The core Java language does not have any I/O methods. For a program to do I/O, it must import an I/O package. These notes (and most programs) use the package java.io.

Data for a program may come from several sources. Data created by a program may be sent to several destinations. The connection between a program and a data source or destination is called a stream.

An input stream handles data flowing into a program. An output stream handles data flowing out of a program.

In the picture, each "O" is a piece of data. The data are streaming from the source into the program. Computed results are streaming from the program to the destination.

There are many types of I/O devices. Some can be a source, others can be a destination, and others can be both source and destination. Often the destination for one stream is the source for another. In the table, decide what role each object can play.

Some objects switch roles depending on what program is running. For example a disk file might be the destination for the output of one program, and later it may be the source for another program.

Character streams are intended exclusively for character data.

Byte streams are intended for general purpose input and output.

Of course, fundamentally all data consists of patterns of bits grouped into 8-bit bytes. So, logically all streams could be called "byte streams". However streams that are intended for bytes that represent characters are called "character streams" and all others are called "byte streams".

Character streams are optimized for character data and perform some other useful character-oriented tasks (more on this later). Often the source or destination of a character stream is a text file, a file that contains bytes that represent characters.

So far in these chapters, programs have done I/O only with character data (usually with the keyboard as the source and the monitor as the destination). However, data sources and destinations often contain non-character data. For example, the bytecode file created by the Java compiler contains machine instructions for the Java virtual machine. These are not intended to represent characters, and input and output of them must use byte streams.

A stream object may be:

• An input stream or an output stream,
• a processing stream or an ordinary stream,
• a character-oriented stream or a byte-oriented stream,
• and may be connected to a variety of sources or destinations.

There are many types of streams. For example, a stream might be an "input, character-oriented, processing stream". Another stream might be an "byte-oriented output stream". Not only are there many types of streams, there are many ways to connect them together. The I/O aspects of a program can get complicated.

This situation is common to all industrial-strength programming languages. I/O is a big topic because of the wide variety of IO devices and the wide variety of data formats. A production language must be able to read data from sources written by any language running on any type of computer.

For example, a program might need to read data from magnetic tapes written 30 years ago by a COBOL program running on an IBM mainframe. (You would likely use byte-oriented streams for this).

Fortunately, if a program is written in Java it will use identical data formats on all computers. If a Java program running on a Macintosh writes a text file, that file can easily be read by a Java program running on a PC or on any other computer.

This is not true of most other languages. A C program running on one type of computer is unlikely to create a file that can easily be read by a C program running on another type of computer. Worse than that, two different C compilers for the same computer might use different data formats!

The diagram shows the top of the hierarchy for the java.io package. The dotted clouds are abstract classes. They act as base classes for specialized streams (to be discussed shortly).

Streams are byte-oriented or character-oriented. Each type has input streams and output streams.

1. Byte-oriented streams.
   • Intended for general-purpose input and output.
   • Data may be primitive data types or raw bytes.
2. Character-oriented streams.
   • Intended for character data.
   • Data is transformed from/to 16 bit Java char used inside programs to the UTF format used externally.

In the diagram, abstract classes are drawn as clouds with a dotted border.

Readers and Writers deal with character streams. These are abstract classes. A program must use classes derived from them. For example, a BufferedReader is a Reader.

Character streams are optimized for handling character data. They also translate between the internal format used by Java programs and an external format used for text files. Inside a Java program character data is represented with the 16-bit char data type. The characters of a String also use this 16-bit code. On a disk file, characters are represented in a format called UTF. This format uses one to four bytes per character and is intended to be a universal format—one format for all text files in any language anywhere in the world.

UTF stands for "Unicode Transformation Format". Usually a UTF text file is identical to an ASCII text file. ASCII is the standard way to represent characters that most computers have used for the past forty years. A file created with a text editor is usually an ASCII text file.

A UTF text file can include non-ASCII characters such as Cyrillic, Greek, and Asian characters. By reading and writing UTF files, Java programs can process text from any of the World's languages.

Readers and Writers deal with character streams. These are abstract classes. A program must use classes derived from them. For example, a BufferedReader is a Reader.

Character streams are optimized for handling character data. They also translate between the internal format used by Java programs and an external format used for text files. Inside a Java program character data is represented with the 16-bit char data type. The characters of a String also use this 16-bit code. On a disk file, characters are represented in a format called UTF. This format uses one to four bytes per character and is intended to be a universal format—one format for all text files in any language anywhere in the world.

UTF stands for "Unicode Transformation Format". Usually a UTF text file is identical to an ASCII text file. ASCII is the standard way to represent characters that most computers have used for the past forty years. A file created with a text editor is usually an ASCII text file.

A UTF text file can include non-ASCII characters such as Cyrillic, Greek, and Asian characters. By reading and writing UTF files, Java programs can process text from any of the World's languages.

Class FileReader is discussed in the following chapters. But now let us look at the classes that descend from Writer, an abstract class from which all character-oriented output streams are derived. All these streams are aimed at receiving 16-bit char data from a program, and sending it to another destination, which may use a different character format (such as UTF format on a disk file).

All these classes are character-oriented output streams. We will be especially interested in FileWriter and PrintWriter classes.

InputStream is an abstract class from which all byte-oriented input streams are derived. Its descendant classes are used for general-purpose input (non-character input). These streams deliver data to a program in groups of 8-bit bytes. The bytes can be grouped into the size necessary for the type of data.

For example, if a disk file contains 32-bit int data, data can be delivered to the program in 4-byte groups in the same format as Java primitive type int.

We will be mostly concerned with DataInputStream and FileInputStream.