jerichoextractorhtml.html
来自「网络爬虫开源代码」· HTML 代码 · 共 477 行 · 第 1/3 页
HTML
477 行
<a name="151" href="#151">151</a> getController().logUriError(e, curi.getUURI(),<a name="152" href="#152">152</a> attrValue);<a name="153" href="#153">153</a> } <strong>else</strong> {<a name="154" href="#154">154</a> logger.info(<span class="string">"Failed set base uri: "</span> + curi + <span class="string">", "</span><a name="155" href="#155">155</a> + attrValue + <span class="string">": "</span> + e.getMessage());<a name="156" href="#156">156</a> }<a name="157" href="#157">157</a> }<a name="158" href="#158">158</a> }<a name="159" href="#159">159</a> }<a name="160" href="#160">160</a> <em class="comment">// ACTION</em><a name="161" href="#161">161</a> <strong>else</strong> <strong>if</strong> (((attr = attributes.get(<span class="string">"action"</span>)) != <strong>null</strong>) &&<a name="162" href="#162">162</a> ((attrValue = attr.getValue()) != <strong>null</strong>)) {<a name="163" href="#163">163</a> <strong>if</strong> (!ignoreFormActions) {<a name="164" href="#164">164</a> CharSequence context = Link.elementContext(elementName, attr<a name="165" href="#165">165</a> .getKey());<a name="166" href="#166">166</a> processLink(curi, attrValue, context);<a name="167" href="#167">167</a> }<a name="168" href="#168">168</a> }<a name="169" href="#169">169</a> <em class="comment">// ON_</em><a name="170" href="#170">170</a> <strong>else</strong> <strong>if</strong> ((attrList = findOnAttributes(attributes)).size() != 0) {<a name="171" href="#171">171</a> <strong>for</strong> (Iterator attrIter = attrList.iterator(); attrIter.hasNext();) {<a name="172" href="#172">172</a> attr = (Attribute) attrIter.next();<a name="173" href="#173">173</a> CharSequence valueSegment = attr.getValueSegment();<a name="174" href="#174">174</a> <strong>if</strong> (valueSegment != <strong>null</strong>)<a name="175" href="#175">175</a> processScriptCode(curi, valueSegment);<a name="176" href="#176">176</a> <a name="177" href="#177">177</a> }<a name="178" href="#178">178</a> }<a name="179" href="#179">179</a> <em class="comment">// SRC atc.</em><a name="180" href="#180">180</a> <strong>else</strong> <strong>if</strong> ((((attr = attributes.get(<span class="string">"src"</span>)) != <strong>null</strong>)<a name="181" href="#181">181</a> || ((attr = attributes.get(<span class="string">"lowsrc"</span>)) != <strong>null</strong>)<a name="182" href="#182">182</a> || ((attr = attributes.get(<span class="string">"background"</span>)) != <strong>null</strong>)<a name="183" href="#183">183</a> || ((attr = attributes.get(<span class="string">"cite"</span>)) != <strong>null</strong>)<a name="184" href="#184">184</a> || ((attr = attributes.get(<span class="string">"longdesc"</span>)) != <strong>null</strong>)<a name="185" href="#185">185</a> || ((attr = attributes.get(<span class="string">"usemap"</span>)) != <strong>null</strong>)<a name="186" href="#186">186</a> || ((attr = attributes.get(<span class="string">"profile"</span>)) != <strong>null</strong>)<a name="187" href="#187">187</a> || ((attr = attributes.get(<span class="string">"datasrc"</span>)) != <strong>null</strong>)) &&<a name="188" href="#188">188</a> ((attrValue = attr.getValue()) != <strong>null</strong>)) {<a name="189" href="#189">189</a> <a name="190" href="#190">190</a> <strong>final</strong> <strong>char</strong> hopType;<a name="191" href="#191">191</a> CharSequence context = Link.elementContext(elementName, attr<a name="192" href="#192">192</a> .getKey());<a name="193" href="#193">193</a> <a name="194" href="#194">194</a> <strong>if</strong> (!framesAsEmbeds<a name="195" href="#195">195</a> && (<span class="string">"frame"</span>.equals(elementName) || <span class="string">"iframe"</span><a name="196" href="#196">196</a> .equals(elementName)))<a name="197" href="#197">197</a> hopType = Link.NAVLINK_HOP;<a name="198" href="#198">198</a> <strong>else</strong><a name="199" href="#199">199</a> hopType = Link.EMBED_HOP;<a name="200" href="#200">200</a> <a name="201" href="#201">201</a> processEmbed(curi, attrValue, context, hopType);<a name="202" href="#202">202</a> }<a name="203" href="#203">203</a> <em class="comment">// CODEBASE</em><a name="204" href="#204">204</a> <strong>else</strong> <strong>if</strong> (((attr = attributes.get(<span class="string">"codebase"</span>)) != <strong>null</strong>) &&<a name="205" href="#205">205</a> ((attrValue = attr.getValue()) != <strong>null</strong>)) {<a name="206" href="#206">206</a> codebase = StringEscapeUtils.unescapeHtml(attrValue);<a name="207" href="#207">207</a> CharSequence context = Link.elementContext(elementName, attr<a name="208" href="#208">208</a> .getKey());<a name="209" href="#209">209</a> processEmbed(curi, codebase, context);<a name="210" href="#210">210</a> }<a name="211" href="#211">211</a> <em class="comment">// CLASSID DATA</em><a name="212" href="#212">212</a> <strong>else</strong> <strong>if</strong> ((((attr = attributes.get(<span class="string">"classid"</span>)) != <strong>null</strong>)<a name="213" href="#213">213</a> || ((attr = attributes.get(<span class="string">"data"</span>)) != <strong>null</strong>)) &&<a name="214" href="#214">214</a> ((attrValue = attr.getValue()) != <strong>null</strong>)) {<a name="215" href="#215">215</a> <strong>if</strong> (resources == <strong>null</strong>)<a name="216" href="#216">216</a> resources = <strong>new</strong> ArrayList<String>();<a name="217" href="#217">217</a> resources.add(attrValue);<a name="218" href="#218">218</a> }<a name="219" href="#219">219</a> <em class="comment">// ARCHIVE</em><a name="220" href="#220">220</a> <strong>else</strong> <strong>if</strong> (((attr = attributes.get(<span class="string">"archive"</span>)) != <strong>null</strong>) &&<a name="221" href="#221">221</a> ((attrValue = attr.getValue()) != <strong>null</strong>)) {<a name="222" href="#222">222</a> <strong>if</strong> (resources == <strong>null</strong>)<a name="223" href="#223">223</a> resources = <strong>new</strong> ArrayList<String>();<a name="224" href="#224">224</a> String[] multi = TextUtils.split(WHITESPACE, attrValue);<a name="225" href="#225">225</a> <strong>for</strong> (<strong>int</strong> i = 0; i < multi.length; i++) {<a name="226" href="#226">226</a> resources.add(multi[i]);<a name="227" href="#227">227</a> }<a name="228" href="#228">228</a> }<a name="229" href="#229">229</a> <em class="comment">// CODE</em><a name="230" href="#230">230</a> <strong>else</strong> <strong>if</strong> (((attr = attributes.get(<span class="string">"code"</span>)) != <strong>null</strong>) &&<a name="231" href="#231">231</a> ((attrValue = attr.getValue()) != <strong>null</strong>)) {<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<String>();<a name="234" href="#234">234</a> <em class="comment">// If element is applet and code value does not end with</em><a name="235" href="#235">235</a> <em class="comment">// '.class' then append '.class' to the code value.</em><a name="236" href="#236">236</a> <strong>if</strong> (APPLET.equals(elementName) && !attrValue.endsWith(CLASSEXT)) {<a name="237" href="#237">237</a> resources.add(attrValue + CLASSEXT);<a name="238" href="#238">238</a> } <strong>else</strong> {<a name="239" href="#239">239</a> resources.add(attrValue);<a name="240" href="#240">240</a> }<a name="241" href="#241">241</a> }<a name="242" href="#242">242</a> <em class="comment">// VALUE</em><a name="243" href="#243">243</a> <strong>else</strong> <strong>if</strong> (((attr = attributes.get(<span class="string">"value"</span>)) != <strong>null</strong>) &&<a name="244" href="#244">244</a> ((attrValue = attr.getValue()) != <strong>null</strong>)) {<a name="245" href="#245">245</a> <strong>if</strong> (TextUtils.matches(LIKELY_URI_PATH, attrValue)<a name="246" href="#246">246</a> && overlyEagerLinkDetection) {<a name="247" href="#247">247</a> CharSequence context = Link.elementContext(elementName, attr<a name="248" href="#248">248</a> .getKey());<a name="249" href="#249">249</a> processLink(curi, attrValue, context);<a name="250" href="#250">250</a> }<a name="251" href="#251">251</a> <a name="252" href="#252">252</a> }<a name="253" href="#253">253</a> <em class="comment">// STYLE</em><a name="254" href="#254">254</a> <strong>else</strong> <strong>if</strong> (((attr = attributes.get(<span class="string">"style"</span>)) != <strong>null</strong>) &&<a name="255" href="#255">255</a> ((attrValue = attr.getValue()) != <strong>null</strong>)) {<a name="256" href="#256">256</a> <em class="comment">// STYLE inline attribute</em><a name="257" href="#257">257</a> <em class="comment">// then, parse for URIs</em><a name="258" href="#258">258</a> <strong>this</strong>.numberOfLinksExtracted += ExtractorCSS.processStyleCode(curi,<a name="259" href="#259">259</a> attrValue, getController());<a name="260" href="#260">260</a> }<a name="261" href="#261">261</a> <a name="262" href="#262">262</a> <em class="comment">// handle codebase/resources</em><a name="263" href="#263">263</a> <strong>if</strong> (resources == <strong>null</strong>)<a name="264" href="#264">264</a> <strong>return</strong>;<a name="265" href="#265">265</a> <a name="266" href="#266">266</a> Iterator<String> iter = resources.iterator();<a name="267" href="#267">267</a> <a href="../../../../org/archive/net/UURI.html">UURI</a> codebaseURI = <strong>null</strong>;<a name="268" href="#268">268</a> String res = <strong>null</strong>;<a name="269" href="#269">269</a> <strong>try</strong> {<a name="270" href="#270">270</a> <strong>if</strong> (codebase != <strong>null</strong>) {<a name="271" href="#271">271</a> <em class="comment">// TODO: Pass in the charset.</em><a name="272" href="#272">272</a> codebaseURI = UURIFactory.getInstance(curi.getUURI(), codebase);<a name="273" href="#273">273</a> }<a name="274" href="#274">274</a> <strong>while</strong> (iter.hasNext()) {<a name="275" href="#275">275</a> res = iter.next();<a name="276" href="#276">276</a> res = StringEscapeUtils.unescapeHtml(res);<a name="277" href="#277">277</a> <strong>if</strong> (codebaseURI != <strong>null</strong>) {<a name="278" href="#278">278</a> res = codebaseURI.resolve(res).toString();<a name="279" href="#279">279</a> }<a name="280" href="#280">280</a> processEmbed(curi, res, element); <em class="comment">// TODO: include attribute</em><a name="281" href="#281">281</a> <em class="comment">// too</em><a name="282" href="#282">282</a> }<a name="283" href="#283">283</a> } <strong>catch</strong> (URIException e) {<a name="284" href="#284">284</a> curi.addLocalizedError(getName(), e, <span class="string">"BAD CODEBASE "</span> + codebase);<a name="285" href="#285">285</a> } <strong>catch</strong> (IllegalArgumentException e) {<a name="286" href="#286">286</a> DevUtils.logger.log(Level.WARNING, <span class="string">"processGeneralTag()\n"</span><a name="287" href="#287">287</a> + <span class="string">"codebase="</span> + codebase + <span class="string">" res="</span> + res + <span class="string">"\n"</span><a name="288" href="#288">288</a> + DevUtils.extraInfo(), e);<a name="289" href="#289">289</a> }<a name="290" href="#290">290</a> }<a name="291" href="#291">291</a> <a name="292" href="#292">292</a> <strong>protected</strong> <strong>boolean</strong> processMeta(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi, Element element) {<a name="293" href="#293">293</a> String name = element.getAttributeValue(<span class="string">"name"</span>);<a name="294" href="#294">294</a> String httpEquiv = element.getAttributeValue(<span class="string">"http-equiv"</span>);<a name="295" href="#295">295</a> String content = element.getAttributeValue(<span class="string">"content"</span>);<a name="296" href="#296">296</a> <a name="297" href="#297">297</a> <strong>if</strong> (<span class="string">"robots"</span>.equals(name) && content != <strong>null</strong>) {<a name="298" href="#298">298</a> curi.putString(A_META_ROBOTS, content);<a name="299" href="#299">299</a> <a href="../../../../org/archive/crawler/datamodel/RobotsHonoringPolicy.html">RobotsHonoringPolicy</a> policy = getSettingsHandler().getOrder()<a name="300" href="#300">300</a> .getRobotsHonoringPolicy();<a name="301" href="#301">301</a> String contentLower = content.toLowerCase();<a name="302" href="#302">302</a> <strong>if</strong> ((policy == <strong>null</strong> || (!policy.isType(curi,<a name="303" href="#303">303</a> RobotsHonoringPolicy.IGNORE) && !policy.isType(curi,<a name="304" href="#304">304</a> RobotsHonoringPolicy.CUSTOM)))<a name="305" href="#305">305</a> && (contentLower.indexOf(<span class="string">"nofollow"</span>) >= 0 || contentLower<a name="306" href="#306">306</a> .indexOf(<span class="string">"none"</span>) >= 0)) {<a name="307" href="#307">307</a> <em class="comment">// if 'nofollow' or 'none' is specified and the</em><a name="308" href="#308">308</a> <em class="comment">// honoring policy is not IGNORE or CUSTOM, end html extraction</em><a name="309" href="#309">309</a> logger.fine(<span class="string">"HTML extraction skipped due to robots meta-tag "</span> +
⌨️ 快捷键说明
复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?