mirrorwriterprocessor.html

用JAVA编写的,在做实验的时候留下来的,本来想删的,但是传上来,大家分享吧

<a name="270" href="#270">270</a>             <strong>new</strong> Integer(DEFAULT_MAX_SEG_LEN)));
<a name="271" href="#271">271</a>         addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_PORT_DIRECTORY,
<a name="272" href="#272">272</a>             <span class="string">"Create a subdirectory named for the port in the URI."</span>,
<a name="273" href="#273">273</a>             Boolean.FALSE));
<a name="274" href="#274">274</a>         addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_SUFFIX_AT_END,
<a name="275" href="#275">275</a>             <span class="string">"If true, the suffix is placed at the end of the path, "</span>
<a name="276" href="#276">276</a>             + <span class="string">"after the query (if any).  If false, the suffix is placed "</span>
<a name="277" href="#277">277</a>             + <span class="string">"before the query."</span>,
<a name="278" href="#278">278</a>             Boolean.TRUE));
<a name="279" href="#279">279</a>         e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_TOO_LONG_DIRECTORY,
<a name="280" href="#280">280</a>             <span class="string">"If all the directories in the URI would exceed, "</span>
<a name="281" href="#281">281</a>             + <span class="string">"or come close to exceeding, the file system maximum "</span>
<a name="282" href="#282">282</a>             + <span class="string">"path length, then they are all replaced by this."</span>,
<a name="283" href="#283">283</a>             DEFAULT_TOO_LONG_DIRECTORY));
<a name="284" href="#284">284</a>         e.addConstraint(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/RegularExpressionConstraint.html">RegularExpressionConstraint</a>(TOO_LONG_DIRECTORY_RE,
<a name="285" href="#285">285</a>             Level.SEVERE, <span class="string">"This must be relative and not empty."</span>));
<a name="286" href="#286">286</a>         addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/StringList.html">StringList</a>(ATTR_UNDERSCORE_SET,
<a name="287" href="#287">287</a>             <span class="string">"If a directory name appears (case-insensitive) in this list "</span>
<a name="288" href="#288">288</a>             + <span class="string">"then an underscore is placed before it.  "</span>
<a name="289" href="#289">289</a>             + <span class="string">"For all file systems except Windows, this is not needed.  "</span>
<a name="290" href="#290">290</a>             + <span class="string">"For Windows, the following is recommended: "</span>
<a name="291" href="#291">291</a>             + <span class="string">"[com1 com2 com3 com4 com5 com6 com7 com8 com9 "</span>
<a name="292" href="#292">292</a>             + <span class="string">"lpt1 lpt2 lpt3 lpt4 lpt5 lpt6 lpt7 lpt8 lpt9 "</span>
<a name="293" href="#293">293</a>             + <span class="string">"con nul prn]."</span>));
<a name="294" href="#294">294</a>     }
<a name="295" href="#295">295</a> 
<a name="296" href="#296">296</a>     <strong>protected</strong> <strong>void</strong> innerProcess(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="297" href="#297">297</a>         <strong>if</strong> (!curi.isSuccess()) {
<a name="298" href="#298">298</a>             <strong>return</strong>;
<a name="299" href="#299">299</a>         }
<a name="300" href="#300">300</a>         <a href="../../../../org/archive/net/UURI.html">UURI</a> uuri = curi.getUURI(); <em class="comment">// Current URI.</em>
<a name="301" href="#301">301</a> 
<a name="302" href="#302">302</a>         <em class="comment">// Only http and https schemes are supported.</em>
<a name="303" href="#303">303</a>         String scheme = uuri.getScheme();
<a name="304" href="#304">304</a>         <strong>if</strong> (!<span class="string">"http"</span>.equalsIgnoreCase(scheme)
<a name="305" href="#305">305</a>                 &amp;&amp; !<span class="string">"https"</span>.equalsIgnoreCase(scheme)) {
<a name="306" href="#306">306</a>             <strong>return</strong>;
<a name="307" href="#307">307</a>         }
<a name="308" href="#308">308</a>         <a href="../../../../org/archive/io/RecordingInputStream.html">RecordingInputStream</a> recis = curi.getHttpRecorder().getRecordedInput();
<a name="309" href="#309">309</a>         <strong>if</strong> (0L == recis.getResponseContentLength()) {
<a name="310" href="#310">310</a>             <strong>return</strong>;
<a name="311" href="#311">311</a>         }
<a name="312" href="#312">312</a> 
<a name="313" href="#313">313</a>         String baseDir = <strong>null</strong>; <em class="comment">// Base directory.</em>
<a name="314" href="#314">314</a>         String baseSeg = <strong>null</strong>; <em class="comment">// ATTR_PATH value.</em>
<a name="315" href="#315">315</a>         <strong>try</strong> {
<a name="316" href="#316">316</a>             baseSeg = (String) getAttribute(ATTR_PATH, curi);
<a name="317" href="#317">317</a>         } <strong>catch</strong> (AttributeNotFoundException e) {
<a name="318" href="#318">318</a>             logger.warning(e.getLocalizedMessage());
<a name="319" href="#319">319</a>             <strong>return</strong>;
<a name="320" href="#320">320</a>         }
<a name="321" href="#321">321</a> 
<a name="322" href="#322">322</a>         <em class="comment">// Trim any trailing File.separatorChar characters from baseSeg.</em>
<a name="323" href="#323">323</a>         <strong>while</strong> ((baseSeg.length() > 1) &amp;&amp; baseSeg.endsWith(File.separator)) {
<a name="324" href="#324">324</a>             baseSeg = baseSeg.substring(0, baseSeg.length() - 1);
<a name="325" href="#325">325</a>         }
<a name="326" href="#326">326</a>         <strong>if</strong> (0 == baseSeg.length()) {
<a name="327" href="#327">327</a>             baseDir = getController().getDisk().getPath();
<a name="328" href="#328">328</a>         } <strong>else</strong> <strong>if</strong> ((<strong>new</strong> File(baseSeg)).isAbsolute()) {
<a name="329" href="#329">329</a>             baseDir = baseSeg;
<a name="330" href="#330">330</a>         } <strong>else</strong> {
<a name="331" href="#331">331</a>             baseDir = getController().getDisk().getPath() + File.separator
<a name="332" href="#332">332</a>                 + baseSeg;
<a name="333" href="#333">333</a>         }
<a name="334" href="#334">334</a> 
<a name="335" href="#335">335</a>         <em class="comment">// Already have a path for this URI.</em>
<a name="336" href="#336">336</a>         <strong>boolean</strong> reCrawl = curi.containsKey(A_MIRROR_PATH);
<a name="337" href="#337">337</a> 
<a name="338" href="#338">338</a>         <em class="comment">/*</em>
<a name="339" href="#339">339</a> <em class="comment">          The file system path, relative to the value of ATTR_PATH, where</em>
<a name="340" href="#340">340</a> <em class="comment">          this resource should be written.  The intent is to</em>
<a name="341" href="#341">341</a> <em class="comment">          add later a persistent mapping from URI to path.</em>
<a name="342" href="#342">342</a> <em class="comment">          This will allow a URI to be re-crawled and updated</em>
<a name="343" href="#343">343</a> <em class="comment">          if it has changed.  If the resource has already been fetched</em>
<a name="344" href="#344">344</a> <em class="comment">          and written to a file before, the path to that file</em>
<a name="345" href="#345">345</a> <em class="comment">          has already been obtained from the persistent mapping</em>
<a name="346" href="#346">346</a> <em class="comment">          and placed on the AList by some other module,</em>
<a name="347" href="#347">347</a> <em class="comment">          such as the frontier.</em>
<a name="348" href="#348">348</a> <em class="comment">        */</em>
<a name="349" href="#349">349</a>         String mps = <strong>null</strong>;
<a name="350" href="#350">350</a>         File destFile = <strong>null</strong>; <em class="comment">// Write resource contents to this file.</em>
<a name="351" href="#351">351</a>         <strong>try</strong> {
<a name="352" href="#352">352</a>             <strong>if</strong> (reCrawl) {
<a name="353" href="#353">353</a>                 mps = curi.getString(A_MIRROR_PATH);
<a name="354" href="#354">354</a>                 destFile = <strong>new</strong> File(baseDir + File.separator + mps);
<a name="355" href="#355">355</a>                 File parent = destFile.getParentFile();
<a name="356" href="#356">356</a>                 <strong>if</strong> (<strong>null</strong> != parent) {
<a name="357" href="#357">357</a>                     IoUtils.ensureWriteableDirectory(parent);
<a name="358" href="#358">358</a>                 }
<a name="359" href="#359">359</a>             } <strong>else</strong> {
<a name="360" href="#360">360</a>                 URIToFileReturn r = <strong>null</strong>; <em class="comment">// Return from uriToFile().</em>
<a name="361" href="#361">361</a>                 <strong>try</strong> {
<a name="362" href="#362">362</a>                      r = uriToFile(baseDir, curi);
<a name="363" href="#363">363</a>                 } <strong>catch</strong> (AttributeNotFoundException e) {
<a name="364" href="#364">364</a>                     logger.warning(e.getLocalizedMessage());
<a name="365" href="#365">365</a>                     <strong>return</strong>;
<a name="366" href="#366">366</a>                 }
<a name="367" href="#367">367</a>                 destFile = r.getFile();
<a name="368" href="#368">368</a>                 mps = r.getRelativePath();
<a name="369" href="#369">369</a>             }
<a name="370" href="#370">370</a>             logger.info(uuri.toString() + <span class="string">" -> "</span> + destFile.getPath());
<a name="371" href="#371">371</a>             writeToPath(recis, destFile);
<a name="372" href="#372">372</a>             <strong>if</strong> (!reCrawl) {
<a name="373" href="#373">373</a>                 curi.putString(A_MIRROR_PATH, mps);
<a name="374" href="#374">374</a>             }
<a name="375" href="#375">375</a>         } <strong>catch</strong> (IOException e) {
<a name="376" href="#376">376</a>             curi.addLocalizedError(<strong>this</strong>.getName(), e, <span class="string">"Mirror"</span>);
<a name="377" href="#377">377</a>         }
<a name="378" href="#378">378</a>     }
<a name="379" href="#379">379</a> 
<a name="380" href="#380">380</a>     <em>/**<em>*</em></em>
<a name="381" href="#381">381</a> <em>       Gets the directory in which the file will reside.</em>
<a name="382" href="#382">382</a> <em>       Any directories needed are created.</em>
<a name="383" href="#383">383</a> <em>       @param baseDir the path to the starting directory</em>
<a name="384" href="#384">384</a> <em>       @param host the host part of the URI, or null if the host name</em>
<a name="385" href="#385">385</a> <em>       should not be part of the returned path</em>
<a name="386" href="#386">386</a> <em>       @param port the port part of the URI, or -1 if the port</em>
<a name="387" href="#387">387</a> <em>       should not be part of the returned path</em>
<a name="388" href="#388">388</a> <em>       @param segs all the segments in the URI</em>
<a name="389" href="#389">389</a> <em>       @param maxLen the maximum path length allowed to the directory;</em>
<a name="390" href="#390">390</a> <em>       this must leave some room for the file itself</em>
<a name="391" href="#391">391</a> <em>       @return the directory, or null if maxLen would be exceeded</em>
<a name="392" href="#392">392</a> <em>       @throws IOException</em>
<a name="393" href="#393">393</a> <em>       if a needed directory could not be created</em>
<a name="394" href="#394">394</a> <em>       @throws IOException</em>
<a name="395" href="#395">395</a> <em>       if a needed directory is not writeable</em>
<a name="396" href="#396">396</a> <em>       @throws IOException</em>
<a name="397" href="#397">397</a> <em>       if a non-directory file exists with the same path as a needed directory</em>
<a name="398" href="#398">398</a> <em>    */</em>
<a name="399" href="#399">399</a>     <strong>private</strong> URIToFileReturn dirPath(String baseDir, String host, <strong>int</strong> port,
<a name="400" href="#400">400</a>                                     PathSegment[] segs, <strong>int</strong> maxLen)
<a name="401" href="#401">401</a>         throws IOException {
<a name="402" href="#402">402</a> 
<a name="403" href="#403">403</a>         <em class="comment">// Return value.</em>
<a name="404" href="#404">404</a>         URIToFileReturn r = <strong>new</strong> URIToFileReturn(baseDir, host, port);
<a name="405" href="#405">405</a>         r.mkdirs();
<a name="406" href="#406">406</a>         <strong>for</strong> (<strong>int</strong> i = 0; (segs.length - 1) != i; ++i) {
<a name="407" href="#407">407</a>             segs[i].addToPath(r);
<a name="408" href="#408">408</a>             <strong>if</strong> (r.longerThan(maxLen)) {