|
Groovy example source code file (XmlSlurper.java)
This example Groovy source code file (XmlSlurper.java) is included in the DevDaily.com
"Java Source Code
Warehouse" project. The intent of this project is to help you "Learn Java by Example" TM.
The Groovy XmlSlurper.java source code
/*
* Copyright 2003-2010 the original author or authors.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package groovy.util;
import groovy.util.slurpersupport.GPathResult;
import groovy.util.slurpersupport.Node;
import groovy.util.slurpersupport.NodeChild;
import groovy.xml.FactorySupport;
import java.io.File;
import java.io.FileInputStream;
import java.io.IOException;
import java.io.InputStream;
import java.io.Reader;
import java.io.StringReader;
import java.net.URL;
import java.util.HashMap;
import java.util.Hashtable;
import java.util.Map;
import java.util.Stack;
import javax.xml.parsers.ParserConfigurationException;
import javax.xml.parsers.SAXParser;
import javax.xml.parsers.SAXParserFactory;
import org.xml.sax.Attributes;
import org.xml.sax.DTDHandler;
import org.xml.sax.EntityResolver;
import org.xml.sax.ErrorHandler;
import org.xml.sax.InputSource;
import org.xml.sax.SAXException;
import org.xml.sax.SAXNotRecognizedException;
import org.xml.sax.SAXNotSupportedException;
import org.xml.sax.XMLReader;
import org.xml.sax.helpers.DefaultHandler;
/**
* <p>Parse XML into a document tree that may be traversed similar to XPath
* expressions. For example:</p>
* <pre>
* def rootNode = new XmlSlurper().parseText(
* '<root><one a1="uno!"/><two>Some text!</two></root>' )
*
* assert rootNode.name() == 'root'
* assert rootNode.one[0].@a1 == 'uno!'
* assert rootNode.two.text() == 'Some text!'
* rootNode.children().each { assert it.name() in ['one','two'] }
* </pre>
*
* <p>Note that in some cases, a 'selector' expression may not resolve to a
* single node. For example: </p>
* <pre>
* def rootNode = new XmlSlurper().parseText(
* '''<root>
* <a>one!</a>
* <a>two!</a>
* </root>''' )
*
* assert rootNode.a.size() == 2
* rootNode.a.each { assert it.text() in ['one!','two!'] }
* </pre>
* @see GPathResult
* @author John Wilson
*/
public class XmlSlurper extends DefaultHandler {
private final XMLReader reader;
private Node currentNode = null;
private final Stack stack = new Stack();
private final StringBuffer charBuffer = new StringBuffer();
private final Map<String, String> namespaceTagHints = new Hashtable();
private boolean keepWhitespace = false;
/**
* Uses the defaults of not validating and namespace aware.
* @throws ParserConfigurationException
* @throws SAXException
*/
public XmlSlurper() throws ParserConfigurationException, SAXException {
this(false, true);
}
public XmlSlurper(final boolean validating, final boolean namespaceAware) throws ParserConfigurationException, SAXException {
SAXParserFactory factory = FactorySupport.createSaxParserFactory();
factory.setNamespaceAware(namespaceAware);
factory.setValidating(validating);
this.reader = factory.newSAXParser().getXMLReader();
}
public XmlSlurper(final XMLReader reader) {
this.reader = reader;
}
public XmlSlurper(final SAXParser parser) throws SAXException {
this(parser.getXMLReader());
}
/**
* @param keepWhitespace
*
* If true then whitespace before elements is kept.
* The default is to discard the whitespace.
*/
public void setKeepWhitespace(boolean keepWhitespace) {
this.keepWhitespace = keepWhitespace;
}
/**
* @return The GPathResult instance created by consuming a stream of SAX events
* Note if one of the parse methods has been called then this returns null
* Note if this is called more than once all calls after the first will return null
*
*/
public GPathResult getDocument() {
try {
return new NodeChild(this.currentNode, null, this.namespaceTagHints);
} finally {
this.currentNode = null;
}
}
/**
* Parse the content of the specified input source into a GPathResult object
*
* @param input
* @return An object which supports GPath expressions
* @throws IOException
* @throws SAXException
*/
public GPathResult parse(final InputSource input) throws IOException, SAXException {
this.reader.setContentHandler(this);
this.reader.parse(input);
return getDocument();
}
/**
* Parses the content of the given file as XML turning it into a GPathResult object
*
* @param file
* @return An object which supports GPath expressions
* @throws IOException
* @throws SAXException
*/
public GPathResult parse(final File file) throws IOException, SAXException {
final InputSource input = new InputSource(new FileInputStream(file));
input.setSystemId("file://" + file.getAbsolutePath());
return parse(input);
}
/**
* Parse the content of the specified input stream into an GPathResult Object.
* Note that using this method will not provide the parser with any URI
* for which to find DTDs etc
*
* @param input
* @return An object which supports GPath expressions
* @throws IOException
* @throws SAXException
*/
public GPathResult parse(final InputStream input) throws IOException, SAXException {
return parse(new InputSource(input));
}
/**
* Parse the content of the specified reader into a GPathResult Object.
* Note that using this method will not provide the parser with any URI
* for which to find DTDs etc
*
* @param in
* @return An object which supports GPath expressions
* @throws IOException
* @throws SAXException
*/
public GPathResult parse(final Reader in) throws IOException, SAXException {
return parse(new InputSource(in));
}
/**
* Parse the content of the specified URI into a GPathResult Object
*
* @param uri
* @return An object which supports GPath expressions
* @throws IOException
* @throws SAXException
*/
public GPathResult parse(final String uri) throws IOException, SAXException {
return parse(new InputSource(uri));
}
/**
* A helper method to parse the given text as XML
*
* @param text
* @return An object which supports GPath expressions
*/
public GPathResult parseText(final String text) throws IOException, SAXException {
return parse(new StringReader(text));
}
// Delegated XMLReader methods
//------------------------------------------------------------------------
/* (non-Javadoc)
* @see org.xml.sax.XMLReader#getDTDHandler()
*/
public DTDHandler getDTDHandler() {
return this.reader.getDTDHandler();
}
/* (non-Javadoc)
* @see org.xml.sax.XMLReader#getEntityResolver()
*/
public EntityResolver getEntityResolver() {
return this.reader.getEntityResolver();
}
/* (non-Javadoc)
* @see org.xml.sax.XMLReader#getErrorHandler()
*/
public ErrorHandler getErrorHandler() {
return this.reader.getErrorHandler();
}
/* (non-Javadoc)
* @see org.xml.sax.XMLReader#getFeature(java.lang.String)
*/
public boolean getFeature(final String uri) throws SAXNotRecognizedException, SAXNotSupportedException {
return this.reader.getFeature(uri);
}
/* (non-Javadoc)
* @see org.xml.sax.XMLReader#getProperty(java.lang.String)
*/
public Object getProperty(final String uri) throws SAXNotRecognizedException, SAXNotSupportedException {
return this.reader.getProperty(uri);
}
/* (non-Javadoc)
* @see org.xml.sax.XMLReader#setDTDHandler(org.xml.sax.DTDHandler)
*/
public void setDTDHandler(final DTDHandler dtdHandler) {
this.reader.setDTDHandler(dtdHandler);
}
/* (non-Javadoc)
* @see org.xml.sax.XMLReader#setEntityResolver(org.xml.sax.EntityResolver)
*/
public void setEntityResolver(final EntityResolver entityResolver) {
this.reader.setEntityResolver(entityResolver);
}
/**
* Resolves entities against using the supplied URL as the base for relative URLs
*
* @param base
* The URL used to resolve relative URLs
*/
public void setEntityBaseUrl(final URL base) {
this.reader.setEntityResolver(new EntityResolver() {
public InputSource resolveEntity(final String publicId, final String systemId) throws IOException {
return new InputSource(new URL(base, systemId).openStream());
}
});
}
/* (non-Javadoc)
* @see org.xml.sax.XMLReader#setErrorHandler(org.xml.sax.ErrorHandler)
*/
public void setErrorHandler(final ErrorHandler errorHandler) {
this.reader.setErrorHandler(errorHandler);
}
/* (non-Javadoc)
* @see org.xml.sax.XMLReader#setFeature(java.lang.String, boolean)
*/
public void setFeature(final String uri, final boolean value) throws SAXNotRecognizedException, SAXNotSupportedException {
this.reader.setFeature(uri, value);
}
/* (non-Javadoc)
* @see org.xml.sax.XMLReader#setProperty(java.lang.String, java.lang.Object)
*/
public void setProperty(final String uri, final Object value) throws SAXNotRecognizedException, SAXNotSupportedException {
this.reader.setProperty(uri, value);
}
// ContentHandler interface
//-------------------------------------------------------------------------
/* (non-Javadoc)
* @see org.xml.sax.ContentHandler#startDocument()
*/
public void startDocument() throws SAXException {
this.currentNode = null;
this.charBuffer.setLength(0);
}
/* (non-Javadoc)
* @see org.xml.sax.helpers.DefaultHandler#startPrefixMapping(java.lang.String, java.lang.String)
*/
public void startPrefixMapping(final String tag, final String uri) throws SAXException {
this.namespaceTagHints.put(tag, uri);
}
/* (non-Javadoc)
* @see org.xml.sax.ContentHandler#startElement(java.lang.String, java.lang.String, java.lang.String, org.xml.sax.Attributes)
*/
public void startElement(final String namespaceURI, final String localName, final String qName, final Attributes atts) throws SAXException {
addCdata();
final Map attributes = new HashMap();
final Map attributeNamespaces = new HashMap();
for (int i = atts.getLength() - 1; i != -1; i--) {
if (atts.getURI(i).length() == 0) {
attributes.put(atts.getQName(i), atts.getValue(i));
} else {
attributes.put(atts.getLocalName(i), atts.getValue(i));
attributeNamespaces.put(atts.getLocalName(i), atts.getURI(i));
}
}
final Node newElement;
if (namespaceURI.length() == 0){
newElement = new Node(this.currentNode, qName, attributes, attributeNamespaces, namespaceURI);
} else {
newElement = new Node(this.currentNode, localName, attributes, attributeNamespaces, namespaceURI);
}
if (this.currentNode != null) {
this.currentNode.addChild(newElement);
}
this.stack.push(this.currentNode);
this.currentNode = newElement;
}
/* (non-Javadoc)
* @see org.xml.sax.ContentHandler#characters(char[], int, int)
*/
public void characters(final char[] ch, final int start, final int length) throws SAXException {
this.charBuffer.append(ch, start, length);
}
/* (non-Javadoc)
* @see org.xml.sax.ContentHandler#endElement(java.lang.String, java.lang.String, java.lang.String)
*/
public void endElement(final String namespaceURI, final String localName, final String qName) throws SAXException {
addCdata();
final Object oldCurrentNode = this.stack.pop();
if (oldCurrentNode != null) {
this.currentNode = (Node)oldCurrentNode;
}
}
/* (non-Javadoc)
* @see org.xml.sax.ContentHandler#endDocument()
*/
public void endDocument() throws SAXException {
}
// Implementation methods
//-------------------------------------------------------------------------
/**
*
*/
private void addCdata() {
if (this.charBuffer.length() != 0) {
//
// This element is preceded by CDATA if keepWhitespace is false (the default setting) and
// it's not whitespace add it to the body
// Note that, according to the XML spec, we should preserve the CDATA if it's all whitespace
// but for the sort of work I'm doing ignoring the whitespace is preferable
//
final String cdata = this.charBuffer.toString();
this.charBuffer.setLength(0);
if (this.keepWhitespace || cdata.trim().length() != 0) {
this.currentNode.addChild(cdata);
}
}
}
}
Other Groovy examples (source code examples)
Here is a short list of links related to this Groovy XmlSlurper.java source code file:
|