A fistfull of readers

After presenting a working InputStream on a ByteBuffer, i have to more readers for you out there.

First, the StringBufferReader to efficient read data from a StringBuffer. One can use new java.io.StringReader(sb.toString()) but that would convert the whole StringBuffer (sb) to a string, loosing a whole lotta memory if the string is just big enough. If you can assure that the StringBuffer don’t need to be modified, use the following code:

import java.io.IOException;
import java.io.Reader;
 
public class StringBufferReader extends Reader {
	private int pos;
	private final StringBuffer sb;	
	private boolean closed;
 
	public StringBufferReader(final StringBuffer sb) {
		this.sb = sb;
		this.pos = 0;
		this.closed = false;
	}
 
	@Override
	public void close() throws IOException {
		this.closed = true;
	}
 
	@Override
	public int read(char[] cbuf, int off, int len) throws IOException {
		if(closed)
			throw new IOException("Reader is closed");
		int _len = Math.min(len, sb.length() - pos);		
		sb.getChars(pos, pos + _len, cbuf, off);
		pos += _len;
		return _len == 0 ? -1 : _len;
	}
}

It uses no intermediate buffer.

In the same context i stumbled upon the Byte Order Mark (BOM) in some UTF8 files (especially ones that were created with tools under Microsoft Windows).

The InputReaders and Streams in the JRE don’t skip the BOM, neither does the org.dom4j.io.SAXReader so parsing of such XML files or strings fails with something like “Content not allowed in prolog”. Enter my simple BOMSkippingReader:

import java.io.IOException;
import java.io.Reader;
 
/**
 * This reader skips a possible Byte Order Marker at the 
 * start of UTF8 files which java doesn't.
 * @author michael
 *
 */
public class BOMSkippingReader extends Reader {
	private final Reader decorated;	
	private final boolean rewrite;
	private final int firstchar;
	private int pos = 0;
 
	public BOMSkippingReader(final Reader decorated) throws IOException {
		this.decorated = decorated;
 
		this.firstchar = decorated.read();						
		this.rewrite = firstchar != 65279;			
	}
 
	@Override
	public void close() throws IOException {
		decorated.close();
	}
 
	@Override
	public int read(char[] cbuf, int off, int len) throws IOException {					
		int redone = 0;					
		while(rewrite && pos < 1) {
			cbuf[off + pos++] = (char) firstchar;
			++redone;
		}
		return redone + decorated.read(cbuf, off + redone, len - redone);				
	}
}

If there’s a BOM in the underlying reader, it’s skipped, otherwise written back to the reader. Works well for me and the small while loop doesn’t have much impact on the performance in my app.

Filed in Java | Comments (0) »

23-Oct-07

Java: Creating a Stream on a ByteBuffer

This example is plain wrong. The InputStream will cause an endless loop as it always returns a value >= 0. Working code as follows:

private static InputStream _newInputStream(final ByteBuffer buf) {
	return new InputStream() {
		public synchronized int read() throws IOException {				
			return buf.hasRemaining() ? buf.get() : -1;
		}
 
		public synchronized int read(byte[] bytes, int off, int len) throws IOException {			
			int rv = Math.min(len, buf.remaining());				
			buf.get(bytes, off, rv);
			return rv == 0 ? -1 : rv;
		}
	};
}

Filed in Java, Shortcuts | Comments (1) »

18-Oct-07

JDBC: Get autogenerated keys on a Oracle DB

With the current and latest Oracle JDBC Drivers it’s possible to retrieve one automatically generated key based on a sequence or any other arbitrary value (autogenerated-keys or identity columns in other databases).

Certainly it isn’t as simple as just use using Statement.html#getGeneratedKeys() as it simply returns an emtpy resultset.

After browsing around, i saw that java.sql.Connection can prepare a statement with a flag to return generated keys like so:

connection.prepareStatement("INSERT INTO FOOBAR VALUES(id.nextval, 1), Statement.RETURN_GENERATED_KEYS);

(By the way, it’s always a good idea to use prepared statements from a performance point of view as they can be reused)

Anyway, it wouldn’t be an Oracle product if something is different like any other products and i still was left alone in the dark with an empty result set.

Follow the path to enlightment:

final String sql = "INSERT INTO foobar(id, b) VALUES (id.nextval, ?)";
stmt = connection.prepareStatement(sql,new String[]{"ID"});			
stmt.setString(1, "bar");
stmt.execute();		
rs = stmt.getGeneratedKeys();
rs.next();
rv = rs.getInt(1);

Telling oracle which column contains the generated value does the trick. It’s really well hidden on their website. Be aware, you cannot refer to the returned keys by name, you need to address them 1 based.

Be aware that this doesn’t work inside a 9.2.x.x or 10.2.x.x Oracle Database as a Java Stored Procedure. Either the driver isn’t JDBC 3 (9.2) or the methods are not supported (10.2). You can work around this problem with 2 statements, first select id.nextval from somewhere, then execute your insert. Lame, but i didn’t find any other solution.

Ran on the client side with the latest JDBC from Oracle on the other hand works just fine.

Filed in Java, Oracle | Comments (1) »

09-Oct-07