Purpose of Java Filter Stream Classes

There are several stream classes in the java.io package that perform different kinds of filtering. You can also write your own classes to do custom filtering. Furthermore, you can chain multiple filters together to get the effect of a combined filter.

The BufferedInputStream and BufferedOutputStream classes buffer reads and writes by reading or writing data first into a buffer.
Thus an application can read or write bytes to the stream without necessarily calling the underlying native method. The data is read from or written into the buffer in blocks; subsequent accesses go straight to the buffer.
This improves performance in many situations. BufferedInputStream also allows the reader to back up and reread data that's already been read.

public class FilterInputStream extends InputStream
public class FilterOutputStream extends OutputStream

protected FilterInputStream(InputStream in)
protected FilterOutputStream(OutputStream out)

protected InputStream in
protected OutputStream out

Since the constructors are protected, filter streams may only be created by subclasses. Each subclass implements a particular filtering operation. Normally, such a pattern suggests that polymorphism is going to be used heavily, with subclasses standing in for the common superclass; however, it is uncommon to use filter streams polymorphically as instances of FilterInputStream or FilterOutputStream. Most of the time, references to a filter stream are either references to a more specific subclass like BufferedInputStream or they are polymorphic references to InputStream or OutputStream with no hint of the filter left.

The DataInputStream and DataOutputStream classes read and write primitive Java data types and strings in a machine independent way.

The DataInputStream and DataOutputStream classes read and write primitive Java data types and Strings in a machine-independent way.
There are three main data formats:

1. Big-endian for integer types
2. IEEE-754 for floats and doubles
3. UTF-8 for Unicode

Big-endian data stores the bytes that make up an integer starting with the most significant byte and counting down to the least-significant byte.
Little-endian architectures store the least-significant byte first with the most-significant byte coming last.
In a big-endian decimal system (as opposed to a big-endian binary system like computers use) the number 12 means (1 * 10) + (2 * 1). In a little-endian system, 12 means (1 * 1) + (2 * 10), or 21. Big-endian format is used in the Java Virtual Machine and in almost every modern CPU except Intel's X86 family.

IEEE-754 is an international standard for the representation of single- and double- precision floating point numbers. It is used by many modern computer architectures, especially ones aimed at engineering and scientific markets. Again the notable exception is the Intel X86 family. In fact, Microsoft has complained about the requirement of IEEE-754 math in Java. It makes Java slower on X86 Windows than it could be because floating point arithmetic has to be emulated in software rather than sent directly to the FPU.

UTF-8 is a compressed form of Unicode that uses only one byte for the ASCII characters, two bytes for the most common non-ASCII Unicode characters, and three bytes for the less-common Unicode characters.
In practice, UTF-8 is much more space-efficient than pure Unicode, especially when working with English text.

You have already encountered the java.io.PrintStream class. It allows very simple printing of primitive values, objects, and string literals. It uses the platform's default character encoding to convert characters into bytes. This class traps all IOExceptions and is primarily intended for debugging. System.out and System.err are instances of the PrintStream class, but you can connect a PrintStream filter to other output streams as well. For example, you can chain a PrintStream to a FileOutputStream to easily write text into the file.

The PushbackInputStream class has a one-byte pushback buffer so a program can unread the last character read. The next time data is read from the stream, the unread character is reread.