httpcontentdigest.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="112" href="#112">112</a>         addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_STRIP_REG_EXPR,
<a name="113" href="#113">113</a>                 <span class="string">"A regular expression that matches those portions of "</span> +
<a name="114" href="#114">114</a>                 <span class="string">"downloaded documents that need to be ignored when "</span> +
<a name="115" href="#115">115</a>                 <span class="string">"calculating the content digest. "</span> +
<a name="116" href="#116">116</a>                 <span class="string">"Segments matching this expression will be rewritten with "</span> +
<a name="117" href="#117">117</a>                 <span class="string">"the blank character for the content digest."</span>,
<a name="118" href="#118">118</a>                 DEFAULT_STRIP_REG_EXPR));
<a name="119" href="#119">119</a>         addElementToDefinition(<strong>new</strong> <a href="../../../../org/archive/crawler/settings/SimpleType.html">SimpleType</a>(ATTR_MAX_SIZE_BYTES,
<a name="120" href="#120">120</a>                 <span class="string">"Maximum size of of documents to recalculate the digest for."</span> +
<a name="121" href="#121">121</a>                 <span class="string">" Documents that exceed this value (bytes) will be ignored."</span> +
<a name="122" href="#122">122</a>                 <span class="string">" Defaults to 1048576 bytes, or 1 MB. "</span> +
<a name="123" href="#123">123</a>                 <span class="string">"-1 denotes unlimited size. A setting of 0 will effectively "</span> +
<a name="124" href="#124">124</a>                 <span class="string">"disable the processor."</span>,
<a name="125" href="#125">125</a>                 DEFAULT_MAX_SIZE_BYTES));
<a name="126" href="#126">126</a>     }
<a name="127" href="#127">127</a> 
<a name="128" href="#128">128</a>     <strong>protected</strong> <strong>void</strong> innerProcess(<a href="../../../../org/archive/crawler/datamodel/CrawlURI.html">CrawlURI</a> curi) throws InterruptedException {
<a name="129" href="#129">129</a>         <strong>if</strong> (!curi.isHttpTransaction()){
<a name="130" href="#130">130</a>             <em class="comment">// Only handles HTTP docsuments.</em>
<a name="131" href="#131">131</a>             <strong>return</strong>;
<a name="132" href="#132">132</a>         }
<a name="133" href="#133">133</a>         <strong>if</strong>(!TextUtils.matches(<span class="string">"^text.*$"</span>, curi.getContentType())){
<a name="134" href="#134">134</a>             <em class="comment">// Only handles text based documents.</em>
<a name="135" href="#135">135</a>             <strong>return</strong>;
<a name="136" href="#136">136</a>         }
<a name="137" href="#137">137</a>         <strong>long</strong> maxsize = DEFAULT_MAX_SIZE_BYTES.<strong>long</strong>Value(); 
<a name="138" href="#138">138</a>         <strong>try</strong> {
<a name="139" href="#139">139</a>             maxsize = ((Long)getAttribute(curi,ATTR_MAX_SIZE_BYTES)).longValue();
<a name="140" href="#140">140</a>         } <strong>catch</strong> (AttributeNotFoundException e) {
<a name="141" href="#141">141</a>             logger.severe(<span class="string">"Missing max-size-bytes attribute when processing "</span> + 
<a name="142" href="#142">142</a>                     curi.toString());
<a name="143" href="#143">143</a>         }
<a name="144" href="#144">144</a>         <strong>if</strong>(maxsize &lt; curi.getContentSize() &amp;&amp; maxsize > -1){
<a name="145" href="#145">145</a>             <em class="comment">// Document too big.</em>
<a name="146" href="#146">146</a>             <strong>return</strong>;
<a name="147" href="#147">147</a>         }
<a name="148" href="#148">148</a>         
<a name="149" href="#149">149</a>         <em class="comment">// Ok, if we got this far we need to calculate the content digest. </em>
<a name="150" href="#150">150</a>         <em class="comment">// Get the regexpr</em>
<a name="151" href="#151">151</a>         String regexpr = <span class="string">""</span>;
<a name="152" href="#152">152</a>         <strong>try</strong> {
<a name="153" href="#153">153</a>             regexpr = (String)getAttribute(curi,ATTR_STRIP_REG_EXPR);
<a name="154" href="#154">154</a>         } <strong>catch</strong> (AttributeNotFoundException e2) {
<a name="155" href="#155">155</a>             logger.severe(<span class="string">"Missing strip-reg-exp when processing "</span> +
<a name="156" href="#156">156</a>                     curi.toString());
<a name="157" href="#157">157</a>             <strong>return</strong>; <em class="comment">// Can't do anything without it.</em>
<a name="158" href="#158">158</a>         }
<a name="159" href="#159">159</a>         
<a name="160" href="#160">160</a>         <em class="comment">// Get a replay of the document character seq.</em>
<a name="161" href="#161">161</a>         <a href="../../../../org/archive/io/ReplayCharSequence.html">ReplayCharSequence</a> cs = <strong>null</strong>;
<a name="162" href="#162">162</a>         
<a name="163" href="#163">163</a>         <strong>try</strong> {
<a name="164" href="#164">164</a>            cs = curi.getHttpRecorder().getReplayCharSequence();
<a name="165" href="#165">165</a>         } <strong>catch</strong>(Exception e) {
<a name="166" href="#166">166</a>             curi.addLocalizedError(<strong>this</strong>.getName(), e,
<a name="167" href="#167">167</a>                 <span class="string">"Failed get of replay char sequence "</span> + curi.toString() +
<a name="168" href="#168">168</a>                     <span class="string">" "</span> + e.getMessage());
<a name="169" href="#169">169</a>             logger.warning(<span class="string">"Failed get of replay char sequence "</span> +
<a name="170" href="#170">170</a>                 curi.toString() + <span class="string">" "</span> + e.getMessage() + <span class="string">" "</span> +
<a name="171" href="#171">171</a>                 Thread.currentThread().getName());
<a name="172" href="#172">172</a>             <strong>return</strong>; <em class="comment">// Can't proceed if this happens.</em>
<a name="173" href="#173">173</a>         }
<a name="174" href="#174">174</a>         
<a name="175" href="#175">175</a>         <em class="comment">// Create a MessageDigest </em>
<a name="176" href="#176">176</a>         MessageDigest digest = <strong>null</strong>;
<a name="177" href="#177">177</a>         
<a name="178" href="#178">178</a>         <em class="comment">// We have a ReplayCharSequence open.  Wrap all in finally so we</em>
<a name="179" href="#179">179</a>         <em class="comment">// for sure close it before we leave.</em>
<a name="180" href="#180">180</a>         <strong>try</strong> {
<a name="181" href="#181">181</a>             <strong>try</strong> {
<a name="182" href="#182">182</a>                 digest = MessageDigest.getInstance(SHA1);
<a name="183" href="#183">183</a>             } <strong>catch</strong> (NoSuchAlgorithmException e1) {
<a name="184" href="#184">184</a>                 e1.printStackTrace();
<a name="185" href="#185">185</a>                 <strong>return</strong>;
<a name="186" href="#186">186</a>             }
<a name="187" href="#187">187</a> 
<a name="188" href="#188">188</a>             digest.reset();
<a name="189" href="#189">189</a> 
<a name="190" href="#190">190</a>             String s = <strong>null</strong>;
<a name="191" href="#191">191</a> 
<a name="192" href="#192">192</a>             <strong>if</strong> (regexpr.length() == 0) {
<a name="193" href="#193">193</a>                 s = cs.toString();
<a name="194" href="#194">194</a>             } <strong>else</strong> {
<a name="195" href="#195">195</a>                 <em class="comment">// Process the document</em>
<a name="196" href="#196">196</a>                 Matcher m = TextUtils.getMatcher(regexpr, cs);
<a name="197" href="#197">197</a>                 s = m.replaceAll(<span class="string">" "</span>);
<a name="198" href="#198">198</a>                 TextUtils.recycleMatcher(m);
<a name="199" href="#199">199</a>             }
<a name="200" href="#200">200</a>             digest.update(s.getBytes());
<a name="201" href="#201">201</a> 
<a name="202" href="#202">202</a>             <em class="comment">// Get the new digest value</em>
<a name="203" href="#203">203</a>             byte[] newDigestValue = digest.digest();
<a name="204" href="#204">204</a> 
<a name="205" href="#205">205</a>             <em class="comment">// Log if needed.</em>
<a name="206" href="#206">206</a>             <strong>if</strong> (logger.isLoggable(Level.FINEST)) {
<a name="207" href="#207">207</a>                 logger.finest(<span class="string">"Recalculated content digest for "</span>
<a name="208" href="#208">208</a>                         + curi.toString() + <span class="string">" old: "</span>
<a name="209" href="#209">209</a>                         + Base32.encode((byte[]) curi.getContentDigest())
<a name="210" href="#210">210</a>                         + <span class="string">", new: "</span> + Base32.encode(newDigestValue));
<a name="211" href="#211">211</a>             }
<a name="212" href="#212">212</a>             <em class="comment">// Save new digest value</em>
<a name="213" href="#213">213</a>             curi.setContentDigest(SHA1, newDigestValue);
<a name="214" href="#214">214</a>         } <strong>finally</strong> {
<a name="215" href="#215">215</a>             <strong>if</strong> (cs != <strong>null</strong>) {
<a name="216" href="#216">216</a>                 <strong>try</strong> {
<a name="217" href="#217">217</a>                     cs.close();
<a name="218" href="#218">218</a>                 } <strong>catch</strong> (IOException ioe) {
<a name="219" href="#219">219</a>                     logger.warning(TextUtils.exceptionToString(
<a name="220" href="#220">220</a>                             <span class="string">"Failed close of ReplayCharSequence."</span>, ioe));
<a name="221" href="#221">221</a>                 }
<a name="222" href="#222">222</a>             }
<a name="223" href="#223">223</a>         }
<a name="224" href="#224">224</a>     }
<a name="225" href="#225">225</a> }
</pre>
<hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/">Maven</a></div></body>
</html>