📄 extractortool.html
字号:
<a name="132" href="#132">132</a> <a name="133" href="#133">133</a> <strong>public</strong> <strong>void</strong> extract(String resource) throws IOException,<a name="134" href="#134">134</a> URIException, InterruptedException {<a name="135" href="#135">135</a> <a href="../../../../org/archive/io/arc/ARCReader.html">ARCReader</a> reader = ARCReaderFactory.get(<strong>new</strong> File(resource));<a name="136" href="#136">136</a> <strong>for</strong> (Iterator i = reader.iterator(); i.hasNext();) {<a name="137" href="#137">137</a> <a href="../../../../org/archive/io/arc/ARCRecord.html">ARCRecord</a> ar = (ARCRecord)i.next();<a name="138" href="#138">138</a> <a href="../../../../org/archive/util/HttpRecorder.html">HttpRecorder</a> hr = HttpRecorder.<a name="139" href="#139">139</a> wrapInputStreamWithHttpRecord(<strong>this</strong>.scratchDir,<a name="140" href="#140">140</a> <strong>this</strong>.getClass().getName(), ar, <strong>null</strong>);<a name="141" href="#141">141</a> <a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi = getCrawlURI(ar, hr);<a name="142" href="#142">142</a> <strong>for</strong> (Iterator ii = <strong>this</strong>.extractors.iterator(); ii.hasNext();) {<a name="143" href="#143">143</a> ((<a href="../../../../org/archive/crawler/framework/Processor.html">Processor</a>)ii.next()).process(curi);<a name="144" href="#144">144</a> }<a name="145" href="#145">145</a> outlinks(curi);<a name="146" href="#146">146</a> }<a name="147" href="#147">147</a> }<a name="148" href="#148">148</a> <a name="149" href="#149">149</a> <strong>protected</strong> <strong>void</strong> outlinks(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {<a name="150" href="#150">150</a> System.out.println(curi.getUURI().toString());<a name="151" href="#151">151</a> <strong>for</strong>(Iterator i = curi.getOutLinks().iterator(); i.hasNext();) {<a name="152" href="#152">152</a> <a href="../../../../org/archive/crawler/extractor/Link.html">Link</a> l = (Link)i.next();<a name="153" href="#153">153</a> System.out.println(<span class="string">" "</span> + l.getDestination() + <span class="string">" "</span> +<a name="154" href="#154">154</a> l.getHopType() + <span class="string">" "</span> + l.getContext());<a name="155" href="#155">155</a> }<a name="156" href="#156">156</a> }<a name="157" href="#157">157</a> <a name="158" href="#158">158</a> <strong>protected</strong> <a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> getCrawlURI(<strong>final</strong> <a href="../../../../org/archive/io/arc/ARCRecord.html">ARCRecord</a> record,<a name="159" href="#159">159</a> <strong>final</strong> <a href="../../../../org/archive/util/HttpRecorder.html">HttpRecorder</a> hr)<a name="160" href="#160">160</a> throws URIException {<a name="161" href="#161">161</a> <a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi = <strong>new</strong> <a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a>(UURIFactory.<a name="162" href="#162">162</a> getInstance(record.getMetaData().getUrl()));<a name="163" href="#163">163</a> curi.setContentSize(record.getMetaData().getLength());<a name="164" href="#164">164</a> curi.setContentType(record.getMetaData().getMimetype());<a name="165" href="#165">165</a> curi.setHttpRecorder(hr);<a name="166" href="#166">166</a> <em class="comment">// Fake out the extractor that this is a legit HTTP transaction.</em><a name="167" href="#167">167</a> <strong>if</strong> (!curi.getUURI().getScheme().equals(<span class="string">"filedesc"</span>)) {<a name="168" href="#168">168</a> curi.putObject(CoreAttributeConstants.A_HTTP_TRANSACTION,<a name="169" href="#169">169</a> <strong>new</strong> <a href="../../../../org/apache/commons/httpclient/HttpMethodBase.html">HttpMethodBase</a>() {<a name="170" href="#170">170</a> <strong>public</strong> String getName() {<a name="171" href="#171">171</a> <strong>return</strong> <strong>this</strong>.getClass().getName() + <span class="string">"_method"</span>;<a name="172" href="#172">172</a> }<a name="173" href="#173">173</a> <a name="174" href="#174">174</a> <strong>public</strong> Header getResponseHeader(String headerName) {<a name="175" href="#175">175</a> String value = (String)record.getMetaData().<a name="176" href="#176">176</a> getHeaderValue(headerName);<a name="177" href="#177">177</a> <strong>return</strong> (value == <strong>null</strong> || value.length() == 0)?<a name="178" href="#178">178</a> <strong>null</strong>: <strong>new</strong> Header(headerName, value);<a name="179" href="#179">179</a> }<a name="180" href="#180">180</a> });<a name="181" href="#181">181</a> String statusCode = record.getMetaData().getStatusCode();<a name="182" href="#182">182</a> curi.setFetchStatus(statusCode == <strong>null</strong>?<a name="183" href="#183">183</a> 200: Integer.parseInt(statusCode));<a name="184" href="#184">184</a> }<a name="185" href="#185">185</a> <strong>return</strong> curi;<a name="186" href="#186">186</a> }<a name="187" href="#187">187</a> <a name="188" href="#188">188</a> <em>/**<em>*</em></em><a name="189" href="#189">189</a> <em> * Format usage message.</em><a name="190" href="#190">190</a> <em> * @param formatter Help formatter instance.</em><a name="191" href="#191">191</a> <em> * @param options Usage options.</em><a name="192" href="#192">192</a> <em> * @param exitCode Exit code.</em><a name="193" href="#193">193</a> <em> */</em><a name="194" href="#194">194</a> <strong>private</strong> <strong>static</strong> <strong>void</strong> usage(HelpFormatter formatter, Options options,<a name="195" href="#195">195</a> <strong>int</strong> exitCode) {<a name="196" href="#196">196</a> formatter.printHelp(<span class="string">"java "</span> + ExtractorTool.<strong>class</strong>.getName() +<a name="197" href="#197">197</a> <span class="string">" //\n[--scratch=DIR] [--extractor=EXTRACTOR1,EXTRACTOR2,...] ARC"</span>, options);<a name="198" href="#198">198</a> System.exit(exitCode);<a name="199" href="#199">199</a> }<a name="200" href="#200">200</a> <a name="201" href="#201">201</a> <strong>public</strong> <strong>static</strong> <strong>void</strong> main(String[] args)<a name="202" href="#202">202</a> throws Exception {<a name="203" href="#203">203</a> Options options = <strong>new</strong> Options();<a name="204" href="#204">204</a> options.addOption(<strong>new</strong> Option(<span class="string">"h"</span>, <span class="string">"help"</span>, false,<a name="205" href="#205">205</a> <span class="string">"Prints this message and exits."</span>));<a name="206" href="#206">206</a> StringBuffer defaultExtractors = <strong>new</strong> StringBuffer();<a name="207" href="#207">207</a> <strong>for</strong> (<strong>int</strong> i = 0; i < DEFAULT_EXTRACTORS.length; i++) {<a name="208" href="#208">208</a> <strong>if</strong> (i > 0) {<a name="209" href="#209">209</a> defaultExtractors.append(<span class="string">", "</span>);<a name="210" href="#210">210</a> }<a name="211" href="#211">211</a> defaultExtractors.append(DEFAULT_EXTRACTORS[i]);<a name="212" href="#212">212</a> }<a name="213" href="#213">213</a> options.addOption(<strong>new</strong> Option(<span class="string">"e"</span>, <span class="string">"extractor"</span>, <strong>true</strong>,<a name="214" href="#214">214</a> <span class="string">"List of comma-separated extractor class names. "</span> +<a name="215" href="#215">215</a> <span class="string">"Run in order listed. "</span> +<a name="216" href="#216">216</a> <span class="string">"If no extractors listed, runs following: "</span> +<a name="217" href="#217">217</a> defaultExtractors.toString() + <span class="string">"."</span>));<a name="218" href="#218">218</a> options.addOption(<strong>new</strong> Option(<span class="string">"s"</span>, <span class="string">"scratch"</span>, <strong>true</strong>,<a name="219" href="#219">219</a> <span class="string">"Directory to write scratch files to. Default: '/tmp'."</span>));<a name="220" href="#220">220</a> PosixParser parser = <strong>new</strong> PosixParser();<a name="221" href="#221">221</a> CommandLine cmdline = parser.parse(options, args, false);<a name="222" href="#222">222</a> List cmdlineArgs = cmdline.getArgList();<a name="223" href="#223">223</a> Option [] cmdlineOptions = cmdline.getOptions();<a name="224" href="#224">224</a> HelpFormatter formatter = <strong>new</strong> HelpFormatter();<a name="225" href="#225">225</a> <em class="comment">// If no args, print help.</em><a name="226" href="#226">226</a> <strong>if</strong> (cmdlineArgs.size() <= 0) {<a name="227" href="#227">227</a> usage(formatter, options, 0);<a name="228" href="#228">228</a> }<a name="229" href="#229">229</a> <a name="230" href="#230">230</a> <em class="comment">// Now look at options passed.</em><a name="231" href="#231">231</a> String [] extractors = DEFAULT_EXTRACTORS;<a name="232" href="#232">232</a> String scratch = <strong>null</strong>;<a name="233" href="#233">233</a> <strong>for</strong> (<strong>int</strong> i = 0; i < cmdlineOptions.length; i++) {<a name="234" href="#234">234</a> <strong>switch</strong>(cmdlineOptions[i].getId()) {<a name="235" href="#235">235</a> <strong>case</strong> 'h':<a name="236" href="#236">236</a> usage(formatter, options, 0);<a name="237" href="#237">237</a> <strong>break</strong>;<a name="238" href="#238">238</a> <a name="239" href="#239">239</a> <strong>case</strong> 'e':<a name="240" href="#240">240</a> String value = cmdlineOptions[i].getValue();<a name="241" href="#241">241</a> <strong>if</strong> (value == <strong>null</strong> || value.length() <= 0) {<a name="242" href="#242">242</a> <em class="comment">// Allow saying NO extractors so we can see</em><a name="243" href="#243">243</a> <em class="comment">// how much it costs just reading through</em><a name="244" href="#244">244</a> <em class="comment">// ARCs.</em><a name="245" href="#245">245</a> extractors = <strong>new</strong> String [0];<a name="246" href="#246">246</a> } <strong>else</strong> {<a name="247" href="#247">247</a> extractors = value.split(<span class="string">","</span>);<a name="248" href="#248">248</a> }<a name="249" href="#249">249</a> <strong>break</strong>;<a name="250" href="#250">250</a> <a name="251" href="#251">251</a> <strong>case</strong> 's':<a name="252" href="#252">252</a> scratch = cmdlineOptions[i].getValue();<a name="253" href="#253">253</a> <strong>break</strong>;<a name="254" href="#254">254</a> <a name="255" href="#255">255</a> <strong>default</strong>:<a name="256" href="#256">256</a> <strong>throw</strong> <strong>new</strong> RuntimeException(<span class="string">"Unexpected option: "</span> +<a name="257" href="#257">257</a> + cmdlineOptions[i].getId());<a name="258" href="#258">258</a> }<a name="259" href="#259">259</a> }<a name="260" href="#260">260</a> <a name="261" href="#261">261</a> <a href="../../../../org/archive/crawler/extractor/ExtractorTool.html">ExtractorTool</a> tool = <strong>new</strong> <a href="../../../../org/archive/crawler/extractor/ExtractorTool.html">ExtractorTool</a>(extractors, scratch);<a name="262" href="#262">262</a> <strong>for</strong> (Iterator i = cmdlineArgs.iterator(); i.hasNext();) {<a name="263" href="#263">263</a> tool.extract((String)i.next());<a name="264" href="#264">264</a> }<a name="265" href="#265">265</a> }<a name="266" href="#266">266</a> }</pre><hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/">Maven</a></div></body></html>
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -