Thank you.

Here is my code and the same out of memory exception.

public synchronized ResultSet executeSQL(String sql) throws DataAdapterException
{
ResultSet result = null;
Statement st = _conn.createStatement(java.sql.ResultSet.TYPE_FORWARD_ONLY,
java.sql.ResultSet.CONCUR_READ_ONLY);
st.setFetchSize(Integer.MIN_VALUE);
setCurrentStatement(st);
st.execute(sql);
result = st.getResultSet();
if ( result == null )
st.close();
}
catch ( OutOfMemoryError e )
{
MessageLog.errorMessage(“execute method, out of memory exception ” + e.getMessage());
}
finally
{
releaseCurrentStatement();
}
return result;
}

java.lang.OutOfMemoryError: Java heap space

then the program crashes. Its ok if it gives the out of memory but should not crash. Thank you

It’s also clear that starting a streaming query puts a lock on the whole table. So trying another SQL operation (e.g. an UPDATE) on the same table using a different Connection won’t work.

I don’t know if this can be worked around by setting some fancy transaction isolation levels. Probably not.

My solution is to put all updated records in a temporary table, and do the real updates afterwards.

Hope this helps others in the same predicament. Cheers!

The documentation talks about concurrent access (see the bottom of the original post) and discusses transactions, etc.

It definitely is not possible to use the same connection as is performing the streaming to do updates, the documentation makes that clear.

What’s not clear to me is whether it is possible to have a second connection also reading and/or writing to the table while another connection is streaming. I haven’t tried that.

I would have to explore further documentation and do a variety of tests with different transaction settings and multiple connections to determine the actual behavior.

If you find this out it would be great if you could report back with that information.

Now I have another problem (in a different application): reading a table using the streaming style locks the entire table. So It seems to be impossible to do updates on some of the read records on the fly. Perhaps I have to put them away somewhere during the read phase. Unless there’s some trick you can do by setting some additional transaction isolation parameters…?

(Did I understand that stmt.setFetchSize(10000) is not a practical option since the server would first have to read all the millions of records anyway, before dishing out the result in chunks?)