/*
* Copyright (c) 1998-2010 Caucho Technology -- all rights reserved
*
* This file is part of Resin(R) Open Source
*
* Each copy or derived work must preserve the copyright notice and this
* notice unmodified.
*
* Resin Open Source 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.
*
* Resin Open Source 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, or any warranty
* of NON-INFRINGEMENT. See the GNU General Public License for more
* details.
*
* You should have received a copy of the GNU General Public License
* along with Resin Open Source; if not, write to the
*
* Free Software Foundation, Inc.
* 59 Temple Place, Suite 330
* Boston, MA 02111-1307 USA
*
* @author Nam Nguyen
*/
package com.caucho.quercus.lib.i18n;
import com.caucho.quercus.env.Env;
import com.caucho.quercus.env.StringValue;
import com.caucho.quercus.env.UnicodeBuilderValue;
abstract public class Decoder {
protected String _charset;
protected CharSequence _replacement;
protected boolean _isIgnoreErrors = false;
protected boolean _isReplaceUnicode = false;
protected boolean _isAllowMalformedOut = false;
protected boolean _hasError;
protected Decoder(String charset) {
_charset = charset;
}
public static Decoder create(String charset) {
if (charset.equalsIgnoreCase("utf8")
|| charset.equalsIgnoreCase("utf-8")) {
return new Utf8Decoder(charset);
} else if (charset.equalsIgnoreCase("big5")
|| charset.equalsIgnoreCase("big-5")) {
return new Big5Decoder(charset);
} else {
return new GenericDecoder(charset);
}
}
public boolean isUtf8() {
return false;
}
public final boolean isIgnoreErrors() {
return _isIgnoreErrors;
}
public final void setIgnoreErrors(boolean isIgnore) {
_isIgnoreErrors = isIgnore;
}
public final boolean hasError() {
return _hasError;
}
public final void setReplacement(CharSequence replacement) {
_replacement = replacement;
}
public final void setReplaceUnicode(boolean isReplaceUnicode) {
_isReplaceUnicode = isReplaceUnicode;
}
public final void setAllowMalformedOut(boolean isAllowMalformedOut) {
_isAllowMalformedOut = isAllowMalformedOut;
}
public void reset() {
_hasError = false;
}
public final CharSequence decode(Env env, StringValue str) {
if (str.isUnicode()) {
return str;
}
return decodeStringBuilder(env, str);
}
public StringBuilder decodeStringBuilder(Env env, StringValue str) {
return decodeImpl(env, str);
}
public StringValue decodeUnicode(Env env, StringValue str) {
UnicodeBuilderValue sb = new UnicodeBuilderValue();
StringBuilder unicodeStr = decodeImpl(env, str);
return sb.append(unicodeStr);
}
abstract public boolean isDecodable(Env env, StringValue str);
abstract protected StringBuilder decodeImpl(Env env, StringValue str);
}