crawljob.html

一个开源的网页爬虫一个开源的网页爬虫一个开源的网页爬虫一个开源的网页爬虫一个开源的网页爬虫一个开源的网页爬虫

<a name="262" href="#262">262</a>         TOTAL_DATA_ATTR, CRAWL_TIME_ATTR, DOC_RATE_ATTR,
<a name="263" href="#263">263</a>         CURRENT_DOC_RATE_ATTR, KB_RATE_ATTR, CURRENT_KB_RATE_ATTR,
<a name="264" href="#264">264</a>         THREAD_COUNT_ATTR, DOWNLOAD_COUNT_ATTR, DISCOVERED_COUNT_ATTR};
<a name="265" href="#265">265</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> List ATTRIBUTE_LIST = Arrays.asList(ATTRIBUTE_ARRAY);
<a name="266" href="#266">266</a>     
<a name="267" href="#267">267</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String IMPORT_URI_OPER = <span class="string">"importUri"</span>;
<a name="268" href="#268">268</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String IMPORT_URIS_OPER = <span class="string">"importUris"</span>;
<a name="269" href="#269">269</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String PAUSE_OPER = <span class="string">"pause"</span>;
<a name="270" href="#270">270</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String RESUME_OPER = <span class="string">"resume"</span>;
<a name="271" href="#271">271</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String FRONTIER_REPORT_OPER = <span class="string">"frontierReport"</span>;
<a name="272" href="#272">272</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String THREADS_REPORT_OPER = <span class="string">"threadsReport"</span>;
<a name="273" href="#273">273</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String SEEDS_REPORT_OPER = <span class="string">"seedsReport"</span>;
<a name="274" href="#274">274</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String CHECKPOINT_OPER = <span class="string">"startCheckpoint"</span>;
<a name="275" href="#275">275</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String PROGRESS_STATISTICS_OPER =
<a name="276" href="#276">276</a>         <span class="string">"progressStatistics"</span>;
<a name="277" href="#277">277</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String PROGRESS_STATISTICS_LEGEND_OPER =
<a name="278" href="#278">278</a>         <span class="string">"progressStatisticsLegend"</span>;
<a name="279" href="#279">279</a>     
<a name="280" href="#280">280</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String PROG_STATS = <span class="string">"progressStatistics"</span>;
<a name="281" href="#281">281</a>     
<a name="282" href="#282">282</a>     <em class="comment">// Same as JEMBeanHelper.OP_DB_STAT</em>
<a name="283" href="#283">283</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> String OP_DB_STAT = <span class="string">"getDatabaseStats"</span>;
<a name="284" href="#284">284</a>     
<a name="285" href="#285">285</a>     <em>/**<em>*</em></em>
<a name="286" href="#286">286</a> <em>     * Don't add the following crawl-order items.</em>
<a name="287" href="#287">287</a> <em>     */</em>
<a name="288" href="#288">288</a>     <strong>private</strong> <strong>final</strong> <strong>static</strong> List ORDER_EXCLUDE;
<a name="289" href="#289">289</a>     <strong>static</strong> {
<a name="290" href="#290">290</a>         ORDER_EXCLUDE = Arrays.asList(<strong>new</strong> String [] {<span class="string">"bdb-cache-percent"</span>,
<a name="291" href="#291">291</a>             <span class="string">"extract-processors"</span>, <span class="string">"DNS"</span>, <span class="string">"uri-included-structure"</span>});
<a name="292" href="#292">292</a>     }
<a name="293" href="#293">293</a>     
<a name="294" href="#294">294</a>     <em>/**<em>*</em></em>
<a name="295" href="#295">295</a> <em>     * Sequence number for jmx notifications.</em>
<a name="296" href="#296">296</a> <em>     */</em>
<a name="297" href="#297">297</a>     <strong>private</strong> <strong>static</strong> <strong>int</strong> notificationsSequenceNumber = 1;
<a name="298" href="#298">298</a>     
<a name="299" href="#299">299</a>     <em>/**<em>*</em></em>
<a name="300" href="#300">300</a> <em>     * A shutdown Constructor.</em>
<a name="301" href="#301">301</a> <em>     */</em>
<a name="302" href="#302">302</a>     <strong>protected</strong> <a href="../../../../org/archive/crawler/admin/CrawlJob.html">CrawlJob</a>() {
<a name="303" href="#303">303</a>         <strong>super</strong>();
<a name="304" href="#304">304</a>     }
<a name="305" href="#305">305</a> 
<a name="306" href="#306">306</a>     <em>/**<em>*</em></em>
<a name="307" href="#307">307</a> <em>     * A constructor for jobs.</em>
<a name="308" href="#308">308</a> <em>     *</em>
<a name="309" href="#309">309</a> <em>     * &lt;p> Create, ready to crawl, jobs.</em>
<a name="310" href="#310">310</a> <em>     * @param UID A unique ID for this job. Typically emitted by the</em>
<a name="311" href="#311">311</a> <em>     *            CrawlJobHandler.</em>
<a name="312" href="#312">312</a> <em>     * @param name The name of the job</em>
<a name="313" href="#313">313</a> <em>     * @param settingsHandler The associated settings</em>
<a name="314" href="#314">314</a> <em>     * @param errorHandler The crawl jobs settings error handler.</em>
<a name="315" href="#315">315</a> <em>     *           &lt;tt>null&lt;/tt> means none is set</em>
<a name="316" href="#316">316</a> <em>     * @param priority job priority.</em>
<a name="317" href="#317">317</a> <em>     * @param dir The directory that is considered this jobs working directory.</em>
<a name="318" href="#318">318</a> <em>     */</em>
<a name="319" href="#319">319</a>     <strong>public</strong> <a href="../../../../org/archive/crawler/admin/CrawlJob.html">CrawlJob</a>(<strong>final</strong> String UID,
<a name="320" href="#320">320</a>             <strong>final</strong> String name, <strong>final</strong> <a href="../../../../org/archive/crawler/settings/XMLSettingsHandler.html">XMLSettingsHandler</a> settingsHandler,
<a name="321" href="#321">321</a>             <strong>final</strong> <a href="../../../../org/archive/crawler/admin/CrawlJobErrorHandler.html">CrawlJobErrorHandler</a> errorHandler, <strong>final</strong> <strong>int</strong> priority,
<a name="322" href="#322">322</a>             <strong>final</strong> File dir) {
<a name="323" href="#323">323</a>         <strong>this</strong>(UID, name, settingsHandler, errorHandler,
<a name="324" href="#324">324</a>                 priority, dir, <strong>null</strong>, false, <strong>true</strong>);
<a name="325" href="#325">325</a>     }
<a name="326" href="#326">326</a> 
<a name="327" href="#327">327</a>     <em>/**<em>*</em></em>
<a name="328" href="#328">328</a> <em>     * A constructor for profiles.</em>
<a name="329" href="#329">329</a> <em>     *</em>
<a name="330" href="#330">330</a> <em>     * &lt;p> Any job created with this constructor will be</em>
<a name="331" href="#331">331</a> <em>     * considered a profile. Profiles are not stored on disk (only their</em>
<a name="332" href="#332">332</a> <em>     * settings files are stored on disk). This is because their data is</em>
<a name="333" href="#333">333</a> <em>     * predictible given any settings files.</em>
<a name="334" href="#334">334</a> <em>     * @param UIDandName A unique ID for this job. For profiles this is the same</em>
<a name="335" href="#335">335</a> <em>     *           as name</em>
<a name="336" href="#336">336</a> <em>     * @param settingsHandler The associated settings</em>
<a name="337" href="#337">337</a> <em>     * @param errorHandler The crawl jobs settings error handler.</em>
<a name="338" href="#338">338</a> <em>     *           &lt;tt>null&lt;/tt> means none is set</em>
<a name="339" href="#339">339</a> <em>     */</em>
<a name="340" href="#340">340</a>     <strong>protected</strong> <a href="../../../../org/archive/crawler/admin/CrawlJob.html">CrawlJob</a>(<strong>final</strong> String UIDandName,
<a name="341" href="#341">341</a>             <strong>final</strong> <a href="../../../../org/archive/crawler/settings/XMLSettingsHandler.html">XMLSettingsHandler</a> settingsHandler,
<a name="342" href="#342">342</a>             <strong>final</strong> <a href="../../../../org/archive/crawler/admin/CrawlJobErrorHandler.html">CrawlJobErrorHandler</a> errorHandler) {
<a name="343" href="#343">343</a>         <strong>this</strong>(UIDandName, UIDandName, settingsHandler, errorHandler,
<a name="344" href="#344">344</a>             PRIORITY_AVERAGE, <strong>null</strong>, STATUS_PROFILE, <strong>true</strong>, false);
<a name="345" href="#345">345</a>     }
<a name="346" href="#346">346</a>     
<a name="347" href="#347">347</a>     <strong>public</strong> <a href="../../../../org/archive/crawler/admin/CrawlJob.html">CrawlJob</a>(<strong>final</strong> String UID,
<a name="348" href="#348">348</a>             <strong>final</strong> String name, <strong>final</strong> <a href="../../../../org/archive/crawler/settings/XMLSettingsHandler.html">XMLSettingsHandler</a> settingsHandler,
<a name="349" href="#349">349</a>             <strong>final</strong> <a href="../../../../org/archive/crawler/admin/CrawlJobErrorHandler.html">CrawlJobErrorHandler</a> errorHandler, <strong>final</strong> <strong>int</strong> priority,
<a name="350" href="#350">350</a>             <strong>final</strong> File dir, <strong>final</strong> String status, <strong>final</strong> <strong>boolean</strong> isProfile,
<a name="351" href="#351">351</a>             <strong>final</strong> <strong>boolean</strong> isNew) {
<a name="352" href="#352">352</a>         <strong>super</strong>();
<a name="353" href="#353">353</a>         <strong>this</strong>.UID = UID;
<a name="354" href="#354">354</a>         <strong>this</strong>.name = name;
<a name="355" href="#355">355</a>         <strong>this</strong>.settingsHandler = settingsHandler;
<a name="356" href="#356">356</a>         <strong>this</strong>.errorHandler = errorHandler;
<a name="357" href="#357">357</a>         <strong>this</strong>.status = status;
<a name="358" href="#358">358</a>         <strong>this</strong>.isProfile = isProfile;
<a name="359" href="#359">359</a>         <strong>this</strong>.isNew = isNew;
<a name="360" href="#360">360</a>         <strong>this</strong>.jobDir = dir;
<a name="361" href="#361">361</a>         <strong>this</strong>.priority = priority;
<a name="362" href="#362">362</a>     }
<a name="363" href="#363">363</a> 
<a name="364" href="#364">364</a>     <em>/**<em>*</em></em>
<a name="365" href="#365">365</a> <em>     * A constructor for reloading jobs from disk. Jobs (not profiles) have</em>
<a name="366" href="#366">366</a> <em>     * their data written to persistent storage in the file system. This method</em>
<a name="367" href="#367">367</a> <em>     * is used to load the job from such storage. This is done by the</em>
<a name="368" href="#368">368</a> <em>     * &lt;code>CrawlJobHandler&lt;/code>.</em>
<a name="369" href="#369">369</a> <em>     * &lt;p></em>
<a name="370" href="#370">370</a> <em>     * Proper structure of a job file (TODO: Maybe one day make this an XML file)</em>
<a name="371" href="#371">371</a> <em>     * Line 1. UID &lt;br></em>
<a name="372" href="#372">372</a> <em>     * Line 2. Job name (string) &lt;br></em>
<a name="373" href="#373">373</a> <em>     * Line 3. Job status (string) &lt;br></em>
<a name="374" href="#374">374</a> <em>     * Line 4. is job read only (true/false) &lt;br></em>
<a name="375" href="#375">375</a> <em>     * Line 5. is job running (true/false) &lt;br></em>
<a name="376" href="#376">376</a> <em>     * Line 6. job priority (int) &lt;br></em>
<a name="377" href="#377">377</a> <em>     * Line 7. number of journal entries &lt;br></em>
<a name="378" href="#378">378</a> <em>     * Line 8. setting file (with path) &lt;br></em>
<a name="379" href="#379">379</a> <em>     * Line 9. statistics tracker file (with path) &lt;br></em>
<a name="380" href="#380">380</a> <em>     * Line 10-?. error message (String, empty for null), can be many lines &lt;br></em>
<a name="381" href="#381">381</a> <em>     * @param jobFile</em>
<a name="382" href="#382">382</a> <em>     *            a file containing information about the job to load.</em>
<a name="383" href="#383">383</a> <em>     * @param errorHandler The crawl jobs settings error handler.</em>
<a name="384" href="#384">384</a> <em>     *            null means none is set</em>
<a name="385" href="#385">385</a> <em>     * @throws InvalidJobFileException</em>
<a name="386" href="#386">386</a> <em>     *            if the specified file does not refer to a valid job file.</em>
<a name="387" href="#387">387</a> <em>     * @throws IOException</em>
<a name="388" href="#388">388</a> <em>     *            if io operations fail</em>
<a name="389" href="#389">389</a> <em>     */</em>
<a name="390" href="#390">390</a>     <strong>protected</strong> <a href="../../../../org/archive/crawler/admin/CrawlJob.html">CrawlJob</a>(<strong>final</strong> File jobFile,
<a name="391" href="#391">391</a>             <strong>final</strong> <a href="../../../../org/archive/crawler/admin/CrawlJobErrorHandler.html">CrawlJobErrorHandler</a> errorHandler)
<a name="392" href="#392">392</a>             throws InvalidJobFileException, IOException {
<a name="393" href="#393">393</a>         <strong>this</strong>(<strong>null</strong>, <strong>null</strong>, <strong>null</strong>, errorHandler,
<a name="394" href="#394">394</a>                 PRIORITY_AVERAGE, <strong>null</strong>, <strong>null</strong>, false, <strong>true</strong>);
<a name="395" href="#395">395</a>         <strong>this</strong>.jobDir = jobFile.getParentFile();
<a name="396" href="#396">396</a>