Re: adding a nullable column of type domain w/ check constraint runs checks?

From: Joe Van Dyk <joe(at)tanga(dot)com>
To: "pgsql-general(at)postgresql(dot)org" <pgsql-general(at)postgresql(dot)org>
Subject: Re: adding a nullable column of type domain w/ check constraint runs checks?
Date: 2014-09-03 00:11:29
Message-ID: CACfv+p+8dToaR7h06_M_YMjpV5Na-CQq7kN=KgJq_k84H7UHWA@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

On Tue, Aug 19, 2014 at 3:20 PM, Joe Van Dyk <joe(at)tanga(dot)com> wrote:

> On Tue, Aug 19, 2014 at 3:16 PM, Joe Van Dyk <joe(at)tanga(dot)com> wrote:
>
>> On Tue, Aug 19, 2014 at 3:10 PM, Joe Van Dyk <joe(at)tanga(dot)com> wrote:
>>
>>> I have a large table that I don't want to lock for more than couple
>>> seconds. I want to add a nullable column to the table, the type of the
>>> column is a domain with a check constraint.
>>>
>>> It appears that the check constraint is being checked for each row, even
>>> though the column can be nullable? Is there a way around this?
>>>
>>> BEGIN
>>> Timing is on.
>>>
>>> create domain test_enum numeric check (value > 0);
>>> CREATE DOMAIN
>>> Time: 1.817 ms
>>>
>>> create table test_enum_table (id serial primary key);
>>> CREATE TABLE
>>> Time: 2.213 ms
>>>
>>> insert into test_enum_table select * from generate_series(1, 2000000);
>>> INSERT 0 2000000
>>> Time: 4299.000 ms
>>>
>>> alter table test_enum_table add column t test_enum;
>>> ALTER TABLE
>>> Time: 3165.869 ms -- Takes 3 seconds in this test example
>>>
>>> Also:
>>
>> alter table test_enum_table add column t1 numeric check (t1 > 0);
>> ALTER TABLE
>> Time: 140.185 ms
>>
>> which is much more reasonable.
>>
>
> johnto on irc says:
>
> "I'm not sure why it's done this way. it seems like it could test the
> domain once against NULL and see whether that's rejected or not. instead
> it just forces a rewrite :-("
>

Would it be possible to check the domain against null and if that works,
then don't check any of the rows?

Joe

In response to

Browse pgsql-general by date

  From Date Subject
Next Message Patrick Dung 2014-09-03 03:03:15 Re: Question about gin index not used on a tsv column
Previous Message David G Johnston 2014-09-02 22:24:12 Re: puzzled by "commit" Logging statement duration