Re: Questions on domain on composite / casts ignoring domains - Mailing list pgsql-hackers

From David G Johnston
Subject Re: Questions on domain on composite / casts ignoring domains
Date
Msg-id 1413854318303-5823763.post@n5.nabble.com
Whole thread Raw
In response to Questions on domain on composite / casts ignoring domains  (Jim Nasby <Jim.Nasby@BlueTreble.com>)
List pgsql-hackers
Jim Nasby-5 wrote
> I'm trying to create what amounts to a new type. This would be rather easy
> if I could perform a CHECK on a composite type, which I could do if I
> could create a domain on top of a composite. Is there any reason in
> particular that hasn't been done?
> 
> As an alternative, I tried accomplishing this with a straight domain. That
> would work, except for this:
> 
> WARNING:  cast will be ignored because the source data type is a domain
> 
> Why do we ignore casts from domains to other data types? I'm guessing
> because it's simply not what domains were meant for?

A domain is a base type with a constraint.  When you cast you already know
the existing value is valid and the system simply uses the cast available
for the base type instead. i.e., You cannot have a domain with a different
cast rule than the base type over which it is defined.

Likely the lack of capability is simply a matter of complexity in the face
of somewhat uncommon usage and limited resources.

David J.




--
View this message in context:
http://postgresql.1045698.n5.nabble.com/Questions-on-domain-on-composite-casts-ignoring-domains-tp5823745p5823763.html
Sent from the PostgreSQL - hackers mailing list archive at Nabble.com.



pgsql-hackers by date:

Previous
From: Jim Nasby
Date:
Subject: Re: Trailing comma support in SELECT statements
Next
From: Jim Nasby
Date:
Subject: Re: Inconsistencies in documentation of row-level locking