[Open-bio-l] BioSQL schema: some questions

[Thomas Down]

On Tue, Apr 30, 2002 at 02:29:21PM -0700, Chris Mungall wrote:
>
> Sorry, deliberate provocation - I really think every adapter should have a
> test suite so that we can make minor schema changes without guessing if it
> will break code.

There is a test script for the BioJava-BioSQL stuff.  Since
it needs database details, it's not currently in the JUnit test
suite (this might change in the future).  You you can do
something like:

   java biosql.SequenceTest jdbc:postgresql://localhost/thomasd_biosql2 thomas \passw0rd

Anyone who needs any help getting this up and running, just
drop me a message.  For committed Java-phobes, I'm quite happy
to test any proposed schema modifications (well, assuming that
the schema isn't changing 10 times a day...) -- again, just
send mail.

I quite like Hilmar's idea of extracting example queries from
each implementation.  However, this can only ever be a partial
solution since:

  - It's possible to get behavioural quirks which are dependant
    on the database connector technology in use.  For example,
    after an apparently minor change of some columns from
    varchar(n) to char(n), we discovered that the Perl and
    Java PostgreSQL drivers treat such columns in rather
    different ways.

  - At least in the BioJava case, not all the SQL is pre-canned.
    There's code which compiles queries in the BioJava 
    FeatureFilter language into SQL.

So we definitely need to do regular testing with the real
adaptor code as well, preferably on all the different databases
we claim to support.

   Thomas.