Horton

Re: information_schema and not-null constraints

Vik Fearing <vik@postgresfriends.org>

View thread
From: Vik Fearing <vik@postgresfriends.org>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: "David G. Johnston" <david.g.johnston@gmail.com>, Alvaro Herrera <alvherre@alvh.no-ip.org>, Pg Hackers <pgsql-hackers@lists.postgresql.org>, Peter Eisentraut <peter@eisentraut.org>, Peter Eisentraut <peter.eisentraut@enterprisedb.com>
Date: 2023-09-06T19:09:20Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Revert structural changes to not-null constraints

  2. Fix inconsistencies in error messages

  3. Disallow direct change of NO INHERIT of not-null constraints

  4. Disallow NO INHERIT not-null constraints on partitioned tables

  5. Better handle indirect constraint drops

  6. Don't try to assign smart names to constraints

  7. Fix restore of not-null constraints with inheritance

  8. ATTACH PARTITION: Don't match a PK with a UNIQUE constraint

  9. Fix propagating attnotnull in multiple inheritance

  10. Check stack depth in new recursive functions

  11. Move privilege check to the right place

  12. Update information_schema definition for not-null constraints

  13. Fix not-null constraint test

  14. Disallow changing NO INHERIT status of a not-null constraint

  15. Catalog not-null constraints

  16. parallel_schedule: add comment on event_trigger test dependency

  17. Revert "Catalog NOT NULL constraints" and fallout

  18. Adjust contrib/sepgsql regression test expected outputs.

  19. Fix table name clash in recently introduced test

  20. Catalog NOT NULL constraints

  21. Change the rules for inherited CHECK constraints to be essentially the same

On 9/6/23 05:40, Tom Lane wrote:
> Vik Fearing <vik@postgresfriends.org> writes:
>> On 9/6/23 02:53, Tom Lane wrote:
>>> What solution do you propose?  Starting to enforce the spec's rather
>>> arbitrary requirement that constraint names be unique per-schema is
>>> a complete nonstarter.  Changing the set of columns in a spec-defined
>>> view is also a nonstarter, or at least we've always taken it as such.
> 
>> I both semi-agree and semi-disagree that these are nonstarters.  One of
>> them has to give.
> 
> [ shrug... ] if you stick to a SQL-compliant schema setup, then the
> information_schema views will serve for introspection.  If you don't,
> they won't, and you'll need to look at Postgres-specific catalog data.


As someone who regularly asks people to cite chapter and verse of the 
standard, do you not see this as a problem?

If there is /one thing/ I wish we were 100% compliant on, it's 
information_schema.


> This compromise has served for twenty years or so, and I'm not in a
> hurry to change it.  


Has it?  Or is this just the first time someone has complained?


> I think the odds of changing to the spec's
> restriction without enormous pushback are nil, and I do not think
> that the benefit could possibly be worth the ensuing pain to users.


That is a valid opinion, and probably one that will win out for quite a 
while.


> (It's not even the absolute pain level that is a problem, so much
> as the asymmetry: the pain would fall exclusively on users who get
> no benefit, because they weren't relying on these views anyway.
> If you think that's an easy sell, you're mistaken.)


I am curious how many people we are selling this to.  In my career as a 
consultant, I have never once come across anyone specifying their own 
constraint names.  That is certainly anecdotal, and by no means means it 
doesn't happen, but my personal experience says that it is very low.

And since our generated names obey the spec (see ChooseConstraintName() 
which even says some apps depend on this), I don't see making this 
change being a big problem in the real world.

Mind you, I am not pushing (right now) to make this change; I am just 
saying that it is the right thing to do.


> It could possibly be a little more palatable to add column(s) to the
> information_schema views, but I'm having a hard time seeing how that
> moves the needle.  The situation would still be precisely describable
> as "if you stick to a SQL-compliant schema setup, then the standard
> columns of the information_schema views will serve for introspection.
> If you don't, they won't, and you'll need to look at Postgres-specific
> columns".  That doesn't seem like a big improvement.  Also, given your
> point about normalization, how would we define the additions exactly?


This is precisely my point.
-- 
Vik Fearing