Strange DOMAIN behavior

Lists: pgsql-sql
From: Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru>
To: pgsql-sql(at)postgresql(dot)org
Subject: Strange DOMAIN behavior
Date: 2015-07-09 10:03:59
Message-ID: 559E470F.6020002@postgrespro.ru
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-sql

Hello everyone!!!
Got strange DOMAIN behavior in the following plpgsql code:

postgres=# DROP DOMAIN lexema_str CASCADE;
DROP DOMAIN
postgres=# CREATE DOMAIN lexema_str TEXT DEFAULT 'abc';
CREATE DOMAIN
postgres=# DO $$
postgres$# DECLARE
postgres$# lex lexema_str;
postgres$# BEGIN
postgres$# RAISE NOTICE 'lex = %', lex;
postgres$# END;
postgres$# $$;
NOTICE: lex = <NULL>
DO

But i expect that lex = abc!

So default value of DOMAIN type is not set in pgplsql block but:

postgres=# DROP DOMAIN lexema_str CASCADE;
DROP DOMAIN
postgres=# CREATE DOMAIN lexema_str TEXT DEFAULT 'abc' NOT NULL;
CREATE DOMAIN
postgres=# DO $$
postgres$# DECLARE
postgres$# lex lexema_str;
postgres$# BEGIN
postgres$# RAISE NOTICE 'lex = %', lex;
postgres$# END;
postgres$# $$;
ERROR: domain lexema_str does not allow null values
CONTEXT: PL/pgSQL function inline_code_block line 4 during statement
block local variable initialization

So constraints in DOMAIN type work in pgplsql !

Is this correct behavior??

--
Alex Ignatov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus


From: "David G(dot) Johnston" <david(dot)g(dot)johnston(at)gmail(dot)com>
To: Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru>
Cc: "pgsql-sql(at)postgresql(dot)org" <pgsql-sql(at)postgresql(dot)org>
Subject: Strange DOMAIN behavior
Date: 2015-07-09 12:49:30
Message-ID: CAKFQuwYOZQ22qNtPe0m4w047DLhnDati8+Yqhcd6S2SOYL2GDQ@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-sql

On Thursday, July 9, 2015, Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru
<javascript:_e(%7B%7D,'cvml','a(dot)ignatov(at)postgrespro(dot)ru');>> wrote:

> Hello everyone!!!
> Got strange DOMAIN behavior in the following plpgsql code:
>
> But i expect that lex = abc!
>
> So default value of DOMAIN type is not set in pgplsql block but:
>
> Is this correct behavior??
>
>
If you read the create domain sql command documentation carefully the
default clause is only used when inserting into a table that has a column
of the domain type that is not explicitly provided a value. Each
language deals with domains differently and the behavior you expect is not
currently implemented in pl/pgsql. If you want a default inside the
procedure you need to declare one explicitly.

David J.


From: Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru>
To: "David G(dot) Johnston" <david(dot)g(dot)johnston(at)gmail(dot)com>
Cc: "pgsql-sql(at)postgresql(dot)org" <pgsql-sql(at)postgresql(dot)org>
Subject: Re: Strange DOMAIN behavior
Date: 2015-07-09 14:10:40
Message-ID: 559E80E0.8000002@postgrespro.ru
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-sql

Thank you for your reply.
In fact i want to use domains in the following case:

DROP DOMAIN lexema_str CASCADE;
CREATE DOMAIN lexema_str TEXT DEFAULT 'abc' NOT NULL;
DROP TYPE lexema_test CASCADE;
CREATE TYPE lexema_test AS (
lex lexema_str,
lex2 BIGINT
);
DROP TABLE ttt;
CREATE TABLE ttt (
a lexema_test,
b BIGINT
);
INSERT INTO ttt (b) VALUES (1);
SELECT *
FROM ttt;
a | b
---+---
| 1
(1 row)

a.lex is null again not 'abc' as I expected like in plpgsql.

All i want is to have default values in composite types. Feature that
Oracle have.
I thought that with domain type it should be possible.

On 09.07.2015 15:49, David G. Johnston wrote:
> On Thursday, July 9, 2015, Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru
> <javascript:_e(%7B%7D,'cvml','a(dot)ignatov(at)postgrespro(dot)ru');>> wrote:
>
> Hello everyone!!!
> Got strange DOMAIN behavior in the following plpgsql code:
>
> But i expect that lex = abc!
>
> So default value of DOMAIN type is not set in pgplsql block but:
>
> Is this correct behavior??
>
>
> If you read the create domain sql command documentation carefully the
> default clause is only used when inserting into a table that has a
> column of the domain type that is not explicitly provided a value.
> Each language deals with domains differently and the behavior you
> expect is not currently implemented in pl/pgsql. If you want a
> default inside the procedure you need to declare one explicitly.
>
> David J.

--
Alex Ignatov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus


From: "David G(dot) Johnston" <david(dot)g(dot)johnston(at)gmail(dot)com>
To: Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru>
Cc: "pgsql-sql(at)postgresql(dot)org" <pgsql-sql(at)postgresql(dot)org>
Subject: Re: Strange DOMAIN behavior
Date: 2015-07-09 14:23:19
Message-ID: CAKFQuwbXQsGKZ6QLun_br+32cT0RUuNg8ct+B0f2fut4S3-K+w@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-sql

On Thu, Jul 9, 2015 at 10:10 AM, Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru>
wrote:

> Thank you for your reply.
> In fact i want to use domains in the following case:
>
> DROP DOMAIN lexema_str CASCADE;
> CREATE DOMAIN lexema_str TEXT DEFAULT 'abc' NOT NULL;
> DROP TYPE lexema_test CASCADE;
> CREATE TYPE lexema_test AS (
> lex lexema_str,
> lex2 BIGINT
> );
> DROP TABLE ttt;
> CREATE TABLE ttt (
> a lexema_test,
> b BIGINT
> );
> INSERT INTO ttt (b) VALUES (1);
> SELECT *
> FROM ttt;
> a | b
> ---+---
> | 1
> (1 row)
>
> a.lex is null again not 'abc' as I expected like in plpgsql.
>
> All i want is to have default values in composite types. Feature that
> Oracle have.
> I thought that with domain type it should be possible.
>
>
Please don't top-post.

So even though ​there is no default specified for column a you expect there
to be a non-null value even though the column was omitted from the insert
statement? I doubt that such a change in behavior would be accepted.

​As far as I know the less-redundant way to accomplish your goal is to
create a constructor function for the type (CREATE FUNCTION default_type()
RETURNS type) and call it where you need to interject a default.

CREATE TABLE ttt ( a lexema_test DEFAULT ROW(default_type(),
NULL)::lexema_test​ )

There likely isn't any hard reasons behind the lack of capabilities wrt.
defaults and composites/domains; its just that no one has been bothered
enough to affect change.

David J.


From: Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru>
To: "David G(dot) Johnston" <david(dot)g(dot)johnston(at)gmail(dot)com>
Cc: "pgsql-sql(at)postgresql(dot)org" <pgsql-sql(at)postgresql(dot)org>
Subject: Re: Strange DOMAIN behavior
Date: 2015-07-09 16:42:24
Message-ID: 559EA470.40201@postgrespro.ru
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-sql

On 09.07.2015 17:23, David G. Johnston wrote:
> On Thu, Jul 9, 2015 at 10:10 AM, Alex Ignatov
> <a(dot)ignatov(at)postgrespro(dot)ru <mailto:a(dot)ignatov(at)postgrespro(dot)ru>>wrote:
>
> Thank you for your reply.
> In fact i want to use domains in the following case:
>
> DROP DOMAIN lexema_str CASCADE;
> CREATE DOMAIN lexema_str TEXT DEFAULT 'abc' NOT NULL;
> DROP TYPE lexema_test CASCADE;
> CREATE TYPE lexema_test AS (
> lex lexema_str,
> lex2 BIGINT
> );
> DROP TABLE ttt;
> CREATE TABLE ttt (
> a lexema_test,
> b BIGINT
> );
> INSERT INTO ttt (b) VALUES (1);
> SELECT *
> FROM ttt;
> a | b
> ---+---
> | 1
> (1 row)
>
> a.lex is null again not 'abc' as I expected like in plpgsql.
>
> All i want is to have default values in composite types. Feature
> that Oracle have.
> I thought that with domain type it should be possible.
>
>
> Please don't top-post.
>
> So even though ​there is no default specified for column a you expect
> there to be a non-null value even though the column was omitted from
> the insert statement? I doubt that such a change in behavior would be
> accepted.
>
> ​As far as I know the less-redundant way to accomplish your goal is to
> create a constructor function for the type (CREATE FUNCTION
> default_type() RETURNS type) and call it where you need to interject a
> default.
>
> CREATE TABLE ttt ( a lexema_test DEFAULT ROW(default_type(),
> NULL)::lexema_test​ )
>
> There likely isn't any hard reasons behind the lack of capabilities
> wrt. defaults and composites/domains; its just that no one has been
> bothered enough to affect change.
>
> David J.
>
Sorry for top-post.
It is sad but constructor function for composite type doesn't work in
declare block. For example:

CREATE OR REPLACE FUNCTION new_lexema()
RETURNS lexema AS $body$
DECLARE
lex lexema;
BEGIN
lex=ROW ('tt',0);
RETURN lex;
END;
$body$
LANGUAGE PLPGSQL
SECURITY DEFINER;

DROP FUNCTION lexema_test( );
CREATE OR REPLACE FUNCTION lexema_test()
RETURNS VOID AS $body$
DECLARE
lex lexema :=new_lexema();
BEGIN
END;
$body$
LANGUAGE PLPGSQL
SECURITY DEFINER;

Then I got:
ERROR: default value for row or record variable is not supported
LINE 17: lex lexema :=new_lexema();

--
Alex Ignatov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus


From: "David G(dot) Johnston" <david(dot)g(dot)johnston(at)gmail(dot)com>
To: Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru>
Cc: "pgsql-sql(at)postgresql(dot)org" <pgsql-sql(at)postgresql(dot)org>
Subject: Re: Strange DOMAIN behavior
Date: 2015-07-09 16:50:53
Message-ID: CAKFQuwZcGygkNLQewcO-YhyMkgwax8WY9-z_4bkomZHOCp28_g@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-sql

On Thu, Jul 9, 2015 at 12:42 PM, Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru>
wrote:

> DROP FUNCTION lexema_test( );
> CREATE OR REPLACE FUNCTION lexema_test()
> RETURNS VOID AS $body$
> DECLARE
> lex lexema :=new_lexema();
> BEGIN
> END;
> $body$
> LANGUAGE PLPGSQL
> SECURITY DEFINER;
>
> Then I got:
> ERROR: default value for row or record variable is not supported
> LINE 17: lex lexema :=new_lexema();
>
>
​Undocumented limitation :(

While slightly more verbose you simply need to:

DECLARE lex lexema;
BEGIN
lex := new_lexema();
[...]

David J.


From: Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru>
To: "David G(dot) Johnston" <david(dot)g(dot)johnston(at)gmail(dot)com>
Cc: "pgsql-sql(at)postgresql(dot)org" <pgsql-sql(at)postgresql(dot)org>
Subject: Re: Strange DOMAIN behavior
Date: 2015-07-09 17:28:27
Message-ID: 559EAF3B.9070900@postgrespro.ru
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-sql

On 09.07.2015 19:50, David G. Johnston wrote:
> On Thu, Jul 9, 2015 at 12:42 PM, Alex Ignatov
> <a(dot)ignatov(at)postgrespro(dot)ru <mailto:a(dot)ignatov(at)postgrespro(dot)ru>>wrote:
>
> DROP FUNCTION lexema_test( );
> CREATE OR REPLACE FUNCTION lexema_test()
> RETURNS VOID AS $body$
> DECLARE
> lex lexema :=new_lexema();
> BEGIN
> END;
> $body$
> LANGUAGE PLPGSQL
> SECURITY DEFINER;
>
> Then I got:
> ERROR: default value for row or record variable is not supported
> LINE 17: lex lexema :=new_lexema();
>
>
>
>
> ​Undocumented limitation :(
>
> While slightly more verbose you simply need to:
>
> DECLARE lex lexema;
> BEGIN
> lex := new_lexema();
> [...]
>
> David J.
>
Thats sad =(. But it is not so gracefully to initialize composite type
in the begin block as it would be in declare block %).
Thank you for yours answers!

--
Alex Ignatov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus