I think this might be a general problem with some InputStream instances
when they are asked to process byte streams. I'm being tentative, because
I found something that looks like the same problem last night, diagnosed
it just now, but haven't tried fixing it yet.

I'm using J2SE 1.4 at present because I want to get this application
running and stable before moving to Java 6.

In my case I'm reading from the InputStream returned by the JavaMail
Message.getContent() method. It returns a PipedInputStream instance
which should transfer the contents of a byte array. I plug this directly
into PreparedStatement.setBinaryStream() to write the message content
into a bytea (binary equivalent of text or varchar) in a Postgres
database, having got the field size from InputStream.available().

As part of my diagnostic code, if debugging is on I execute the
following before loading data into the database:
- I call Message.getContent() and read the stream a byte at a time
- convert each byte to a char and append it to a StringBuffer.
- the StringBuffer is displayed by writing it to System.err.

The diagnostic display is ALWAYS a complete version of the message
content, but the database field content varies depending on whether the
message content is MIME or non-MIME:
- non-MIME: (plaintext UTF-8 or bit-7) the database field is always
  correct.

- MIME: the database field is sometimes empty and sometimes incomplete.

It looks to me as if, for a PipedInputStream, the value returned by
available() can be invalidated by the content of the byte array but that
 read() will always correctly recognise the end of the byte array.

I'm about to modify my code to check this assumption and will let you
know the result.

stc, 29.11.2007 15:35:

I'm not sure that getBytes() works for other JDBC drivers as well. I
think get/setBinaryStream() was the only reliable (cross-dbms) way that
I found.

What I do, to read the BLOB (I removed the error handling obviously)

InputStream in = rs.getBinaryStream(1);
ByteArrayOutputStream baos = new ByteArrayOutputStream();
byte[] buffer = new byte[1024];

int bytesRead = in.read(buff);

while (bytesRead > -1)
{
  baos.write(buff, 0, bytesRead);
  bytesRead = in.read(buff);
}

in.close();
baos.close();

Now you can e.g. use baos.toByteArray() to access the raw data as a byte
array.

Note that using ByteArrayOutputStream.toByteArray() will create a copy
of its internal buffer so that will double the amount of memory that you
need. That's the reason I wrote my own dynamic byte array (where I can
access the byte array without copying it around)

Thomas

That's probably it. I used available() because it gave me what I needed when
called on ByteArrayInputStream so I didn't bother to check the javadocs.
I'll try with a fixed-size buffer tomorrow and see what happens but there's
no reason not to work :-)

Regarding the getBytes(), I can confirm that it works with PostgreSQL 8.1.4
(with 8.1.407 JDBC3 driver).

Thanks for the help...