/* This file is part of Libresonic. Libresonic is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. Libresonic is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with Libresonic. If not, see <http://www.gnu.org/licenses/>. Copyright 2016 (C) Libresonic Authors Based upon Subsonic, Copyright 2009 (C) Sindre Mehus */ package org.libresonic.player.util; import org.apache.commons.codec.binary.Hex; import org.apache.commons.io.IOUtils; import org.apache.commons.lang.StringUtils; import java.io.*; import java.net.MalformedURLException; import java.net.URL; import java.net.URLDecoder; import java.net.URLEncoder; import java.security.MessageDigest; import java.text.*; import java.util.ArrayList; import java.util.Date; import java.util.List; import java.util.Locale; import java.util.regex.Matcher; import java.util.regex.Pattern; /** * Miscellaneous string utility methods. * * @author Sindre Mehus */ public final class StringUtil { public static final String ENCODING_LATIN = "ISO-8859-1"; public static final String ENCODING_UTF8 = "UTF-8"; private static final DateFormat ISO_8601_DATE_FORMAT = new SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss"); private static final String[][] HTML_SUBSTITUTIONS = { {"&", "&"}, {"<", "<"}, {">", ">"}, {"'", "'"}, {"\"", """}, }; private static final String[][] MIME_TYPES = { {"mp3", "audio/mpeg"}, {"ogg", "audio/ogg"}, {"oga", "audio/ogg"}, {"opus", "audio/ogg"}, {"ogx", "application/ogg"}, {"aac", "audio/mp4"}, {"m4a", "audio/mp4"}, {"flac", "audio/flac"}, {"wav", "audio/x-wav"}, {"wma", "audio/x-ms-wma"}, {"ape", "audio/x-monkeys-audio"}, {"mpc", "audio/x-musepack"}, {"shn", "audio/x-shn"}, {"flv", "video/x-flv"}, {"avi", "video/avi"}, {"mpg", "video/mpeg"}, {"mpeg", "video/mpeg"}, {"mp4", "video/mp4"}, {"m4v", "video/x-m4v"}, {"mkv", "video/x-matroska"}, {"mov", "video/quicktime"}, {"wmv", "video/x-ms-wmv"}, {"ogv", "video/ogg"}, {"divx", "video/divx"}, {"m2ts", "video/MP2T"}, {"ts", "video/MP2T"}, {"webm", "video/webm"}, {"gif", "image/gif"}, {"jpg", "image/jpeg"}, {"jpeg", "image/jpeg"}, {"png", "image/png"}, {"bmp", "image/bmp"}, }; private static final String[] FILE_SYSTEM_UNSAFE = {"/", "\\", "..", ":", "\"", "?", "*", "|"}; /** * Disallow external instantiation. */ private StringUtil() { } /** * Returns the specified string converted to a format suitable for * HTML. All single-quote, double-quote, greater-than, less-than and * ampersand characters are replaces with their corresponding HTML * Character Entity code. * * @param s the string to convert * @return the converted string */ public static String toHtml(String s) { if (s == null) { return null; } for (String[] substitution : HTML_SUBSTITUTIONS) { if (s.contains(substitution[0])) { s = s.replaceAll(substitution[0], substitution[1]); } } return s; } /** * Formats the given date to a ISO-8601 date/time format, and UTC timezone. * <p/> * The returned date uses the following format: 2007-12-17T14:57:17 * * @param date The date to format * @return The corresponding ISO-8601 formatted string. */ public static String toISO8601(Date date) { if (date == null) { return null; } synchronized (ISO_8601_DATE_FORMAT) { return ISO_8601_DATE_FORMAT.format(date); } } /** * Removes the suffix (the substring after the last dot) of the given string. The dot is * also removed. * * @param s The string in question, e.g., "foo.mp3". * @return The string without the suffix, e.g., "foo". */ public static String removeSuffix(String s) { int index = s.lastIndexOf('.'); return index == -1 ? s : s.substring(0, index); } /** * Returns the proper MIME type for the given suffix. * * @param suffix The suffix, e.g., "mp3" or ".mp3". * @return The corresponding MIME type, e.g., "audio/mpeg". If no MIME type is found, * <code>application/octet-stream</code> is returned. */ public static String getMimeType(String suffix) { for (String[] map : MIME_TYPES) { if (map[0].equalsIgnoreCase(suffix) || ('.' + map[0]).equalsIgnoreCase(suffix)) { return map[1]; } } return "application/octet-stream"; } public static String getMimeType(String suffix, boolean sonos) { String result = getMimeType(suffix); // Sonos doesn't work with "audio/mp4" but needs "audio/aac" for ALAC and AAC (in MP4 container) return sonos && "audio/mp4".equals(result) ? "audio/aac" : result; } public static String getSuffix(String mimeType) { for (String[] map : MIME_TYPES) { if (map[1].equalsIgnoreCase(mimeType)) { return map[0]; } } return null; } /** * Converts a byte-count to a formatted string suitable for display to the user. * For instance: * <ul> * <li><code>format(918)</code> returns <em>"918 B"</em>.</li> * <li><code>format(98765)</code> returns <em>"96 KB"</em>.</li> * <li><code>format(1238476)</code> returns <em>"1.2 MB"</em>.</li> * </ul> * This method assumes that 1 KB is 1024 bytes. * * @param byteCount The number of bytes. * @param locale The locale used for formatting. * @return The formatted string. */ public static synchronized String formatBytes(long byteCount, Locale locale) { // More than 1 TB? if (byteCount >= 1024L * 1024 * 1024 * 1024) { NumberFormat teraByteFormat = new DecimalFormat("0.00 TB", new DecimalFormatSymbols(locale)); return teraByteFormat.format( ((double) byteCount ) / ((double) 1024 * 1024 * 1024 * 1024)); } // More than 1 GB? if (byteCount >= 1024L * 1024 * 1024) { NumberFormat gigaByteFormat = new DecimalFormat("0.00 GB", new DecimalFormatSymbols(locale)); return gigaByteFormat.format((double) byteCount / ((double) 1024 * 1024 * 1024)); } // More than 1 MB? if (byteCount >= 1024L * 1024) { NumberFormat megaByteFormat = new DecimalFormat("0.0 MB", new DecimalFormatSymbols(locale)); return megaByteFormat.format((double) byteCount / ((double) 1024 * 1024)); } // More than 1 KB? if (byteCount >= 1024L) { NumberFormat kiloByteFormat = new DecimalFormat("0 KB", new DecimalFormatSymbols(locale)); return kiloByteFormat.format((double) byteCount / ((double) 1024)); } return byteCount + " B"; } /** * Formats a duration with minutes and seconds, e.g., "93:45" */ public static String formatDuration(int seconds) { int minutes = seconds / 60; int secs = seconds % 60; StringBuilder builder = new StringBuilder(6); builder.append(minutes).append(":"); if (secs < 10) { builder.append("0"); } builder.append(secs); return builder.toString(); } /** * Splits the input string. White space is interpreted as separator token. Double quotes * are interpreted as grouping operator. <br/> * For instance, the input <code>"u2 rem "greatest hits""</code> will return an array with * three elements: <code>{"u2", "rem", "greatest hits"}</code> * * @param input The input string. * @return Array of elements. */ public static String[] split(String input) { if (input == null) { return new String[0]; } Pattern pattern = Pattern.compile("\".*?\"|\\S+"); Matcher matcher = pattern.matcher(input); List<String> result = new ArrayList<String>(); while (matcher.find()) { String element = matcher.group(); if (element.startsWith("\"") && element.endsWith("\"") && element.length() > 1) { element = element.substring(1, element.length() - 1); } result.add(element); } return result.toArray(new String[result.size()]); } /** * Reads lines from the given input stream. All lines are trimmed. Empty lines and lines starting * with "#" are skipped. The input stream is always closed by this method. * * @param in The input stream to read from. * @return Array of lines. * @throws IOException If an I/O error occurs. */ public static String[] readLines(InputStream in) throws IOException { BufferedReader reader = null; try { reader = new BufferedReader(new InputStreamReader(in)); List<String> result = new ArrayList<String>(); for (String line = reader.readLine(); line != null; line = reader.readLine()) { line = line.trim(); if (!line.startsWith("#") && line.length() > 0) { result.add(line); } } return result.toArray(new String[result.size()]); } finally { IOUtils.closeQuietly(in); IOUtils.closeQuietly(reader); } } /** * Converts the given string of whitespace-separated integers to an <code>int</code> array. * * @param s String consisting of integers separated by whitespace. * @return The corresponding array of ints. * @throws NumberFormatException If string contains non-parseable text. */ public static int[] parseInts(String s) { if (s == null) { return new int[0]; } String[] strings = StringUtils.split(s); int[] ints = new int[strings.length]; for (int i = 0; i < strings.length; i++) { ints[i] = Integer.parseInt(strings[i]); } return ints; } /** * Determines whether a is equal to b, taking null into account. * * @return Whether a and b are equal, or both null. */ public static boolean isEqual(Object a, Object b) { return a == null ? b == null : a.equals(b); } /** * Parses a locale from the given string. * * @param s The locale string. Should be formatted as per the documentation in {@link Locale#toString()}. * @return The locale. */ public static Locale parseLocale(String s) { if (s == null) { return null; } String[] elements = s.split("_"); if (elements.length == 0) { return new Locale(s, "", ""); } if (elements.length == 1) { return new Locale(elements[0], "", ""); } if (elements.length == 2) { return new Locale(elements[0], elements[1], ""); } return new Locale(elements[0], elements[1], elements[2]); } /** * URL-encodes the input value using UTF-8. */ public static String urlEncode(String s) { try { return URLEncoder.encode(s, StringUtil.ENCODING_UTF8); } catch (UnsupportedEncodingException x) { throw new RuntimeException(x); } } /** * URL-decodes the input value using UTF-8. */ public static String urlDecode(String s) { try { return URLDecoder.decode(s, StringUtil.ENCODING_UTF8); } catch (UnsupportedEncodingException x) { throw new RuntimeException(x); } } /** * Encodes the given string by using the hexadecimal representation of its UTF-8 bytes. * * @param s The string to encode. * @return The encoded string. */ public static String utf8HexEncode(String s) { if (s == null) { return null; } byte[] utf8; try { utf8 = s.getBytes(ENCODING_UTF8); } catch (UnsupportedEncodingException x) { throw new RuntimeException(x); } return String.valueOf(Hex.encodeHex(utf8)); } /** * Decodes the given string by using the hexadecimal representation of its UTF-8 bytes. * * @param s The string to decode. * @return The decoded string. * @throws Exception If an error occurs. */ public static String utf8HexDecode(String s) throws Exception { if (s == null) { return null; } return new String(Hex.decodeHex(s.toCharArray()), ENCODING_UTF8); } /** * Calculates the MD5 digest and returns the value as a 32 character hex string. * * @param s Data to digest. * @return MD5 digest as a hex string. */ public static String md5Hex(String s) { if (s == null) { return null; } try { MessageDigest md5 = MessageDigest.getInstance("MD5"); return new String(Hex.encodeHex(md5.digest(s.getBytes(ENCODING_UTF8)))); } catch (Exception x) { throw new RuntimeException(x.getMessage(), x); } } /** * Returns the file part of an URL. For instance: * <p/> * <code> * getUrlFile("http://archive.ncsa.uiuc.edu:80/SDG/Software/Mosaic/Demo/url-primer.html") * </code> * <p/> * will return "url-primer.html". * * @param url The URL in question. * @return The file part, or <code>null</code> if no file can be resolved. */ public static String getUrlFile(String url) { try { String path = new URL(url).getPath(); if (StringUtils.isBlank(path) || path.endsWith("/")) { return null; } File file = new File(path); String filename = file.getName(); if (StringUtils.isBlank(filename)) { return null; } return filename; } catch (MalformedURLException x) { return null; } } /** * Makes a given filename safe by replacing special characters like slashes ("/" and "\") * with dashes ("-"). * * @param filename The filename in question. * @return The filename with special characters replaced by underscores. */ public static String fileSystemSafe(String filename) { for (String s : FILE_SYSTEM_UNSAFE) { filename = filename.replace(s, "-"); } return filename; } public static String removeMarkup(String s) { if (s == null) { return null; } return s.replaceAll("<.*?>", ""); } }