Character set issues with Oracle Gateways, SQL Server, and Application Express

I am migrating data from a Oracle on VMS that accesses data on SQL Server using heterogeneous services (over ODBC) to Oracle on AIX accessing the SQL Server via Oracle Gateways (dg4msql). The Oracle VMS database used the WE8ISO8859P1 character set. The AIX database uses WE8MSWIN1252. The SQL Server database uses "Latin1-General, c开发者_如何学JAVAase-insensitive, accent-sensitive, kanatype-insensitive, width-insensitive for Unicode Data, SQL Server Sort Order 52 on Code Page 1252 for non-Unicode Data" according to sp_helpsort. The SQL Server databases uses nchar/nvarchar or all string columns.

In Application Express, extra characters are appearing in some cases, for example 123 shows up as %001%002%003. In sqlplus, things look ok but if I use Oracle functions like initcap, I see what appear as spaces between each letter of a string when I query the sql server database (using a database link). This did not occur under the old configuration.

I'm assuming the issue is that an nchar has extra bytes in it and the character set in Oracle can't convert it. It appears that the ODBC solution didn't support nchars so must have just cast them back to char and they showed up ok. I only need to view the sql server data so I'm open to any solution such as casting, but I haven't found anything that works.

Any ideas on how to deal with this? Should I be using a different character set in Oracle and if so, does that apply to all schemas since I only care about one of them.

Update: I think I can simplify this question. SQL Server table uses nchar. select dump(column) from table returns Typ=1 Len=6: 0,67,0,79,0,88 when the value is 'COX' whether I select from a remote link to sql server, cast the literal 'COX' to an nvarchar, or copy into an Oracle table as an nvarchar. But when I select the column itself it appears with extra spaces only when selecting from the remote sql server link. I can't understand why dump would return the same thing but not using dump would show different values. Any help is appreciated.

There is an incompatibility between Oracle Gateways and nchar on that particular version of SQL Server. The solution was to create views on the SQL Server side casting the nchars to varchars. Then I could select from the views via gateways and it handled the character sets correctly.

You might be interested in the Oracle NLS Lang FAQ