Java tutorial
/* Copyright 2004 Sun Microsystems, Inc. * * 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 nl.minbzk.dwr.zoeken.enricher.util; import java.io.BufferedInputStream; import java.io.BufferedReader; import java.io.ByteArrayInputStream; import java.io.File; import java.io.FileInputStream; import java.io.IOException; import java.io.InputStream; import java.io.InputStreamReader; import java.io.Reader; import java.net.HttpURLConnection; import java.net.URL; import java.net.URLConnection; import java.text.MessageFormat; import java.util.ArrayList; import java.util.List; import java.util.regex.Matcher; import java.util.regex.Pattern; import org.apache.commons.lang.StringUtils; import org.slf4j.Logger; import org.slf4j.LoggerFactory; /** * Character stream that handles (or at least attemtps to) all the necessary Voodo to figure out the charset encoding of both XML and HTML documents within a given stream. * * IMPORTANT: This class is not related in any way to the org.xml.sax.XMLReader. This one IS a character stream. * * All this has to be done without consuming characters from the stream, if not the XML parser will not recognized the document as a valid XML. This is not 100% true, but it's close enough (UTF-8 BOM is not handled by all parsers right now, * XmlReader handles it and things work in all parsers). * * The XmlReader class handles the charset encoding of XML documents in Files, raw streams and HTTP streams by offering a wide set of constructors. * * By default the charset encoding detection is lenient, the constructor with the lenient flag can be used for an script (following HTTP MIME and XML specifications). * * All this is nicely explained by Mark Pilgrim in his blog, <a href="http://diveintomark.org/archives/2004/02/13/xml-media-types"> Determining the character encoding of a feed</a>. * * In addition to XML-based detection, HTML-encoding detection from the META-tag is performed when the given content-type match an HTML-based type. * * @author Alejandro Abdelnur * @author Jasper van Veghel <j.veghel@rijksoverheid.nl> */ public class XmlHtmlReader extends Reader { /** * The logger. */ private static final Logger logger = LoggerFactory.getLogger(XmlHtmlReader.class); /** * Constants. */ private static final int BUFFER_SIZE = 4096; private static final String UTF_8 = "UTF-8"; private static final String US_ASCII = "US-ASCII"; private static final String UTF_16BE = "UTF-16BE"; private static final String UTF_16LE = "UTF-16LE"; private static final String UTF_16 = "UTF-16"; private static String _staticDefaultEncoding = null; private Reader _reader; private String _encoding; private List<String> _languages; private final String _defaultEncoding; /** * Sets the default encoding to use if none is set in HTTP content-type, XML prolog and the rules based on content-type are not adequate. * <p/> * If it is set to NULL the content-type based rules are used. * <p/> * By default it is NULL. * <p/> * * @param encoding * charset encoding to default to. */ public static void setDefaultEncoding(final String encoding) { _staticDefaultEncoding = encoding; } /** * Returns the default encoding to use if none is set in HTTP content-type, XML prolog and the rules based on content-type are not adequate. * <p/> * If it is NULL the content-type based rules are used. * <p/> * * @return the default encoding to use. */ public static String getDefaultEncoding() { return _staticDefaultEncoding; } /** * Creates a Reader for a File. * <p> * It looks for the UTF-8 BOM first, if none sniffs the XML prolog charset, if this is also missing defaults to UTF-8. * <p> * It does a lenient charset encoding detection, check the constructor with the lenient parameter for details. * <p> * * @param file * File to create a Reader from. * @throws IOException * thrown if there is a problem reading the file. * */ public XmlHtmlReader(final File file) throws IOException { this(new FileInputStream(file)); } /** * Creates a Reader for a raw InputStream. * <p> * It follows the same logic used for files. * <p> * It does a lenient charset encoding detection, check the constructor with the lenient parameter for details. * <p> * * @param is * InputStream to create a Reader from. * @throws IOException * thrown if there is a problem reading the stream. * */ public XmlHtmlReader(final InputStream is) throws IOException { this(is, true); } /** * Creates a Reader for a raw InputStream and uses the provided default encoding if none is determined. * <p> * It follows the same logic used for files. * <p> * If lenient detection is indicated and the detection above fails as per specifications it then attempts the following: * <p> * If the content type was 'text/html' it replaces it with 'text/xml' and tries the detection again. * <p> * Else if the XML prolog had a charset encoding that encoding is used. * <p> * Else if the content type had a charset encoding that encoding is used. * <p> * Else 'UTF-8' is used. * <p> * If lenient detection is indicated an XmlReaderException is never thrown. * <p> * * @param is * InputStream to create a Reader from. * @param lenient * indicates if the charset encoding detection should be relaxed. * @param defaultEncoding * default encoding to use if one cannot be detected. * @throws IOException * thrown if there is a problem reading the stream. * @throws XmlReaderException * thrown if the charset encoding could not be determined according to the specs. * */ public XmlHtmlReader(final InputStream is, final boolean lenient, final String defaultEncoding) throws IOException, XmlReaderException { _defaultEncoding = defaultEncoding == null ? _staticDefaultEncoding : defaultEncoding; try { doRawStream(is, lenient); } catch (XmlReaderException ex) { if (!lenient) { throw ex; } else { doLenientDetection(null, ex); } } } /** * Creates a Reader for a raw InputStream. * <p> * It follows the same logic used for files. * <p> * If lenient detection is indicated and the detection above fails as per specifications it then attempts the following: * <p> * If the content type was 'text/html' it replaces it with 'text/xml' and tries the detection again. * <p> * Else if the XML prolog had a charset encoding that encoding is used. * <p> * Else if the content type had a charset encoding that encoding is used. * <p> * Else 'UTF-8' is used. * <p> * If lenient detection is indicated an XmlReaderException is never thrown. * <p> * * @param is * InputStream to create a Reader from. * @param lenient * indicates if the charset encoding detection should be relaxed. * @throws IOException * thrown if there is a problem reading the stream. * @throws XmlReaderException * thrown if the charset encoding could not be determined according to the specs. * */ public XmlHtmlReader(final InputStream is, final boolean lenient) throws IOException, XmlReaderException { this(is, lenient, null); } /** * Creates a Reader using the InputStream of a URL. * <p> * If the URL is not of type HTTP and there is not 'content-type' header in the fetched data it uses the same logic used for Files. * <p> * If the URL is a HTTP Url or there is a 'content-type' header in the fetched data it uses the same logic used for an InputStream with content-type. * <p> * It does a lenient charset encoding detection, check the constructor with the lenient parameter for details. * <p> * * @param url * URL to create a Reader from. * @throws IOException * thrown if there is a problem reading the stream of the URL. * */ public XmlHtmlReader(final URL url) throws IOException { this(url.openConnection()); } /** * Creates a Reader using the InputStream of a URLConnection. * <p> * If the URLConnection is not of type HttpURLConnection and there is not 'content-type' header in the fetched data it uses the same logic used for files. * <p> * If the URLConnection is a HTTP Url or there is a 'content-type' header in the fetched data it uses the same logic used for an InputStream with content-type. * <p> * It does a lenient charset encoding detection, check the constructor with the lenient parameter for details. * <p> * * @param conn * URLConnection to create a Reader from. * @throws IOException * thrown if there is a problem reading the stream of the URLConnection. * */ public XmlHtmlReader(final URLConnection conn) throws IOException { _defaultEncoding = _staticDefaultEncoding; boolean lenient = true; if (conn instanceof HttpURLConnection) { try { doHttpStream(conn.getInputStream(), conn.getContentType(), lenient); } catch (XmlReaderException ex) { doLenientDetection(conn.getContentType(), ex); } } else if (conn.getContentType() != null) { try { doHttpStream(conn.getInputStream(), conn.getContentType(), lenient); } catch (XmlReaderException ex) { doLenientDetection(conn.getContentType(), ex); } } else { try { doRawStream(conn.getInputStream(), lenient); } catch (XmlReaderException ex) { doLenientDetection(null, ex); } } } /** * Creates a Reader using an InputStream and the associated content-type header. * <p> * First it checks if the stream has BOM. If there is not BOM checks the content-type encoding. If there is not content-type encoding checks the XML prolog encoding. If there is not XML prolog encoding uses the default encoding * mandated by the content-type MIME type. * <p> * It does a lenient charset encoding detection, check the constructor with the lenient parameter for details. * <p> * * @param is * InputStream to create the reader from. * @param httpContentType * content-type header to use for the resolution of the charset encoding. * @throws IOException * thrown if there is a problem reading the file. * */ public XmlHtmlReader(final InputStream is, final String httpContentType) throws IOException { this(is, httpContentType, true); } /** * Creates a Reader using an InputStream and the associated content-type header. * <p> * First it checks if the stream has BOM. If there is not BOM checks the content-type encoding. If there is not content-type encoding checks the XML prolog encoding. If there is not XML prolog encoding checks the META Content-Type * encoding. If there is no META Content-Type encoding uses the default encoding mandated by the content-type MIME type. * <p> * If lenient detection is indicated and the detection above fails as per specifications it then attempts the following: * <p> * If the content type was 'text/html' it replaces it with 'text/xml' and tries the detection again. * <p> * Else if the XML prolog had a charset encoding that encoding is used. * <p> * Else if the content type had a charset encoding that encoding is used. * <p> * Else 'UTF-8' is used. * <p> * If lenient detection is indicated and XmlReaderException is never thrown. * <p> * * @param is * InputStream to create the reader from. * @param httpContentType * content-type header to use for the resolution of the charset encoding. * @param lenient * indicates if the charset encoding detection should be relaxed. * @param defaultEncoding * default encoding to use if one cannot be detected. * @throws IOException * thrown if there is a problem reading the file. * @throws XmlReaderException * thrown if the charset encoding could not be determined according to the specs. * */ public XmlHtmlReader(final InputStream is, final String httpContentType, final boolean lenient, final String defaultEncoding) throws IOException, XmlReaderException { _defaultEncoding = defaultEncoding == null ? _staticDefaultEncoding : defaultEncoding; try { doHttpStream(is, httpContentType, lenient); } catch (XmlReaderException ex) { if (!lenient) { throw ex; } else { doLenientDetection(httpContentType, ex); } } } /** * Creates a Reader using an InputStream and the associated content-type header. * <p> * First it checks if the stream has BOM. If there is not BOM checks the content-type encoding. If there is not content-type encoding checks the XML prolog encoding. If there is not XML prolog encoding uses the default encoding * mandated by the content-type MIME type. * <p> * If lenient detection is indicated and the detection above fails as per specifications it then attempts the following: * <p> * If the content type was 'text/html' it replaces it with 'text/xml' and tries the detection again. * <p> * Else if the XML prolog had a charset encoding that encoding is used. * <p> * Else if the content type had a charset encoding that encoding is used. * <p> * Else 'UTF-8' is used. * <p> * If lenient detection is indicated and XmlReaderException is never thrown. * <p> * * @param is * InputStream to create the reader from. * @param httpContentType * content-type header to use for the resolution of the charset encoding. * @param lenient * indicates if the charset encoding detection should be relaxed. * @throws IOException * thrown if there is a problem reading the file. * @throws XmlReaderException * thrown if the charset encoding could not be determined according to the specs. * */ public XmlHtmlReader(final InputStream is, final String httpContentType, final boolean lenient) throws IOException, XmlReaderException { this(is, httpContentType, lenient, null); } private void doLenientDetection(String httpContentType, XmlReaderException ex) throws IOException { if (httpContentType != null) { if (httpContentType.startsWith("text/html")) { httpContentType = httpContentType.substring("text/html".length()); httpContentType = "text/xml" + httpContentType; try { doHttpStream(ex.getInputStream(), httpContentType, true); ex = null; } catch (XmlReaderException ex2) { ex = ex2; } } } if (ex != null) { String encoding = ex.getXmlEncoding(); if (encoding == null) { encoding = ex.getContentTypeEncoding(); } if (encoding == null) { encoding = _defaultEncoding == null ? UTF_8 : _defaultEncoding; } prepareReader(ex.getInputStream(), encoding); } } /** * Returns the charset encoding of the XmlReader. * <p> * * @return charset encoding. * */ public String getEncoding() { return _encoding; } /** * Returns the language of the XmlReader. * <p> * * @return languages. * */ public List<String> getLanguages() { return _languages; } @Override public int read(final char[] buf, final int offset, final int len) throws IOException { return _reader.read(buf, offset, len); } /** * Closes the XmlReader stream. * <p> * * @throws IOException * thrown if there was a problem closing the stream. * */ @Override public void close() throws IOException { _reader.close(); } private void doRawStream(final InputStream is, final boolean lenient) throws IOException { BufferedInputStream pis = new BufferedInputStream(is, BUFFER_SIZE); String bomEnc = getBOMEncoding(pis); String xmlGuessEnc = getXMLGuessEncoding(pis); String xmlEnc = getXmlProlog(pis, xmlGuessEnc != null ? xmlGuessEnc : "UTF-8"); String encoding = calculateRawEncoding(bomEnc, xmlGuessEnc, xmlEnc, pis); prepareReader(pis, encoding); } private void doHttpStream(final InputStream is, final String httpContentType, final boolean lenient) throws IOException { BufferedInputStream pis = new BufferedInputStream(is, BUFFER_SIZE); String cTMime = getContentTypeMime(httpContentType); String cTEnc = getContentTypeEncoding(httpContentType); String bomEnc = getBOMEncoding(pis); String xmlGuessEnc = getXMLGuessEncoding(pis); String xmlEnc = getXmlProlog(pis, xmlGuessEnc != null ? xmlGuessEnc : "UTF-8"); // Always detect the HTML meta anyway, since it also detects and sets the Content-Language String htmlMeta = httpContentType != null && (httpContentType.startsWith("text/html") || httpContentType.startsWith("application/xhtml+xml")) ? getHtmlMeta(pis, xmlGuessEnc != null ? xmlGuessEnc : "UTF-8") : null; // Always prefer the reference encoding /and/ the prolog encoding to the META encoding if (xmlEnc == null && cTEnc == null && htmlMeta != null) xmlEnc = htmlMeta; String encoding = calculateHttpEncoding(cTMime, cTEnc, bomEnc, xmlGuessEnc, xmlEnc, pis, lenient); prepareReader(pis, encoding); } private void prepareReader(final InputStream is, final String encoding) throws IOException { _reader = new InputStreamReader(is, encoding); _encoding = encoding; } // InputStream is passed for XmlReaderException creation only private String calculateRawEncoding(final String bomEnc, final String xmlGuessEnc, final String xmlEnc, final InputStream is) throws IOException { String encoding; if (bomEnc == null) { if (xmlGuessEnc == null || xmlEnc == null) { encoding = _defaultEncoding == null ? UTF_8 : _defaultEncoding; } else if (xmlEnc.equals(UTF_16) && (xmlGuessEnc.equals(UTF_16BE) || xmlGuessEnc.equals(UTF_16LE))) { encoding = xmlGuessEnc; } else { encoding = xmlEnc; } } else if (bomEnc.equals(UTF_8)) { if (xmlGuessEnc != null && !xmlGuessEnc.equals(UTF_8)) { throw new XmlReaderException(RAW_EX_1.format(new Object[] { bomEnc, xmlGuessEnc, xmlEnc }), bomEnc, xmlGuessEnc, xmlEnc, is); } if (xmlEnc != null && !xmlEnc.equals(UTF_8)) { throw new XmlReaderException(RAW_EX_1.format(new Object[] { bomEnc, xmlGuessEnc, xmlEnc }), bomEnc, xmlGuessEnc, xmlEnc, is); } encoding = UTF_8; } else if (bomEnc.equals(UTF_16BE) || bomEnc.equals(UTF_16LE)) { if (xmlGuessEnc != null && !xmlGuessEnc.equals(bomEnc)) { throw new IOException(RAW_EX_1.format(new Object[] { bomEnc, xmlGuessEnc, xmlEnc })); } if (xmlEnc != null && !xmlEnc.equals(UTF_16) && !xmlEnc.equals(bomEnc)) { throw new XmlReaderException(RAW_EX_1.format(new Object[] { bomEnc, xmlGuessEnc, xmlEnc }), bomEnc, xmlGuessEnc, xmlEnc, is); } encoding = bomEnc; } else { throw new XmlReaderException(RAW_EX_2.format(new Object[] { bomEnc, xmlGuessEnc, xmlEnc }), bomEnc, xmlGuessEnc, xmlEnc, is); } return encoding; } // InputStream is passed for XmlReaderException creation only private String calculateHttpEncoding(final String cTMime, final String cTEnc, final String bomEnc, final String xmlGuessEnc, final String xmlEnc, final InputStream is, final boolean lenient) throws IOException { String encoding; if (lenient & xmlEnc != null) { encoding = xmlEnc; } else { boolean appXml = isAppXml(cTMime); boolean textXml = isTextXml(cTMime); if (appXml || textXml) { if (cTEnc == null) { if (appXml) { encoding = calculateRawEncoding(bomEnc, xmlGuessEnc, xmlEnc, is); } else { encoding = _defaultEncoding == null ? US_ASCII : _defaultEncoding; } } else if (bomEnc != null && (cTEnc.equals(UTF_16BE) || cTEnc.equals(UTF_16LE))) { throw new XmlReaderException( HTTP_EX_1.format(new Object[] { cTMime, cTEnc, bomEnc, xmlGuessEnc, xmlEnc }), cTMime, cTEnc, bomEnc, xmlGuessEnc, xmlEnc, is); } else if (cTEnc.equals(UTF_16)) { if (bomEnc != null && bomEnc.startsWith(UTF_16)) { encoding = bomEnc; } else { throw new XmlReaderException( HTTP_EX_2.format(new Object[] { cTMime, cTEnc, bomEnc, xmlGuessEnc, xmlEnc }), cTMime, cTEnc, bomEnc, xmlGuessEnc, xmlEnc, is); } } else { encoding = cTEnc; } } else { throw new XmlReaderException( HTTP_EX_3.format(new Object[] { cTMime, cTEnc, bomEnc, xmlGuessEnc, xmlEnc }), cTMime, cTEnc, bomEnc, xmlGuessEnc, xmlEnc, is); } } return encoding; } // returns MIME type or NULL if httpContentType is NULL private static String getContentTypeMime(final String httpContentType) { String mime = null; if (httpContentType != null) { int i = httpContentType.indexOf(";"); mime = (i == -1 ? httpContentType : httpContentType.substring(0, i)).trim(); } return mime; } private static final Pattern CHARSET_PATTERN = Pattern.compile("charset=([.[^; ]]*)"); // returns charset parameter value, NULL if not present, NULL if httpContentType is NULL private static String getContentTypeEncoding(final String httpContentType) { String encoding = null; if (httpContentType != null) { int i = httpContentType.indexOf(";"); if (i > -1) { String postMime = httpContentType.substring(i + 1); Matcher m = CHARSET_PATTERN.matcher(postMime); encoding = m.find() ? m.group(1) : null; encoding = encoding != null ? encoding.toUpperCase() : null; } if (encoding != null && (encoding.startsWith("\"") && encoding.endsWith("\"") || encoding.startsWith("'") && encoding.endsWith("'"))) { encoding = encoding.substring(1, encoding.length() - 1); } } return encoding; } // returns the BOM in the stream, NULL if not present, // if there was BOM the in the stream it is consumed private static String getBOMEncoding(final BufferedInputStream is) throws IOException { String encoding = null; int[] bytes = new int[3]; is.mark(3); bytes[0] = is.read(); bytes[1] = is.read(); bytes[2] = is.read(); if (bytes[0] == 0xFE && bytes[1] == 0xFF) { encoding = UTF_16BE; is.reset(); is.read(); is.read(); } else if (bytes[0] == 0xFF && bytes[1] == 0xFE) { encoding = UTF_16LE; is.reset(); is.read(); is.read(); } else if (bytes[0] == 0xEF && bytes[1] == 0xBB && bytes[2] == 0xBF) { encoding = UTF_8; } else { is.reset(); } return encoding; } // returns the best guess for the encoding by looking the first bytes of the stream, '<?' private static String getXMLGuessEncoding(final BufferedInputStream is) throws IOException { String encoding = null; int[] bytes = new int[4]; is.mark(4); bytes[0] = is.read(); bytes[1] = is.read(); bytes[2] = is.read(); bytes[3] = is.read(); is.reset(); if (bytes[0] == 0x00 && bytes[1] == 0x3C && bytes[2] == 0x00 && bytes[3] == 0x3F) { encoding = UTF_16BE; } else if (bytes[0] == 0x3C && bytes[1] == 0x00 && bytes[2] == 0x3F && bytes[3] == 0x00) { encoding = UTF_16LE; } else if (bytes[0] == 0x3C && bytes[1] == 0x3F && bytes[2] == 0x78 && bytes[3] == 0x6D) { encoding = UTF_8; } return encoding; } private static final Pattern ENCODING_PATTERN = Pattern .compile("<\\?xml.*encoding[\\s]*=[\\s]*((?:\".[^\"]*\")|(?:'.[^']*'))", Pattern.MULTILINE); /** * Returns the encoding declared in <?xml encoding=...?>, or NULL if none. * * @param is * @param guessedEnc * @return String * @throws IOException */ private static String getXmlProlog(final BufferedInputStream is, final String guessedEnc) { try { String encoding = null; if (guessedEnc != null) { byte[] bytes = new byte[BUFFER_SIZE]; is.mark(BUFFER_SIZE); int offset = 0; int max = BUFFER_SIZE; int c = is.read(bytes, offset, max); int firstGT = -1; while (c != -1 && firstGT == -1 && offset < BUFFER_SIZE) { offset += c; max -= c; try { c = is.read(bytes, offset, max); } catch (IOException e) { if (offset > 0) is.reset(); throw e; } firstGT = new String(bytes, 0, offset).indexOf(">"); } if (firstGT == -1) { if (offset > 0) is.reset(); if (c == -1) { throw new IOException("Unexpected end of XML stream"); } else { throw new IOException("XML prolog or ROOT element not found on first " + offset + " bytes"); } } int bytesRead = offset; if (bytesRead > 0) { is.reset(); Reader reader = new InputStreamReader(new ByteArrayInputStream(bytes, 0, firstGT + 1), guessedEnc); BufferedReader bReader = new BufferedReader(reader); StringBuffer prolog = new StringBuffer(); String line = bReader.readLine(); while (line != null) { prolog.append(line); line = bReader.readLine(); } Matcher m = ENCODING_PATTERN.matcher(prolog); if (m.find()) { encoding = m.group(1).toUpperCase(); encoding = encoding.substring(1, encoding.length() - 1); } } } return encoding; } catch (IOException e) { if (logger.isDebugEnabled()) logger.debug("Could not retrieve the XML prolog from the given input: " + e.getMessage()); return null; } } private static final Pattern META_CONTENT_TYPE_PATTERN = Pattern.compile( "<meta.*http-equiv=\"content-type\"[\\ ]*content=\"([^\"]*)\".*>|<meta.*content=\"([^\"]*)\"[\\ ]*http-equiv=\"content-type\".*>", Pattern.MULTILINE | Pattern.CASE_INSENSITIVE); private static final Pattern META_CONTENT_LANGUAGE_PATTERN = Pattern.compile( "<meta.*http-equiv=\"content-language\"[\\ ]*content=\"([^\"]*)\".*>|<meta.*content=\"([^\"]*)\"[\\ ]*http-equiv=\"content-language\".*>", Pattern.MULTILINE | Pattern.CASE_INSENSITIVE); /** * Returns the encoding declared in the <meta http-equiv="content-type" ..>, or NULL if none. * * @param is * @param guessedEnc * @return String * @throws IOException */ private String getHtmlMeta(final BufferedInputStream is, final String guessedEnc) throws IOException { String encoding = null; if (guessedEnc != null) { // Mark the stream up until the maximum bytes we'd like to read (4096 * 2), which should be enough to account for double-byte encodings like UTF-16 is.mark(BUFFER_SIZE * 2); // Rather than read until we run into the meta tag, we just read out 4096 * 4 bytes byte[] buffer = new byte[BUFFER_SIZE * 2]; int count = is.read(buffer); if (count > 0) { String content = new String(buffer, 0, count, guessedEnc); if (content.indexOf("<") != -1 && content.indexOf(">") != -1) { Matcher contentTypeMatcher = META_CONTENT_TYPE_PATTERN.matcher(content); Matcher contentLanguageMatcher = META_CONTENT_LANGUAGE_PATTERN.matcher(content); // Either the first or second group contains the match, the format of should be equivalent to the Content-Type header if (contentTypeMatcher.find()) encoding = getContentTypeEncoding( (StringUtils.isEmpty(contentTypeMatcher.group(1)) ? contentTypeMatcher.group(2) : contentTypeMatcher.group(1))); // Likewise for the language, except store it directly within the class if (contentLanguageMatcher.find()) { String[] languages = (StringUtils.isEmpty(contentLanguageMatcher.group(1)) ? contentLanguageMatcher.group(2) : contentLanguageMatcher.group(1)).split(","); _languages = new ArrayList<String>(); for (String language : languages) _languages.add(language.split("-")[0].trim()); } } is.reset(); } } return encoding; } // indicates if the MIME type belongs to the APPLICATION XML family private static boolean isAppXml(final String mime) { return mime != null && (mime.equals("application/xml") || mime.equals("application/xml-dtd") || mime.equals("application/xml-external-parsed-entity") || mime.startsWith("application/") && mime.endsWith("+xml")); } // indicates if the MIME type belongs to the TEXT XML family private static boolean isTextXml(final String mime) { return mime != null && (mime.equals("text/xml") || mime.equals("text/xml-external-parsed-entity") || mime.startsWith("text/") && mime.endsWith("+xml")); } private static final MessageFormat RAW_EX_1 = new MessageFormat( "Invalid encoding, BOM [{0}] XML guess [{1}] XML prolog [{2}] encoding mismatch"); private static final MessageFormat RAW_EX_2 = new MessageFormat( "Invalid encoding, BOM [{0}] XML guess [{1}] XML prolog [{2}] unknown BOM"); private static final MessageFormat HTTP_EX_1 = new MessageFormat( "Invalid encoding, CT-MIME [{0}] CT-Enc [{1}] BOM [{2}] XML guess [{3}] XML prolog [{4}], BOM must be NULL"); private static final MessageFormat HTTP_EX_2 = new MessageFormat( "Invalid encoding, CT-MIME [{0}] CT-Enc [{1}] BOM [{2}] XML guess [{3}] XML prolog [{4}], encoding mismatch"); private static final MessageFormat HTTP_EX_3 = new MessageFormat( "Invalid encoding, CT-MIME [{0}] CT-Enc [{1}] BOM [{2}] XML guess [{3}] XML prolog [{4}], Invalid MIME"); }