Modified: incubator/tika/site/xref/org/apache/tika/parser/xml/XMLParser.html URL: http://svn.apache.org/viewvc/incubator/tika/site/xref/org/apache/tika/parser/xml/XMLParser.html?rev=609491&r1=609490&r2=609491&view=diff ============================================================================== --- incubator/tika/site/xref/org/apache/tika/parser/xml/XMLParser.html (original) +++ incubator/tika/site/xref/org/apache/tika/parser/xml/XMLParser.html Sun Jan 6 19:06:26 2008 @@ -35,186 +35,187 @@ <a name="25" href="#25">25</a> <strong class="jxr_keyword">import</strong> org.apache.tika.exception.TikaException; <a name="26" href="#26">26</a> <strong class="jxr_keyword">import</strong> org.apache.tika.metadata.Metadata; <a name="27" href="#27">27</a> <strong class="jxr_keyword">import</strong> org.apache.tika.parser.Parser; -<a name="28" href="#28">28</a> <strong class="jxr_keyword">import</strong> org.apache.tika.sax.XHTMLContentHandler; -<a name="29" href="#29">29</a> <strong class="jxr_keyword">import</strong> org.apache.tika.utils.Utils; -<a name="30" href="#30">30</a> -<a name="31" href="#31">31</a> <strong class="jxr_keyword">import</strong> org.apache.commons.lang.StringUtils; -<a name="32" href="#32">32</a> <strong class="jxr_keyword">import</strong> org.apache.log4j.Logger; -<a name="33" href="#33">33</a> <strong class="jxr_keyword">import</strong> org.jaxen.JaxenException; -<a name="34" href="#34">34</a> <strong class="jxr_keyword">import</strong> org.jaxen.SimpleNamespaceContext; -<a name="35" href="#35">35</a> <strong class="jxr_keyword">import</strong> org.jaxen.jdom.JDOMXPath; -<a name="36" href="#36">36</a> <strong class="jxr_keyword">import</strong> org.jdom.Attribute; -<a name="37" href="#37">37</a> <strong class="jxr_keyword">import</strong> org.jdom.Comment; -<a name="38" href="#38">38</a> <strong class="jxr_keyword">import</strong> org.jdom.Document; -<a name="39" href="#39">39</a> <strong class="jxr_keyword">import</strong> org.jdom.Element; -<a name="40" href="#40">40</a> <strong class="jxr_keyword">import</strong> org.jdom.EntityRef; -<a name="41" href="#41">41</a> <strong class="jxr_keyword">import</strong> org.jdom.Namespace; -<a name="42" href="#42">42</a> <strong class="jxr_keyword">import</strong> org.jdom.ProcessingInstruction; -<a name="43" href="#43">43</a> <strong class="jxr_keyword">import</strong> org.jdom.Text; -<a name="44" href="#44">44</a> <strong class="jxr_keyword">import</strong> org.xml.sax.ContentHandler; -<a name="45" href="#45">45</a> <strong class="jxr_keyword">import</strong> org.xml.sax.SAXException; -<a name="46" href="#46">46</a> -<a name="47" href="#47">47</a> <em class="jxr_javadoccomment">/**</em> -<a name="48" href="#48">48</a> <em class="jxr_javadoccomment"> * XML parser</em> -<a name="49" href="#49">49</a> <em class="jxr_javadoccomment"> */</em> -<a name="50" href="#50">50</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">class</strong> <a href="../../../../../org/apache/tika/parser/xml/XMLParser.html">XMLParser</a> implements <a href="../../../../../org/apache/tika/parser/Parser.html">Parser</a> { -<a name="51" href="#51">51</a> -<a name="52" href="#52">52</a> <strong class="jxr_keyword">static</strong> Logger logger = Logger.getRootLogger(); -<a name="53" href="#53">53</a> -<a name="54" href="#54">54</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">void</strong> parse( -<a name="55" href="#55">55</a> InputStream stream, ContentHandler handler, <a href="../../../../../org/apache/tika/metadata/Metadata.html">Metadata</a> metadata) -<a name="56" href="#56">56</a> <strong class="jxr_keyword">throws</strong> IOException, SAXException, <a href="../../../../../org/apache/tika/exception/TikaException.html">TikaException</a> { -<a name="57" href="#57">57</a> Document xmlDoc = Utils.parse(stream); -<a name="58" href="#58">58</a> -<a name="59" href="#59">59</a> extractContent(xmlDoc, Metadata.TITLE, <span class="jxr_string">"//dc:title"</span>, metadata); -<a name="60" href="#60">60</a> extractContent(xmlDoc, Metadata.SUBJECT, <span class="jxr_string">"//dc:subject"</span>, metadata); -<a name="61" href="#61">61</a> extractContent(xmlDoc, Metadata.CREATOR, <span class="jxr_string">"//dc:creator"</span>, metadata); -<a name="62" href="#62">62</a> extractContent(xmlDoc, Metadata.DESCRIPTION, <span class="jxr_string">"//dc:description"</span>, metadata); -<a name="63" href="#63">63</a> extractContent(xmlDoc, Metadata.PUBLISHER, <span class="jxr_string">"//dc:publisher"</span>, metadata); -<a name="64" href="#64">64</a> extractContent(xmlDoc, Metadata.CONTRIBUTOR, <span class="jxr_string">"//dc:contributor"</span>, metadata); -<a name="65" href="#65">65</a> extractContent(xmlDoc, Metadata.TYPE, <span class="jxr_string">"//dc:type"</span>, metadata); -<a name="66" href="#66">66</a> extractContent(xmlDoc, Metadata.FORMAT, <span class="jxr_string">"//dc:format"</span>, metadata); -<a name="67" href="#67">67</a> extractContent(xmlDoc, Metadata.IDENTIFIER, <span class="jxr_string">"//dc:identifier"</span>, metadata); -<a name="68" href="#68">68</a> extractContent(xmlDoc, Metadata.LANGUAGE, <span class="jxr_string">"//dc:language"</span>, metadata); -<a name="69" href="#69">69</a> extractContent(xmlDoc, Metadata.RIGHTS, <span class="jxr_string">"//dc:rights"</span>, metadata); -<a name="70" href="#70">70</a> -<a name="71" href="#71">71</a> <a href="../../../../../org/apache/tika/sax/XHTMLContentHandler.html">XHTMLContentHandler</a> xhtml = <strong class="jxr_keyword">new</strong> <a href="../../../../../org/apache/tika/sax/XHTMLContentHandler.html">XHTMLContentHandler</a>(handler, metadata); -<a name="72" href="#72">72</a> xhtml.startDocument(); -<a name="73" href="#73">73</a> xhtml.element(<span class="jxr_string">"p"</span>, concatOccurrence(xmlDoc, <span class="jxr_string">"//*"</span>, <span class="jxr_string">" "</span>)); -<a name="74" href="#74">74</a> xhtml.endDocument(); -<a name="75" href="#75">75</a> } -<a name="76" href="#76">76</a> -<a name="77" href="#77">77</a> <strong class="jxr_keyword">public</strong> String concatOccurrence(Object xmlDoc, String xpath, String concatSep) { -<a name="78" href="#78">78</a> -<a name="79" href="#79">79</a> StringBuilder chaineConcat = <strong class="jxr_keyword">new</strong> StringBuilder(); -<a name="80" href="#80">80</a> <strong class="jxr_keyword">try</strong> { -<a name="81" href="#81">81</a> JDOMXPath xp = <strong class="jxr_keyword">new</strong> JDOMXPath(xpath); -<a name="82" href="#82">82</a> List ls = xp.selectNodes(xmlDoc); -<a name="83" href="#83">83</a> Iterator i = ls.iterator(); -<a name="84" href="#84">84</a> <strong class="jxr_keyword">int</strong> j = 0; -<a name="85" href="#85">85</a> <strong class="jxr_keyword">while</strong> (i.hasNext()) { -<a name="86" href="#86">86</a> j++; -<a name="87" href="#87">87</a> String text = <span class="jxr_string">""</span>; -<a name="88" href="#88">88</a> Object obj = i.next(); -<a name="89" href="#89">89</a> <strong class="jxr_keyword">if</strong> (obj instanceof Element) { -<a name="90" href="#90">90</a> Element elem = (Element) obj; -<a name="91" href="#91">91</a> text = elem.getText().trim(); -<a name="92" href="#92">92</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (obj instanceof Attribute) { -<a name="93" href="#93">93</a> Attribute att = (Attribute) obj; -<a name="94" href="#94">94</a> text = att.getValue().trim(); -<a name="95" href="#95">95</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (obj instanceof Text) { -<a name="96" href="#96">96</a> Text txt = (Text) obj; -<a name="97" href="#97">97</a> text = txt.getText().trim(); -<a name="98" href="#98">98</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (obj instanceof Comment) { -<a name="99" href="#99">99</a> Comment com = (Comment) obj; -<a name="100" href="#100">100</a> text = com.getText().trim(); -<a name="101" href="#101">101</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (obj instanceof ProcessingInstruction) { -<a name="102" href="#102">102</a> ProcessingInstruction pi = (ProcessingInstruction) obj; -<a name="103" href="#103">103</a> text = pi.getData().trim(); -<a name="104" href="#104">104</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (obj instanceof EntityRef) { -<a name="105" href="#105">105</a> EntityRef er = (EntityRef) obj; -<a name="106" href="#106">106</a> text = er.toString().trim(); -<a name="107" href="#107">107</a> } -<a name="108" href="#108">108</a> <strong class="jxr_keyword">if</strong> (StringUtils.isNotEmpty(text)) { -<a name="109" href="#109">109</a> chaineConcat.append(text); -<a name="110" href="#110">110</a> <strong class="jxr_keyword">if</strong> (ls.size() == 1) { -<a name="111" href="#111">111</a> <strong class="jxr_keyword">return</strong> chaineConcat.toString().trim(); -<a name="112" href="#112">112</a> } <strong class="jxr_keyword">else</strong> { -<a name="113" href="#113">113</a> <strong class="jxr_keyword">if</strong> (ls.size() != j) { -<a name="114" href="#114">114</a> chaineConcat.append(' ') -<a name="115" href="#115">115</a> .append(concatSep) -<a name="116" href="#116">116</a> .append(' '); -<a name="117" href="#117">117</a> } -<a name="118" href="#118">118</a> } -<a name="119" href="#119">119</a> } -<a name="120" href="#120">120</a> } -<a name="121" href="#121">121</a> } <strong class="jxr_keyword">catch</strong> (JaxenException j) { -<a name="122" href="#122">122</a> logger.error(j.getMessage()); -<a name="123" href="#123">123</a> } -<a name="124" href="#124">124</a> <strong class="jxr_keyword">return</strong> chaineConcat.toString().trim(); -<a name="125" href="#125">125</a> } -<a name="126" href="#126">126</a> -<a name="127" href="#127">127</a> <strong class="jxr_keyword">public</strong> List getAllDocumentNs(org.jdom.Document doc) { -<a name="128" href="#128">128</a> List ls = <strong class="jxr_keyword">new</strong> ArrayList(); -<a name="129" href="#129">129</a> processChildren(doc.getRootElement(), ls); -<a name="130" href="#130">130</a> <strong class="jxr_keyword">return</strong> ls; -<a name="131" href="#131">131</a> } -<a name="132" href="#132">132</a> -<a name="133" href="#133">133</a> <strong class="jxr_keyword">private</strong> <strong class="jxr_keyword">boolean</strong> exist(List nsLs, String nsUri) { -<a name="134" href="#134">134</a> <strong class="jxr_keyword">if</strong> (nsLs.isEmpty()) -<a name="135" href="#135">135</a> <strong class="jxr_keyword">return</strong> false; -<a name="136" href="#136">136</a> <strong class="jxr_keyword">for</strong> (Object nsL : nsLs) { -<a name="137" href="#137">137</a> <strong class="jxr_keyword">if</strong> (nsL.equals(nsUri)) { -<a name="138" href="#138">138</a> <strong class="jxr_keyword">return</strong> <strong class="jxr_keyword">true</strong>; -<a name="139" href="#139">139</a> } -<a name="140" href="#140">140</a> } -<a name="141" href="#141">141</a> <strong class="jxr_keyword">return</strong> false; -<a name="142" href="#142">142</a> } -<a name="143" href="#143">143</a> -<a name="144" href="#144">144</a> <strong class="jxr_keyword">private</strong> <strong class="jxr_keyword">void</strong> processChildren(Element elem, List ns) { -<a name="145" href="#145">145</a> Namespace nsCourent = elem.getNamespace(); -<a name="146" href="#146">146</a> String nsUri = (nsCourent.getURI()); -<a name="147" href="#147">147</a> <strong class="jxr_keyword">if</strong> (!exist(ns, nsUri)) { -<a name="148" href="#148">148</a> ns.add(nsUri.trim()); -<a name="149" href="#149">149</a> } -<a name="150" href="#150">150</a> List additionalNs = elem.getAdditionalNamespaces(); -<a name="151" href="#151">151</a> <strong class="jxr_keyword">if</strong> (!additionalNs.isEmpty()) -<a name="152" href="#152">152</a> copyNsList(additionalNs, ns); -<a name="153" href="#153">153</a> <strong class="jxr_keyword">if</strong> (elem.getChildren().size() > 0) { -<a name="154" href="#154">154</a> List elemChildren = elem.getChildren(); -<a name="155" href="#155">155</a> <strong class="jxr_keyword">for</strong> (Object anElemChildren : elemChildren) { -<a name="156" href="#156">156</a> processChildren((Element) anElemChildren, ns); -<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> -<a name="161" href="#161">161</a> <strong class="jxr_keyword">private</strong> <strong class="jxr_keyword">void</strong> copyNsList(List nsElem, List nsRes) { -<a name="162" href="#162">162</a> <strong class="jxr_keyword">for</strong> (Object aNsElem : nsElem) { -<a name="163" href="#163">163</a> nsRes.add(((Namespace) aNsElem).getURI().trim()); -<a name="164" href="#164">164</a> } -<a name="165" href="#165">165</a> } -<a name="166" href="#166">166</a> -<a name="167" href="#167">167</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">void</strong> extractContent( -<a name="168" href="#168">168</a> Document xmlDoc, String name, String xpath, <a href="../../../../../org/apache/tika/metadata/Metadata.html">Metadata</a> metadata) { -<a name="169" href="#169">169</a> <strong class="jxr_keyword">try</strong> { -<a name="170" href="#170">170</a> JDOMXPath xp = <strong class="jxr_keyword">new</strong> JDOMXPath(xpath); -<a name="171" href="#171">171</a> SimpleNamespaceContext context = <strong class="jxr_keyword">new</strong> SimpleNamespaceContext(); -<a name="172" href="#172">172</a> context.addNamespace(<span class="jxr_string">"dc"</span>, <span class="jxr_string">"http://purl.org/dc/elements/1.1/";</span>); -<a name="173" href="#173">173</a> context.addNamespace(<span class="jxr_string">"meta"</span>, <span class="jxr_string">"urn:oasis:names:tc:opendocument:xmlns:meta:1.0"</span>); -<a name="174" href="#174">174</a> xp.setNamespaceContext(context); -<a name="175" href="#175">175</a> List selectNodes = xp.selectNodes(xmlDoc); -<a name="176" href="#176">176</a> Iterator nodes = selectNodes.iterator(); -<a name="177" href="#177">177</a> <strong class="jxr_keyword">while</strong> (nodes.hasNext()) { -<a name="178" href="#178">178</a> Object node = nodes.next(); -<a name="179" href="#179">179</a> <strong class="jxr_keyword">if</strong> (node instanceof Element) { -<a name="180" href="#180">180</a> Element elem = (Element) node; -<a name="181" href="#181">181</a> <strong class="jxr_keyword">if</strong> (StringUtils.isNotBlank(elem.getText())) { -<a name="182" href="#182">182</a> metadata.add(name, elem.getText().trim()); -<a name="183" href="#183">183</a> } -<a name="184" href="#184">184</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (node instanceof Attribute) { -<a name="185" href="#185">185</a> Attribute att = (Attribute) node; -<a name="186" href="#186">186</a> metadata.add(name, att.getValue()); -<a name="187" href="#187">187</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (node instanceof Text) { -<a name="188" href="#188">188</a> Text text = (Text) node; -<a name="189" href="#189">189</a> metadata.add(name, text.getText()); -<a name="190" href="#190">190</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (node instanceof Comment) { -<a name="191" href="#191">191</a> Comment com = (Comment) node; -<a name="192" href="#192">192</a> metadata.add(name, com.getText()); -<a name="193" href="#193">193</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (node instanceof ProcessingInstruction) { -<a name="194" href="#194">194</a> ProcessingInstruction pi = (ProcessingInstruction) node; -<a name="195" href="#195">195</a> metadata.add(name, pi.getData()); -<a name="196" href="#196">196</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (node instanceof EntityRef) { -<a name="197" href="#197">197</a> EntityRef er = (EntityRef) node; -<a name="198" href="#198">198</a> metadata.add(name, er.toString()); -<a name="199" href="#199">199</a> } -<a name="200" href="#200">200</a> } -<a name="201" href="#201">201</a> } <strong class="jxr_keyword">catch</strong> (JaxenException e) { -<a name="202" href="#202">202</a> logger.error(e.getMessage()); -<a name="203" href="#203">203</a> } -<a name="204" href="#204">204</a> -<a name="205" href="#205">205</a> } -<a name="206" href="#206">206</a> -<a name="207" href="#207">207</a> } +<a name="28" href="#28">28</a> <strong class="jxr_keyword">import</strong> org.apache.tika.sax.AppendableAdaptor; +<a name="29" href="#29">29</a> <strong class="jxr_keyword">import</strong> org.apache.tika.sax.XHTMLContentHandler; +<a name="30" href="#30">30</a> <strong class="jxr_keyword">import</strong> org.apache.tika.utils.Utils; +<a name="31" href="#31">31</a> +<a name="32" href="#32">32</a> <strong class="jxr_keyword">import</strong> org.apache.commons.lang.StringUtils; +<a name="33" href="#33">33</a> <strong class="jxr_keyword">import</strong> org.apache.log4j.Logger; +<a name="34" href="#34">34</a> <strong class="jxr_keyword">import</strong> org.jaxen.JaxenException; +<a name="35" href="#35">35</a> <strong class="jxr_keyword">import</strong> org.jaxen.SimpleNamespaceContext; +<a name="36" href="#36">36</a> <strong class="jxr_keyword">import</strong> org.jaxen.jdom.JDOMXPath; +<a name="37" href="#37">37</a> <strong class="jxr_keyword">import</strong> org.jdom.Attribute; +<a name="38" href="#38">38</a> <strong class="jxr_keyword">import</strong> org.jdom.Comment; +<a name="39" href="#39">39</a> <strong class="jxr_keyword">import</strong> org.jdom.Document; +<a name="40" href="#40">40</a> <strong class="jxr_keyword">import</strong> org.jdom.Element; +<a name="41" href="#41">41</a> <strong class="jxr_keyword">import</strong> org.jdom.EntityRef; +<a name="42" href="#42">42</a> <strong class="jxr_keyword">import</strong> org.jdom.Namespace; +<a name="43" href="#43">43</a> <strong class="jxr_keyword">import</strong> org.jdom.ProcessingInstruction; +<a name="44" href="#44">44</a> <strong class="jxr_keyword">import</strong> org.jdom.Text; +<a name="45" href="#45">45</a> <strong class="jxr_keyword">import</strong> org.xml.sax.ContentHandler; +<a name="46" href="#46">46</a> <strong class="jxr_keyword">import</strong> org.xml.sax.SAXException; +<a name="47" href="#47">47</a> +<a name="48" href="#48">48</a> <em class="jxr_javadoccomment">/**</em> +<a name="49" href="#49">49</a> <em class="jxr_javadoccomment"> * XML parser</em> +<a name="50" href="#50">50</a> <em class="jxr_javadoccomment"> */</em> +<a name="51" href="#51">51</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">class</strong> <a href="../../../../../org/apache/tika/parser/xml/XMLParser.html">XMLParser</a> implements <a href="../../../../../org/apache/tika/parser/Parser.html">Parser</a> { +<a name="52" href="#52">52</a> +<a name="53" href="#53">53</a> <strong class="jxr_keyword">static</strong> Logger logger = Logger.getRootLogger(); +<a name="54" href="#54">54</a> +<a name="55" href="#55">55</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">void</strong> parse( +<a name="56" href="#56">56</a> InputStream stream, ContentHandler handler, <a href="../../../../../org/apache/tika/metadata/Metadata.html">Metadata</a> metadata) +<a name="57" href="#57">57</a> <strong class="jxr_keyword">throws</strong> IOException, SAXException, <a href="../../../../../org/apache/tika/exception/TikaException.html">TikaException</a> { +<a name="58" href="#58">58</a> Document xmlDoc = Utils.parse(stream); +<a name="59" href="#59">59</a> +<a name="60" href="#60">60</a> extractContent(xmlDoc, Metadata.TITLE, <span class="jxr_string">"//dc:title"</span>, metadata); +<a name="61" href="#61">61</a> extractContent(xmlDoc, Metadata.SUBJECT, <span class="jxr_string">"//dc:subject"</span>, metadata); +<a name="62" href="#62">62</a> extractContent(xmlDoc, Metadata.CREATOR, <span class="jxr_string">"//dc:creator"</span>, metadata); +<a name="63" href="#63">63</a> extractContent(xmlDoc, Metadata.DESCRIPTION, <span class="jxr_string">"//dc:description"</span>, metadata); +<a name="64" href="#64">64</a> extractContent(xmlDoc, Metadata.PUBLISHER, <span class="jxr_string">"//dc:publisher"</span>, metadata); +<a name="65" href="#65">65</a> extractContent(xmlDoc, Metadata.CONTRIBUTOR, <span class="jxr_string">"//dc:contributor"</span>, metadata); +<a name="66" href="#66">66</a> extractContent(xmlDoc, Metadata.TYPE, <span class="jxr_string">"//dc:type"</span>, metadata); +<a name="67" href="#67">67</a> extractContent(xmlDoc, Metadata.FORMAT, <span class="jxr_string">"//dc:format"</span>, metadata); +<a name="68" href="#68">68</a> extractContent(xmlDoc, Metadata.IDENTIFIER, <span class="jxr_string">"//dc:identifier"</span>, metadata); +<a name="69" href="#69">69</a> extractContent(xmlDoc, Metadata.LANGUAGE, <span class="jxr_string">"//dc:language"</span>, metadata); +<a name="70" href="#70">70</a> extractContent(xmlDoc, Metadata.RIGHTS, <span class="jxr_string">"//dc:rights"</span>, metadata); +<a name="71" href="#71">71</a> +<a name="72" href="#72">72</a> <a href="../../../../../org/apache/tika/sax/XHTMLContentHandler.html">XHTMLContentHandler</a> xhtml = <strong class="jxr_keyword">new</strong> <a href="../../../../../org/apache/tika/sax/XHTMLContentHandler.html">XHTMLContentHandler</a>(handler, metadata); +<a name="73" href="#73">73</a> xhtml.startDocument(); +<a name="74" href="#74">74</a> xhtml.startElement(<span class="jxr_string">"p"</span>); +<a name="75" href="#75">75</a> concatOccurrence(xmlDoc, <span class="jxr_string">"//*"</span>, <span class="jxr_string">" "</span>, <strong class="jxr_keyword">new</strong> <a href="../../../../../org/apache/tika/sax/AppendableAdaptor.html">AppendableAdaptor</a>(xhtml)); +<a name="76" href="#76">76</a> xhtml.endElement(<span class="jxr_string">"p"</span>); +<a name="77" href="#77">77</a> xhtml.endDocument(); +<a name="78" href="#78">78</a> } +<a name="79" href="#79">79</a> +<a name="80" href="#80">80</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">void</strong> concatOccurrence(Object xmlDoc, String xpath, String concatSep, Appendable chaineConcat) <strong class="jxr_keyword">throws</strong> IOException { +<a name="81" href="#81">81</a> +<a name="82" href="#82">82</a> <strong class="jxr_keyword">try</strong> { +<a name="83" href="#83">83</a> JDOMXPath xp = <strong class="jxr_keyword">new</strong> JDOMXPath(xpath); +<a name="84" href="#84">84</a> List ls = xp.selectNodes(xmlDoc); +<a name="85" href="#85">85</a> Iterator i = ls.iterator(); +<a name="86" href="#86">86</a> <strong class="jxr_keyword">int</strong> j = 0; +<a name="87" href="#87">87</a> <strong class="jxr_keyword">while</strong> (i.hasNext()) { +<a name="88" href="#88">88</a> j++; +<a name="89" href="#89">89</a> String text = <span class="jxr_string">""</span>; +<a name="90" href="#90">90</a> Object obj = i.next(); +<a name="91" href="#91">91</a> <strong class="jxr_keyword">if</strong> (obj instanceof Element) { +<a name="92" href="#92">92</a> Element elem = (Element) obj; +<a name="93" href="#93">93</a> text = elem.getText().trim(); +<a name="94" href="#94">94</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (obj instanceof Attribute) { +<a name="95" href="#95">95</a> Attribute att = (Attribute) obj; +<a name="96" href="#96">96</a> text = att.getValue().trim(); +<a name="97" href="#97">97</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (obj instanceof Text) { +<a name="98" href="#98">98</a> Text txt = (Text) obj; +<a name="99" href="#99">99</a> text = txt.getText().trim(); +<a name="100" href="#100">100</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (obj instanceof Comment) { +<a name="101" href="#101">101</a> Comment com = (Comment) obj; +<a name="102" href="#102">102</a> text = com.getText().trim(); +<a name="103" href="#103">103</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (obj instanceof ProcessingInstruction) { +<a name="104" href="#104">104</a> ProcessingInstruction pi = (ProcessingInstruction) obj; +<a name="105" href="#105">105</a> text = pi.getData().trim(); +<a name="106" href="#106">106</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (obj instanceof EntityRef) { +<a name="107" href="#107">107</a> EntityRef er = (EntityRef) obj; +<a name="108" href="#108">108</a> text = er.toString().trim(); +<a name="109" href="#109">109</a> } +<a name="110" href="#110">110</a> <strong class="jxr_keyword">if</strong> (StringUtils.isNotEmpty(text)) { +<a name="111" href="#111">111</a> chaineConcat.append(text); +<a name="112" href="#112">112</a> <strong class="jxr_keyword">if</strong> (ls.size() == 1) { +<a name="113" href="#113">113</a> <strong class="jxr_keyword">return</strong>; +<a name="114" href="#114">114</a> } <strong class="jxr_keyword">else</strong> { +<a name="115" href="#115">115</a> <strong class="jxr_keyword">if</strong> (ls.size() != j) { +<a name="116" href="#116">116</a> chaineConcat.append(' ') +<a name="117" href="#117">117</a> .append(concatSep) +<a name="118" href="#118">118</a> .append(' '); +<a name="119" href="#119">119</a> } +<a name="120" href="#120">120</a> } +<a name="121" href="#121">121</a> } +<a name="122" href="#122">122</a> } +<a name="123" href="#123">123</a> } <strong class="jxr_keyword">catch</strong> (JaxenException j) { +<a name="124" href="#124">124</a> logger.error(j.getMessage()); +<a name="125" href="#125">125</a> } +<a name="126" href="#126">126</a> } +<a name="127" href="#127">127</a> +<a name="128" href="#128">128</a> <strong class="jxr_keyword">public</strong> List getAllDocumentNs(org.jdom.Document doc) { +<a name="129" href="#129">129</a> List ls = <strong class="jxr_keyword">new</strong> ArrayList(); +<a name="130" href="#130">130</a> processChildren(doc.getRootElement(), ls); +<a name="131" href="#131">131</a> <strong class="jxr_keyword">return</strong> ls; +<a name="132" href="#132">132</a> } +<a name="133" href="#133">133</a> +<a name="134" href="#134">134</a> <strong class="jxr_keyword">private</strong> <strong class="jxr_keyword">boolean</strong> exist(List nsLs, String nsUri) { +<a name="135" href="#135">135</a> <strong class="jxr_keyword">if</strong> (nsLs.isEmpty()) +<a name="136" href="#136">136</a> <strong class="jxr_keyword">return</strong> false; +<a name="137" href="#137">137</a> <strong class="jxr_keyword">for</strong> (Object nsL : nsLs) { +<a name="138" href="#138">138</a> <strong class="jxr_keyword">if</strong> (nsL.equals(nsUri)) { +<a name="139" href="#139">139</a> <strong class="jxr_keyword">return</strong> <strong class="jxr_keyword">true</strong>; +<a name="140" href="#140">140</a> } +<a name="141" href="#141">141</a> } +<a name="142" href="#142">142</a> <strong class="jxr_keyword">return</strong> false; +<a name="143" href="#143">143</a> } +<a name="144" href="#144">144</a> +<a name="145" href="#145">145</a> <strong class="jxr_keyword">private</strong> <strong class="jxr_keyword">void</strong> processChildren(Element elem, List ns) { +<a name="146" href="#146">146</a> Namespace nsCourent = elem.getNamespace(); +<a name="147" href="#147">147</a> String nsUri = (nsCourent.getURI()); +<a name="148" href="#148">148</a> <strong class="jxr_keyword">if</strong> (!exist(ns, nsUri)) { +<a name="149" href="#149">149</a> ns.add(nsUri.trim()); +<a name="150" href="#150">150</a> } +<a name="151" href="#151">151</a> List additionalNs = elem.getAdditionalNamespaces(); +<a name="152" href="#152">152</a> <strong class="jxr_keyword">if</strong> (!additionalNs.isEmpty()) +<a name="153" href="#153">153</a> copyNsList(additionalNs, ns); +<a name="154" href="#154">154</a> <strong class="jxr_keyword">if</strong> (elem.getChildren().size() > 0) { +<a name="155" href="#155">155</a> List elemChildren = elem.getChildren(); +<a name="156" href="#156">156</a> <strong class="jxr_keyword">for</strong> (Object anElemChildren : elemChildren) { +<a name="157" href="#157">157</a> processChildren((Element) anElemChildren, ns); +<a name="158" href="#158">158</a> } +<a name="159" href="#159">159</a> } +<a name="160" href="#160">160</a> } +<a name="161" href="#161">161</a> +<a name="162" href="#162">162</a> <strong class="jxr_keyword">private</strong> <strong class="jxr_keyword">void</strong> copyNsList(List nsElem, List nsRes) { +<a name="163" href="#163">163</a> <strong class="jxr_keyword">for</strong> (Object aNsElem : nsElem) { +<a name="164" href="#164">164</a> nsRes.add(((Namespace) aNsElem).getURI().trim()); +<a name="165" href="#165">165</a> } +<a name="166" href="#166">166</a> } +<a name="167" href="#167">167</a> +<a name="168" href="#168">168</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">void</strong> extractContent( +<a name="169" href="#169">169</a> Document xmlDoc, String name, String xpath, <a href="../../../../../org/apache/tika/metadata/Metadata.html">Metadata</a> metadata) { +<a name="170" href="#170">170</a> <strong class="jxr_keyword">try</strong> { +<a name="171" href="#171">171</a> JDOMXPath xp = <strong class="jxr_keyword">new</strong> JDOMXPath(xpath); +<a name="172" href="#172">172</a> SimpleNamespaceContext context = <strong class="jxr_keyword">new</strong> SimpleNamespaceContext(); +<a name="173" href="#173">173</a> context.addNamespace(<span class="jxr_string">"dc"</span>, <span class="jxr_string">"http://purl.org/dc/elements/1.1/";</span>); +<a name="174" href="#174">174</a> context.addNamespace(<span class="jxr_string">"meta"</span>, <span class="jxr_string">"urn:oasis:names:tc:opendocument:xmlns:meta:1.0"</span>); +<a name="175" href="#175">175</a> xp.setNamespaceContext(context); +<a name="176" href="#176">176</a> List selectNodes = xp.selectNodes(xmlDoc); +<a name="177" href="#177">177</a> Iterator nodes = selectNodes.iterator(); +<a name="178" href="#178">178</a> <strong class="jxr_keyword">while</strong> (nodes.hasNext()) { +<a name="179" href="#179">179</a> Object node = nodes.next(); +<a name="180" href="#180">180</a> <strong class="jxr_keyword">if</strong> (node instanceof Element) { +<a name="181" href="#181">181</a> Element elem = (Element) node; +<a name="182" href="#182">182</a> <strong class="jxr_keyword">if</strong> (StringUtils.isNotBlank(elem.getText())) { +<a name="183" href="#183">183</a> metadata.add(name, elem.getText().trim()); +<a name="184" href="#184">184</a> } +<a name="185" href="#185">185</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (node instanceof Attribute) { +<a name="186" href="#186">186</a> Attribute att = (Attribute) node; +<a name="187" href="#187">187</a> metadata.add(name, att.getValue()); +<a name="188" href="#188">188</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (node instanceof Text) { +<a name="189" href="#189">189</a> Text text = (Text) node; +<a name="190" href="#190">190</a> metadata.add(name, text.getText()); +<a name="191" href="#191">191</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (node instanceof Comment) { +<a name="192" href="#192">192</a> Comment com = (Comment) node; +<a name="193" href="#193">193</a> metadata.add(name, com.getText()); +<a name="194" href="#194">194</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (node instanceof ProcessingInstruction) { +<a name="195" href="#195">195</a> ProcessingInstruction pi = (ProcessingInstruction) node; +<a name="196" href="#196">196</a> metadata.add(name, pi.getData()); +<a name="197" href="#197">197</a> } <strong class="jxr_keyword">else</strong> <strong class="jxr_keyword">if</strong> (node instanceof EntityRef) { +<a name="198" href="#198">198</a> EntityRef er = (EntityRef) node; +<a name="199" href="#199">199</a> metadata.add(name, er.toString()); +<a name="200" href="#200">200</a> } +<a name="201" href="#201">201</a> } +<a name="202" href="#202">202</a> } <strong class="jxr_keyword">catch</strong> (JaxenException e) { +<a name="203" href="#203">203</a> logger.error(e.getMessage()); +<a name="204" href="#204">204</a> } +<a name="205" href="#205">205</a> +<a name="206" href="#206">206</a> } +<a name="207" href="#207">207</a> +<a name="208" href="#208">208</a> } </pre> <hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/";>Maven</a></div></body> </html>
Modified: incubator/tika/site/xref/org/apache/tika/parser/xml/package-frame.html URL: http://svn.apache.org/viewvc/incubator/tika/site/xref/org/apache/tika/parser/xml/package-frame.html?rev=609491&r1=609490&r2=609491&view=diff ============================================================================== --- incubator/tika/site/xref/org/apache/tika/parser/xml/package-frame.html (original) +++ incubator/tika/site/xref/org/apache/tika/parser/xml/package-frame.html Sun Jan 6 19:06:26 2008 @@ -3,7 +3,7 @@ <html xml:lang="en" lang="en"> <head> <meta http-equiv="content-type" content="text/html; charset=ISO-8859-1" /> - <title>Apache Tika 0.1-SNAPSHOT Reference Package org.apache.tika.parser.xml</title> + <title>Apache Tika 0.2-SNAPSHOT Reference Package org.apache.tika.parser.xml</title> <link rel="stylesheet" type="text/css" href="../../../../../stylesheet.css" title="style" /> </head> <body> Modified: incubator/tika/site/xref/org/apache/tika/parser/xml/package-summary.html URL: http://svn.apache.org/viewvc/incubator/tika/site/xref/org/apache/tika/parser/xml/package-summary.html?rev=609491&r1=609490&r2=609491&view=diff ============================================================================== --- incubator/tika/site/xref/org/apache/tika/parser/xml/package-summary.html (original) +++ incubator/tika/site/xref/org/apache/tika/parser/xml/package-summary.html Sun Jan 6 19:06:26 2008 @@ -3,7 +3,7 @@ <html xml:lang="en" lang="en"> <head> <meta http-equiv="content-type" content="text/html; charset=ISO-8859-1" /> - <title>Apache Tika 0.1-SNAPSHOT Reference Package org.apache.tika.parser.xml</title> + <title>Apache Tika 0.2-SNAPSHOT Reference Package org.apache.tika.parser.xml</title> <link rel="stylesheet" type="text/css" href="../../../../../stylesheet.css" title="style" /> </head> <body> @@ -62,6 +62,6 @@ </ul> </div> <hr /> - Copyright © 2007 The Apache Software Foundation. All Rights Reserved. + Copyright © 2008 The Apache Software Foundation. All Rights Reserved. </body> </html> Modified: incubator/tika/site/xref/org/apache/tika/sax/package-frame.html URL: http://svn.apache.org/viewvc/incubator/tika/site/xref/org/apache/tika/sax/package-frame.html?rev=609491&r1=609490&r2=609491&view=diff ============================================================================== --- incubator/tika/site/xref/org/apache/tika/sax/package-frame.html (original) +++ incubator/tika/site/xref/org/apache/tika/sax/package-frame.html Sun Jan 6 19:06:26 2008 @@ -3,7 +3,7 @@ <html xml:lang="en" lang="en"> <head> <meta http-equiv="content-type" content="text/html; charset=ISO-8859-1" /> - <title>Apache Tika 0.1-SNAPSHOT Reference Package org.apache.tika.sax</title> + <title>Apache Tika 0.2-SNAPSHOT Reference Package org.apache.tika.sax</title> <link rel="stylesheet" type="text/css" href="../../../../stylesheet.css" title="style" /> </head> <body> @@ -16,6 +16,9 @@ <ul> <li> + <a href="AppendableAdaptor.html" target="classFrame">AppendableAdaptor</a> + </li> + <li> <a href="ContentHandlerDecorator.html" target="classFrame">ContentHandlerDecorator</a> </li> <li> Modified: incubator/tika/site/xref/org/apache/tika/sax/package-summary.html URL: http://svn.apache.org/viewvc/incubator/tika/site/xref/org/apache/tika/sax/package-summary.html?rev=609491&r1=609490&r2=609491&view=diff ============================================================================== --- incubator/tika/site/xref/org/apache/tika/sax/package-summary.html (original) +++ incubator/tika/site/xref/org/apache/tika/sax/package-summary.html Sun Jan 6 19:06:26 2008 @@ -3,7 +3,7 @@ <html xml:lang="en" lang="en"> <head> <meta http-equiv="content-type" content="text/html; charset=ISO-8859-1" /> - <title>Apache Tika 0.1-SNAPSHOT Reference Package org.apache.tika.sax</title> + <title>Apache Tika 0.2-SNAPSHOT Reference Package org.apache.tika.sax</title> <link rel="stylesheet" type="text/css" href="../../../../stylesheet.css" title="style" /> </head> <body> @@ -37,6 +37,11 @@ <tbody> <tr> <td> + <a href="AppendableAdaptor.html" target="classFrame">AppendableAdaptor</a> + </td> + </tr> + <tr> + <td> <a href="ContentHandlerDecorator.html" target="classFrame">ContentHandlerDecorator</a> </td> </tr> @@ -77,6 +82,6 @@ </ul> </div> <hr /> - Copyright © 2007 The Apache Software Foundation. All Rights Reserved. + Copyright © 2008 The Apache Software Foundation. All Rights Reserved. </body> </html> Modified: incubator/tika/site/xref/org/apache/tika/utils/RegexUtils.html URL: http://svn.apache.org/viewvc/incubator/tika/site/xref/org/apache/tika/utils/RegexUtils.html?rev=609491&r1=609490&r2=609491&view=diff ============================================================================== --- incubator/tika/site/xref/org/apache/tika/utils/RegexUtils.html (original) +++ incubator/tika/site/xref/org/apache/tika/utils/RegexUtils.html Sun Jan 6 19:06:26 2008 @@ -27,55 +27,53 @@ <a name="17" href="#17">17</a> <strong class="jxr_keyword">package</strong> org.apache.tika.utils; <a name="18" href="#18">18</a> <a name="19" href="#19">19</a> <strong class="jxr_keyword">import</strong> java.util.ArrayList; -<a name="20" href="#20">20</a> <strong class="jxr_keyword">import</strong> java.util.List; -<a name="21" href="#21">21</a> -<a name="22" href="#22">22</a> <strong class="jxr_keyword">import</strong> org.apache.log4j.Logger; -<a name="23" href="#23">23</a> <strong class="jxr_keyword">import</strong> org.apache.oro.text.regex.MalformedPatternException; -<a name="24" href="#24">24</a> <strong class="jxr_keyword">import</strong> org.apache.oro.text.regex.MatchResult; -<a name="25" href="#25">25</a> <strong class="jxr_keyword">import</strong> org.apache.oro.text.regex.Pattern; -<a name="26" href="#26">26</a> <strong class="jxr_keyword">import</strong> org.apache.oro.text.regex.PatternCompiler; -<a name="27" href="#27">27</a> <strong class="jxr_keyword">import</strong> org.apache.oro.text.regex.PatternMatcher; -<a name="28" href="#28">28</a> <strong class="jxr_keyword">import</strong> org.apache.oro.text.regex.PatternMatcherInput; -<a name="29" href="#29">29</a> <strong class="jxr_keyword">import</strong> org.apache.oro.text.regex.Perl5Compiler; -<a name="30" href="#30">30</a> <strong class="jxr_keyword">import</strong> org.apache.oro.text.regex.Perl5Matcher; -<a name="31" href="#31">31</a> -<a name="32" href="#32">32</a> <em class="jxr_javadoccomment">/**</em> -<a name="33" href="#33">33</a> <em class="jxr_javadoccomment"> * Inspired from Nutch code class OutlinkExtractor. Apply regex to extract</em> -<a name="34" href="#34">34</a> <em class="jxr_javadoccomment"> * content</em> -<a name="35" href="#35">35</a> <em class="jxr_javadoccomment"> * </em> -<a name="36" href="#36">36</a> <em class="jxr_javadoccomment"> * </em> -<a name="37" href="#37">37</a> <em class="jxr_javadoccomment"> */</em> -<a name="38" href="#38">38</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">class</strong> <a href="../../../../org/apache/tika/utils/RegexUtils.html">RegexUtils</a> { -<a name="39" href="#39">39</a> -<a name="40" href="#40">40</a> <strong class="jxr_keyword">static</strong> Logger logger = Logger.getRootLogger(); -<a name="41" href="#41">41</a> -<a name="42" href="#42">42</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">static</strong> List<String> extract(String content, String regex) -<a name="43" href="#43">43</a> <strong class="jxr_keyword">throws</strong> MalformedPatternException { -<a name="44" href="#44">44</a> -<a name="45" href="#45">45</a> List<String> extractions = <strong class="jxr_keyword">new</strong> ArrayList<String>(); -<a name="46" href="#46">46</a> <strong class="jxr_keyword">final</strong> PatternCompiler cp = <strong class="jxr_keyword">new</strong> Perl5Compiler(); -<a name="47" href="#47">47</a> <strong class="jxr_keyword">final</strong> Pattern pattern = cp.compile(regex, -<a name="48" href="#48">48</a> Perl5Compiler.CASE_INSENSITIVE_MASK -<a name="49" href="#49">49</a> | Perl5Compiler.READ_ONLY_MASK -<a name="50" href="#50">50</a> | Perl5Compiler.MULTILINE_MASK); -<a name="51" href="#51">51</a> <strong class="jxr_keyword">final</strong> PatternMatcher matcher = <strong class="jxr_keyword">new</strong> Perl5Matcher(); -<a name="52" href="#52">52</a> -<a name="53" href="#53">53</a> <strong class="jxr_keyword">final</strong> PatternMatcherInput input = <strong class="jxr_keyword">new</strong> PatternMatcherInput(content); -<a name="54" href="#54">54</a> -<a name="55" href="#55">55</a> MatchResult result; -<a name="56" href="#56">56</a> String extractedContent; +<a name="20" href="#20">20</a> <strong class="jxr_keyword">import</strong> java.util.Collections; +<a name="21" href="#21">21</a> <strong class="jxr_keyword">import</strong> java.util.List; +<a name="22" href="#22">22</a> <strong class="jxr_keyword">import</strong> java.util.regex.Matcher; +<a name="23" href="#23">23</a> <strong class="jxr_keyword">import</strong> java.util.regex.Pattern; +<a name="24" href="#24">24</a> +<a name="25" href="#25">25</a> <em class="jxr_javadoccomment">/**</em> +<a name="26" href="#26">26</a> <em class="jxr_javadoccomment"> * Inspired from Nutch code class OutlinkExtractor. Apply regex to extract</em> +<a name="27" href="#27">27</a> <em class="jxr_javadoccomment"> * content</em> +<a name="28" href="#28">28</a> <em class="jxr_javadoccomment"> * </em> +<a name="29" href="#29">29</a> <em class="jxr_javadoccomment"> * </em> +<a name="30" href="#30">30</a> <em class="jxr_javadoccomment"> */</em> +<a name="31" href="#31">31</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">class</strong> <a href="../../../../org/apache/tika/utils/RegexUtils.html">RegexUtils</a> { +<a name="32" href="#32">32</a> +<a name="33" href="#33">33</a> <em class="jxr_javadoccomment">/**</em> +<a name="34" href="#34">34</a> <em class="jxr_javadoccomment"> * Regex pattern to get URLs within a plain text.</em> +<a name="35" href="#35">35</a> <em class="jxr_javadoccomment"> * </em> +<a name="36" href="#36">36</a> <em class="jxr_javadoccomment"> * @see <a</em> +<a name="37" href="#37">37</a> <em class="jxr_javadoccomment"> * href="<a href="http://www.truerwords.net/articles/ut/urlactivation.html"; target="alexandria_uri">http://www.truerwords.net/articles/ut/urlactivation.html</a>"><a href="http://www.truerwords.net/articles/ut/urlactivation.html"; target="alexandria_uri">http://www.truerwords.net/articles/ut/urlactivation.html</a></em> +<a name="38" href="#38">38</a> <em class="jxr_javadoccomment"> * </a></em> +<a name="39" href="#39">39</a> <em class="jxr_javadoccomment"> */</em> +<a name="40" href="#40">40</a> <strong class="jxr_keyword">private</strong> <strong class="jxr_keyword">static</strong> <strong class="jxr_keyword">final</strong> String LINKS_REGEX = +<a name="41" href="#41">41</a> <span class="jxr_string">"([A-Za-z][A-Za-z0-9+.-]{1,120}:"</span> +<a name="42" href="#42">42</a> + <span class="jxr_string">"[A-Za-z0-9/](([A-Za-z0-9$_.+!*,;/?:@&~=-])|%[A-Fa-f0-9]{2}){1,333}"</span> +<a name="43" href="#43">43</a> + <span class="jxr_string">"(#([a-zA-Z0-9][a-zA-Z0-9$_.+!*,;/?:@&~=%-]{0,1000}))?)"</span>; +<a name="44" href="#44">44</a> +<a name="45" href="#45">45</a> <strong class="jxr_keyword">private</strong> <strong class="jxr_keyword">static</strong> <strong class="jxr_keyword">final</strong> Pattern LINKS_PATTERN = Pattern.compile(LINKS_REGEX, Pattern.CASE_INSENSITIVE + Pattern.MULTILINE); +<a name="46" href="#46">46</a> +<a name="47" href="#47">47</a> <em class="jxr_javadoccomment">/**</em> +<a name="48" href="#48">48</a> <em class="jxr_javadoccomment"> * Extract urls from plain text.</em> +<a name="49" href="#49">49</a> <em class="jxr_javadoccomment"> *</em> +<a name="50" href="#50">50</a> <em class="jxr_javadoccomment"> * @param content The plain text content to examine</em> +<a name="51" href="#51">51</a> <em class="jxr_javadoccomment"> * @return List of urls within found in the plain text</em> +<a name="52" href="#52">52</a> <em class="jxr_javadoccomment"> */</em> +<a name="53" href="#53">53</a> <strong class="jxr_keyword">public</strong> <strong class="jxr_keyword">static</strong> List<String> extractLinks(String content) { +<a name="54" href="#54">54</a> <strong class="jxr_keyword">if</strong> (content == <strong class="jxr_keyword">null</strong> || content.length() == 0) { +<a name="55" href="#55">55</a> <strong class="jxr_keyword">return</strong> Collections.emptyList(); +<a name="56" href="#56">56</a> } <a name="57" href="#57">57</a> -<a name="58" href="#58">58</a> <strong class="jxr_keyword">while</strong> (matcher.contains(input, pattern)) { -<a name="59" href="#59">59</a> result = matcher.getMatch(); -<a name="60" href="#60">60</a> extractedContent = result.group(0); -<a name="61" href="#61">61</a> extractions.add(extractedContent); +<a name="58" href="#58">58</a> List<String> extractions = <strong class="jxr_keyword">new</strong> ArrayList<String>(); +<a name="59" href="#59">59</a> <strong class="jxr_keyword">final</strong> Matcher matcher = LINKS_PATTERN.matcher(content); +<a name="60" href="#60">60</a> <strong class="jxr_keyword">while</strong> (matcher.find()) { +<a name="61" href="#61">61</a> extractions.add(matcher.group()); <a name="62" href="#62">62</a> } -<a name="63" href="#63">63</a> -<a name="64" href="#64">64</a> <strong class="jxr_keyword">return</strong> extractions; -<a name="65" href="#65">65</a> -<a name="66" href="#66">66</a> } -<a name="67" href="#67">67</a> -<a name="68" href="#68">68</a> } +<a name="63" href="#63">63</a> <strong class="jxr_keyword">return</strong> extractions; +<a name="64" href="#64">64</a> +<a name="65" href="#65">65</a> } +<a name="66" href="#66">66</a> } </pre> <hr/><div id="footer">This page was automatically generated by <a href="http://maven.apache.org/";>Maven</a></div></body> </html> Modified: incubator/tika/site/xref/org/apache/tika/utils/package-frame.html URL: http://svn.apache.org/viewvc/incubator/tika/site/xref/org/apache/tika/utils/package-frame.html?rev=609491&r1=609490&r2=609491&view=diff ============================================================================== --- incubator/tika/site/xref/org/apache/tika/utils/package-frame.html (original) +++ incubator/tika/site/xref/org/apache/tika/utils/package-frame.html Sun Jan 6 19:06:26 2008 @@ -3,7 +3,7 @@ <html xml:lang="en" lang="en"> <head> <meta http-equiv="content-type" content="text/html; charset=ISO-8859-1" /> - <title>Apache Tika 0.1-SNAPSHOT Reference Package org.apache.tika.utils</title> + <title>Apache Tika 0.2-SNAPSHOT Reference Package org.apache.tika.utils</title> <link rel="stylesheet" type="text/css" href="../../../../stylesheet.css" title="style" /> </head> <body> Modified: incubator/tika/site/xref/org/apache/tika/utils/package-summary.html URL: http://svn.apache.org/viewvc/incubator/tika/site/xref/org/apache/tika/utils/package-summary.html?rev=609491&r1=609490&r2=609491&view=diff ============================================================================== --- incubator/tika/site/xref/org/apache/tika/utils/package-summary.html (original) +++ incubator/tika/site/xref/org/apache/tika/utils/package-summary.html Sun Jan 6 19:06:26 2008 @@ -3,7 +3,7 @@ <html xml:lang="en" lang="en"> <head> <meta http-equiv="content-type" content="text/html; charset=ISO-8859-1" /> - <title>Apache Tika 0.1-SNAPSHOT Reference Package org.apache.tika.utils</title> + <title>Apache Tika 0.2-SNAPSHOT Reference Package org.apache.tika.utils</title> <link rel="stylesheet" type="text/css" href="../../../../stylesheet.css" title="style" /> </head> <body> @@ -82,6 +82,6 @@ </ul> </div> <hr /> - Copyright © 2007 The Apache Software Foundation. All Rights Reserved. + Copyright © 2008 The Apache Software Foundation. All Rights Reserved. </body> </html> Modified: incubator/tika/site/xref/overview-frame.html URL: http://svn.apache.org/viewvc/incubator/tika/site/xref/overview-frame.html?rev=609491&r1=609490&r2=609491&view=diff ============================================================================== --- incubator/tika/site/xref/overview-frame.html (original) +++ incubator/tika/site/xref/overview-frame.html Sun Jan 6 19:06:26 2008 @@ -3,7 +3,7 @@ <html xml:lang="en" lang="en"> <head> <meta http-equiv="content-type" content="text/html; charset=ISO-8859-1" /> - <title>Apache Tika 0.1-SNAPSHOT Reference</title> + <title>Apache Tika 0.2-SNAPSHOT Reference</title> <link rel="stylesheet" type="text/css" href="stylesheet.css" title="style" /> </head> <body> Modified: incubator/tika/site/xref/overview-summary.html URL: http://svn.apache.org/viewvc/incubator/tika/site/xref/overview-summary.html?rev=609491&r1=609490&r2=609491&view=diff ============================================================================== --- incubator/tika/site/xref/overview-summary.html (original) +++ incubator/tika/site/xref/overview-summary.html Sun Jan 6 19:06:26 2008 @@ -3,7 +3,7 @@ <html xml:lang="en" lang="en"> <head> <meta http-equiv="content-type" content="text/html; charset=ISO-8859-1" /> - <title>Apache Tika 0.1-SNAPSHOT Reference</title> + <title>Apache Tika 0.2-SNAPSHOT Reference</title> <link rel="stylesheet" type="text/css" href="stylesheet.css" title="style" /> </head> <body> @@ -24,7 +24,7 @@ </ul> </div> - <h2>Apache Tika 0.1-SNAPSHOT Reference</h2> + <h2>Apache Tika 0.2-SNAPSHOT Reference</h2> <table class="summary"> <thead> @@ -124,6 +124,6 @@ </div> <hr /> - Copyright © 2007 The Apache Software Foundation. All Rights Reserved. + Copyright © 2008 The Apache Software Foundation. All Rights Reserved. </body> </html>
