The APIs of NIO were designed to provide access to the low-level I/O operations of modern operating systems. Although the APIs are themselves relatively high-level, the intent is to facilitate an implementation that can directly use the most efficient operations of the underlying platform.
The Java NIO APIs are provided in the package and its subpackages. The documentation by Sun Microsystems identifies these features.
Integer.MAX_SIZE bytes (2 GB)NIO data transfer is based on buffers (and related classes). These classes represent a contiguous extent of memory, together with a small number of data transfer operations. Although theoretically these are general-purpose data structures, the implementation may select memory for alignment or paging characteristics, which are not otherwise accessible in Java. Typically, this would be used to allow the buffer contents to occupy the same physical memory used by the underlying operating system for its native I/O operations, thus allowing the most direct transfer mechanism, and eliminating the need for any additional copying. In most operating systems, provided the particular area of memory has the right properties, transfer can take place without using the CPU at all. The NIO buffer is intentionally limited in features in order to support these goals.
There are buffer classes for all of Java's primitive types except boolean, which can share memory with byte buffers and allow arbitrary interpretation of the underlying bytes.
flip() method, rather than performing a "flip" or paging function in the canonical sense, moves the position pointer to the origin of the underlying array (if any) and the limit pointer to the former position of the position pointer.get() methods are supplied for transferring data out of a NIO buffer. The bulk implementation, rather than performing a "get" in the traditional sense, "puts" the data into a specified array. The "offset" argument supplied to this method refers not to the offset from within the buffer from which to read, nor an offset from the position pointer, but rather the offset from 0 within the target array.get() and put() methods, any get() or put() is conducted from the position pointer. Should one need to read from a different position within the underlying array, whilst not adjusting the writing position, the mark() and reset() methods have been supplied.mark() method effectively stores the position of the position pointer by setting the mark pointer to the position of the position pointer. The reset() method causes the position pointer to move to the mark pointer's position. clear() method and the flip() method both return the mark pointer to 0. clear() method does not ensure zero-ing of the buffer, but does return the limit pointer to the upper boundary of the underlying array, and the position pointer to the index.put() and get() operations for NIO buffers are not thread safe.get() operation is executed for a NIO buffer that would cause the position pointer to pass the limit pointer, the executing thread will block until the limit pointer adjusts outward to accommodate the position pointer.get() operation is executed into the underlying array of a readonly NIO buffer, the executing thread will block indefinitely.map() a from a up to Integer.MAX_VALUE in size (2GB); regions beyond this limit can be accessed using an offset greater than zero.Channels (classes implementing the interface ) are designed to provide for bulk data transfers to and from NIO buffers. This is a low-level data transfer mechanism that exists in parallel with the classes of the higher-level I/O library (packages and ). A channel implementation can be obtained from a high-level data transfer class such as , , or , and vice versa.
File channels can use arbitrary buffers but can also establish a buffer directly mapped to file contents using memory-mapped I/O. They can also interact with file system locks. Similarly, socket channels (and ) allow for data transfer between sockets and NIO buffers.
FileChannel can be used to do a file copy, which is potentially far more efficient than using old read/write with a byte array. The typical code for this is:
// Getting file channels
FileChannel in = new FileInputStream(source).getChannel();
FileChannel out = new FileOutputStream(target).getChannel();
// JavaVM does its best to do this as native I/O operations.
in.transferTo (0, in.size(), out);
// Closing file channels will close corresponding stream objects as well.
out.close();
in.close();
A selector (and subclasses) provides a mechanism for waiting on channels and recognizing when one or more become available for data transfer. When a number of channels are registered with the selector, it enables blocking of the program flow until at least one channel is ready for use, or until an interruption condition occurs.
Although this multiplexing behavior could be implemented with Java threads, the selector can provide a significantly more efficient implementation using native platform threads or, more likely, even lower-level operating system constructs. A POSIX-compliant operating system, for example, would have direct representations of these concepts, select(). A notable application of this design would be the common paradigm in server software which involves simultaneously waiting for responses on a number of sessions.
In Java, a character set is a mapping between Unicode characters (or a subset of them) and bytes. The package of NIO provides facilities for identifying character sets and providing encoding and decoding algorithms for new mappings.
The regular expression library in the java.util.regex package provides a powerful search facility for character data based on regular expression matching.
The following example was adopted from the NIO API guide examples, where there are more examples.
import java.io.*;
import java.nio.*;
import java.nio.channels.*;
import java.nio.charset.*;
import java.util.regex.*;
public class Grep {
// Charset and decoder for ISO-8859-15
private static Charset charset = Charset.forName("ISO-8859-15");
private static CharsetDecoder decoder = charset.newDecoder();
// Pattern used to parse lines
private static Pattern linePattern = Pattern.compile(".*r?n");
// The input pattern that we're looking for
private static Pattern pattern;
// Compile the pattern from the command line
private static void compile(String pat) {
try {
pattern = Pattern.compile(pat);
} catch (PatternSyntaxException x) {
System.err.println(x.getMessage());
System.exit(1);
}
}
// Use the linePattern to break the given CharBuffer into lines, applying
// the input pattern to each line to see if we have a match
private static void grep(File f, CharBuffer cb) {
Matcher lm = linePattern.matcher(cb); // Line matcher
Matcher pm = null; // Pattern matcher
int lines = 0;
while (lm.find()) {
lines++;
CharSequence cs = lm.group(); // The current line
if (pm == null)
pm = pattern.matcher(cs);
else
pm.reset(cs);
if (pm.find())
System.out.print(f + ":" + lines + ":" + cs);
if (lm.end() == cb.limit())
break;
}
}
// Search for occurrences of the input pattern in the given file
private static void grep(File f) throws IOException {
// Open the file and then get a channel from the stream
FileInputStream fis = new FileInputStream(f);
FileChannel fc = fis.getChannel();
// Get the file's size and then map it into memory
int sz = (int)fc.size();
MappedByteBuffer bb = fc.map(FileChannel.MapMode.READ_ONLY, 0, sz);
// Decode the file into a char buffer
CharBuffer cb = decoder.decode(bb);
// Perform the search
grep(f, cb);
// Close the channel and the stream
fc.close();
}
public static void main(String[] args) {
if (args.length < 2) {
System.err.println("Usage: java Grep pattern file...");
return;
}
compile(args[0]);
for (int i = 1; i < args.length; i++) {
File f = new File(args[i]);
try {
grep(f);
} catch (IOException x) {
System.err.println(f + ": " + x);
}
}
}
}