Re: [HACKERS] RI generic trigger procs

Hannu Krosing <hannu@trust.ee>

From: Hannu Krosing <hannu@trust.ee>
To: Jan Wieck <wieck@debis.com>
Cc: PostgreSQL HACKERS <pgsql-hackers@postgreSQL.org>
Date: 1999-09-29T20:37:50Z
Lists: pgsql-hackers
Jan Wieck wrote:
> 
>     Any combination of attributes in a table referenced to by one
>     or  more  FOREIGN  KEY  ...  REFERENCES constraint of another
>     table shall have a UNIQUE and NOT NULL constraint.
...
>  So we assume here that any PK is unique and cannot contain NULL's.

What is the reasoning behind requiring this ?

I can't see anything that would mandate this -
  * NULLs are'nt equal anyway and ar even disregarded under your 
    current description. 
    Or are you just protecting yourself against the case where the 
    foreign key field is set to null - could this be handled the 
    same as deleting for cascaded constraints ?
  * UNIQUE would save us the check for existing other possible 
    referenced values - is this mandated by SQL spec ?

-------------
Hannu