linksscoper.html

来自「网络爬虫开源代码」· HTML 代码 · 共 310 行 · 第 1/2 页

<a name="147" href="#147">147</a> 
<a name="148" href="#148">148</a>         <strong>final</strong> <strong>boolean</strong> redirectsNewSeeds = ((Boolean)getUncheckedAttribute(curi,
<a name="149" href="#149">149</a>             ATTR_SEED_REDIRECTS_NEW_SEEDS)).booleanValue();
<a name="150" href="#150">150</a>         <strong>int</strong> preferenceDepthHops = ((Integer)getUncheckedAttribute(curi,
<a name="151" href="#151">151</a>             ATTR_PREFERENCE_DEPTH_HOPS)).intValue();
<a name="152" href="#152">152</a>         Collection&lt;CandidateURI> inScopeLinks = <strong>new</strong> HashSet&lt;CandidateURI>();
<a name="153" href="#153">153</a>         <strong>for</strong> (<strong>final</strong> Iterator i = curi.getOutObjects().iterator(); i.hasNext();) {
<a name="154" href="#154">154</a>             Object o = i.next();
<a name="155" href="#155">155</a>             <strong>if</strong>(o instanceof <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a>){
<a name="156" href="#156">156</a>                 <strong>final</strong> <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a> wref = (Link)o;
<a name="157" href="#157">157</a>                 <strong>try</strong> {
<a name="158" href="#158">158</a>                     <strong>final</strong> <strong>int</strong> directive = getSchedulingFor(curi, wref, 
<a name="159" href="#159">159</a>                         preferenceDepthHops);
<a name="160" href="#160">160</a>                     <strong>final</strong> <a href="../../../../org/archive/crawler/datamodel/CandidateURI.html">CandidateURI</a> caURI =
<a name="161" href="#161">161</a>                         curi.createCandidateURI(curi.getBaseURI(), wref, 
<a name="162" href="#162">162</a>                             directive, 
<a name="163" href="#163">163</a>                             considerAsSeed(curi, wref, redirectsNewSeeds));
<a name="164" href="#164">164</a>                     <strong>if</strong> (isInScope(caURI)) {
<a name="165" href="#165">165</a>                         inScopeLinks.add(caURI);
<a name="166" href="#166">166</a>                     }
<a name="167" href="#167">167</a>                 } <strong>catch</strong> (URIException e) {
<a name="168" href="#168">168</a>                     getController().logUriError(e, curi.getUURI(), 
<a name="169" href="#169">169</a>                         wref.getDestination().toString());
<a name="170" href="#170">170</a>                 }
<a name="171" href="#171">171</a>             } <strong>else</strong> <strong>if</strong>(o instanceof <a href="../../../../org/archive/crawler/datamodel/CandidateURI.html">CandidateURI</a>){
<a name="172" href="#172">172</a>                 <a href="../../../../org/archive/crawler/datamodel/CandidateURI.html">CandidateURI</a> caURI = (CandidateURI)o;
<a name="173" href="#173">173</a>                 <strong>if</strong>(isInScope(caURI)){
<a name="174" href="#174">174</a>                     inScopeLinks.add(caURI);
<a name="175" href="#175">175</a>                 }
<a name="176" href="#176">176</a>             } <strong>else</strong> {
<a name="177" href="#177">177</a>                 LOGGER.severe(<span class="string">"Unexpected type: "</span> + o);
<a name="178" href="#178">178</a>             }
<a name="179" href="#179">179</a>         }
<a name="180" href="#180">180</a>         <em class="comment">// Replace current links collection w/ inscopeLinks.  May be</em>
<a name="181" href="#181">181</a>         <em class="comment">// an empty collection.</em>
<a name="182" href="#182">182</a>         curi.replaceOutlinks(inScopeLinks);
<a name="183" href="#183">183</a>     }
<a name="184" href="#184">184</a>     
<a name="185" href="#185">185</a>     <em>/**<em>*</em></em>
<a name="186" href="#186">186</a> <em>     * The CrawlURI has a prerequisite; apply scoping and update</em>
<a name="187" href="#187">187</a> <em>     * Link to CandidateURI in manner analogous to outlink handling. </em>
<a name="188" href="#188">188</a> <em>     * @param curi CrawlURI with prereq to consider</em>
<a name="189" href="#189">189</a> <em>     */</em>
<a name="190" href="#190">190</a>     <strong>protected</strong> <strong>void</strong> handlePrerequisite(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="191" href="#191">191</a>         <strong>try</strong> {
<a name="192" href="#192">192</a>             <em class="comment">// Create prerequisite CandidateURI</em>
<a name="193" href="#193">193</a>             <a href="../../../../org/archive/crawler/datamodel/CandidateURI.html">CandidateURI</a> caUri =
<a name="194" href="#194">194</a>                 curi.createCandidateURI(curi.getBaseURI(),
<a name="195" href="#195">195</a>                     (<a href="../../../../org/archive/crawler/extractor/Link.html">Link</a>) curi.getPrerequisiteUri());
<a name="196" href="#196">196</a>             <strong>int</strong> prereqPriority = curi.getSchedulingDirective() - 1;
<a name="197" href="#197">197</a>             <strong>if</strong> (prereqPriority &lt; 0) {
<a name="198" href="#198">198</a>                 prereqPriority = 0;
<a name="199" href="#199">199</a>                 LOGGER.severe(<span class="string">"Unable to promote prerequisite "</span> + caUri +
<a name="200" href="#200">200</a>                     <span class="string">" above "</span> + curi);
<a name="201" href="#201">201</a>             }
<a name="202" href="#202">202</a>             caUri.setSchedulingDirective(prereqPriority);
<a name="203" href="#203">203</a>             caUri.setForceFetch(<strong>true</strong>);
<a name="204" href="#204">204</a>             <strong>if</strong>(isInScope(caUri)) {
<a name="205" href="#205">205</a>                 <em class="comment">// replace link with CandidateURI</em>
<a name="206" href="#206">206</a>                 curi.setPrerequisiteUri(caUri);
<a name="207" href="#207">207</a>             } <strong>else</strong> {
<a name="208" href="#208">208</a>                 <em class="comment">// prerequisite is out-of-scope; mark CrawlURI as error,</em>
<a name="209" href="#209">209</a>                 <em class="comment">// preventinting normal S_DEFERRED handling</em>
<a name="210" href="#210">210</a>                 curi.setFetchStatus(S_PREREQUISITE_UNSCHEDULABLE_FAILURE);
<a name="211" href="#211">211</a>             }
<a name="212" href="#212">212</a>        } <strong>catch</strong> (URIException ex) {
<a name="213" href="#213">213</a>             Object[] array = {curi, curi.getPrerequisiteUri()};
<a name="214" href="#214">214</a>             getController().uriErrors.log(Level.INFO,ex.getMessage(), array);
<a name="215" href="#215">215</a>         } <strong>catch</strong> (NumberFormatException e) {
<a name="216" href="#216">216</a>             <em class="comment">// UURI.createUURI will occasionally throw this error.</em>
<a name="217" href="#217">217</a>             Object[] array = {curi, curi.getPrerequisiteUri()};
<a name="218" href="#218">218</a>             getController().uriErrors.log(Level.INFO,e.getMessage(), array);
<a name="219" href="#219">219</a>         }
<a name="220" href="#220">220</a>     }
<a name="221" href="#221">221</a> 
<a name="222" href="#222">222</a>     <strong>protected</strong> <strong>void</strong> outOfScope(<a href="../../../../org/archive/crawler/datamodel/CandidateURI.html">CandidateURI</a> caUri) {
<a name="223" href="#223">223</a>         <strong>super</strong>.outOfScope(caUri);
<a name="224" href="#224">224</a>         <strong>if</strong> (!LOGGER.isLoggable(Level.INFO)) {
<a name="225" href="#225">225</a>             <strong>return</strong>;
<a name="226" href="#226">226</a>         }
<a name="227" href="#227">227</a>         <em class="comment">// TODO: Fix filters so work on CandidateURI.</em>
<a name="228" href="#228">228</a>         <a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi = (caUri instanceof CrawlURI)?
<a name="229" href="#229">229</a>             (<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a>)caUri:
<a name="230" href="#230">230</a>             <strong>new</strong> <a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a>(caUri.getUURI());
<a name="231" href="#231">231</a>         <strong>if</strong> (rulesAccept(getRejectLogRules(curi), curi)) {
<a name="232" href="#232">232</a>             LOGGER.info(curi.getUURI().toString());
<a name="233" href="#233">233</a>         }
<a name="234" href="#234">234</a>     }
<a name="235" href="#235">235</a>     
<a name="236" href="#236">236</a>     <strong>protected</strong> <a href="../../../../org/archive/crawler/deciderules/DecideRule.html">DecideRule</a> getRejectLogRules(Object o) {
<a name="237" href="#237">237</a>         <strong>try</strong> {
<a name="238" href="#238">238</a>             <strong>return</strong> (DecideRule)getAttribute(o, ATTR_REJECTLOG_DECIDE_RULES);
<a name="239" href="#239">239</a>         } <strong>catch</strong> (AttributeNotFoundException e) {
<a name="240" href="#240">240</a>             <strong>throw</strong> <strong>new</strong> RuntimeException(e);
<a name="241" href="#241">241</a>         }
<a name="242" href="#242">242</a>     }
<a name="243" href="#243">243</a>     
<a name="244" href="#244">244</a>     <strong>private</strong> <strong>boolean</strong> considerAsSeed(<strong>final</strong> <a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi, <strong>final</strong> <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a> wref,
<a name="245" href="#245">245</a>             <strong>final</strong> <strong>boolean</strong> redirectsNewSeeds) {
<a name="246" href="#246">246</a>         <em class="comment">// Check if this is a seed with a 301 or 302.</em>
<a name="247" href="#247">247</a>         <strong>if</strong> (curi.isSeed()
<a name="248" href="#248">248</a>                 &amp;&amp; (curi.getFetchStatus() == 301 ||
<a name="249" href="#249">249</a>                     curi.getFetchStatus() == 302)
<a name="250" href="#250">250</a>                 &amp;&amp; wref.getHopType() == Link.REFER_HOP) {
<a name="251" href="#251">251</a>             <em class="comment">// Check if redirects from seeds should be treated as seeds.</em>
<a name="252" href="#252">252</a>             <strong>if</strong> (redirectsNewSeeds) {
<a name="253" href="#253">253</a>                 <strong>return</strong> <strong>true</strong>;
<a name="254" href="#254">254</a>             }
<a name="255" href="#255">255</a>         }
<a name="256" href="#256">256</a>         <strong>return</strong> false;
<a name="257" href="#257">257</a>     }
<a name="258" href="#258">258</a>     
<a name="259" href="#259">259</a>     <em>/**<em>*</em></em>
<a name="260" href="#260">260</a> <em>     * Determine scheduling for the  &lt;code>curi&lt;/code>.</em>
<a name="261" href="#261">261</a> <em>     * As with the LinksScoper in general, this only handles extracted links,</em>
<a name="262" href="#262">262</a> <em>     * seeds do not pass through here, but are given MEDIUM priority.  </em>
<a name="263" href="#263">263</a> <em>     * Imports into the frontier similarly do not pass through here, </em>
<a name="264" href="#264">264</a> <em>     * but are given NORMAL priority.</em>
<a name="265" href="#265">265</a> <em>     */</em>
<a name="266" href="#266">266</a>     <strong>protected</strong> <strong>int</strong> getSchedulingFor(<strong>final</strong> <a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi, <strong>final</strong> <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a> wref,
<a name="267" href="#267">267</a>             <strong>final</strong> <strong>int</strong> preferenceDepthHops) {
<a name="268" href="#268">268</a>         <strong>final</strong> <strong>char</strong> c = wref.getHopType();
<a name="269" href="#269">269</a>         <strong>if</strong> (LOGGER.isLoggable(Level.FINEST)) {
<a name="270" href="#270">270</a>             LOGGER.finest(curi + <span class="string">" with path="</span> + curi.getPathFromSeed() +
<a name="271" href="#271">271</a>                 <span class="string">" isSeed="</span> + curi.isSeed() + <span class="string">" with fetchStatus="</span> +
<a name="272" href="#272">272</a>                 curi.getFetchStatus() + <span class="string">" -> "</span> + wref.getDestination() +
<a name="273" href="#273">273</a>                 <span class="string">" type "</span> + c + <span class="string">" with context="</span> + wref.getContext());
<a name="274" href="#274">274</a>         }
<a name="275" href="#275">275</a> 
<a name="276" href="#276">276</a>         <strong>switch</strong> (c) {
<a name="277" href="#277">277</a>             <strong>case</strong> Link.REFER_HOP:
<a name="278" href="#278">278</a>                 <em class="comment">// Treat redirects somewhat urgently</em>
<a name="279" href="#279">279</a>                 <em class="comment">// This also ensures seed redirects remain seed priority</em>
<a name="280" href="#280">280</a>                 <strong>return</strong> (preferenceDepthHops >= 0 ? CandidateURI.HIGH :
<a name="281" href="#281">281</a>                     CandidateURI.MEDIUM);
<a name="282" href="#282">282</a>             <strong>default</strong>:
<a name="283" href="#283">283</a>                 <strong>if</strong> (preferenceDepthHops == 0)
<a name="284" href="#284">284</a>                     <strong>return</strong> CandidateURI.HIGH;
<a name="285" href="#285">285</a>                     <em class="comment">// this implies seed redirects are treated as path</em>
<a name="286" href="#286">286</a>                     <em class="comment">// length 1, which I belive is standard.</em>
<a name="287" href="#287">287</a>                     <em class="comment">// curi.getPathFromSeed() can never be null here, because</em>
<a name="288" href="#288">288</a>                     <em class="comment">// we're processing a link extracted from curi</em>
<a name="289" href="#289">289</a>                 <strong>if</strong> (preferenceDepthHops > 0 &amp;&amp; 
<a name="290" href="#290">290</a>                     curi.getPathFromSeed().length() + 1 &lt;= preferenceDepthHops)
<a name="291" href="#291">291</a>                     <strong>return</strong> CandidateURI.HIGH;
<a name="292" href="#292">292</a>                 <em class="comment">// Everything else normal (at least for now)</em>
<a name="293" href="#293">293</a>                 <strong>return</strong> CandidateURI.NORMAL;
<a name="294" href="#294">294</a>         }
<a name="295" href="#295">295</a>     }
<a name="296" href="#296">296</a> }
</pre>
<hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/">Maven</a></div></body>
</html>