Here you can find the source of languageByIntCode(int languageCode, int subLanguageCode)
public static String languageByIntCode(int languageCode, int subLanguageCode)
//package com.java2s; /*//from w w w .ja v a 2 s . co m * Copyright (C) 2007-2012 Geometer Plus <contact@geometerplus.com> * * This program 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 2 of the License, or * (at your option) any later version. * * This program 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 this program; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA * 02110-1301, USA. */ public class Main { public static String languageByIntCode(int languageCode, int subLanguageCode) { switch (languageCode) { default: return null; case 0x01: return "ar"; // Arabic case 0x02: return "bg"; // Bulgarian case 0x03: return "ca"; // Catalan case 0x04: return "zh"; // Chinese case 0x05: return "cs"; // Czech case 0x06: return "da"; // Danish case 0x07: return "de"; // German case 0x08: return "el"; // Greek case 0x09: return "en"; // English case 0x0A: return "es"; // Spanish case 0x0B: return "fi"; // Finnish case 0x0C: return "fr"; // French case 0x0D: return "he"; // Hebrew case 0x0E: return "hu"; // Hungarian case 0x0F: return "is"; // Icelandic case 0x10: return "it"; // Italian case 0x11: return "ja"; // Japanese case 0x12: return "ko"; // Korean case 0x13: return "nl"; // Dutch case 0x14: return "no"; // Norwegian case 0x15: return "pl"; // Polish case 0x16: return "pt"; // Portuguese case 0x17: return "rm"; // Romansh case 0x18: return "ro"; // Romanian case 0x19: return "ru"; // Russian case 0x1A: switch (subLanguageCode) { default: return "sr"; // Serbian case 0x04: case 0x10: return "hr"; // Croatian case 0x14: case 0x20: case 0x78: return "bs"; // Bosnian } case 0x1B: return "sk"; // Slovak case 0x1C: return "sq"; // Albanian case 0x1D: return "sv"; // Swedish case 0x1E: return "th"; // Thai case 0x1F: return "tr"; // Turkish case 0x20: return "ur"; // Urdu case 0x21: return "id"; // Indonesian case 0x22: return "uk"; // Ukrainian case 0x23: return "be"; // Belarusian case 0x24: return "sl"; // Slovenian case 0x25: return "et"; // Estonian case 0x26: return "lv"; // Latvian case 0x27: return "lt"; // Lithuanian case 0x28: return "tg"; // Tajik case 0x29: return "fa"; // Persian (Farsi) case 0x2A: return "vi"; // Vietnamese case 0x2B: return "hy"; // Armenian case 0x2C: return "az"; // Azeri case 0x2D: return "eu"; // Basque case 0x2E: return (subLanguageCode == 0x08) ? "dsb" // Lower Sorbian : "wen"; // Upper Sorbian case 0x2F: return "mk"; // Makedonian case 0x32: return "tn"; // Setswana/Tswana case 0x34: return "xh"; // Xhosa/isiXhosa case 0x35: return "zu"; // Zulu/isiZulu case 0x36: return "af"; // Afrikaans case 0x37: return "ka"; // Georgian case 0x38: return "fo"; // Faeroese case 0x39: return "hi"; // Hindi case 0x3A: return "mt"; // Maltese case 0x3B: return "se"; // Sami case 0x3C: return "ga"; // Irish case 0x3E: return "ms"; // Malay case 0x3F: return "kk"; // Kazak case 0x40: return "ky"; // Kyrgyz case 0x41: return "sw"; // Swahili case 0x42: return "tk"; // Turkmen case 0x43: return "uz"; // Uzbek case 0x44: return "tt"; // Tatar case 0x45: return "bn"; // Bengali case 0x46: return "pa"; // Punjabi case 0x47: return "gu"; // Gujaratu case 0x48: return "or"; // Oriya case 0x49: return "ta"; // Tamil case 0x4A: return "te"; // Telugi case 0x4B: return "kn"; // Kannada case 0x4C: return "ml"; // Malayalam case 0x4D: return "as"; // Assamese case 0x4E: return "mr"; // Marathi case 0x4F: return "sa"; // Sanskrit case 0x50: return "mn"; // Mongolian case 0x51: return "bo"; // Tibetian case 0x52: return "cy"; // Welsh case 0x53: return "kh"; // Khmer case 0x54: return "lo"; // Lao case 0x56: return "gl"; // Galician case 0x57: return "kok"; // Konkani case 0x58: return "mni"; // Manipuri case 0x59: return "sd"; // Sindhi case 0x5A: return "syr"; // Syriac case 0x5B: return "si"; // Sinhala case 0x5D: return "iu"; // Inuktitut case 0x5E: return "am"; // Amharic case 0x5F: return "tzm"; // Tamazight case 0x60: return "ks"; // Kashmiri case 0x61: return "ne"; // Nepali case 0x62: return "fy"; // Frisian case 0x63: return "ps"; // Pashto case 0x64: return "fil"; // Filipino case 0x65: return "dv"; // Divehi case 0x68: return "ha"; // Hausa case 0x6A: return "yo"; // Yoruba case 0x6B: return "quz"; // Quechua case 0x6C: return "ns"; // Northern Sotho case 0x6D: return "ba"; // Bashkir case 0x6E: return "lb"; // Luxemburgish case 0x6F: return "kl"; // Greenlandic case 0x70: return "ig"; // Igbo case 0x73: return "ti"; // Tigrinya case 0x78: return "yi"; // Yi case 0x7A: return "arn"; // Mapudungun case 0x7C: return "moh"; // Mohawk case 0x7E: return "be"; // Breton case 0x80: return "ug"; // Uighur case 0x81: return "mi"; // Maori case 0x82: return "oc"; // Occitan case 0x83: return "co"; // Corsican case 0x84: return "gsw"; // Alsatian case 0x85: return "sah"; // Yakut case 0x86: return "qut"; // K'iche case 0x87: return "rw"; // Kinyarwanda case 0x88: return "wo"; // Wolof case 0x8C: return "prs"; // Dari case 0x8D: return "mg"; // Malagasy } } }