Character streams in JDK 1.1

Version 1.1 of the Java Development Kit introduces support for character streams to the package.

Prior to this release, the standard I/O facilities supported only byte streams, via the InputStream and OutputStream classes and their subclasses. Character streams are like byte streams, but they contain 16-bit Unicode characters rather than eight-bit bytes. They are implemented by the Reader and Writer classes and their subclasses. Readers and Writers support essentially the same operations as InputStreams and OutputStreams, except that where byte-stream methods operate on bytes or byte arrays, character-stream methods operate on characters, character arrays, or strings.

Most of the functionality available for byte streams is also provided for character streams. This is reflected in the name of each character-stream class, whose prefix is usually shared with the name of the corresponding byte-stream class. For example, there is a PushbackReader class that provides the same functionality for character streams that is provided by PushbackInputStream for byte streams.

Why use character streams?

The primary advantage of character streams is that they make it easy to write programs that are not dependent upon a specific character encoding, and are therefore easy to internationalize.

Java stores strings in Unicode, an international standard character encoding that is capable of representing most of the world's written languages. Typical user-readable text files, however, use encodings that are not necessarily related to Unicode, or even to ASCII, and there are many such encodings. Character streams hide the complexity of dealing with these encodings by providing two classes that serve as bridges between byte streams and character streams. The InputStreamReader class implements a character-input stream that reads bytes from a byte-input stream and converts them to characters according to a specified encoding. Similarly, the OutputStreamWriter class implements a character-output stream that converts characters into bytes according a specified encoding and writes them to a byte-output stream.

A second advantage of character streams is that they are potentially much more efficient than byte streams. The implementations of many of Java's original byte streams are oriented around byte-at-a-time read and write operations. The character-stream classes, in contrast, are oriented around buffer-at-a-time read and write operations. This difference, in combination with a more efficient locking scheme, allows the character stream classes to make up for the added overhead of encoding conversion in many cases.

API overview

The new character-stream classes have been designed to parallel the existing byte-stream classes in the package. As noted above, the name of each character-stream class ends in Reader or Writer, as appropriate, while its prefix is usually shared with the corresponding byte-stream class, if any. The following table summarizes the new classes; in the left column, indentation indicates subclass relationships.

Character-stream class Description Corresponding byte class
Reader Abstract class for character-input streams InputStream
    BufferedReader Buffers input, parses lines BufferedInputStream
        LineNumberReader     Keeps track of line numbers LineNumberInputStream
    CharArrayReader Reads from a character array ByteArrayInputStream
    InputStreamReader Translates a byte stream into a character stream (none)
        FileReader Translates bytes from a file into a character stream     FileInputStream
    FilterReader Abstract class for filtered character input FilterInputStream
        PushbackReader Allows characters to be pushed back PushbackInputStream
    PipedReader Reads from a PipedWriter PipedInputStream
    StringReader Reads from a String StringBufferInputStream
Writer Abstract class for character-output streams OutputStream
    BufferedWriter Buffers output, uses platform's line separator BufferedOutputStream
    CharArrayWriter Writes to a character array ByteArrayOutputStream
    FilterWriter Abstract class for filtered character output FilterOutputStream
    OutputStreamWriter Translates a character stream into a byte stream (none)
        FileWriter Translates a character stream into a byte file FileOutputStream
    PrintWriter Prints values and objects to a Writer PrintStream
    PipedWriter Writes to a PipedReader PipedOutputStream
    StringWriter Writes to a String (none)

Related changes


The PrintStream class has been modified to use the platform's default character encoding and the platform's default line terminator. Thus each PrintStream incorporates an OutputStreamWriter, and it passes all characters through this writer to produce bytes for output. The println methods use the platform's default line terminator, which is defined by the system property line.separator and is not necessarily a single newline character ('\n'). Bytes and byte arrays written via the existing write methods are not passed through the writer.

The primary motivation for changing the PrintStream class is that it will make System.out and System.err more useful to people writing Java programs on platforms where the local encoding is something other than ASCII. PrintStream is, in other words, provided primarily for use in debugging and for compatibility with existing code. Therefore its two constructors have been deprecated:

   PrintStream(OutputStream, boolean)

Deprecating the constructors rather than the entire class allows existing uses of System.out and System.err to be compiled without generating deprecation warnings. Thus programmers just learning Java, or programmers inserting System.err.println calls for debugging purposes, will not be bothered by such warnings. Programmers writing code that explicitly constructs a PrintStream, on the other hand, will see a deprecation warning when that code is compiled. Code that produces textual output should use the new PrintWriter class, which allows the character encoding to be specified or the default encoding to be accepted. For convenience, the PrintWriter class provides constructors that take an OutputStream object and create an intermediate OutputStreamWriter object that uses the default encoding.

Other classes

Minor changes have been made to several existing classes. The following constructors and methods have been deprecated because they do not properly convert between bytes and characters:

String   DataInputStream.readLine()
InputStream   Runtime.getLocalizedInputStream(InputStream)
OutputStream   Runtime.getLocalizedOutputStream(OutputStream)
   String(byte ascii[], int hibyte, int offset, int count)
   String(byte ascii[], int hibyte)
void   String.getBytes(int srcBegin, int srcEnd, byte dst[], int dstBegin)

Finally, the following constructor and methods have been added:

byte[]   String.getBytes()
void   Throwable.printStackTrace(PrintWriter)

Comments to: Mark Reinhold / 4 February 1997