href="#224">224</a>         }
<a name="225" href="#225">225</a>         <strong>if</strong> (<strong>this</strong>.hostsLastFinished != <strong>null</strong>) {
<a name="226" href="#226">226</a>             <strong>this</strong>.hostsLastFinished.clear();
<a name="227" href="#227">227</a>             <strong>this</strong>.hostsLastFinished = <strong>null</strong>;
<a name="228" href="#228">228</a>         }
<a name="229" href="#229">229</a>         <strong>if</strong> (<strong>this</strong>.processedSeedsRecords != <strong>null</strong>) {
<a name="230" href="#230">230</a>             <strong>this</strong>.processedSeedsRecords.clear();
<a name="231" href="#231">231</a>             <strong>this</strong>.processedSeedsRecords = <strong>null</strong>;
<a name="232" href="#232">232</a>         }
<a name="233" href="#233">233</a>         <strong>if</strong> (<strong>this</strong>.sourceHostDistribution != <strong>null</strong>) {
<a name="234" href="#234">234</a>             <strong>this</strong>.sourceHostDistribution.clear();
<a name="235" href="#235">235</a>             <strong>this</strong>.sourceHostDistribution = <strong>null</strong>;
<a name="236" href="#236">236</a>         }
<a name="237" href="#237">237</a> 
<a name="238" href="#238">238</a>     }
<a name="239" href="#239">239</a> 
<a name="240" href="#240">240</a>     <strong>protected</strong> <strong>synchronized</strong> <strong>void</strong> progressStatisticsEvent(<strong>final</strong> EventObject e) {
<a name="241" href="#241">241</a>         <em class="comment">// This method loads "snapshot" data.</em>
<a name="242" href="#242">242</a>         discoveredUriCount = discoveredUriCount();
<a name="243" href="#243">243</a>         downloadedUriCount = successfullyFetchedCount();
<a name="244" href="#244">244</a>         finishedUriCount = finishedUriCount();
<a name="245" href="#245">245</a>         queuedUriCount = queuedUriCount();
<a name="246" href="#246">246</a>         downloadFailures = failedFetchAttempts();
<a name="247" href="#247">247</a>         downloadDisregards = disregardedFetchAttempts();
<a name="248" href="#248">248</a>         totalProcessedBytes = totalBytesCrawled();
<a name="249" href="#249">249</a>         congestionRatio = congestionRatio();
<a name="250" href="#250">250</a>         deepestUri = deepestUri();
<a name="251" href="#251">251</a>         averageDepth = averageDepth();
<a name="252" href="#252">252</a>         
<a name="253" href="#253">253</a>         <strong>if</strong> (finishedUriCount() == 0) {
<a name="254" href="#254">254</a>             docsPerSecond = 0;
<a name="255" href="#255">255</a>             totalKBPerSec = 0;
<a name="256" href="#256">256</a>         } <strong>else</strong> <strong>if</strong> (getCrawlerTotalElapsedTime() &lt; 1000) {
<a name="257" href="#257">257</a>             <strong>return</strong>; <em class="comment">// Not enough time has passed for a decent snapshot.</em>
<a name="258" href="#258">258</a>         } <strong>else</strong> {
<a name="259" href="#259">259</a>             docsPerSecond = (<strong>double</strong>) downloadedUriCount /
<a name="260" href="#260">260</a>                 (<strong>double</strong>)(getCrawlerTotalElapsedTime() / 1000);
<a name="261" href="#261">261</a>             <em class="comment">// Round to nearest long.</em>
<a name="262" href="#262">262</a>             totalKBPerSec = (<strong>long</strong>)(((totalProcessedBytes / 1024) /
<a name="263" href="#263">263</a>                  ((getCrawlerTotalElapsedTime()) / 1000)) + .5 );
<a name="264" href="#264">264</a>         }
<a name="265" href="#265">265</a> 
<a name="266" href="#266">266</a>         busyThreads = activeThreadCount();
<a name="267" href="#267">267</a> 
<a name="268" href="#268">268</a>         <strong>if</strong>(shouldrun ||
<a name="269" href="#269">269</a>             (System.currentTimeMillis() - lastLogPointTime) >= 1000) {
<a name="270" href="#270">270</a>             <em class="comment">// If shouldrun is false there is a chance that the time interval</em>
<a name="271" href="#271">271</a>             <em class="comment">// since last time is too small for a good sample.  We only want</em>
<a name="272" href="#272">272</a>             <em class="comment">// to update "current" data when the interval is long enough or</em>
<a name="273" href="#273">273</a>             <em class="comment">// shouldrun is true.</em>
<a name="274" href="#274">274</a>             currentDocsPerSecond = 0;
<a name="275" href="#275">275</a>             currentKBPerSec = 0;
<a name="276" href="#276">276</a> 
<a name="277" href="#277">277</a>             <em class="comment">// Note time.</em>
<a name="278" href="#278">278</a>             <strong>long</strong> currentTime = System.currentTimeMillis();
<a name="279" href="#279">279</a>             <strong>long</strong> sampleTime = currentTime - lastLogPointTime;
<a name="280" href="#280">280</a> 
<a name="281" href="#281">281</a>             <em class="comment">// if we haven't done anyting or there isn't a reasonable sample</em>
<a name="282" href="#282">282</a>             <em class="comment">// size give up.</em>
<a name="283" href="#283">283</a>             <strong>if</strong> (sampleTime >= 1000) {
<a name="284" href="#284">284</a>                 <em class="comment">// Update docs/sec snapshot</em>
<a name="285" href="#285">285</a>                 <strong>long</strong> currentPageCount = successfullyFetchedCount();
<a name="286" href="#286">286</a>                 <strong>long</strong> samplePageCount = currentPageCount - lastPagesFetchedCount;
<a name="287" href="#287">287</a>