crawlorder.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="246" href="#246">246</a>                 <span class="string">"'Mozilla/5.0 (compatible; loc-crawler/0.11.0 "</span> +
<a name="247" href="#247">247</a>                 <span class="string">"+http://loc.gov)'. "</span> +
<a name="248" href="#248">248</a>                 <span class="string">"Note, you must preserve the '+' before the 'http'."</span>,
<a name="249" href="#249">249</a>           <span class="string">"Mozilla/5.0 (compatible; heritrix/@VERSION@ +PROJECT_URL_HERE)"</span>));
<a name="250" href="#250">250</a> 
<a name="251" href="#251">251</a>         e = httpHeaders.addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_FROM,
<a name="252" href="#252">252</a>                 <span class="string">"Contact information. This field must contain a valid "</span> +
<a name="253" href="#253">253</a>                 <span class="string">"e-mail address for the person or organization responsible"</span> +
<a name="254" href="#254">254</a>                 <span class="string">"for this crawl: e.g. '[e-mail]'"</span>,
<a name="255" href="#255">255</a>                 <span class="string">"CONTACT_EMAIL_ADDRESS_HERE"</span>));
<a name="256" href="#256">256</a> 
<a name="257" href="#257">257</a>         addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/datamodel/RobotsHonoringPolicy.html">RobotsHonoringPolicy</a>());
<a name="258" href="#258">258</a> 
<a name="259" href="#259">259</a>         e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/ModuleType.html">ModuleType</a>(
<a name="260" href="#260">260</a>                 Frontier.ATTR_NAME, <span class="string">"Frontier"</span>));
<a name="261" href="#261">261</a>         e.setLegalValueType(Frontier.<strong>class</strong>);
<a name="262" href="#262">262</a> 
<a name="263" href="#263">263</a>         e = (MapType) addElementToDefinition(<strong>new</strong> MapType(ATTR_RULES,
<a name="264" href="#264">264</a>             <span class="string">"Ordered list of url canonicalization rules. "</span> +
<a name="265" href="#265">265</a>             <span class="string">"Rules are applied in the order listed from top to bottom."</span>,
<a name="266" href="#266">266</a>             BaseRule.<strong>class</strong>));
<a name="267" href="#267">267</a>         e.setOverrideable(<strong>true</strong>);
<a name="268" href="#268">268</a>         e.setExpertSetting(<strong>true</strong>);
<a name="269" href="#269">269</a>         
<a name="270" href="#270">270</a>         e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/MapType.html">MapType</a>(
<a name="271" href="#271">271</a>                 ATTR_PRE_FETCH_PROCESSORS, <span class="string">"Processors to run prior to"</span> +
<a name="272" href="#272">272</a>                         <span class="string">" fetching anything from the network."</span>,
<a name="273" href="#273">273</a>                         Processor.<strong>class</strong>));
<a name="274" href="#274">274</a>         e.setOverrideable(false);
<a name="275" href="#275">275</a> 
<a name="276" href="#276">276</a>         e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/MapType.html">MapType</a>(
<a name="277" href="#277">277</a>                 ATTR_FETCH_PROCESSORS, <span class="string">"Processors that fetch documents."</span>
<a name="278" href="#278">278</a>                 , Processor.<strong>class</strong>));
<a name="279" href="#279">279</a>         e.setOverrideable(false);
<a name="280" href="#280">280</a> 
<a name="281" href="#281">281</a>         e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/MapType.html">MapType</a>(
<a name="282" href="#282">282</a>                 ATTR_EXTRACT_PROCESSORS, <span class="string">"Processors that extract new URIs"</span> +
<a name="283" href="#283">283</a>                         <span class="string">" from fetched documents."</span>, Processor.<strong>class</strong>));
<a name="284" href="#284">284</a>         e.setOverrideable(false);
<a name="285" href="#285">285</a> 
<a name="286" href="#286">286</a>         e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/MapType.html">MapType</a>(
<a name="287" href="#287">287</a>                 ATTR_WRITE_PROCESSORS, <span class="string">"Processors that write documents"</span> +
<a name="288" href="#288">288</a>                         <span class="string">" to archives."</span>, Processor.<strong>class</strong>));
<a name="289" href="#289">289</a>         e.setOverrideable(false);
<a name="290" href="#290">290</a> 
<a name="291" href="#291">291</a>         e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/MapType.html">MapType</a>(
<a name="292" href="#292">292</a>                 ATTR_POST_PROCESSORS, <span class="string">"Processors that do cleanup and feed"</span> +
<a name="293" href="#293">293</a>                         <span class="string">" the frontier with new URIs."</span>, Processor.<strong>class</strong>));
<a name="294" href="#294">294</a>         e.setOverrideable(false);
<a name="295" href="#295">295</a> 
<a name="296" href="#296">296</a>         loggers = (MapType) addElementToDefinition(<strong>new</strong> MapType(ATTR_LOGGERS,
<a name="297" href="#297">297</a>                 <span class="string">"Statistics tracking modules. Any number of specialized "</span> +
<a name="298" href="#298">298</a>                 <span class="string">"statistics tracker that monitor a crawl and write logs, "</span> +
<a name="299" href="#299">299</a>                 <span class="string">"reports and/or provide information to the user interface."</span>));
<a name="300" href="#300">300</a> 
<a name="301" href="#301">301</a>         e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_RECOVER_PATH,
<a name="302" href="#302">302</a>                 <span class="string">"Optional. Points at recover log (or recover.gz log) OR "</span> +
<a name="303" href="#303">303</a>                 <span class="string">"the checkpoint directory to use recovering a crawl."</span>, <span class="string">""</span>));
<a name="304" href="#304">304</a>         e.setOverrideable(false);
<a name="305" href="#305">305</a>         e.setExpertSetting(<strong>true</strong>);
<a name="306" href="#306">306</a>         
<a name="307" href="#307">307</a>         e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(
<a name="308" href="#308">308</a>             ATTR_CHECKPOINT_COPY_BDBJE_LOGS,
<a name="309" href="#309">309</a>             <span class="string">"When true, on a checkpoint, we copy off the bdbje log files to "</span> +
<a name="310" href="#310">310</a>             <span class="string">"the checkpoint directory. To recover a checkpoint, just "</span> +
<a name="311" href="#311">311</a>             <span class="string">"set the "</span> + ATTR_RECOVER_PATH + <span class="string">" to point at the checkpoint "</span> +
<a name="312" href="#312">312</a>             <span class="string">"directory to recover.  This is default setting. "</span> +
<a name="313" href="#313">313</a>             <span class="string">"But if crawl is large, "</span> +
<a name="314" href="#314">314</a>             <span class="string">"copying bdbje log files can take tens of minutes and even "</span> +
<a name="315" href="#315">315</a>             <span class="string">"upwards of an hour (Copying bdbje log files will consume bulk "</span> +
<a name="316" href="#316">316</a>             <span class="string">"of time checkpointing). If this setting is false, we do NOT copy "</span> +
<a name="317" href="#317">317</a>             <span class="string">"bdbje logs on checkpoint AND we set bdbje to NEVER delete log "</span> +
<a name="318" href="#318">318</a>             <span class="string">"files (instead we have it rename files-to-delete with a '.del'"</span> +
<a name="319" href="#319">319</a>             <span class="string">"extension). Assumption is that when this setting is false, "</span> +
<a name="320" href="#320">320</a>             <span class="string">"an external process is managing the removal of bdbje log files "</span> +
<a name="321" href="#321">321</a>             <span class="string">"and that come time to recover from a checkpoint, the files that "</span> +
<a name="322" href="#322">322</a>             <span class="string">"comprise a checkpoint are manually assembled. This is an expert "</span> +
<a name="323" href="#323">323</a>             <span class="string">"setting."</span>,
<a name="324" href="#324">324</a>             DEFAULT_CHECKPOINT_COPY_BDBJE_LOGS));
<a name="325" href="#325">325</a>         e.setOverrideable(false);
<a name="326" href="#326">326</a>         e.setExpertSetting(<strong>true</strong>);
<a name="327" href="#327">327</a> 
<a name="328" href="#328">328</a>         e = addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_RECOVER_RETAIN_FAILURES,
<a name="329" href="#329">329</a>                 <span class="string">"When recovering via the recover.log, should failures "</span> +
<a name="330" href="#330">330</a>                 <span class="string">"in the log be retained in the recovered crawl, "</span> +
<a name="331" href="#331">331</a>                 <span class="string">"preventing the corresponding URIs from being retried. "</span> +
<a name="332" href="#332">332</a>                 <span class="string">"Default is false, meaning failures are forgotten, and "</span> +
<a name="333" href="#333">333</a>                 <span class="string">"the corresponding URIs will be retried in the recovered "</span> +
<a name="334" href="#334">334</a>                 <span class="string">"crawl."</span>, Boolean.FALSE));
<a name="335" href="#335">335</a>         e.setOverrideable(false);
<a name="336" href="#336">336</a>         e.setExpertSetting(<strong>true</strong>);
<a name="337" href="#337">337</a>         
<a name="338" href="#338">338</a>         e = addElementToDefinition(
<a name="339" href="#339">339</a>            <strong>new</strong> <a href="../../../../org/archive/crawler/datamodel/CredentialStore.html">CredentialStore</a>(CredentialStore.ATTR_NAME));
<a name="340" href="#340">340</a>         e.setOverrideable(<strong>true</strong>);
<a name="341" href="#341">341</a>         e.setExpertSetting(<strong>true</strong>);
<a name="342" href="#342">342</a>     }
<a name="343" href="#343">343</a> 
<a name="344" href="#344">344</a>     <em>/**<em>*</em></em>
<a name="345" href="#345">345</a> <em>     * @param curi</em>
<a name="346" href="#346">346</a> <em>     * @return user-agent header value to use</em>
<a name="347" href="#347">347</a> <em>     */</em>
<a name="348" href="#348">348</a>     <strong>public</strong> String getUserAgent(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="349" href="#349">349</a>         <strong>return</strong> ((String) httpHeaders.getUncheckedAttribute(curi, ATTR_USER_AGENT));
<a name="350" href="#350">350</a>     }
<a name="351" href="#351">351</a> 
<a name="352" href="#352">352</a>     <em>/**<em>*</em></em>
<a name="353" href="#353">353</a> <em>     * @param curi</em>
<a name="354" href="#354">354</a> <em>     * @return from header value to use</em>
<a name="355" href="#355">355</a> <em>     */</em>
<a name="356" href="#356">356</a>     <strong>public</strong> String getFrom(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) {
<a name="357" href="#357">357</a>         String res = <strong>null</strong>;
<a name="358" href="#358">358</a>         <strong>try</strong> {
<a name="359" href="#359">359</a>             res = (String) httpHeaders.getAttribute(ATTR_FROM, curi);
<a name="360" href="#360">360</a>         } <strong>catch</strong> (AttributeNotFoundException e) {
<a name="361" href="#361">361</a>             logger.severe(e.getMessage());
<a name="362" href="#362">362</a>         }
<a name="363" href="#363">363</a>         <strong>return</strong> res;
<a name="364" href="#364">364</a>     }
<a name="365" href="#365">365</a> 
<a name="366" href="#366">366</a>     <em>/**<em>*</em></em>
<a name="367" href="#367">367</a> <em>     * Returns the set number of maximum toe threads.</em>
<a name="368" href="#368">368</a> <em>     * @return Number of maximum toe threads</em>
<a name="369" href="#369">369</a> <em>     */</em>
<a name="370" href="#370">370</a>     <strong>public</strong> <strong>int</strong> getMaxToes() {
<a name="371" href="#371">371</a>         Integer res = <strong>null</strong>;
<a name="372" href="#372">372</a>         <strong>try</strong> {