Ah, it looks like the problem is occurring in DefaultDatabaseCollection and in StringHelper, and it's ultimately caused by the fact that MetaMatrix allows table names to contain "." in the names.

DefaultDatabaseCollection creates a qualified list of tables by splitting the fully-qualified name on the last occurrence of ".". It considers everything before the last dot to be the "qualifier", which is used uniquely identify each schema and its set of tables in a Map.

Since MetaMatrix has a NULL catalog name, but also has table names that include dots, the end result is that DefaultDatabaseCollection builds a fully-qualified name that is misinterpreted as if it had a catalog name.

For example:
Catalog: null
Schema: POC
Table: CDB_CIM.GLOBAL_SIZE

Fully-qualified name created by Table.qualify: POC.CDB_CIM.GLOBAL_SIZE
Qualifier, as determined by StringHelper.qualifier(key): POC.CDB_CIM <-- this is "incorrect", should be POC, as CDB_CIM is part of the table name

Later, TableFilterView code gets a TableContainer that appears to have the name of CATALOG.SCHEMA. It splits this string on ".", which results in a catalog=POC and schema=CDB_CIM. Using those values, it produces the incorrect filter string, shown earlier in this thread.

So, the first question is whether table names with dots are legal according to SQL92 and JDBC specifications.

If it is legal, then perhaps DefaultDatabaseCollection's logic could be changed, such that it constructs the qualifier by concatenating the catalog + "." + schema name, rather than by using the StringHelper.qualifier() trick.

In fact, I made this code change, and it does now work as expected. I'd be happy to share it if you guys think it's worth fixing.

Also took a quick look through the SQL92 specification, along with a trip to Mimer SQL validator, and it appears that dots in table names are technically legal.

create a jira with a patch (not the whole file please ;) so we can see.



it all depends on quoting I believe.

If your table got "." in them how do you distinguish SCH1.TAB1 from SCH1.TAB1 without quoting ?

I'd be happy to build a patch, but the process is not working for me, at least not using a freshly checked out copy of HibernateExt from SVN.

I notice there is an ant task for creating a patch in common-build.xml, but it references CVS. I originally checked out this code using SVN, per vyemialyanchyk's instructions, so it's a bit odd to see CVS being used here.

When I try to run the ant "patch" task, it complains that CVS is not setup properly. I can modify the task to add the proper CVSROOT in SourceForge, but still get the complaint that I need to checkout the entire project using 'cvs checkout' first. And when I try to checkout the entire project using CVS, I receive a CVS server bug: "cvs checkout: dying gasps from hibernate.cvs.sourceforge.net unexpected".

Am I missing something obvious here? Have you modified your build script in order to build a patch successfully?

if you use eclipse there is a simple Team > Create Patch in the context sensitive menu.

otherwise just use svn diff <directory> to get a patch

Thanks Max.

I created a new JIRA for the table filter wizard issue, and attached my patch file. You can find the JIRA here:
https://jira.jboss.org/browse/JBIDE-6396

I've also attached a patch file to the previous JIRA, dealing with mixed-case table types.

Thanks for all your help and I hope these changes become useful to others.