Improve documentation for current_setting function

On Wed, Feb 20, 2019 at 12:59:45PM +0000, PG Doc comments form wrote:

The following documentation comment has been logged on the website:

Page: https://www.postgresql.org/docs/11/functions-admin.html
Description:

Hi.

Doc is not clear about what is returned if current_setting(
'app.not_existent', true ) is called.

Does it return empty string '', because return type is 'text' or it return
NULL, because 'app.not_existent' does not exist

Testing shows:

SELECT current_setting('asdf', true);
current_setting
-----------------

\pset null (null)

SELECT current_setting('asdf', true);
current_setting
-----------------
--> (null)

How do you like the attached patch, which clarifies this?

--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ As you are, so once was I.  As I am, so you will be. +
+                      Ancient Roman grave inscription +

Hello Bruce,

How do you like the attached patch, which clarifies this?

in which case a missing parameter returns
! NULL.

not always. It can be NULL or empty string. It depends if the current
session is new or after finished transaction which makes 'local' changes
to setting

Take into account these steps: /messages/by-id/15646-0d3caee9f6266f60@postgresql.org

From my user point of view. if the boot value before transaction was
NULL then after transaction it MUST stay NULL and not empty string.

This will be consistent with that if I open new parallel session it
will return NULL.

Also I see NULL value (if setting was not defined) is useful.

Also this:
set_config( 'my.name', NULL )
must set 'my.name' to NULL or must issue warning about NULL is turned
into empty string.

Personally I dislike that NULL is converted to empty string. This
looks ugly. =(

--
Best regards,
Eugen Konkov

On Fri, Feb 22, 2019 at 11:43:54AM +0200, Eugen Konkov wrote:

Hello Bruce,

How do you like the attached patch, which clarifies this?

in which case a missing parameter returns
! NULL.

not always. It can be NULL or empty string. It depends if the current
session is new or after finished transaction which makes 'local' changes
to setting

Take into account these steps: /messages/by-id/15646-0d3caee9f6266f60@postgresql.org

Yes, I saw that recent thread.

From my user point of view. if the boot value before transaction was
NULL then after transaction it MUST stay NULL and not empty string.

This will be consistent with that if I open new parallel session it
will return NULL.

Also I see NULL value (if setting was not defined) is useful.

Also this:
set_config( 'my.name', NULL )
must set 'my.name' to NULL or must issue warning about NULL is turned
into empty string.

Personally I dislike that NULL is converted to empty string. This
looks ugly. =(

Understood.

--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ As you are, so once was I.  As I am, so you will be. +
+                      Ancient Roman grave inscription +