Querying CHAR columns with character length semantics unreliable

Question

Hi again,
It appears that there is a bug in the JDBC drivers whereby it is highly unlikely that the values of CHAR columns that use character length semantics can be accurately queried using ResultSet.getString(). Instead, the drivers return the value padded with space (0x#20) characters out to a number of bytes equal to the number of characters multiplied by 4. The number of bytes varies depending on the number and size of any non-ascii characters stored in the column.
For instance, if I have a CHAR(1) column, a value of 'a' will return 'a ' (4 characters/bytes are returned), a value of '\u00E0' will return '\u00E0 ' (3 characters / 4 bytes), and a value of '\uE000' will return '\uE000 ' (2 characters / 4 bytes).
I'm currently using version 9.2.0.3 of the standalone drivers (ojdbc.jar) with JDK 1.4.1_04 on Redhat Linux 9, connecting to Oracle 9.2.0.2.0 running on Solaris.
The following sample code can be used to demonstrate the problem (where the DDL at the top of the file must be executed first):
import java.sql.*;
import java.util.*;
This sample generates another bug in the Oracle JDBC drivers where it is not
possible to query the values of CHAR columns that use character length semantics
and are NOT full of non-ascii characters. The inclusion of the VARCHAR2 column
is just a control.
CREATE TABLE TMP2
TMP_ID NUMBER(10) NOT NULL PRIMARY KEY,
TMP_CHAR CHAR(10 CHAR),
TMP_VCHAR VARCHAR2(10 CHAR)
public class ClsCharSelection
private static String createString(char character, int length)
char characters[] = new char[length];
Arrays.fill(characters, character);
return new String(characters);
} // private static String createString(char, int)
private static void insertRow(PreparedStatement ps,
int key, char character)
throws SQLException
ps.setInt(1, key);
ps.setString(2, createString(character, 10));
ps.setString(3, createString(character, 10));
ps.executeUpdate();
} // private static String insertRow(PreparedStatement, int, char)
private static void analyseResults(PreparedStatement ps, int key)
throws SQLException
ps.setInt(1, key);
ResultSet results = ps.executeQuery();
results.next();
String tmpChar = results.getString(1);
String tmpVChar = results.getString(2);
System.out.println(key + ", " + tmpChar.length() + ", '" + tmpChar + "'");
System.out.println(key + ", " + tmpVChar.length() + ", '" + tmpVChar + "'");
results.close();
} // private static void analyseResults(PreparedStatement, int)
public static void main(String argv[])
throws Exception
Driver driver = (Driver)Class.forName(
"oracle.jdbc.driver.OracleDriver").newInstance();
DriverManager.registerDriver(driver);
Connection connection = DriverManager.getConnection(
argv[0], argv[1], argv[2]);
PreparedStatement ps = null;
try
ps = connection.prepareStatement(
"DELETE FROM tmp2");
ps.executeUpdate();
ps.close();
ps = connection.prepareStatement(
"INSERT INTO tmp2 ( tmp_id, tmp_char, tmp_vchar " +
") VALUES ( ?, ?, ? )");
insertRow(ps, 1, 'a');
insertRow(ps, 2, '\u00E0');
insertRow(ps, 3, '\uE000');
ps.close();
ps = connection.prepareStatement(
"SELECT tmp_char, tmp_vchar FROM tmp2 WHERE tmp_id = ?");
analyseResults(ps, 1);
analyseResults(ps, 2);
analyseResults(ps, 3);
ps.close();
connection.commit();
catch (SQLException e)
e.printStackTrace();
connection.close();
} // public static void main(String[])
} // public class ClsColumnInsertion

Answer 1

FYI, this has been mentioned as early as November last year:
String with length 1 became 4 when nls_lang_semantics=CHAR
and was also brought up in Feburary:
JDBC thin driver pads CHAR col to byte size when NLS_LENGTH_SEMANTICS=CHAR