regexphtmllinkextractor.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="150" href="#150">150</a>     <em class="comment">// 3: ACTION - single URI relative to doc base, or occasionally javascript:</em>
<a name="151" href="#151">151</a>     <em class="comment">// 4: ON[WHATEVER] - script handler</em>
<a name="152" href="#152">152</a>     <em class="comment">// 5: SRC,LOWSRC,BACKGROUND,CITE,LONGDESC,USEMAP,PROFILE,DATASRC, or FOR</em>
<a name="153" href="#153">153</a>     <em class="comment">//    single URI relative to doc base</em>
<a name="154" href="#154">154</a>     <em class="comment">// 6: CODEBASE - a single URI relative to doc base, affecting other</em>
<a name="155" href="#155">155</a>     <em class="comment">//    attributes</em>
<a name="156" href="#156">156</a>     <em class="comment">// 7: CLASSID, DATA - a single URI relative to CODEBASE (if supplied)</em>
<a name="157" href="#157">157</a>     <em class="comment">// 8: ARCHIVE - one or more space-delimited URIs relative to CODEBASE</em>
<a name="158" href="#158">158</a>     <em class="comment">//    (if supplied)</em>
<a name="159" href="#159">159</a>     <em class="comment">// 9: CODE - a single URI relative to the CODEBASE (is specified).</em>
<a name="160" href="#160">160</a>     <em class="comment">// 10: VALUE - often includes a uri path on forms</em>
<a name="161" href="#161">161</a>     <em class="comment">// 11: any other attribute</em>
<a name="162" href="#162">162</a>     <em class="comment">// 12: double-quote delimited attr value</em>
<a name="163" href="#163">163</a>     <em class="comment">// 13: single-quote delimited attr value</em>
<a name="164" href="#164">164</a>     <em class="comment">// 14: space-delimited attr value</em>
<a name="165" href="#165">165</a> 
<a name="166" href="#166">166</a> 
<a name="167" href="#167">167</a>     <em class="comment">// much like the javascript likely-URI extractor, but</em>
<a name="168" href="#168">168</a>     <em class="comment">// without requiring quotes -- this can indicate whether</em>
<a name="169" href="#169">169</a>     <em class="comment">// an HTML tag attribute that isn't definitionally a</em>
<a name="170" href="#170">170</a>     <em class="comment">// URI might be one anyway, as in form-tag VALUE attributes</em>
<a name="171" href="#171">171</a>     <strong>static</strong> <strong>final</strong> String LIKELY_URI_PATH =
<a name="172" href="#172">172</a>      <span class="string">"(&#47;&#47;.{0,2}[^&#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;']+)+)"</span>;
<a name="173" href="#173">173</a>     <strong>static</strong> <strong>final</strong> String ESCAPED_AMP = <span class="string">"&amp;amp;"</span>;
<a name="174" href="#174">174</a>     <strong>static</strong> <strong>final</strong> String AMP =<span class="string">"&amp;"</span>;
<a name="175" href="#175">175</a>     <strong>static</strong> <strong>final</strong> String WHITESPACE = <span class="string">"&#47;&#47;s"</span>;
<a name="176" href="#176">176</a>     <strong>static</strong> <strong>final</strong> String CLASSEXT =<span class="string">".class"</span>;
<a name="177" href="#177">177</a>     <strong>static</strong> <strong>final</strong> String APPLET = <span class="string">"applet"</span>;
<a name="178" href="#178">178</a>     <strong>static</strong> <strong>final</strong> String BASE = <span class="string">"base"</span>;
<a name="179" href="#179">179</a>     <strong>static</strong> <strong>final</strong> String LINK = <span class="string">"link"</span>;
<a name="180" href="#180">180</a> 
<a name="181" href="#181">181</a>     <strong>protected</strong> <strong>boolean</strong> processGeneralTag(CharSequence element, CharSequence cs) {
<a name="182" href="#182">182</a> 
<a name="183" href="#183">183</a>         Matcher attr = TextUtils.getMatcher(EACH_ATTRIBUTE_EXTRACTOR, cs);
<a name="184" href="#184">184</a> 
<a name="185" href="#185">185</a>         <em class="comment">// Just in case it's an OBJECT or APPLET tag</em>
<a name="186" href="#186">186</a>         String codebase = <strong>null</strong>;
<a name="187" href="#187">187</a>         ArrayList resources = <strong>null</strong>;
<a name="188" href="#188">188</a>         <strong>long</strong> tally = next.size();
<a name="189" href="#189">189</a> 
<a name="190" href="#190">190</a>         <strong>while</strong> (attr.find()) {
<a name="191" href="#191">191</a>             <strong>int</strong> valueGroup =
<a name="192" href="#192">192</a>                 (attr.start(12) > -1) ? 12 : (attr.start(13) > -1) ? 13 : 14;
<a name="193" href="#193">193</a>             <strong>int</strong> start = attr.start(valueGroup);
<a name="194" href="#194">194</a>             <strong>int</strong> end = attr.end(valueGroup);
<a name="195" href="#195">195</a>             CharSequence value = cs.subSequence(start, end);
<a name="196" href="#196">196</a>             <strong>if</strong> (attr.start(2) > -1) {
<a name="197" href="#197">197</a>                 <em class="comment">// HREF</em>
<a name="198" href="#198">198</a>                 CharSequence context = Link.elementContext(element, attr.group(2));
<a name="199" href="#199">199</a>                 <strong>if</strong>(element.toString().equalsIgnoreCase(LINK)) {
<a name="200" href="#200">200</a>                     <em class="comment">// &lt;LINK> elements treated as embeds (css, ico, etc)</em>
<a name="201" href="#201">201</a>                     processEmbed(value, context);
<a name="202" href="#202">202</a>                 } <strong>else</strong> {
<a name="203" href="#203">203</a>                     <strong>if</strong> (element.toString().equalsIgnoreCase(BASE)) {
<a name="204" href="#204">204</a>                         <strong>try</strong> {
<a name="205" href="#205">205</a>                             base = UURIFactory.getInstance(value.toString());
<a name="206" href="#206">206</a>                         } <strong>catch</strong> (URIException e) {
<a name="207" href="#207">207</a>                             extractErrorListener.noteExtractError(e,source,value);
<a name="208" href="#208">208</a>                         }
<a name="209" href="#209">209</a>                     }
<a name="210" href="#210">210</a>                     <em class="comment">// other HREFs treated as links</em>
<a name="211" href="#211">211</a>                     processLink(value, context);
<a name="212" href="#212">212</a>                 }
<a name="213" href="#213">213</a>             } <strong>else</strong> <strong>if</strong> (attr.start(3) > -1) {
<a name="214" href="#214">214</a>                 <em class="comment">// ACTION</em>
<a name="215" href="#215">215</a>                 CharSequence context = Link.elementContext(element, attr.group(3));
<a name="216" href="#216">216</a>                 processLink(value, context);
<a name="217" href="#217">217</a>             } <strong>else</strong> <strong>if</strong> (attr.start(4) > -1) {
<a name="218" href="#218">218</a>                 <em class="comment">// ON____</em>
<a name="219" href="#219">219</a>                 processScriptCode(value); <em class="comment">// TODO: context?</em>
<a name="220" href="#220">220</a>             } <strong>else</strong> <strong>if</strong> (attr.start(5) > -1) {
<a name="221" href="#221">221</a>                 <em class="comment">// SRC etc.</em>
<a name="222" href="#222">222</a>                 CharSequence context = Link.elementContext(element, attr.group(5));
<a name="223" href="#223">223</a>                 processEmbed(value, context);
<a name="224" href="#224">224</a>             } <strong>else</strong> <strong>if</strong> (attr.start(6) > -1) {
<a name="225" href="#225">225</a>                 <em class="comment">// CODEBASE</em>
<a name="226" href="#226">226</a>                 <em class="comment">// TODO: more HTML deescaping?</em>
<a name="227" href="#227">227</a>                 codebase = TextUtils.replaceAll(ESCAPED_AMP, value, AMP);
<a name="228" href="#228">228</a>                 CharSequence context = Link.elementContext(element,attr.group(6));
<a name="229" href="#229">229</a>                 processEmbed(codebase, context);
<a name="230" href="#230">230</a>             } <strong>else</strong> <strong>if</strong> (attr.start(7) > -1) {
<a name="231" href="#231">231</a>                 <em class="comment">// CLASSID, DATA</em>
<a name="232" href="#232">232</a>                 <strong>if</strong> (resources == <strong>null</strong>) {
<a name="233" href="#233">233</a>                     resources = <strong>new</strong> ArrayList();
<a name="234" href="#234">234</a>                 }
<a name="235" href="#235">235</a>                 resources.add(value.toString());
<a name="236" href="#236">236</a>             } <strong>else</strong> <strong>if</strong> (attr.start(8) > -1) {
<a name="237" href="#237">237</a>                 <em class="comment">// ARCHIVE</em>
<a name="238" href="#238">238</a>                 <strong>if</strong> (resources==<strong>null</strong>) {
<a name="239" href="#239">239</a>                     resources = <strong>new</strong> ArrayList();
<a name="240" href="#240">240</a>                 }
<a name="241" href="#241">241</a>                 String[] multi = TextUtils.split(WHITESPACE, value);
<a name="242" href="#242">242</a>                 <strong>for</strong>(<strong>int</strong> i = 0; i &lt; multi.length; i++ ) {
<a name="243" href="#243">243</a>                     resources.add(multi[i]);
<a name="244" href="#244">244</a>                 }
<a name="245" href="#245">245</a>             } <strong>else</strong> <strong>if</strong> (attr.start(9) > -1) {
<a name="246" href="#246">246</a>                 <em class="comment">// CODE</em>
<a name="247" href="#247">247</a>                 <strong>if</strong> (resources==<strong>null</strong>) {
<a name="248" href="#248">248</a>                     resources = <strong>new</strong> ArrayList();
<a name="249" href="#249">249</a>                 }
<a name="250" href="#250">250</a>                 <em class="comment">// If element is applet and code value does not end with</em>
<a name="251" href="#251">251</a>                 <em class="comment">// '.class' then append '.class' to the code value.</em>
<a name="252" href="#252">252</a>                 <strong>if</strong> (element.toString().toLowerCase().equals(APPLET) &amp;&amp;
<a name="253" href="#253">253</a>                         !value.toString().toLowerCase().endsWith(CLASSEXT)) {
<a name="254" href="#254">254</a>                     resources.add(value.toString() + CLASSEXT);
<a name="255" href="#255">255</a>                 } <strong>else</strong> {
<a name="256" href="#256">256</a>                     resources.add(value.toString());
<a name="257" href="#257">257</a>                 }
<a name="258" href="#258">258</a> 
<a name="259" href="#259">259</a>             } <strong>else</strong> <strong>if</strong> (attr.start(10) > -1) {
<a name="260" href="#260">260</a>                 <em class="comment">// VALUE</em>
<a name="261" href="#261">261</a>                 <strong>if</strong>(TextUtils.matches(LIKELY_URI_PATH, value)) {
<a name="262" href="#262">262</a>                     CharSequence context = Link.elementContext(element, attr.group(10));
<a name="263" href="#263">263</a>                     processLink(value, context);
<a name="264" href="#264">264</a>                 }
<a name="265" href="#265">265</a> 
<a name="266" href="#266">266</a>             } <strong>else</strong> <strong>if</strong> (attr.start(11) > -1) {
<a name="267" href="#267">267</a>                 <em class="comment">// any other attribute</em>
<a name="268" href="#268">268</a>                 <em class="comment">// ignore for now</em>
<a name="269" href="#269">269</a>                 <em class="comment">// could probe for path- or script-looking strings, but</em>
<a name="270" href="#270">270</a>                 <em class="comment">// those should be vanishingly rare in other attributes,</em>
<a name="271" href="#271">271</a>                 <em class="comment">// and/or symptomatic of page bugs</em>
<a name="272" href="#272">272</a>             }
<a name="273" href="#273">273</a>         }
<a name="274" href="#274">274</a>         TextUtils.recycleMatcher(attr);
<a name="275" href="#275">275</a> 
<a name="276" href="#276">276</a>         <em class="comment">// handle codebase/resources</em>
<a name="277" href="#277">277</a>         <strong>if</strong> (resources == <strong>null</strong>) {
<a name="278" href="#278">278</a>             <strong>return</strong> (tally-next.size())>0;
<a name="279" href="#279">279</a>         }
<a name="280" href="#280">280</a>         Iterator iter = resources.iterator();
<a name="281" href="#281">281</a>         <a href="../../../org/archive/net/UURI.html">UURI</a> codebaseURI = <strong>null</strong>;
<a name="282" href="#282">282</a>         String res = <strong>null</strong>;
<a name="283" href="#283">283</a>         <strong>try</strong> {
<a name="284" href="#284">284</a>             <strong>if</strong> (codebase != <strong>null</strong>) {
<a name="285" href="#285">285</a>                 <em class="comment">// TODO: Pass in the charset.</em>
<a name="286" href="#286">286</a>                 codebaseURI = UURIFactory.getInstance(base, codebase);
<a name="287" href="#287">287</a>             }
<a name="288" href="#288">288</a>             <strong>while</strong>(iter.hasNext()) {
<a name="289" href="#289">289</a>                 res = iter.next().toString();
<a name="290" href="#290">290</a>                 <em class="comment">// TODO: more HTML deescaping?</em>
<a name="291" href="#291">291</a>                 res = TextUtils.replaceAll(ESCAPED_AMP, res, AMP);
<a name="292" href="#292">292</a>                 <strong>if</strong> (codebaseURI != <strong>null</strong>) {
<a name="293" href="#293">293</a>                     res = codebaseURI.resolve(res).toString();
<a name="294" href="#294">294</a>                 }
<a name="295" href="#295">295</a>                 processEmbed(res, element); <em class="comment">// TODO: include attribute too</em>
<a name="296" href="#296">296</a>             }
<a name="297" href="#297">297</a>         } <strong>catch</strong> (URIException e) {
<a name="298" href="#298">298</a>             extractErrorListener.noteExtractError(e,source,codebase);
<a name="299" href="#299">299</a>         } <strong>catch</strong> (IllegalArgumentException e) {
<a name="300" href="#300">300</a>             DevUtils.logger.log(Level.WARNING, <span class="string">"processGeneralTag()\n"</span> +
<a name="301" href="#301">301</a>                 <span class="string">"codebase="</span> + codebase + <span class="string">" res="</span> + res + <span class="string">"\n"</span> +
<a name="302" href="#302">302</a>                 DevUtils.extraInfo(), e);
<a name="303" href="#303">303</a>         }
<a name="304" href="#304">304</a>         <strong>return</strong> (tally-next.size())>0;
<a name="305" href="#305">305</a>     }
<a name="306" href="#306">306</a> 
<a name="307" href="#307">307</a>     <em>/**<em>*</em></em>