[BioSQL-l] genbank, references, and crc's

[Bryan Cardillo]

On Tue, Apr 10, 2007 at 12:09:43PM -0400, Hilmar Lapp wrote:
> thanks for tracking this down - great, I've committed it.
> 
> The 'correct' condition, as defined by the schema, would actually be  
> test for author or title being specified, because location must be  
> non-empty, according to the schema.
> 
> I.e., at least theoretically, the condition will now always be true,  
> unless you removed the NOT NULL constraint locally on  
> reference.location.
> 
> Would you mind testing whether removing the location() part from the  
> if clause will still solve the issue?

        you are correct, the test for location doesn't seem to be
        necessary.

        from a theoretically point of view, I'm not sure I agree
        with removing the location test though.  it seems to me that
        if you have a field (ie, location) which is used in
        generating a unique identifier (crc64), then you should
        consult that field when determining what the unique
        identifier is for a particular object.

        to put it another way, a reference instance with no authors,
        no title, and a location can have a valid crc.  so why should
        the adaptor ignore this case?

        all that being said, my understanding of how all this goes
        together is still pretty shallow, so I'll defer to you as to
        which solution is best ;)

        --Bryan