extractorhtml.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="268" href="#268">268</a>                 }
<a name="269" href="#269">269</a>             } <strong>else</strong> <strong>if</strong> (attr.start(3) > -1) {
<a name="270" href="#270">270</a>                 <em class="comment">// ACTION</em>
<a name="271" href="#271">271</a>                 <strong>if</strong> (!ignoreFormActions) {
<a name="272" href="#272">272</a>                     CharSequence context = Link.elementContext(element,
<a name="273" href="#273">273</a>                         attr.group(3));
<a name="274" href="#274">274</a>                     processLink(curi, value, context);
<a name="275" href="#275">275</a>                 }
<a name="276" href="#276">276</a>             } <strong>else</strong> <strong>if</strong> (attr.start(4) > -1) {
<a name="277" href="#277">277</a>                 <em class="comment">// ON____</em>
<a name="278" href="#278">278</a>                 processScriptCode(curi, value); <em class="comment">// TODO: context?</em>
<a name="279" href="#279">279</a>             } <strong>else</strong> <strong>if</strong> (attr.start(5) > -1) {
<a name="280" href="#280">280</a>                 <em class="comment">// SRC etc.</em>
<a name="281" href="#281">281</a>                 CharSequence context = Link.elementContext(element,
<a name="282" href="#282">282</a>                     attr.group(5));
<a name="283" href="#283">283</a>                 
<a name="284" href="#284">284</a>                 <em class="comment">// true, if we expect another HTML page instead of an image etc.</em>
<a name="285" href="#285">285</a>                 <strong>final</strong> <strong>char</strong> hopType;
<a name="286" href="#286">286</a>                 
<a name="287" href="#287">287</a>                 <strong>if</strong>(!framesAsEmbeds
<a name="288" href="#288">288</a>                     &amp;&amp; (elementStr.equalsIgnoreCase(FRAME) || elementStr
<a name="289" href="#289">289</a>                         .equalsIgnoreCase(IFRAME))) {
<a name="290" href="#290">290</a>                     hopType = Link.NAVLINK_HOP;
<a name="291" href="#291">291</a>                 } <strong>else</strong> {
<a name="292" href="#292">292</a>                     hopType = Link.EMBED_HOP;
<a name="293" href="#293">293</a>                 }
<a name="294" href="#294">294</a>                 processEmbed(curi, value, context, hopType);
<a name="295" href="#295">295</a>             } <strong>else</strong> <strong>if</strong> (attr.start(6) > -1) {
<a name="296" href="#296">296</a>                 <em class="comment">// CODEBASE</em>
<a name="297" href="#297">297</a>                 codebase = (value instanceof String)?
<a name="298" href="#298">298</a>                     (String)value: value.toString();
<a name="299" href="#299">299</a>                 CharSequence context = Link.elementContext(element,
<a name="300" href="#300">300</a>                     attr.group(6));
<a name="301" href="#301">301</a>                 processEmbed(curi, codebase, context);
<a name="302" href="#302">302</a>             } <strong>else</strong> <strong>if</strong> (attr.start(7) > -1) {
<a name="303" href="#303">303</a>                 <em class="comment">// CLASSID, DATA</em>
<a name="304" href="#304">304</a>                 <strong>if</strong> (resources == <strong>null</strong>) {
<a name="305" href="#305">305</a>                     resources = <strong>new</strong> ArrayList();
<a name="306" href="#306">306</a>                 }
<a name="307" href="#307">307</a>                 resources.add(value.toString());
<a name="308" href="#308">308</a>             } <strong>else</strong> <strong>if</strong> (attr.start(8) > -1) {
<a name="309" href="#309">309</a>                 <em class="comment">// ARCHIVE</em>
<a name="310" href="#310">310</a>                 <strong>if</strong> (resources==<strong>null</strong>) {
<a name="311" href="#311">311</a>                     resources = <strong>new</strong> ArrayList();
<a name="312" href="#312">312</a>                 }
<a name="313" href="#313">313</a>                 String[] multi = TextUtils.split(WHITESPACE, value);
<a name="314" href="#314">314</a>                 <strong>for</strong>(<strong>int</strong> i = 0; i &lt; multi.length; i++ ) {
<a name="315" href="#315">315</a>                     resources.add(multi[i]);
<a name="316" href="#316">316</a>                 }
<a name="317" href="#317">317</a>             } <strong>else</strong> <strong>if</strong> (attr.start(9) > -1) {
<a name="318" href="#318">318</a>                 <em class="comment">// CODE</em>
<a name="319" href="#319">319</a>                 <strong>if</strong> (resources==<strong>null</strong>) {
<a name="320" href="#320">320</a>                     resources = <strong>new</strong> ArrayList();
<a name="321" href="#321">321</a>                 }
<a name="322" href="#322">322</a>                 <em class="comment">// If element is applet and code value does not end with</em>
<a name="323" href="#323">323</a>                 <em class="comment">// '.class' then append '.class' to the code value.</em>
<a name="324" href="#324">324</a>                 <strong>if</strong> (elementStr.equalsIgnoreCase(APPLET) &amp;&amp;
<a name="325" href="#325">325</a>                         !value.toString().toLowerCase().endsWith(CLASSEXT)) {
<a name="326" href="#326">326</a>                     resources.add(value.toString() + CLASSEXT);
<a name="327" href="#327">327</a>                 } <strong>else</strong> {
<a name="328" href="#328">328</a>                     resources.add(value.toString());
<a name="329" href="#329">329</a>                 }
<a name="330" href="#330">330</a>             } <strong>else</strong> <strong>if</strong> (attr.start(10) > -1) {
<a name="331" href="#331">331</a>                 <em class="comment">// VALUE, with possibility of URI</em>
<a name="332" href="#332">332</a>                 <strong>if</strong> (TextUtils.matches(LIKELY_URI_PATH, value)
<a name="333" href="#333">333</a>                         &amp;&amp; overlyEagerLinkDetection) {
<a name="334" href="#334">334</a>                     CharSequence context = Link.elementContext(element,
<a name="335" href="#335">335</a>                         attr.group(10));
<a name="336" href="#336">336</a>                     processLink(curi,value, context);
<a name="337" href="#337">337</a>                 }
<a name="338" href="#338">338</a> 
<a name="339" href="#339">339</a>             } <strong>else</strong> <strong>if</strong> (attr.start(11) > -1) {
<a name="340" href="#340">340</a>                 <em class="comment">// STYLE inline attribute</em>
<a name="341" href="#341">341</a>                 <em class="comment">// then, parse for URIs</em>
<a name="342" href="#342">342</a>                 <strong>this</strong>.numberOfLinksExtracted += ExtractorCSS.processStyleCode(
<a name="343" href="#343">343</a>                     curi, value, getController());
<a name="344" href="#344">344</a>                 
<a name="345" href="#345">345</a>             } <strong>else</strong> <strong>if</strong> (attr.start(12) > -1) {
<a name="346" href="#346">346</a>                 <em class="comment">// any other attribute</em>
<a name="347" href="#347">347</a>                 <em class="comment">// ignore for now</em>
<a name="348" href="#348">348</a>                 <em class="comment">// could probe for path- or script-looking strings, but</em>
<a name="349" href="#349">349</a>                 <em class="comment">// those should be vanishingly rare in other attributes,</em>
<a name="350" href="#350">350</a>                 <em class="comment">// and/or symptomatic of page bugs</em>
<a name="351" href="#351">351</a>             }
<a name="352" href="#352">352</a>         }
<a name="353" href="#353">353</a>         TextUtils.recycleMatcher(attr);
<a name="354" href="#354">354</a> 
<a name="355" href="#355">355</a>         <em class="comment">// handle codebase/resources</em>
<a name="356" href="#356">356</a>         <strong>if</strong> (resources == <strong>null</strong>) {
<a name="357" href="#357">357</a>             <strong>return</strong>;
<a name="358" href="#358">358</a>         }
<a name="359" href="#359">359</a>         Iterator iter = resources.iterator();
<a name="360" href="#360">360</a>         <a href="../../../../org/archive/net/UURI.html">UURI</a> codebaseURI = <strong>null</strong>;
<a name="361" href="#361">361</a>         String res = <strong>null</strong>;
<a name="362" href="#362">362</a>         <strong>try</strong> {
<a name="363" href="#363">363</a>             <strong>if</strong> (codebase != <strong>null</strong>) {
<a name="364" href="#364">364</a>                 <em class="comment">// TODO: Pass in the charset.</em>
<a name="365" href="#365">365</a>                 codebaseURI = UURIFactory.
<a name="366" href="#366">366</a>                     getInstance(curi.getUURI(), codebase);
<a name="367" href="#367">367</a>             }
<a name="368" href="#368">368</a>             <strong>while</strong>(iter.hasNext()) {
<a name="369" href="#369">369</a>                 res = iter.next().toString();
<a name="370" href="#370">370</a>                 res = (String) TextUtils.unescapeHtml(res);
<a name="371" href="#371">371</a>                 <strong>if</strong> (codebaseURI != <strong>null</strong>) {
<a name="372" href="#372">372</a>                     res = codebaseURI.resolve(res).toString();
<a name="373" href="#373">373</a>                 }
<a name="374" href="#374">374</a>                 processEmbed(curi, res, element); <em class="comment">// TODO: include attribute too</em>
<a name="375" href="#375">375</a>             }
<a name="376" href="#376">376</a>         } <strong>catch</strong> (URIException e) {
<a name="377" href="#377">377</a>             curi.addLocalizedError(getName(), e, <span class="string">"BAD CODEBASE "</span> + codebase);
<a name="378" href="#378">378</a>         } <strong>catch</strong> (IllegalArgumentException e) {
<a name="379" href="#379">379</a>             DevUtils.logger.log(Level.WARNING, <span class="string">"processGeneralTag()\n"</span> +
<a name="380" href="#380">380</a>                 <span class="string">"codebase="</span> + codebase + <span class="string">" res="</span> + res + <span class="string">"\n"</span> +
<a name="381" href="#381">381</a>                 DevUtils.extraInfo(), e);
<a name="382" href="#382">382</a>         }
<a name="383" href="#383">383</a>     }
<a name="384" href="#384">384</a> 
<a name="385" href="#385">385</a>     <em class="comment">// finds strings in javascript likely to be URIs/paths</em>
<a name="386" href="#386">386</a>     <em class="comment">// guessing based on '.' in string, so if highly likely to</em>
<a name="387" href="#387">387</a>     <em class="comment">// get gifs/etc, unable to get many other paths</em>
<a name="388" href="#388">388</a>     <em class="comment">// will find false positives</em>
<a name="389" href="#389">389</a>     <em class="comment">// TODO: add '/' check, suppress strings being concatenated via '+'?</em>
<a name="390" href="#390">390</a>     <strong>static</strong> <strong>final</strong> String JAVASCRIPT_LIKELY_URI_EXTRACTOR =
<a name="391" href="#391">391</a>      <span class="string">"(&#47;&#47;&#47;&#47;{0,8}+\&quot;|&#47;&#47;&#47;&#47;{0,8}+\')(&#47;&#47;.{0,2}[^+&#47;&#47;.&#47;&#47;n&#47;&#47;r&#47;&#47;s\&quot;\']+[^&#47;&#47;.&#47;&#47;n&#47;&#47;r&#47;&#47;s\&quot;\']*(&#47;&#47;.[^&#47;&#47;.&#47;&#47;n&#47;&#47;r&#47;&#47;s\&quot;\']+)+)(&#47;&#47;1)"</span>;
<a name="392" href="#392">392</a> 
<a name="393" href="#393">393</a>     <em>/**<em>*</em></em>
<a name="394" href="#394">394</a> <em>     * @param curi</em>
<a name="395" href="#395">395</a> <em>     * @param cs</em>
<a name="396" href="#396">396</a> <em>     */</em>
<a name="397" href="#397">397</a>     <strong>protected</strong> <strong>void</strong> processScriptCode(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi, CharSequence cs) {
<a name="398" href="#398">398</a>         <strong>this</strong>.numberOfLinksExtracted +=
<a name="399" href="#399">399</a>             ExtractorJS.considerStrings(curi, cs, getController(), false);
<a name="400" href="#400">400</a>     }
<a name="401" href="#401">401</a> 
<a name="402" href="#402">402</a>     <strong>static</strong> <strong>final</strong> String JAVASCRIPT = <span class="string">"(?i)^javascript:.*"</span>;
<a name="403" href="#403">403</a> 
<a name="404" href="#404">404</a>     <em>/**<em>*</em></em>
<a name="405" href="#405">405</a> <em>     * Handle generic HREF cases.</em>