[news.eclipse.dtp] Re: JDBCSchemaLoader, JDBCTableLoader, etc. broken fo

["Brian Fitzpatrick" <brianf@xxxxxxxxxx>]

Hi Chris...

Well, we're always on the lookout for improvements in how we handle the 
generics.

Go ahead and register a BZ enhancement request with some information about 
how you think it should be done and we can consider it for the next major 
release. This kind of wholesale change is too much for either of the 
maintenance releases coming up, but we could look at it for Galileo (June 
2009).

It might help to have some code examples of how you'd like to see things 
done as well.

Thanks! Looking forward to more of your input!

--Fitz

"Chris Hines" <chris@xxxxxxxxxxxxxxxxxxxxxxx> wrote in message 
news:g77ont$i2$1@xxxxxxxxxxxxxxxxxxxx
>I have been researching the feasibility of using DTP as an enabling tool 
>for a general purpose database metadata reporting/management application. 
>We already use EMF, so bringing in DTP seems quite natural.  I have 
>encountered a significant roadblock using the GenericJDBC infrastructure.
>
> I have discovered that the code in the 
> org.eclipse.datatools.connectivity.sqm.loader package processes JDBC 
> ResultSets in a non-portable fashion, making it fail when using many JDBC 
> drivers, including the JDBC-ODBC bridge supplied as part of Sun's JVM.
>
> The root problem is that DTP attempts to read the same column from a 
> ResultSet object more than once (for the same row), which is not 
> guaranteed to work as expected.  The javadoc for ResultSet says:  "For 
> maximum portability, result set columns within each row should be read in 
> left-to-right order, and each column should be read only once."
>
> However, DTP uses the following code pattern:
>
> [  Code snippet from JDBCTableLoader.loadTables(List containmentList, 
> Collection existingTables)  ]
>
> while (rs.next()) {
>    String tableName = rs.getString(COLUMN_TABLE_NAME);      // first time 
> to retrieve table_name
>    if (tableName == null || isFiltered(tableName)) {
>        continue;
>    }
>    Table table = (Table) getAndRemoveSQLObject(existingTables, tableName);
>    if (table == null) {
>        table = processRow(rs);                 // eventually calls 
> tableFactory.initialize(table, rs) (see below)
>        if (table != null) {
>            containmentList.add(table);
>        }
>    }
>    else {
>        ITableFactory tableFactory = getTableFactory(rs
>            .getString(COLUMN_TABLE_TYPE));
>        if (tableFactory != null) {
>            tableFactory.initialize(table, rs);     // retrieves table_name 
> a second time
>        }
>        containmentList.add(table);
>        if (table instanceof ICatalogObject) {
>            ((ICatalogObject) table).refresh();
>        }
>    }
> }
>
> This pattern is pervasive throughout the 
> org.eclipse.datatools.connectivity.sqm.loader package.
>
> Does anyone have any suggestions about how to work around this problem? 
> Should I enter this as a bugzilla entry?
>
> Thanks,
> Chris
>