db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Daniel John Debrunner <...@apache.org>
Subject Re: comparisons between system and user table columns with TERRITORY_BASED collation
Date Tue, 26 Jun 2007 14:22:24 GMT

> Mike Matrigali <mikem_app@sbcglobal.net> writes:
>> This sounds like an easy solution to this issue, but will change the
>> error message for all comparison mismatches that are not territory
>> related.  Does that sound ok?

I would have thought that there would be a different error message.
Looking at the comments in DERBY-2668 it seems to says a single message 
is used because DataTypeDescriptor.comparable() is used to determine if 
the collations are compatible for comparision. However, I'm not sure how 
this can be implementing the required logic behind collation 
determination, which requires inspection of all the types involved, not 
just two. I think this goes back to my concern that the required complex 
logic for determining collation is not in a single method, rather spread 
out across several locations, each with different logic. Even the logic 
in DTD.comparable does not match what is required by the SQL spec 
(section 9.13 I think).


View raw message