warcreader.html

来自「网络爬虫开源代码」· HTML 代码 · 共 304 行 · 第 1/2 页

<a name="144" href="#144">144</a>         System.exit(exitCode);
<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>     * Write out the arcfile.</em>
<a name="149" href="#149">149</a> <em>     * </em>
<a name="150" href="#150">150</a> <em>     * @param reader</em>
<a name="151" href="#151">151</a> <em>     * @param format Format to use outputting.</em>
<a name="152" href="#152">152</a> <em>     * @throws IOException</em>
<a name="153" href="#153">153</a> <em>     * @throws java.text.ParseException</em>
<a name="154" href="#154">154</a> <em>     */</em>
<a name="155" href="#155">155</a>     <strong>protected</strong> <strong>static</strong> <strong>void</strong> output(<a href="../../../../org/archive/io/warc/WARCReader.html">WARCReader</a> reader, String format)
<a name="156" href="#156">156</a>     throws IOException, java.text.ParseException {
<a name="157" href="#157">157</a>     	<strong>if</strong> (!reader.output(format)) {
<a name="158" href="#158">158</a>             <strong>throw</strong> <strong>new</strong> IOException(<span class="string">"Unsupported format: "</span> + format);
<a name="159" href="#159">159</a>     	}
<a name="160" href="#160">160</a>     }
<a name="161" href="#161">161</a> 
<a name="162" href="#162">162</a>     <em>/**<em>*</em></em>
<a name="163" href="#163">163</a> <em>     * Generate a CDX index file for an ARC file.</em>
<a name="164" href="#164">164</a> <em>     *</em>
<a name="165" href="#165">165</a> <em>     * @param urlOrPath The ARC file to generate a CDX index for</em>
<a name="166" href="#166">166</a> <em>     * @throws IOException</em>
<a name="167" href="#167">167</a> <em>     * @throws java.text.ParseException</em>
<a name="168" href="#168">168</a> <em>     */</em>
<a name="169" href="#169">169</a>     <strong>public</strong> <strong>static</strong> <strong>void</strong> createCDXIndexFile(String urlOrPath)
<a name="170" href="#170">170</a>     throws IOException, java.text.ParseException {
<a name="171" href="#171">171</a>     	WARCReader r = WARCReaderFactory.get(urlOrPath);
<a name="172" href="#172">172</a>     	r.setStrict(false);
<a name="173" href="#173">173</a>     	r.setDigest(<strong>true</strong>);
<a name="174" href="#174">174</a>     	output(r, CDX_FILE);
<a name="175" href="#175">175</a>     }
<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>     * Command-line interface to WARCReader.</em>
<a name="179" href="#179">179</a> <em>     *</em>
<a name="180" href="#180">180</a> <em>     * Here is the command-line interface:</em>
<a name="181" href="#181">181</a> <em>     * &lt;pre></em>
<a name="182" href="#182">182</a> <em>     * usage: java org.archive.io.arc.WARCReader [--offset=#] ARCFILE</em>
<a name="183" href="#183">183</a> <em>     *  -h,--help      Prints this message and exits.</em>
<a name="184" href="#184">184</a> <em>     *  -o,--offset    Outputs record at this offset into arc file.&lt;/pre></em>
<a name="185" href="#185">185</a> <em>     *</em>
<a name="186" href="#186">186</a> <em>     * &lt;p>Outputs using a pseudo-CDX format as described here:</em>
<a name="187" href="#187">187</a> <em>     * &lt;a href="<a href="http://www.archive.org/web/researcher/cdx_legend.php" target="alexandria_uri">http://www.archive.org/web/researcher/cdx_legend.php</a>">CDX</em>
<a name="188" href="#188">188</a> <em>     * Legent&lt;/a> and here</em>
<a name="189" href="#189">189</a> <em>     * &lt;a href="<a href="http://www.archive.org/web/researcher/example_cdx.php" target="alexandria_uri">http://www.archive.org/web/researcher/example_cdx.php</a>">Example&lt;/a>.</em>
<a name="190" href="#190">190</a> <em>     * Legend used in below is: 'CDX b e a m s c V (or v if uncompressed) n g'.</em>
<a name="191" href="#191">191</a> <em>     * Hash is hard-coded straight SHA-1 hash of content.</em>
<a name="192" href="#192">192</a> <em>     *</em>
<a name="193" href="#193">193</a> <em>     * @param args Command-line arguments.</em>
<a name="194" href="#194">194</a> <em>     * @throws ParseException Failed parse of the command line.</em>
<a name="195" href="#195">195</a> <em>     * @throws IOException</em>
<a name="196" href="#196">196</a> <em>     * @throws java.text.ParseException</em>
<a name="197" href="#197">197</a> <em>     */</em>
<a name="198" href="#198">198</a>     <strong>public</strong> <strong>static</strong> <strong>void</strong> main(String [] args)
<a name="199" href="#199">199</a>     throws ParseException, IOException, java.text.ParseException {
<a name="200" href="#200">200</a>         Options options = getOptions();
<a name="201" href="#201">201</a>         PosixParser parser = <strong>new</strong> PosixParser();
<a name="202" href="#202">202</a>         CommandLine cmdline = parser.parse(options, args, false);
<a name="203" href="#203">203</a>         List cmdlineArgs = cmdline.getArgList();
<a name="204" href="#204">204</a>         Option [] cmdlineOptions = cmdline.getOptions();
<a name="205" href="#205">205</a>         HelpFormatter formatter = <strong>new</strong> HelpFormatter();
<a name="206" href="#206">206</a> 
<a name="207" href="#207">207</a>         <em class="comment">// If no args, print help.</em>
<a name="208" href="#208">208</a>         <strong>if</strong> (cmdlineArgs.size() &lt;= 0) {
<a name="209" href="#209">209</a>             usage(formatter, options, 0);
<a name="210" href="#210">210</a>         }
<a name="211" href="#211">211</a> 
<a name="212" href="#212">212</a>         <em class="comment">// Now look at options passed.</em>
<a name="213" href="#213">213</a>         <strong>long</strong> offset = -1;
<a name="214" href="#214">214</a>         <strong>boolean</strong> digest = false;
<a name="215" href="#215">215</a>         <strong>boolean</strong> strict = false;
<a name="216" href="#216">216</a>         String format = CDX;
<a name="217" href="#217">217</a>         <strong>for</strong> (<strong>int</strong> i = 0; i &lt; cmdlineOptions.length; i++) {
<a name="218" href="#218">218</a>             <strong>switch</strong>(cmdlineOptions[i].getId()) {
<a name="219" href="#219">219</a>                 <strong>case</strong> 'h':
<a name="220" href="#220">220</a>                     usage(formatter, options, 0);
<a name="221" href="#221">221</a>                     <strong>break</strong>;
<a name="222" href="#222">222</a> 
<a name="223" href="#223">223</a>                 <strong>case</strong> 'o':
<a name="224" href="#224">224</a>                     offset =
<a name="225" href="#225">225</a>                         Long.parseLong(cmdlineOptions[i].getValue());
<a name="226" href="#226">226</a>                     <strong>break</strong>;
<a name="227" href="#227">227</a>                     
<a name="228" href="#228">228</a>                 <strong>case</strong> 's':
<a name="229" href="#229">229</a>                     strict = <strong>true</strong>;
<a name="230" href="#230">230</a>                     <strong>break</strong>;
<a name="231" href="#231">231</a>                     
<a name="232" href="#232">232</a>                 <strong>case</strong> 'd':
<a name="233" href="#233">233</a>                 	digest = getTrueOrFalse(cmdlineOptions[i].getValue());
<a name="234" href="#234">234</a>                     <strong>break</strong>;
<a name="235" href="#235">235</a>                     
<a name="236" href="#236">236</a>                 <strong>case</strong> 'f':
<a name="237" href="#237">237</a>                     format = cmdlineOptions[i].getValue().toLowerCase();
<a name="238" href="#238">238</a>                     <strong>boolean</strong> match = false;
<a name="239" href="#239">239</a>                     <em class="comment">// List of supported formats.</em>
<a name="240" href="#240">240</a>                     <strong>final</strong> String [] supportedFormats =
<a name="241" href="#241">241</a>                 		{CDX, DUMP, GZIP_DUMP, CDX_FILE};
<a name="242" href="#242">242</a>                     <strong>for</strong> (<strong>int</strong> ii = 0; ii &lt; supportedFormats.length; ii++) {
<a name="243" href="#243">243</a>                         <strong>if</strong> (supportedFormats[ii].equals(format)) {
<a name="244" href="#244">244</a>                             match = <strong>true</strong>;
<a name="245" href="#245">245</a>                             <strong>break</strong>;
<a name="246" href="#246">246</a>                         }
<a name="247" href="#247">247</a>                     }
<a name="248" href="#248">248</a>                     <strong>if</strong> (!match) {
<a name="249" href="#249">249</a>                         usage(formatter, options, 1);
<a name="250" href="#250">250</a>                     }
<a name="251" href="#251">251</a>                     <strong>break</strong>;
<a name="252" href="#252">252</a> 
<a name="253" href="#253">253</a>                 <strong>default</strong>:
<a name="254" href="#254">254</a>                     <strong>throw</strong> <strong>new</strong> RuntimeException(<span class="string">"Unexpected option: "</span> +
<a name="255" href="#255">255</a>                         + cmdlineOptions[i].getId());
<a name="256" href="#256">256</a>             }
<a name="257" href="#257">257</a>         }
<a name="258" href="#258">258</a>         
<a name="259" href="#259">259</a>         <strong>if</strong> (offset >= 0) {
<a name="260" href="#260">260</a>             <strong>if</strong> (cmdlineArgs.size() != 1) {
<a name="261" href="#261">261</a>                 System.out.println(<span class="string">"Error: Pass one arcfile only."</span>);
<a name="262" href="#262">262</a>                 usage(formatter, options, 1);
<a name="263" href="#263">263</a>             }
<a name="264" href="#264">264</a>             <a href="../../../../org/archive/io/warc/WARCReader.html">WARCReader</a> r = WARCReaderFactory.get(
<a name="265" href="#265">265</a>             	<strong>new</strong> File((String)cmdlineArgs.get(0)), offset);
<a name="266" href="#266">266</a>             r.setStrict(strict);
<a name="267" href="#267">267</a>             outputRecord(r, format);
<a name="268" href="#268">268</a>         } <strong>else</strong> {
<a name="269" href="#269">269</a>             <strong>for</strong> (Iterator i = cmdlineArgs.iterator(); i.hasNext();) {
<a name="270" href="#270">270</a>                 String urlOrPath = (String)i.next();
<a name="271" href="#271">271</a>                 <strong>try</strong> {
<a name="272" href="#272">272</a>                 	WARCReader r = WARCReaderFactory.get(urlOrPath);
<a name="273" href="#273">273</a>                 	r.setStrict(strict);
<a name="274" href="#274">274</a>                 	r.setDigest(digest);
<a name="275" href="#275">275</a>                     output(r, format);
<a name="276" href="#276">276</a>                 } <strong>catch</strong> (RuntimeException e) {
<a name="277" href="#277">277</a>                     <em class="comment">// Write out name of file we failed on to help with</em>
<a name="278" href="#278">278</a>                     <em class="comment">// debugging.  Then print stack trace and try to keep</em>
<a name="279" href="#279">279</a>                     <em class="comment">// going.  We do this for case where we're being fed</em>
<a name="280" href="#280">280</a>                     <em class="comment">// a bunch of ARCs; just note the bad one and move</em>
<a name="281" href="#281">281</a>                     <em class="comment">// on to the next.</em>
<a name="282" href="#282">282</a>                     System.err.println(<span class="string">"Exception processing "</span> + urlOrPath +
<a name="283" href="#283">283</a>                         <span class="string">": "</span> + e.getMessage());
<a name="284" href="#284">284</a>                     e.printStackTrace(System.err);
<a name="285" href="#285">285</a>                     System.exit(1);
<a name="286" href="#286">286</a>                 }
<a name="287" href="#287">287</a>             }
<a name="288" href="#288">288</a>         }
<a name="289" href="#289">289</a>     } 
<a name="290" href="#290">290</a> }
</pre>
<hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/">Maven</a></div></body>
</html>