jerichoextractorhtml.html

来自「网络爬虫开源代码」· HTML 代码 · 共 477 行 · 第 1/3 页

<a name="310" href="#310">310</a>                     <span class="string">"for: "</span> + curi.toString());
<a name="311" href="#311">311</a>                 <strong>return</strong> <strong>true</strong>;
<a name="312" href="#312">312</a>             }
<a name="313" href="#313">313</a>         }
<a name="314" href="#314">314</a>         <strong>if</strong> (<span class="string">"refresh"</span>.equals(httpEquiv) &amp;&amp; content != <strong>null</strong>) {
<a name="315" href="#315">315</a>             String refreshUri = content.substring(content.indexOf(<span class="string">"="</span>) + 1);
<a name="316" href="#316">316</a>             <strong>try</strong> {
<a name="317" href="#317">317</a>                 curi.createAndAddLinkRelativeToBase(refreshUri, <span class="string">"meta"</span>,
<a name="318" href="#318">318</a>                         Link.REFER_HOP);
<a name="319" href="#319">319</a>             } <strong>catch</strong> (URIException e) {
<a name="320" href="#320">320</a>                 <strong>if</strong> (getController() != <strong>null</strong>) {
<a name="321" href="#321">321</a>                     getController().logUriError(e, curi.getUURI(), refreshUri);
<a name="322" href="#322">322</a>                 } <strong>else</strong> {
<a name="323" href="#323">323</a>                     logger.info(<span class="string">"Failed createAndAddLinkRelativeToBase "</span> + curi
<a name="324" href="#324">324</a>                             + <span class="string">", "</span> + element.toString() + <span class="string">", "</span> + refreshUri
<a name="325" href="#325">325</a>                             + <span class="string">": "</span> + e);
<a name="326" href="#326">326</a>                 }
<a name="327" href="#327">327</a>             }
<a name="328" href="#328">328</a>         }
<a name="329" href="#329">329</a>         <strong>return</strong> false;
<a name="330" href="#330">330</a>     }
<a name="331" href="#331">331</a> 
<a name="332" href="#332">332</a>     <strong>protected</strong> <strong>void</strong> processScript(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi, Element element) {
<a name="333" href="#333">333</a>         <em class="comment">// first, get attributes of script-open tag</em>
<a name="334" href="#334">334</a>         <em class="comment">// as per any other tag</em>
<a name="335" href="#335">335</a>         processGeneralTag(curi, element, element.getAttributes());
<a name="336" href="#336">336</a> 
<a name="337" href="#337">337</a>         <em class="comment">// then, apply best-effort string-analysis heuristics</em>
<a name="338" href="#338">338</a>         <em class="comment">// against any code present (false positives are OK)</em>
<a name="339" href="#339">339</a>         processScriptCode(curi, element.getContent());
<a name="340" href="#340">340</a> 
<a name="341" href="#341">341</a>     }
<a name="342" href="#342">342</a> 
<a name="343" href="#343">343</a>     <strong>protected</strong> <strong>void</strong> processStyle(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi, Element element) {
<a name="344" href="#344">344</a>         <em class="comment">// First, get attributes of script-open tag as per any other tag.</em>
<a name="345" href="#345">345</a>         processGeneralTag(curi, element, element.getAttributes());
<a name="346" href="#346">346</a> 
<a name="347" href="#347">347</a>         <em class="comment">// then, parse for URIs</em>
<a name="348" href="#348">348</a>         <strong>this</strong>.numberOfLinksExtracted += ExtractorCSS.processStyleCode(curi,
<a name="349" href="#349">349</a>                 element.getContent(), getController());
<a name="350" href="#350">350</a>     }
<a name="351" href="#351">351</a> 
<a name="352" href="#352">352</a>     <strong>protected</strong> <strong>void</strong> processForm(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi, Element element) {
<a name="353" href="#353">353</a>         String action = element.getAttributeValue(<span class="string">"action"</span>);
<a name="354" href="#354">354</a>         String name = element.getAttributeValue(<span class="string">"name"</span>);
<a name="355" href="#355">355</a>         String queryURL = <span class="string">""</span>;
<a name="356" href="#356">356</a> 
<a name="357" href="#357">357</a>         <strong>final</strong> <strong>boolean</strong> ignoreFormActions = ((Boolean) getUncheckedAttribute(
<a name="358" href="#358">358</a>                 curi, ATTR_IGNORE_FORM_ACTION_URLS)).booleanValue();
<a name="359" href="#359">359</a> 
<a name="360" href="#360">360</a>         <strong>if</strong> (ignoreFormActions)
<a name="361" href="#361">361</a>             <strong>return</strong>;
<a name="362" href="#362">362</a> 
<a name="363" href="#363">363</a>         numberOfFormsProcessed++;
<a name="364" href="#364">364</a> 
<a name="365" href="#365">365</a>         <em class="comment">// get all form fields</em>
<a name="366" href="#366">366</a>         FormFields formFields = element.findFormFields();
<a name="367" href="#367">367</a>         <strong>for</strong> (Iterator fieldsIter = <strong>for</strong>mFields.iterator(); fieldsIter.hasNext();) {
<a name="368" href="#368">368</a>             <em class="comment">// for each form field</em>
<a name="369" href="#369">369</a>             FormField formField = (FormField) fieldsIter.next();
<a name="370" href="#370">370</a> 
<a name="371" href="#371">371</a>             <em class="comment">// for each form control</em>
<a name="372" href="#372">372</a>             <strong>for</strong> (Iterator controlIter = <strong>for</strong>mField.getFormControls().iterator();
<a name="373" href="#373">373</a>                 controlIter.hasNext();) {
<a name="374" href="#374">374</a>                 FormControl formControl = (FormControl) controlIter.next();
<a name="375" href="#375">375</a> 
<a name="376" href="#376">376</a>                 <em class="comment">// get name of control element (and URLEncode it)</em>
<a name="377" href="#377">377</a>                 String controlName = formControl.getName();
<a name="378" href="#378">378</a> 
<a name="379" href="#379">379</a>                 <em class="comment">// retrieve list of values - submit needs special handling</em>
<a name="380" href="#380">380</a>                 Collection controlValues;
<a name="381" href="#381">381</a>                 <strong>if</strong> (!(formControl.getFormControlType() ==
<a name="382" href="#382">382</a>                         FormControlType.SUBMIT)) {
<a name="383" href="#383">383</a>                     controlValues = formControl.getValues();
<a name="384" href="#384">384</a>                 } <strong>else</strong> {
<a name="385" href="#385">385</a>                     controlValues = formControl.getPredefinedValues();
<a name="386" href="#386">386</a>                 }
<a name="387" href="#387">387</a> 
<a name="388" href="#388">388</a>                 <strong>if</strong> (controlValues.size() > 0) {
<a name="389" href="#389">389</a>                     <em class="comment">// for each value set</em>
<a name="390" href="#390">390</a>                     <strong>for</strong> (Iterator valueIter = controlValues.iterator();
<a name="391" href="#391">391</a>                             valueIter.hasNext();) {
<a name="392" href="#392">392</a>                         String value = (String) valueIter.next();
<a name="393" href="#393">393</a>                         queryURL += <span class="string">"&amp;"</span> + controlName + <span class="string">"="</span> + value;
<a name="394" href="#394">394</a>                     }
<a name="395" href="#395">395</a>                 } <strong>else</strong> {
<a name="396" href="#396">396</a>                     queryURL += <span class="string">"&amp;"</span> + controlName + <span class="string">"="</span>;
<a name="397" href="#397">397</a>                 }
<a name="398" href="#398">398</a>             }
<a name="399" href="#399">399</a>         }
<a name="400" href="#400">400</a> 
<a name="401" href="#401">401</a>         <em class="comment">// clean up url</em>
<a name="402" href="#402">402</a>         <strong>if</strong> (action == <strong>null</strong>) {
<a name="403" href="#403">403</a>             queryURL = queryURL.replaceFirst(<span class="string">"&amp;"</span>, <span class="string">"?"</span>);
<a name="404" href="#404">404</a>         } <strong>else</strong> {
<a name="405" href="#405">405</a>             <strong>if</strong> (!action.contains(<span class="string">"?"</span>))
<a name="406" href="#406">406</a>                 queryURL = queryURL.replaceFirst(<span class="string">"&amp;"</span>, <span class="string">"?"</span>);
<a name="407" href="#407">407</a>             queryURL = action + queryURL;
<a name="408" href="#408">408</a>         }
<a name="409" href="#409">409</a> 
<a name="410" href="#410">410</a>         CharSequence context = Link.elementContext(element.getName(),
<a name="411" href="#411">411</a>             <span class="string">"name="</span> + name);
<a name="412" href="#412">412</a>         processLink(curi, queryURL, context);
<a name="413" href="#413">413</a> 
<a name="414" href="#414">414</a>     }
<a name="415" href="#415">415</a> 
<a name="416" href="#416">416</a>     <em>/**<em>*</em></em>
<a name="417" href="#417">417</a> <em>     * Run extractor. This method is package visible to ease testing.</em>
<a name="418" href="#418">418</a> <em>     * </em>
<a name="419" href="#419">419</a> <em>     * @param curi</em>
<a name="420" href="#420">420</a> <em>     *            CrawlURI we're processing.</em>
<a name="421" href="#421">421</a> <em>     * @param cs</em>
<a name="422" href="#422">422</a> <em>     *            Sequence from underlying ReplayCharSequence.</em>
<a name="423" href="#423">423</a> <em>     */</em>
<a name="424" href="#424">424</a>     <strong>void</strong> extract(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi, CharSequence cs) {
<a name="425" href="#425">425</a>         Source source = <strong>new</strong> Source(cs);
<a name="426" href="#426">426</a>         List elements = source.findAllElements(StartTagType.NORMAL);
<a name="427" href="#427">427</a>         <strong>for</strong> (Iterator elementIter = elements.iterator();
<a name="428" href="#428">428</a>                 elementIter.hasNext();) {
<a name="429" href="#429">429</a>             Element element = (Element) elementIter.next();
<a name="430" href="#430">430</a>             String elementName = element.getName();
<a name="431" href="#431">431</a>             Attributes attributes;
<a name="432" href="#432">432</a>             <strong>if</strong> (elementName.equals(HTMLElementName.META)) {
<a name="433" href="#433">433</a>                 <strong>if</strong> (processMeta(curi, element)) {
<a name="434" href="#434">434</a>                     <em class="comment">// meta tag included NOFOLLOW; abort processing</em>
<a name="435" href="#435">435</a>                     <strong>break</strong>;
<a name="436" href="#436">436</a>                 }
<a name="437" href="#437">437</a>             } <strong>else</strong> <strong>if</strong> (elementName.equals(HTMLElementName.SCRIPT)) {
<a name="438" href="#438">438</a>                 processScript(curi, element);
<a name="439" href="#439">439</a>             } <strong>else</strong> <strong>if</strong> (elementName.equals(HTMLElementName.STYLE)) {
<a name="440" href="#440">440</a>                 processStyle(curi, element);
<a name="441" href="#441">441</a>             } <strong>else</strong> <strong>if</strong> (elementName.equals(HTMLElementName.FORM)) {
<a name="442" href="#442">442</a>                 processForm(curi, element);
<a name="443" href="#443">443</a>             } <strong>else</strong> <strong>if</strong> (!(attributes = element.getAttributes()).isEmpty()) {
<a name="444" href="#444">444</a>                 processGeneralTag(curi, element, attributes);
<a name="445" href="#445">445</a>             }
<a name="446" href="#446">446</a>         }
<a name="447" href="#447">447</a>     }
<a name="448" href="#448">448</a> 
<a name="449" href="#449">449</a>     <em class="comment">/*</em>
<a name="450" href="#450">450</a> <em class="comment">     * (non-Javadoc)</em>
<a name="451" href="#451">451</a> <em class="comment">     * </em>
<a name="452" href="#452">452</a> <em class="comment">     * @see org.archive.crawler.framework.Processor#report()</em>
<a name="453" href="#453">453</a> <em class="comment">     */</em>
<a name="454" href="#454">454</a>     <strong>public</strong> String report() {
<a name="455" href="#455">455</a>         StringBuffer ret = <strong>new</strong> StringBuffer();
<a name="456" href="#456">456</a>         ret.append(<span class="string">"Processor: org.archive.crawler.extractor.JerichoExtractorHTML\n"</span>);
<a name="457" href="#457">457</a>         ret.append(<span class="string">"  Function:          Link extraction on HTML documents\n"</span>);
<a name="458" href="#458">458</a>         ret.append(<span class="string">"  CrawlURIs handled: "</span> + <strong>this</strong>.numberOfCURIsHandled + <span class="string">"\n"</span>);
<a name="459" href="#459">459</a>         ret.append(<span class="string">"  Forms processed:   "</span> + <strong>this</strong>.numberOfFormsProcessed + <span class="string">"\n"</span>);
<a name="460" href="#460">460</a>         ret.append(<span class="string">"  Links extracted:   "</span> + <strong>this</strong>.numberOfLinksExtracted + <span class="string">"\n\n"</span>);
<a name="461" href="#461">461</a>         <strong>return</strong> ret.toString();
<a name="462" href="#462">462</a>     }
<a name="463" href="#463">463</a> }
</pre>
<hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/">Maven</a></div></body>
</html>