kw3writerprocessor.html

网络爬虫开源代码

<a name="152" href="#152">152</a>   <strong>private</strong> File arcsDir;
<a name="153" href="#153">153</a>   
<a name="154" href="#154">154</a>   <strong>private</strong> <strong>boolean</strong> chmod;
<a name="155" href="#155">155</a>   
<a name="156" href="#156">156</a>   <strong>private</strong> String chmodValue;
<a name="157" href="#157">157</a>   
<a name="158" href="#158">158</a>   <strong>private</strong> <strong>int</strong> maxSize;
<a name="159" href="#159">159</a>   
<a name="160" href="#160">160</a>   <strong>private</strong> String collection;
<a name="161" href="#161">161</a>   
<a name="162" href="#162">162</a>   <strong>private</strong> String harvester;
<a name="163" href="#163">163</a>   
<a name="164" href="#164">164</a>   
<a name="165" href="#165">165</a>   <em>/**<em>*</em></em>
<a name="166" href="#166">166</a> <em>   * @param name Name of this processor.</em>
<a name="167" href="#167">167</a> <em>   */</em>
<a name="168" href="#168">168</a>   <strong>public</strong> Kw3WriterProcessor(String name) {
<a name="169" href="#169">169</a>       <strong>super</strong>(name, <span class="string">"Kw3Writer processor. "</span> +
<a name="170" href="#170">170</a>           <span class="string">"A writer that writes files in the MIME format of The "</span> +
<a name="171" href="#171">171</a>           <span class="string">"Swedish National Library.  See this class's javadoc for"</span> +
<a name="172" href="#172">172</a>           <span class="string">"format exposition."</span>);
<a name="173" href="#173">173</a>       <a href="../../../../org/archive/crawler/settings/Type.html">Type</a> e; 
<a name="174" href="#174">174</a>       e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_PATH,
<a name="175" href="#175">175</a>               <span class="string">"Top-level directory for archive files."</span>, DEFAULT_PATH));
<a name="176" href="#176">176</a>       e.setOverrideable(false);
<a name="177" href="#177">177</a>       e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_COLLECTION,
<a name="178" href="#178">178</a>               <span class="string">"Name of collection."</span>, DEFAULT_COLLECTION_VALUE));
<a name="179" href="#179">179</a>       e.setOverrideable(false);
<a name="180" href="#180">180</a>       e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_HARVESTER,
<a name="181" href="#181">181</a>               <span class="string">"Name of the harvester that is used for the web harvesting."</span>,
<a name="182" href="#182">182</a>               DEFAULT_HARVESTER_VALUE));
<a name="183" href="#183">183</a>       e.setOverrideable(false);
<a name="184" href="#184">184</a>       e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_MAX_SIZE_BYTES, 
<a name="185" href="#185">185</a>               <span class="string">"Max size of each file"</span>, <strong>new</strong> Integer(DEFAULT_MAX_FILE_SIZE)));
<a name="186" href="#186">186</a>       e.setOverrideable(false);
<a name="187" href="#187">187</a>       e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_CHMOD, 
<a name="188" href="#188">188</a>               <span class="string">"Should permissions be changed for the newly created dirs"</span>,
<a name="189" href="#189">189</a>               <strong>new</strong> Boolean(<strong>true</strong>)));
<a name="190" href="#190">190</a>       e.setOverrideable(false);
<a name="191" href="#191">191</a>       e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_CHMOD_VALUE, 
<a name="192" href="#192">192</a>               <span class="string">"What should the permissions be set to."</span> +
<a name="193" href="#193">193</a>               <span class="string">" Given as three octal digits, as to the UNIX 'chmod' command."</span> +
<a name="194" href="#194">194</a>               <span class="string">" Ex. 777 for all permissions to everyone."</span>,
<a name="195" href="#195">195</a>               DEFAULT_CHMOD_VALUE));
<a name="196" href="#196">196</a>       e.setOverrideable(false);
<a name="197" href="#197">197</a>   }
<a name="198" href="#198">198</a> 
<a name="199" href="#199">199</a>   <strong>protected</strong> <strong>void</strong> initialTasks () {
<a name="200" href="#200">200</a>       <strong>try</strong> {
<a name="201" href="#201">201</a>           String arcsDirPath = (String) getAttribute(ATTR_PATH);
<a name="202" href="#202">202</a>           <strong>this</strong>.arcsDir = <strong>new</strong> File(arcsDirPath);
<a name="203" href="#203">203</a>           <strong>if</strong> (!<strong>this</strong>.arcsDir.isAbsolute()) 
<a name="204" href="#204">204</a>               <strong>this</strong>.arcsDir = <strong>new</strong> File(getController().getDisk(), arcsDirPath);
<a name="205" href="#205">205</a>           
<a name="206" href="#206">206</a>           <strong>this</strong>.collection = (String) getAttribute(ATTR_COLLECTION);
<a name="207" href="#207">207</a>           <strong>this</strong>.harvester = (String) getAttribute(ATTR_HARVESTER);
<a name="208" href="#208">208</a>           <strong>this</strong>.chmod = (Boolean) getAttribute(ATTR_CHMOD);
<a name="209" href="#209">209</a>           <strong>this</strong>.chmodValue = (String) getAttribute(ATTR_CHMOD_VALUE);            
<a name="210" href="#210">210</a>           <strong>this</strong>.maxSize = (Integer) getAttribute(ATTR_MAX_SIZE_BYTES);          
<a name="211" href="#211">211</a>       } <strong>catch</strong> (AttributeNotFoundException e) {
<a name="212" href="#212">212</a>           logger.log(Level.WARNING, <span class="string">"attribute error"</span>, e);
<a name="213" href="#213">213</a>       } <strong>catch</strong> (MBeanException e) {
<a name="214" href="#214">214</a>           logger.log(Level.WARNING, <span class="string">"attribute error"</span>, e);
<a name="215" href="#215">215</a>       } <strong>catch</strong> (ReflectionException e) {
<a name="216" href="#216">216</a>           logger.log(Level.WARNING, <span class="string">"attribute error"</span>, e);
<a name="217" href="#217">217</a>       }     
<a name="218" href="#218">218</a>   }      
<a name="219" href="#219">219</a>   
<a name="220" href="#220">220</a>   <strong>protected</strong> <strong>void</strong> innerProcess(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="221" href="#221">221</a>       <em class="comment">// Only successful fetches are written.</em>
<a name="222" href="#222">222</a>       <strong>if</strong> (!curi.isSuccess()) 
<a name="223" href="#223">223</a>           <strong>return</strong>;
<a name="224" href="#224">224</a>       <em class="comment">// Only http and https schemes are supported.</em>
<a name="225" href="#225">225</a>       String scheme = curi.getUURI().getScheme().toLowerCase();
<a name="226" href="#226">226</a>       <strong>if</strong> (!<span class="string">"http"</span>.equalsIgnoreCase(scheme) &amp;&amp; !<span class="string">"https"</span>.equalsIgnoreCase(scheme))
<a name="227" href="#227">227</a>           <strong>return</strong>;        
<a name="228" href="#228">228</a>       
<a name="229" href="#229">229</a>       <em class="comment">// Write the MIME-file</em>
<a name="230" href="#230">230</a>       <strong>try</strong> {
<a name="231" href="#231">231</a>           writeMimeFile(curi);
<a name="232" href="#232">232</a>       } <strong>catch</strong> (IOException e) {
<a name="233" href="#233">233</a>           logger.log(Level.WARNING, <span class="string">"i/o error"</span>, e);
<a name="234" href="#234">234</a>       }      
<a name="235" href="#235">235</a>   }
<a name="236" href="#236">236</a>   
<a name="237" href="#237">237</a>   <em class="comment">/*</em>
<a name="238" href="#238">238</a> <em class="comment">   * The actual writing of the Kulturarw3 MIME-file.</em>
<a name="239" href="#239">239</a> <em class="comment">   * </em>
<a name="240" href="#240">240</a> <em class="comment">   * The MIME-file consists of three parts:</em>
<a name="241" href="#241">241</a> <em class="comment">   * 1. ArchiveInfo - Metadata about the file and its content.</em>
<a name="242" href="#242">242</a> <em class="comment">   * 2. Header - The HTTP response header.</em>
<a name="243" href="#243">243</a> <em class="comment">   * 3. Content - The HTTP response content, plus content-type.</em>
<a name="244" href="#244">244</a> <em class="comment">   * </em>
<a name="245" href="#245">245</a> <em class="comment">   * For more on this format, see '?'.</em>
<a name="246" href="#246">246</a> <em class="comment">   */</em>
<a name="247" href="#247">247</a>   <strong>protected</strong> <strong>void</strong> writeMimeFile(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) throws IOException {
<a name="248" href="#248">248</a>       <a href="../../../../org/archive/io/ReplayInputStream.html">ReplayInputStream</a> ris = <strong>null</strong>;
<a name="249" href="#249">249</a>       OutputStream out = <strong>null</strong>;
<a name="250" href="#250">250</a>                 
<a name="251" href="#251">251</a>       <strong>try</strong> {
<a name="252" href="#252">252</a>           String boundary = BOUNDARY_START + stringToMD5(curi.toString());
<a name="253" href="#253">253</a>           ris = curi.getHttpRecorder().getRecordedInput().
<a name="254" href="#254">254</a>               getReplayInputStream();
<a name="255" href="#255">255</a>           out = initOutputStream(curi);
<a name="256" href="#256">256</a>           
<a name="257" href="#257">257</a>           <em class="comment">// Part 1: Archive info</em>
<a name="258" href="#258">258</a>           writeArchiveInfoPart(boundary, curi, ris, out);
<a name="259" href="#259">259</a> 
<a name="260" href="#260">260</a>           <em class="comment">// Part 2: Header info + HTTP header</em>
<a name="261" href="#261">261</a>           writeHeaderPart(boundary, ris, out);
<a name="262" href="#262">262</a> 
<a name="263" href="#263">263</a>           <em class="comment">// Part 3: Content info + HTTP content</em>
<a name="264" href="#264">264</a>           writeContentPart(boundary, curi, ris, out);
<a name="265" href="#265">265</a> 
<a name="266" href="#266">266</a>           <em class="comment">// And finally the terminator string</em>
<a name="267" href="#267">267</a>           String terminator = <span class="string">"\n--"</span> + boundary + <span class="string">"--\n"</span>;
<a name="268" href="#268">268</a>           out.write(terminator.getBytes());
<a name="269" href="#269">269</a>       } <strong>finally</strong> {
<a name="270" href="#270">270</a>           <strong>if</strong> (ris != <strong>null</strong>)
<a name="271" href="#271">271</a>               ris.close();
<a name="272" href="#272">272</a>           <strong>if</strong> (out != <strong>null</strong>)
<a name="273" href="#273">273</a>               out.close();
<a name="274" href="#274">274</a>       }
<a name="275" href="#275">275</a>   }
<a name="276" href="#276">276</a>   
<a name="277" href="#277">277</a>   <em class="comment">/*</em>
<a name="278" href="#278">278</a> <em class="comment">   * Get the OutputStream for the file to write to.</em>
<a name="279" href="#279">279</a> <em class="comment">   * </em>
<a name="280" href="#280">280</a> <em class="comment">   * It has a path consisting of:</em>
<a name="281" href="#281">281</a> <em class="comment">   * 1. A dir named with the first two chars of the website's md5.</em>
<a name="282" href="#282">282</a> <em class="comment">   * 2. A dir named after the website.</em>
<a name="283" href="#283">283</a> <em class="comment">   * 3. 'current' - a dir indicating that this is the directory being written</em>
<a name="284" href="#284">284</a> <em class="comment">   *                to by the ongoing crawl. </em>
<a name="285" href="#285">285</a> <em class="comment">   * 4. A file on the format &lt;md5 of url>.&lt;fetchtime in seconds></em>
<a name="286" href="#286">286</a> <em class="comment">   * </em>
<a name="287" href="#287">287</a> <em class="comment">   * Example: '/53/www.kb.se/current/6879ad79c0ccf886ee8ca55d80e5d6a1.1169211837'            </em>
<a name="288" href="#288">288</a> <em class="comment">   */</em>
<a name="289" href="#289">289</a>   <strong>protected</strong> OutputStream initOutputStream(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) throws IOException {
<a name="290" href="#290">290</a>       String uri = curi.toString();
<a name="291" href="#291">291</a>       <strong>int</strong> port = curi.getUURI().getPort();
<a name="292" href="#292">292</a>       String host = (port == 80 || port &lt;= 0) ?
<a name="293" href="#293">293</a>               curi.getUURI().getHost() : curi.getUURI().getHost() + <span class="string">":"</span> + port;
<a name="294" href="#294">294</a>       <strong>long</strong> fetchTime = curi.getLong(A_FETCH_BEGAN_TIME) / 1000;
<a name="295" href="#295">295</a>              
<a name="296" href="#296">296</a>       String md5 = stringToMD5(host);
<a name="297" href="#297">297</a>       File dir = <strong>new</strong> File(<strong>this</strong>.arcsDir, md5.substring(0, 2) + <span class="string">"/"</span> + host +
<a name="298" href="#298">298</a>               <span class="string">"/current"</span>);
<a name="299" href="#299">299</a>       <strong>if</strong> (!dir.exists()) {
<a name="300" href="#300">300</a>           dir.mkdirs();
<a name="301" href="#301">301</a>           <strong>if</strong> (<strong>this</strong>.chmod)
<a name="302" href="#302">302</a>               chmods(dir, <strong>this</strong>.arcsDir);
<a name="303" href="#303">303</a>       }
<a name="304" href="#304">304</a>       md5 = stringToMD5(uri);
<a name="305" href="#305">305</a>       File arcFile = <strong>new</strong> File(dir, md5 + <span class="string">"."</span> + fetchTime);
<a name="306" href="#306">306</a>       <strong>return</strong> <strong>new</strong> FastBufferedOutputStream(<strong>new</strong> FileOutputStream(arcFile));       
<a name="307" href="#307">307</a>   }
<a name="308" href="#308">308</a>   
<a name="309" href="#309">309</a>   <strong>protected</strong> <strong>void</strong> writeArchiveInfoPart(String boundary, <a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi,
<a name="310" href="#310">310</a>           <a href="../../../../org/archive/io/ReplayInputStream.html">ReplayInputStream</a> ris, OutputStream out)
<a name="311" href="#311">311</a>           throws IOException {