Dag H. Wanvik wrote: > Hi Daniel, > > >>>>>>"DJD" == Daniel John Debrunner <[EMAIL PROTECTED]> wrote: > >
> DJD> Not sure it's the job of Derby to recognize syntax it does not > DJD> support/implement. How far would you take it, complete SQL standard, any > DJD> other common SQL? > DJD> (ALTER TABLE is probably not covered by the standard) > > It is covered, even in SQL-92, but I agree covering the full > standard in the parser is quite a task and probably not worth the > effort. Would you recommend that we only recommend the syntax we do > support at any one time? I notice we do throw a few > SQLState.NOT_IMPLEMENTED exceptions in the parser today, mainly > related to national character set. Are there some guidelines which > specify which "not implemented" syntax is currently recognized, but > rejected? No. :-) National Character types were partially removed from Derby, or really disabled. Since there was a chance it would be added back in we didn't just delete all the code, but left it in for future folks to work on. One of the reasons it was disabled is that it wasn't clear that the existing Cloudscape implementation was in line with the SQL standard, especially with conversions to and from date time values. Rather than allow a possible non-standard implementation, that would have to be supported for several releases, it was disabled, with the assumption that if it was someone's itch they could resolve any SQL standard issues before re-enabling it. Given JDBC 4.0 finally adds these types to JDBC, it might have been the best solution, otherwise we would have had changes in behaviour, e.g. metadata changing from Types.CHAR to Types.NATIONAL_CHAR. Dan.
