How to write / update Oracle blob in a reliable way?

I'm trying to write and update a pdf document in a blob column but I'm just able to update the blob only writing more data than the previous stored data. If I try to update the blob column with a smaller document data I get only a corrupted pdf.

First the blob column has been initialized using empty_blob() function. I wrote the sample Java class below to test this behaviour. I run it the first time with 'true' as first parameter of the main method so in the first row there's stored a document of about 31kB and in the second row there's a document of 278kB. Then I run it with 'false' as parameter, in this way the two rows should be updated swapping the documents. The result is that I get a correct result only when I write more data than the existing one.

How is it possible to write a method that writes and updates a blob in a reliable way without worring about binary data's size?

import static org.apache.commons.io.IOUtils.copy;

import java.io.FileInputStream;
import java.io.OutputStream;
import java.sql.Connection;
import java.sql.DriverManager;
import java.sql.PreparedStatement;
import java.sql.ResultSet;
import java.sql.SQLException;

import oracle.jdbc.OracleDriver;
import oracle.jdbc.OracleResultSet;
import oracle.sql.BLOB;

import org.apache.commons.lang.ArrayUtils;
/**
 * Prerequisites:
 * 1) a table named 'x' must exists [create table x (i number, j blob);] 
 * 2) that table should have two columns [insert into x (i, j) values (1, empty_blob()); insert into x (i, j) values (2, empty_blob()); commit;]
 * 3) download lsp.pdf from http://www.objectmentor.com/resources/articles/lsp.pdf
 * 4) download dotguide.pdf from http://www.graphviz.org/Documentation/dotguide.pdf
 */
public class UpdateBlob {
    public static void main(String[] args) throws Exception {
        processFiles(new String[]{"lsp.pdf", "dotguide.pdf"}, Boolean.valueOf(args[0]));
    }

    public static void processFiles(String [] fileNames, boolean forward) throws Exception {
      if(!forward){
        ArrayUtils.reverse(a);
      }
      int idx = 1;
      for(String fname : fileNames){
        insert(idx++, fname);
      }
  }

    private static void insert(int idx, String fname) throws Exception{
        Connection conn = null;
        PreparedStatement ps = null;
        ResultSet rs = null;
        try {
            DriverManager.registerDriver(new OracleDriver());
            conn = DriverManager.getConnection("jdbc:oracle:thin:@"+db+":"+port+":"+sid, user, pwd);
            ps = conn.prepareStatement("select j from x where i = ? for update");
            ps.setLong(1, idx);

            rs = ps.executeQuery();

            if (rs.next()) {
                FileInputStream instream = new FileInputStream(fname);
                BLOB blob = ((OracleResultSet)rs).getBLOB(1);
                OutputStream outstream = blob.setBinaryStream(1L);
                copy(instream, outstream);
                instream.close();
                outstream.close();
            }
            rs.close();
            ps.close();
            conn.close();
        } catch (SQLException e) {
            e.printStackTrace();
            throw new Exception(e);
        }
    }
}

Oracle version: 11.1.0.7.0 - 64bit

I even tried the standard JDBC API without using Oracle's specific one (like in the example above) without any success.

It's a lot easier:

PreparedStatement pstmt =
  conn.prepareStatement("update blob_table set blob = ? where id = ?");
File blob = new File("/path/to/picture.png");
FileInputStream in = new FileInputStream(blob);

// the cast to int is necessary because with JDBC 4 there is 
// also a version of this method with a (int, long) 
// but that is not implemented by Oracle
pstmt.setBinaryStream(1, in, (int)blob.length()); 

pstmt.setInt(2, 42);  // set the PK value
pstmt.executeUpdate();
conn.commit();
pstmt.close();

It works the same when using an INSERT statement. No need for empty_blob() and a second update statement.

In addition to a_horse_with_no_name's answer (which relies on PreparedStatement.setBinaryStream(...) API), there're at least two more options for BLOBs, and 3 more for CLOBs and NCLOBs:

1. Explicitly create a LOB, write to it, and use PreparedStatement.setBlob(int, Blob):

   int insertBlobViaSetBlob(final Connection conn, final String tableName, final int id, final byte value[])
   throws SQLException, IOException {
       try (final PreparedStatement pstmt = conn.prepareStatement(String.format("INSERT INTO %s (ID, VALUE) VALUES (?, ?)", tableName))) {
           final Blob blob = conn.createBlob();
           try (final OutputStream out = new BufferedOutputStream(blob.setBinaryStream(1L))) {
               out.write(value);
           }
   
           pstmt.setInt(1, id);
           pstmt.setBlob(2, blob);
           return pstmt.executeUpdate();
       }
   }
   

2. Update an empty LOB (inserted via DBMS_LOB.EMPTY_BLOB() or DBMS_LOB.EMPTY_CLOB()) via SELECT ... FOR UPDATE. This is Oracle-specific and requires two statements executed instead of one. Additionally, this is what you were trying to accomplish in the first place:

   void insertBlobViaSelectForUpdate(final Connection conn, final String tableName, final int id, final byte value[])
   throws SQLException, IOException {
       try (final PreparedStatement pstmt = conn.prepareStatement(String.format("INSERT INTO %s (ID, VALUE) VALUES (?, EMPTY_BLOB())", tableName))) {
           pstmt.setInt(1, id);
           pstmt.executeUpdate();
       }
   
       try (final PreparedStatement pstmt = conn.prepareStatement(String.format("SELECT VALUE FROM %s WHERE ID = ? FOR UPDATE", tableName))) {
           pstmt.setInt(1, id);
           try (final ResultSet rset = pstmt.executeQuery()) {
               while (rset.next()) {
                   final Blob blob = rset.getBlob(1);
                   try (final OutputStream out = new BufferedOutputStream(blob.setBinaryStream(1L))) {
                       out.write(value);
                   }
               }
           }
       }
   }
   

3. For CLOBs and NCLOBs, you can additionally use PreparedStatement.setString() and setNString(), respectively.

FWIW, for something that fits in memory, I found I could simply pass in a byte array as the prepared statement parameter, rather than going through the "stream" rigor morale (or worse Oracle specific/suggested things)

Using a Spring "JDBC template" wrapper (org.springframework.jdbc.core.JdbcTemplate) to put the contents of a "large" (or not) string into a BLOB column, the code is something like the following:

jdbc.update( "insert into a_table ( clob_col ) values ( ? )", largeStr.getBytes() );

There is no step 2.