aoqi@0: /*
aoqi@0:  * Copyright (c) 2009, Oracle and/or its affiliates. All rights reserved.
aoqi@0:  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
aoqi@0:  *
aoqi@0:  * This code is free software; you can redistribute it and/or modify it
aoqi@0:  * under the terms of the GNU General Public License version 2 only, as
aoqi@0:  * published by the Free Software Foundation.  Oracle designates this
aoqi@0:  * particular file as subject to the "Classpath" exception as provided
aoqi@0:  * by Oracle in the LICENSE file that accompanied this code.
aoqi@0:  *
aoqi@0:  * This code is distributed in the hope that it will be useful, but WITHOUT
aoqi@0:  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
aoqi@0:  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
aoqi@0:  * version 2 for more details (a copy is included in the LICENSE file that
aoqi@0:  * accompanied this code).
aoqi@0:  *
aoqi@0:  * You should have received a copy of the GNU General Public License version
aoqi@0:  * 2 along with this work; if not, write to the Free Software Foundation,
aoqi@0:  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
aoqi@0:  *
aoqi@0:  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
aoqi@0:  * or visit www.oracle.com if you need additional information or have any
aoqi@0:  * questions.
aoqi@0:  */
aoqi@0: 
aoqi@0: package com.sun.xml.internal.dtdparser;
aoqi@0: 
aoqi@0: import java.io.ByteArrayInputStream;
aoqi@0: import java.io.CharConversionException;
aoqi@0: import java.io.IOException;
aoqi@0: import java.io.InputStream;
aoqi@0: import java.io.InputStreamReader;
aoqi@0: import java.io.PushbackInputStream;
aoqi@0: import java.io.Reader;
aoqi@0: import java.util.Hashtable;
aoqi@0: 
aoqi@0: 
aoqi@0: // NOTE:  Add I18N support to this class when JDK gets the ability to
aoqi@0: // defer selection of locale for exception messages ... use the same
aoqi@0: // technique for both.
aoqi@0: 
aoqi@0: 
aoqi@0: /**
aoqi@0:  * This handles several XML-related tasks that normal java.io Readers
aoqi@0:  * don't support, inluding use of IETF standard encoding names and
aoqi@0:  * automatic detection of most XML encodings.  The former is needed
aoqi@0:  * for interoperability; the latter is needed to conform with the XML
aoqi@0:  * spec.  This class also optimizes reading some common encodings by
aoqi@0:  * providing low-overhead unsynchronized Reader support.
aoqi@0:  * <p/>
aoqi@0:  * <P> Note that the autodetection facility should be used only on
aoqi@0:  * data streams which have an unknown character encoding.  For example,
aoqi@0:  * it should never be used on MIME text/xml entities.
aoqi@0:  * <p/>
aoqi@0:  * <P> Note that XML processors are only required to support UTF-8 and
aoqi@0:  * UTF-16 character encodings.  Autodetection permits the underlying Java
aoqi@0:  * implementation to provide support for many other encodings, such as
aoqi@0:  * US-ASCII, ISO-8859-5, Shift_JIS, EUC-JP, and ISO-2022-JP.
aoqi@0:  *
aoqi@0:  * @author David Brownell
aoqi@0:  * @author Janet Koenig
aoqi@0:  * @version 1.3 00/02/24
aoqi@0:  */
aoqi@0: // package private
aoqi@0: final class XmlReader extends Reader {
aoqi@0:     private static final int MAXPUSHBACK = 512;
aoqi@0: 
aoqi@0:     private Reader in;
aoqi@0:     private String assignedEncoding;
aoqi@0:     private boolean closed;
aoqi@0: 
aoqi@0:     //
aoqi@0:     // This class always delegates I/O to a reader, which gets
aoqi@0:     // its data from the very beginning of the XML text.  It needs
aoqi@0:     // to use a pushback stream since (a) autodetection can read
aoqi@0:     // partial UTF-8 characters which need to be fully processed,
aoqi@0:     // (b) the "Unicode" readers swallow characters that they think
aoqi@0:     // are byte order marks, so tests fail if they don't see the
aoqi@0:     // real byte order mark.
aoqi@0:     //
aoqi@0:     // It's got do this efficiently:  character I/O is solidly on the
aoqi@0:     // critical path.  (So keep buffer length over 2 Kbytes to avoid
aoqi@0:     // excess buffering. Many URL handlers stuff a BufferedInputStream
aoqi@0:     // between here and the real data source, and larger buffers keep
aoqi@0:     // that from slowing you down.)
aoqi@0:     //
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Constructs the reader from an input stream, autodetecting
aoqi@0:      * the encoding to use according to the heuristic specified
aoqi@0:      * in the XML 1.0 recommendation.
aoqi@0:      *
aoqi@0:      * @param in the input stream from which the reader is constructed
aoqi@0:      * @throws IOException on error, such as unrecognized encoding
aoqi@0:      */
aoqi@0:     public static Reader createReader(InputStream in) throws IOException {
aoqi@0:         return new XmlReader(in);
aoqi@0:     }
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Creates a reader supporting the given encoding, mapping
aoqi@0:      * from standard encoding names to ones that understood by
aoqi@0:      * Java where necessary.
aoqi@0:      *
aoqi@0:      * @param in       the input stream from which the reader is constructed
aoqi@0:      * @param encoding the IETF standard name of the encoding to use;
aoqi@0:      *                 if null, autodetection is used.
aoqi@0:      * @throws IOException on error, including unrecognized encoding
aoqi@0:      */
aoqi@0:     public static Reader createReader(InputStream in, String encoding)
aoqi@0:             throws IOException {
aoqi@0:         if (encoding == null)
aoqi@0:             return new XmlReader(in);
aoqi@0:         if ("UTF-8".equalsIgnoreCase(encoding)
aoqi@0:                 || "UTF8".equalsIgnoreCase(encoding))
aoqi@0:             return new Utf8Reader(in);
aoqi@0:         if ("US-ASCII".equalsIgnoreCase(encoding)
aoqi@0:                 || "ASCII".equalsIgnoreCase(encoding))
aoqi@0:             return new AsciiReader(in);
aoqi@0:         if ("ISO-8859-1".equalsIgnoreCase(encoding)
aoqi@0:         // plus numerous aliases ...
aoqi@0:         )
aoqi@0:             return new Iso8859_1Reader(in);
aoqi@0: 
aoqi@0:         //
aoqi@0:         // What we really want is an administerable resource mapping
aoqi@0:         // encoding names/aliases to classnames.  For example a property
aoqi@0:         // file resource, "readers/mapping.props", holding and a set
aoqi@0:         // of readers in that (sub)package... defaulting to this call
aoqi@0:         // only if no better choice is available.
aoqi@0:         //
aoqi@0:         return new InputStreamReader(in, std2java(encoding));
aoqi@0:     }
aoqi@0: 
aoqi@0:     //
aoqi@0:     // JDK doesn't know all of the standard encoding names, and
aoqi@0:     // in particular none of the EBCDIC ones IANA defines (and
aoqi@0:     // which IBM encourages).
aoqi@0:     //
aoqi@0:     static private final Hashtable charsets = new Hashtable(31);
aoqi@0: 
aoqi@0:     static {
aoqi@0:         charsets.put("UTF-16", "Unicode");
aoqi@0:         charsets.put("ISO-10646-UCS-2", "Unicode");
aoqi@0: 
aoqi@0:         // NOTE: no support for ISO-10646-UCS-4 yet.
aoqi@0: 
aoqi@0:         charsets.put("EBCDIC-CP-US", "cp037");
aoqi@0:         charsets.put("EBCDIC-CP-CA", "cp037");
aoqi@0:         charsets.put("EBCDIC-CP-NL", "cp037");
aoqi@0:         charsets.put("EBCDIC-CP-WT", "cp037");
aoqi@0: 
aoqi@0:         charsets.put("EBCDIC-CP-DK", "cp277");
aoqi@0:         charsets.put("EBCDIC-CP-NO", "cp277");
aoqi@0:         charsets.put("EBCDIC-CP-FI", "cp278");
aoqi@0:         charsets.put("EBCDIC-CP-SE", "cp278");
aoqi@0: 
aoqi@0:         charsets.put("EBCDIC-CP-IT", "cp280");
aoqi@0:         charsets.put("EBCDIC-CP-ES", "cp284");
aoqi@0:         charsets.put("EBCDIC-CP-GB", "cp285");
aoqi@0:         charsets.put("EBCDIC-CP-FR", "cp297");
aoqi@0: 
aoqi@0:         charsets.put("EBCDIC-CP-AR1", "cp420");
aoqi@0:         charsets.put("EBCDIC-CP-HE", "cp424");
aoqi@0:         charsets.put("EBCDIC-CP-BE", "cp500");
aoqi@0:         charsets.put("EBCDIC-CP-CH", "cp500");
aoqi@0: 
aoqi@0:         charsets.put("EBCDIC-CP-ROECE", "cp870");
aoqi@0:         charsets.put("EBCDIC-CP-YU", "cp870");
aoqi@0:         charsets.put("EBCDIC-CP-IS", "cp871");
aoqi@0:         charsets.put("EBCDIC-CP-AR2", "cp918");
aoqi@0: 
aoqi@0:         // IANA also defines two that JDK 1.2 doesn't handle:
aoqi@0:         //    EBCDIC-CP-GR        --> CP423
aoqi@0:         //    EBCDIC-CP-TR        --> CP905
aoqi@0:     }
aoqi@0: 
aoqi@0:     // returns an encoding name supported by JDK >= 1.1.6
aoqi@0:     // for some cases required by the XML spec
aoqi@0:     private static String std2java(String encoding) {
aoqi@0:         String temp = encoding.toUpperCase();
aoqi@0:         temp = (String) charsets.get(temp);
aoqi@0:         return temp != null ? temp : encoding;
aoqi@0:     }
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Returns the standard name of the encoding in use
aoqi@0:      */
aoqi@0:     public String getEncoding() {
aoqi@0:         return assignedEncoding;
aoqi@0:     }
aoqi@0: 
aoqi@0:     private XmlReader(InputStream stream) throws IOException {
aoqi@0:         super(stream);
aoqi@0: 
aoqi@0:         PushbackInputStream pb;
aoqi@0:         byte buf [];
aoqi@0:         int len;
aoqi@0: 
aoqi@0:         if (stream instanceof PushbackInputStream)
aoqi@0:             pb = (PushbackInputStream) stream;
aoqi@0:         else
aoqi@0:             pb = new PushbackInputStream(stream, MAXPUSHBACK);
aoqi@0: 
aoqi@0:         //
aoqi@0:         // See if we can figure out the character encoding used
aoqi@0:         // in this file by peeking at the first few bytes.
aoqi@0:         //
aoqi@0:         buf = new byte[4];
aoqi@0:         len = pb.read(buf);
aoqi@0:         if (len > 0)
aoqi@0:             pb.unread(buf, 0, len);
aoqi@0: 
aoqi@0:         if (len == 4)
aoqi@0:             switch (buf[0] & 0x0ff) {
aoqi@0:             case 0:
aoqi@0:                 // 00 3c 00 3f == illegal UTF-16 big-endian
aoqi@0:                 if (buf[1] == 0x3c && buf[2] == 0x00 && buf[3] == 0x3f) {
aoqi@0:                     setEncoding(pb, "UnicodeBig");
aoqi@0:                     return;
aoqi@0:                 }
aoqi@0:                 // else it's probably UCS-4
aoqi@0:                 break;
aoqi@0: 
aoqi@0:             case '<':      // 0x3c: the most common cases!
aoqi@0:                 switch (buf[1] & 0x0ff) {
aoqi@0:                 // First character is '<'; could be XML without
aoqi@0:                 // an XML directive such as "<hello>", "<!-- ...",
aoqi@0:                 // and so on.
aoqi@0:                 default:
aoqi@0:                     break;
aoqi@0: 
aoqi@0:                     // 3c 00 3f 00 == illegal UTF-16 little endian
aoqi@0:                 case 0x00:
aoqi@0:                     if (buf[2] == 0x3f && buf[3] == 0x00) {
aoqi@0:                         setEncoding(pb, "UnicodeLittle");
aoqi@0:                         return;
aoqi@0:                     }
aoqi@0:                     // else probably UCS-4
aoqi@0:                     break;
aoqi@0: 
aoqi@0:                     // 3c 3f 78 6d == ASCII and supersets '<?xm'
aoqi@0:                 case '?':
aoqi@0:                     if (buf[2] != 'x' || buf[3] != 'm')
aoqi@0:                         break;
aoqi@0:                     //
aoqi@0:                     // One of several encodings could be used:
aoqi@0:                     // Shift-JIS, ASCII, UTF-8, ISO-8859-*, etc
aoqi@0:                     //
aoqi@0:                     useEncodingDecl(pb, "UTF8");
aoqi@0:                     return;
aoqi@0:                 }
aoqi@0:                 break;
aoqi@0: 
aoqi@0:                 // 4c 6f a7 94 ... some EBCDIC code page
aoqi@0:             case 0x4c:
aoqi@0:                 if (buf[1] == 0x6f
aoqi@0:                         && (0x0ff & buf[2]) == 0x0a7
aoqi@0:                         && (0x0ff & buf[3]) == 0x094) {
aoqi@0:                     useEncodingDecl(pb, "CP037");
aoqi@0:                     return;
aoqi@0:                 }
aoqi@0:                 // whoops, treat as UTF-8
aoqi@0:                 break;
aoqi@0: 
aoqi@0:                 // UTF-16 big-endian
aoqi@0:             case 0xfe:
aoqi@0:                 if ((buf[1] & 0x0ff) != 0xff)
aoqi@0:                     break;
aoqi@0:                 setEncoding(pb, "UTF-16");
aoqi@0:                 return;
aoqi@0: 
aoqi@0:                 // UTF-16 little-endian
aoqi@0:             case 0xff:
aoqi@0:                 if ((buf[1] & 0x0ff) != 0xfe)
aoqi@0:                     break;
aoqi@0:                 setEncoding(pb, "UTF-16");
aoqi@0:                 return;
aoqi@0: 
aoqi@0:                 // default ... no XML declaration
aoqi@0:             default:
aoqi@0:                 break;
aoqi@0:             }
aoqi@0: 
aoqi@0:         //
aoqi@0:         // If all else fails, assume XML without a declaration, and
aoqi@0:         // using UTF-8 encoding.
aoqi@0:         //
aoqi@0:         setEncoding(pb, "UTF-8");
aoqi@0:     }
aoqi@0: 
aoqi@0:     /*
aoqi@0:      * Read the encoding decl on the stream, knowing that it should
aoqi@0:      * be readable using the specified encoding (basically, ASCII or
aoqi@0:      * EBCDIC).  The body of the document may use a wider range of
aoqi@0:      * characters than the XML/Text decl itself, so we switch to use
aoqi@0:      * the specified encoding as soon as we can.  (ASCII is a subset
aoqi@0:      * of UTF-8, ISO-8859-*, ISO-2022-JP, EUC-JP, and more; EBCDIC
aoqi@0:      * has a variety of "code pages" that have these characters as
aoqi@0:      * a common subset.)
aoqi@0:      */
aoqi@0:     private void useEncodingDecl(PushbackInputStream pb, String encoding)
aoqi@0:             throws IOException {
aoqi@0:         byte buffer [] = new byte[MAXPUSHBACK];
aoqi@0:         int len;
aoqi@0:         Reader r;
aoqi@0:         int c;
aoqi@0: 
aoqi@0:         //
aoqi@0:         // Buffer up a bunch of input, and set up to read it in
aoqi@0:         // the specified encoding ... we can skip the first four
aoqi@0:         // bytes since we know that "<?xm" was read to determine
aoqi@0:         // what encoding to use!
aoqi@0:         //
aoqi@0:         len = pb.read(buffer, 0, buffer.length);
aoqi@0:         pb.unread(buffer, 0, len);
aoqi@0:         r = new InputStreamReader(new ByteArrayInputStream(buffer, 4, len),
aoqi@0:                 encoding);
aoqi@0: 
aoqi@0:         //
aoqi@0:         // Next must be "l" (and whitespace) else we conclude
aoqi@0:         // error and choose UTF-8.
aoqi@0:         //
aoqi@0:         if ((c = r.read()) != 'l') {
aoqi@0:             setEncoding(pb, "UTF-8");
aoqi@0:             return;
aoqi@0:         }
aoqi@0: 
aoqi@0:         //
aoqi@0:         // Then, we'll skip any
aoqi@0:         //     S version="..."     [or single quotes]
aoqi@0:         // bit and get any subsequent
aoqi@0:         //     S encoding="..."     [or single quotes]
aoqi@0:         //
aoqi@0:         // We put an arbitrary size limit on how far we read; lots
aoqi@0:         // of space will break this algorithm.
aoqi@0:         //
aoqi@0:         StringBuffer buf = new StringBuffer();
aoqi@0:         StringBuffer keyBuf = null;
aoqi@0:         String key = null;
aoqi@0:         boolean sawEq = false;
aoqi@0:         char quoteChar = 0;
aoqi@0:         boolean sawQuestion = false;
aoqi@0: 
aoqi@0:         XmlDecl:
aoqi@0:         for (int i = 0; i < MAXPUSHBACK - 5; ++i) {
aoqi@0:             if ((c = r.read()) == -1)
aoqi@0:                 break;
aoqi@0: 
aoqi@0:             // ignore whitespace before/between "key = 'value'"
aoqi@0:             if (c == ' ' || c == '\t' || c == '\n' || c == '\r')
aoqi@0:                 continue;
aoqi@0: 
aoqi@0:             // ... but require at least a little!
aoqi@0:             if (i == 0)
aoqi@0:                 break;
aoqi@0: 
aoqi@0:             // terminate the loop ASAP
aoqi@0:             if (c == '?')
aoqi@0:                 sawQuestion = true;
aoqi@0:             else if (sawQuestion) {
aoqi@0:                 if (c == '>')
aoqi@0:                     break;
aoqi@0:                 sawQuestion = false;
aoqi@0:             }
aoqi@0: 
aoqi@0:             // did we get the "key =" bit yet?
aoqi@0:             if (key == null || !sawEq) {
aoqi@0:                 if (keyBuf == null) {
aoqi@0:                     if (Character.isWhitespace((char) c))
aoqi@0:                         continue;
aoqi@0:                     keyBuf = buf;
aoqi@0:                     buf.setLength(0);
aoqi@0:                     buf.append((char) c);
aoqi@0:                     sawEq = false;
aoqi@0:                 } else if (Character.isWhitespace((char) c)) {
aoqi@0:                     key = keyBuf.toString();
aoqi@0:                 } else if (c == '=') {
aoqi@0:                     if (key == null)
aoqi@0:                         key = keyBuf.toString();
aoqi@0:                     sawEq = true;
aoqi@0:                     keyBuf = null;
aoqi@0:                     quoteChar = 0;
aoqi@0:                 } else
aoqi@0:                     keyBuf.append((char) c);
aoqi@0:                 continue;
aoqi@0:             }
aoqi@0: 
aoqi@0:             // space before quoted value
aoqi@0:             if (Character.isWhitespace((char) c))
aoqi@0:                 continue;
aoqi@0:             if (c == '"' || c == '\'') {
aoqi@0:                 if (quoteChar == 0) {
aoqi@0:                     quoteChar = (char) c;
aoqi@0:                     buf.setLength(0);
aoqi@0:                     continue;
aoqi@0:                 } else if (c == quoteChar) {
aoqi@0:                     if ("encoding".equals(key)) {
aoqi@0:                         assignedEncoding = buf.toString();
aoqi@0: 
aoqi@0:                         // [81] Encname ::= [A-Za-z] ([A-Za-z0-9._]|'-')*
aoqi@0:                         for (i = 0; i < assignedEncoding.length(); i++) {
aoqi@0:                             c = assignedEncoding.charAt(i);
aoqi@0:                             if ((c >= 'A' && c <= 'Z')
aoqi@0:                                     || (c >= 'a' && c <= 'z'))
aoqi@0:                                 continue;
aoqi@0:                             if (i == 0)
aoqi@0:                                 break XmlDecl;
aoqi@0:                             if (i > 0 && (c == '-'
aoqi@0:                                     || (c >= '0' && c <= '9')
aoqi@0:                                     || c == '.' || c == '_'))
aoqi@0:                                 continue;
aoqi@0:                             // map illegal names to UTF-8 default
aoqi@0:                             break XmlDecl;
aoqi@0:                         }
aoqi@0: 
aoqi@0:                         setEncoding(pb, assignedEncoding);
aoqi@0:                         return;
aoqi@0: 
aoqi@0:                     } else {
aoqi@0:                         key = null;
aoqi@0:                         continue;
aoqi@0:                     }
aoqi@0:                 }
aoqi@0:             }
aoqi@0:             buf.append((char) c);
aoqi@0:         }
aoqi@0: 
aoqi@0:         setEncoding(pb, "UTF-8");
aoqi@0:     }
aoqi@0: 
aoqi@0:     private void setEncoding(InputStream stream, String encoding)
aoqi@0:             throws IOException {
aoqi@0:         assignedEncoding = encoding;
aoqi@0:         in = createReader(stream, encoding);
aoqi@0:     }
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Reads the number of characters read into the buffer, or -1 on EOF.
aoqi@0:      */
aoqi@0:     public int read(char buf [], int off, int len) throws IOException {
aoqi@0:         int val;
aoqi@0: 
aoqi@0:         if (closed)
aoqi@0:             return -1;        // throw new IOException ("closed");
aoqi@0:         val = in.read(buf, off, len);
aoqi@0:         if (val == -1)
aoqi@0:             close();
aoqi@0:         return val;
aoqi@0:     }
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Reads a single character.
aoqi@0:      */
aoqi@0:     public int read() throws IOException {
aoqi@0:         int val;
aoqi@0: 
aoqi@0:         if (closed)
aoqi@0:             throw new IOException("closed");
aoqi@0:         val = in.read();
aoqi@0:         if (val == -1)
aoqi@0:             close();
aoqi@0:         return val;
aoqi@0:     }
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Returns true iff the reader supports mark/reset.
aoqi@0:      */
aoqi@0:     public boolean markSupported() {
aoqi@0:         return in == null ? false : in.markSupported();
aoqi@0:     }
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Sets a mark allowing a limited number of characters to
aoqi@0:      * be "peeked", by reading and then resetting.
aoqi@0:      *
aoqi@0:      * @param value how many characters may be "peeked".
aoqi@0:      */
aoqi@0:     public void mark(int value) throws IOException {
aoqi@0:         if (in != null) in.mark(value);
aoqi@0:     }
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Resets the current position to the last marked position.
aoqi@0:      */
aoqi@0:     public void reset() throws IOException {
aoqi@0:         if (in != null) in.reset();
aoqi@0:     }
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Skips a specified number of characters.
aoqi@0:      */
aoqi@0:     public long skip(long value) throws IOException {
aoqi@0:         return in == null ? 0 : in.skip(value);
aoqi@0:     }
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Returns true iff input characters are known to be ready.
aoqi@0:      */
aoqi@0:     public boolean ready() throws IOException {
aoqi@0:         return in == null ? false : in.ready();
aoqi@0:     }
aoqi@0: 
aoqi@0:     /**
aoqi@0:      * Closes the reader.
aoqi@0:      */
aoqi@0:     public void close() throws IOException {
aoqi@0:         if (closed)
aoqi@0:             return;
aoqi@0:         in.close();
aoqi@0:         in = null;
aoqi@0:         closed = true;
aoqi@0:     }
aoqi@0: 
aoqi@0:     //
aoqi@0:     // Delegating to a converter module will always be slower than
aoqi@0:     // direct conversion.  Use a similar approach for any other
aoqi@0:     // readers that need to be particularly fast; only block I/O
aoqi@0:     // speed matters to this package.  For UTF-16, separate readers
aoqi@0:     // for big and little endian streams make a difference, too;
aoqi@0:     // fewer conditionals in the critical path!
aoqi@0:     //
aoqi@0:     static abstract class BaseReader extends Reader {
aoqi@0:         protected InputStream instream;
aoqi@0:         protected byte buffer [];
aoqi@0:         protected int start, finish;
aoqi@0: 
aoqi@0:         BaseReader(InputStream stream) {
aoqi@0:             super(stream);
aoqi@0: 
aoqi@0:             instream = stream;
aoqi@0:             buffer = new byte[8192];
aoqi@0:         }
aoqi@0: 
aoqi@0:         public boolean ready() throws IOException {
aoqi@0:             return instream == null
aoqi@0:                     || (finish - start) > 0
aoqi@0:                     || instream.available() != 0;
aoqi@0:         }
aoqi@0: 
aoqi@0:         // caller shouldn't read again
aoqi@0:         public void close() throws IOException {
aoqi@0:             if (instream != null) {
aoqi@0:                 instream.close();
aoqi@0:                 start = finish = 0;
aoqi@0:                 buffer = null;
aoqi@0:                 instream = null;
aoqi@0:             }
aoqi@0:         }
aoqi@0:     }
aoqi@0: 
aoqi@0:     //
aoqi@0:     // We want this reader, to make the default encoding be as fast
aoqi@0:     // as we can make it.  JDK's "UTF8" (not "UTF-8" till JDK 1.2)
aoqi@0:     // InputStreamReader works, but 20+% slower speed isn't OK for
aoqi@0:     // the default/primary encoding.
aoqi@0:     //
aoqi@0:     static final class Utf8Reader extends BaseReader {
aoqi@0:         // 2nd half of UTF-8 surrogate pair
aoqi@0:         private char nextChar;
aoqi@0: 
aoqi@0:         Utf8Reader(InputStream stream) {
aoqi@0:             super(stream);
aoqi@0:         }
aoqi@0: 
aoqi@0:         public int read(char buf [], int offset, int len) throws IOException {
aoqi@0:             int i = 0, c = 0;
aoqi@0: 
aoqi@0:             if (len <= 0)
aoqi@0:                 return 0;
aoqi@0: 
aoqi@0:             // Consume remaining half of any surrogate pair immediately
aoqi@0:             if (nextChar != 0) {
aoqi@0:                 buf[offset + i++] = nextChar;
aoqi@0:                 nextChar = 0;
aoqi@0:             }
aoqi@0: 
aoqi@0:             while (i < len) {
aoqi@0:                 // stop or read data if needed
aoqi@0:                 if (finish <= start) {
aoqi@0:                     if (instream == null) {
aoqi@0:                         c = -1;
aoqi@0:                         break;
aoqi@0:                     }
aoqi@0:                     start = 0;
aoqi@0:                     finish = instream.read(buffer, 0, buffer.length);
aoqi@0:                     if (finish <= 0) {
aoqi@0:                         this.close();
aoqi@0:                         c = -1;
aoqi@0:                         break;
aoqi@0:                     }
aoqi@0:                 }
aoqi@0: 
aoqi@0:                 //
aoqi@0:                 // RFC 2279 describes UTF-8; there are six encodings.
aoqi@0:                 // Each encoding takes a fixed number of characters
aoqi@0:                 // (1-6 bytes) and is flagged by a bit pattern in the
aoqi@0:                 // first byte.  The five and six byte-per-character
aoqi@0:                 // encodings address characters which are disallowed
aoqi@0:                 // in XML documents, as do some four byte ones.
aoqi@0:                 //
aoqi@0: 
aoqi@0:                 //
aoqi@0:                 // Single byte == ASCII.  Common; optimize.
aoqi@0:                 //
aoqi@0:                 c = buffer[start] & 0x0ff;
aoqi@0:                 if ((c & 0x80) == 0x00) {
aoqi@0:                     // 0x0000 <= c <= 0x007f
aoqi@0:                     start++;
aoqi@0:                     buf[offset + i++] = (char) c;
aoqi@0:                     continue;
aoqi@0:                 }
aoqi@0: 
aoqi@0:                 //
aoqi@0:                 // Multibyte chars -- check offsets optimistically,
aoqi@0:                 // ditto the "10xx xxxx" format for subsequent bytes
aoqi@0:                 //
aoqi@0:                 int off = start;
aoqi@0: 
aoqi@0:                 try {
aoqi@0:                     // 2 bytes
aoqi@0:                     if ((buffer[off] & 0x0E0) == 0x0C0) {
aoqi@0:                         c = (buffer[off++] & 0x1f) << 6;
aoqi@0:                         c += buffer[off++] & 0x3f;
aoqi@0: 
aoqi@0:                         // 0x0080 <= c <= 0x07ff
aoqi@0: 
aoqi@0:                         // 3 bytes
aoqi@0:                     } else if ((buffer[off] & 0x0F0) == 0x0E0) {
aoqi@0:                         c = (buffer[off++] & 0x0f) << 12;
aoqi@0:                         c += (buffer[off++] & 0x3f) << 6;
aoqi@0:                         c += buffer[off++] & 0x3f;
aoqi@0: 
aoqi@0:                         // 0x0800 <= c <= 0xffff
aoqi@0: 
aoqi@0:                         // 4 bytes
aoqi@0:                     } else if ((buffer[off] & 0x0f8) == 0x0F0) {
aoqi@0:                         c = (buffer[off++] & 0x07) << 18;
aoqi@0:                         c += (buffer[off++] & 0x3f) << 12;
aoqi@0:                         c += (buffer[off++] & 0x3f) << 6;
aoqi@0:                         c += buffer[off++] & 0x3f;
aoqi@0: 
aoqi@0:                         // 0x0001 0000  <= c  <= 0x001f ffff
aoqi@0: 
aoqi@0:                         // Unicode supports c <= 0x0010 ffff ...
aoqi@0:                         if (c > 0x0010ffff)
aoqi@0:                             throw new CharConversionException("UTF-8 encoding of character 0x00"
aoqi@0:                                     + Integer.toHexString(c)
aoqi@0:                                     + " can't be converted to Unicode.");
aoqi@0: 
aoqi@0:                         // Convert UCS-4 char to surrogate pair (UTF-16)
aoqi@0:                         c -= 0x10000;
aoqi@0:                         nextChar = (char) (0xDC00 + (c & 0x03ff));
aoqi@0:                         c = 0xD800 + (c >> 10);
aoqi@0: 
aoqi@0:                         // 5 and 6 byte versions are XML WF errors, but
aoqi@0:                         // typically come from mislabeled encodings
aoqi@0:                     } else
aoqi@0:                         throw new CharConversionException("Unconvertible UTF-8 character"
aoqi@0:                                 + " beginning with 0x"
aoqi@0:                                 + Integer.toHexString(buffer[start] & 0xff));
aoqi@0: 
aoqi@0:                 } catch (ArrayIndexOutOfBoundsException e) {
aoqi@0:                     // off > length && length >= buffer.length
aoqi@0:                     c = 0;
aoqi@0:                 }
aoqi@0: 
aoqi@0:                 //
aoqi@0:                 // if the buffer held only a partial character,
aoqi@0:                 // compact it and try to read the rest of the
aoqi@0:                 // character.  worst case involves three
aoqi@0:                 // single-byte reads -- quite rare.
aoqi@0:                 //
aoqi@0:                 if (off > finish) {
aoqi@0:                     System.arraycopy(buffer, start,
aoqi@0:                             buffer, 0, finish - start);
aoqi@0:                     finish -= start;
aoqi@0:                     start = 0;
aoqi@0:                     off = instream.read(buffer, finish,
aoqi@0:                             buffer.length - finish);
aoqi@0:                     if (off < 0) {
aoqi@0:                         this.close();
aoqi@0:                         throw new CharConversionException("Partial UTF-8 char");
aoqi@0:                     }
aoqi@0:                     finish += off;
aoqi@0:                     continue;
aoqi@0:                 }
aoqi@0: 
aoqi@0:                 //
aoqi@0:                 // check the format of the non-initial bytes
aoqi@0:                 //
aoqi@0:                 for (start++; start < off; start++) {
aoqi@0:                     if ((buffer[start] & 0xC0) != 0x80) {
aoqi@0:                         this.close();
aoqi@0:                         throw new CharConversionException("Malformed UTF-8 char -- "
aoqi@0:                                 + "is an XML encoding declaration missing?");
aoqi@0:                     }
aoqi@0:                 }
aoqi@0: 
aoqi@0:                 //
aoqi@0:                 // If this needed a surrogate pair, consume ASAP
aoqi@0:                 //
aoqi@0:                 buf[offset + i++] = (char) c;
aoqi@0:                 if (nextChar != 0 && i < len) {
aoqi@0:                     buf[offset + i++] = nextChar;
aoqi@0:                     nextChar = 0;
aoqi@0:                 }
aoqi@0:             }
aoqi@0:             if (i > 0)
aoqi@0:                 return i;
aoqi@0:             return (c == -1) ? -1 : 0;
aoqi@0:         }
aoqi@0:     }
aoqi@0: 
aoqi@0:     //
aoqi@0:     // We want ASCII and ISO-8859 Readers since they're the most common
aoqi@0:     // encodings in the US and Europe, and we don't want performance
aoqi@0:     // regressions for them.  They're also easy to implement efficiently,
aoqi@0:     // since they're bitmask subsets of UNICODE.
aoqi@0:     //
aoqi@0:     // XXX haven't benchmarked these readers vs what we get out of JDK.
aoqi@0:     //
aoqi@0:     static final class AsciiReader extends BaseReader {
aoqi@0:         AsciiReader(InputStream in) {
aoqi@0:             super(in);
aoqi@0:         }
aoqi@0: 
aoqi@0:         public int read(char buf [], int offset, int len) throws IOException {
aoqi@0:             int i, c;
aoqi@0: 
aoqi@0:             if (instream == null)
aoqi@0:                 return -1;
aoqi@0: 
aoqi@0:             for (i = 0; i < len; i++) {
aoqi@0:                 if (start >= finish) {
aoqi@0:                     start = 0;
aoqi@0:                     finish = instream.read(buffer, 0, buffer.length);
aoqi@0:                     if (finish <= 0) {
aoqi@0:                         if (finish <= 0)
aoqi@0:                             this.close();
aoqi@0:                         break;
aoqi@0:                     }
aoqi@0:                 }
aoqi@0:                 c = buffer[start++];
aoqi@0:                 if ((c & 0x80) != 0)
aoqi@0:                     throw new CharConversionException("Illegal ASCII character, 0x"
aoqi@0:                             + Integer.toHexString(c & 0xff));
aoqi@0:                 buf[offset + i] = (char) c;
aoqi@0:             }
aoqi@0:             if (i == 0 && finish <= 0)
aoqi@0:                 return -1;
aoqi@0:             return i;
aoqi@0:         }
aoqi@0:     }
aoqi@0: 
aoqi@0:     static final class Iso8859_1Reader extends BaseReader {
aoqi@0:         Iso8859_1Reader(InputStream in) {
aoqi@0:             super(in);
aoqi@0:         }
aoqi@0: 
aoqi@0:         public int read(char buf [], int offset, int len) throws IOException {
aoqi@0:             int i;
aoqi@0: 
aoqi@0:             if (instream == null)
aoqi@0:                 return -1;
aoqi@0: 
aoqi@0:             for (i = 0; i < len; i++) {
aoqi@0:                 if (start >= finish) {
aoqi@0:                     start = 0;
aoqi@0:                     finish = instream.read(buffer, 0, buffer.length);
aoqi@0:                     if (finish <= 0) {
aoqi@0:                         if (finish <= 0)
aoqi@0:                             this.close();
aoqi@0:                         break;
aoqi@0:                     }
aoqi@0:                 }
aoqi@0:                 buf[offset + i] = (char) (0x0ff & buffer[start++]);
aoqi@0:             }
aoqi@0:             if (i == 0 && finish <= 0)
aoqi@0:                 return -1;
aoqi@0:             return i;
aoqi@0:         }
aoqi@0:     }
aoqi@0: }