Re: Overhaul of type attributes modification

On 9 July 2011 13:26, Thom Brown <thom(at)linux(dot)com> wrote:
> On 9 July 2011 13:25, Guillaume Lelarge <guillaume(at)lelarge(dot)info> wrote:
>> On Sat, 2011-07-09 at 13:14 +0100, Thom Brown wrote:
>>> > The major issue is why we use the "drop all, add all" method.  Let's say
>>> > I have a composite type declared like this:
>>> >
>>> > CREATE TYPE s1.ty2 AS
>>> >   (c1 integer,
>>> >    c2 integer,
>>> >    c3 text,
>>> >    c4 integer);
>>> >
>>> > If I change c3's type (but it could be c1 or c2), I get this SQL query:
>>> >
>>> > ALTER TYPE s1.ty2
>>> >  ADD ATTRIBUTE c3 xml;
>>> > ALTER TYPE s1.ty2
>>> >  DROP ATTRIBUTE c3;
>>> >
>>> > Remember that, on pgAdmin, it shows the list this way:
>>> >    c1 integer,
>>> >    c2 integer,
>>> >    c3 xml,
>>> >    c4 integer
>>> >
>>> > I click OK, and get back to the properties dialog. pgAdmin now shows the
>>> > list this way:
>>> >    c1 integer,
>>> >    c2 integer,
>>> >    c4 integer
>>> >    c3 xml,
>>> >
>>> > Which is true. We drop the attribute and add another one, which will be
>>> > at the end of the list.
>>>
>>> I have a solution.  It seems I overlooked the alter attribute
>>> capabilities.  We can just do:
>>>
>>> ALTER TYPE s1.ty2
>>> ALTER ATTRIBUTE c3 TYPE xml;
>>>
>>> That will preserve its position without ever having to drop it.  I'm
>>> not sure why I didn't see it before.
>>>
>>
>> Can you fix your patch to also use this syntax?
>
> Yes, I'll try to fix both issues.  Thanks for testing it.  I'll be
> back with a patch shortly.

Okay, looks like it needed more work than I realised. There are quite
a few extra changes in this one. I've moved the various checks (such
as ensuring there are at least 2 attributes) out of the block for
creating a type so that the same rules apply to modifying types. This
is because you shouldn't be able to modify a type so that there are
less than 2 attributes.

But please try to break this. I've tried various combinations of
actions (adding, renaming and changing type, just renaming, just
changing type, removing an original attribute, removing all and adding
back in, removing none and just adding additional ones) and can't find
anything wrong now. But maybe there's something I missed.

Revised patch attached.

--
Thom Brown
Twitter: @darkixion
IRC (freenode): dark_ixion
Registered Linux user: #516935

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise PostgreSQL Company