Обсуждение: Niladic functions

Поиск
Список
Период
Сортировка

Niladic functions

От
"D'Arcy" "J.M." Cain
Дата:
The "Integrated Document" on the web page has the following paragraph in
the section "CREATE TABLE" in chapter 20.

In the current release (v6.4), Postgres evaluates all default expressions at
the time the table is defined. Hence, functions which are "non-cacheable"
such as CURRENT_TIMESTAMP may not produce the desired effect. For the
particular case of date/time types, one can work around this behavior by
using "DEFAULT TEXT 'now'" instead of "DEFAULT 'now'" or "DEFAULT
CURRENT_TIMESTAMP". This forces Postgres to consider the constant a string
type and then to convert the value to timestamp at runtime.

This appears to be untrue.  Is this a change since 6.4 or is there
some cases where using CURRENT_TIMESTAMP will not do the expected thing?

Also, the title of the document (The PostgreSQL Development Team) seems
to be incorrect.

-- 
D'Arcy J.M. Cain <darcy@{druid|vex}.net>   |  Democracy is three wolves
http://www.druid.net/darcy/                |  and a sheep voting on
+1 416 424 2871     (DoD#0082)    (eNTP)   |  what's for dinner.


Re: [HACKERS] Niladic functions

От
"Thomas G. Lockhart"
Дата:
> The "Integrated Document" on the web page has the following paragraph 
> in the section "CREATE TABLE" in chapter 20.
> In the current release (v6.4), Postgres evaluates all default 
> expressions at the time the table is defined. Hence, functions which 
> are "non-cacheable" such as CURRENT_TIMESTAMP may not produce the 
> desired effect. For the particular case of date/time types, one can 
> work around this behavior by using "DEFAULT TEXT 'now'" instead of 
> "DEFAULT 'now'" or "DEFAULT CURRENT_TIMESTAMP". This forces Postgres 
> to consider the constant a string type and then to convert the value 
> to timestamp at runtime.
> 
> This appears to be untrue.  Is this a change since 6.4 or is there
> some cases where using CURRENT_TIMESTAMP will not do the expected 
> thing?

Sorry for being slow, but exactly which of the several assertions in the
above paragraph are not true? Just the DEFAULT CURRENT_TIMESTAMP
behavior?

> Also, the title of the document (The PostgreSQL Development Team) 
> seems to be incorrect.

My browser shows that as the second line, just under the title
"PostgreSQL", as I intended. What browser are you using?
                     - Tom


Re: [HACKERS] Niladic functions

От
"D'Arcy" "J.M." Cain
Дата:
Thus spake Thomas G. Lockhart
> > The "Integrated Document" on the web page has the following paragraph 
> > in the section "CREATE TABLE" in chapter 20.
> > In the current release (v6.4), Postgres evaluates all default 
> > expressions at the time the table is defined. Hence, functions which 
> > are "non-cacheable" such as CURRENT_TIMESTAMP may not produce the 
> > desired effect. For the particular case of date/time types, one can 
> > work around this behavior by using "DEFAULT TEXT 'now'" instead of 
> > "DEFAULT 'now'" or "DEFAULT CURRENT_TIMESTAMP". This forces Postgres 
> > to consider the constant a string type and then to convert the value 
> > to timestamp at runtime.
> > 
> > This appears to be untrue.  Is this a change since 6.4 or is there
> > some cases where using CURRENT_TIMESTAMP will not do the expected 
> > thing?
> 
> Sorry for being slow, but exactly which of the several assertions in the
> above paragraph are not true? Just the DEFAULT CURRENT_TIMESTAMP
> behavior?

Yes.  This paragraph says that CURRENT_TIMESTAMP is evaluated at the time
the table is defined yet I tested this with a current version and it
actually gets evaluated at INSERT time.

> > Also, the title of the document (The PostgreSQL Development Team) 
> > seems to be incorrect.
> 
> My browser shows that as the second line, just under the title
> "PostgreSQL", as I intended. What browser are you using?

Sorry, my misunderstanding.  I didn't realize that PostgreSQL was the
title of this particular document.  Anyway, the link to this page is
"Integrated Document" so there is a small inconsistency.

-- 
D'Arcy J.M. Cain <darcy@{druid|vex}.net>   |  Democracy is three wolves
http://www.druid.net/darcy/                |  and a sheep voting on
+1 416 424 2871     (DoD#0082)    (eNTP)   |  what's for dinner.


Re: [HACKERS] Niladic functions

От
"Thomas G. Lockhart"
Дата:
> > > This appears to be untrue.  Is this a change since 6.4 or is there
> > > some cases where using CURRENT_TIMESTAMP will not do the expected
> > > thing?
> This paragraph says that CURRENT_TIMESTAMP is evaluated at the time
> the table is defined yet I tested this with a current version and it
> actually gets evaluated at INSERT time.

You are right; I had changed the parser to force a string into the parse
tree rather than a constant. Would you be interested in sending patches
for the docs? :)

> > > Also, the title of the document (The PostgreSQL Development Team)
> > > seems to be incorrect.
> > My browser shows that as the second line, just under the title
> > "PostgreSQL", as I intended. What browser are you using?
> Sorry, my misunderstanding.  I didn't realize that PostgreSQL was the
> title of this particular document.  Anyway, the link to this page is
> "Integrated Document" so there is a small inconsistency.

The problem is that this is a document integrated from all of the other
documents for purposes of HTML formatting and browsing. I wanted to make
clear that all of the information appears somewhere in the other
documents.
                    - Tom


Re: [HACKERS] Niladic functions

От
"D'Arcy" "J.M." Cain
Дата:
Thus spake Thomas G. Lockhart
> > This paragraph says that CURRENT_TIMESTAMP is evaluated at the time
> > the table is defined yet I tested this with a current version and it
> > actually gets evaluated at INSERT time.
> 
> You are right; I had changed the parser to force a string into the parse
> tree rather than a constant. Would you be interested in sending patches
> for the docs? :)

I can't find the source doc in the distribution but it's pretty simple.
Just remove that whole paragraph.

> > Sorry, my misunderstanding.  I didn't realize that PostgreSQL was the
> > title of this particular document.  Anyway, the link to this page is
> > "Integrated Document" so there is a small inconsistency.
> 
> The problem is that this is a document integrated from all of the other
> documents for purposes of HTML formatting and browsing. I wanted to make
> clear that all of the information appears somewhere in the other
> documents.

Hence the title on the referring page.  How about changing the title on
the page itself to reflect that as well?  Something like;
 "PostgreSQL Integrated Development Document"

or some such.  No biggie but it makes it easier to remember how you got
to the document you are reading.

-- 
D'Arcy J.M. Cain <darcy@{druid|vex}.net>   |  Democracy is three wolves
http://www.druid.net/darcy/                |  and a sheep voting on
+1 416 424 2871     (DoD#0082)    (eNTP)   |  what's for dinner.


Re: [HACKERS] Niladic functions

От
Bruce Momjian
Дата:
Can someone comment on this?


> The "Integrated Document" on the web page has the following paragraph in
> the section "CREATE TABLE" in chapter 20.
> 
> In the current release (v6.4), Postgres evaluates all default expressions at
> the time the table is defined. Hence, functions which are "non-cacheable"
> such as CURRENT_TIMESTAMP may not produce the desired effect. For the
> particular case of date/time types, one can work around this behavior by
> using "DEFAULT TEXT 'now'" instead of "DEFAULT 'now'" or "DEFAULT
> CURRENT_TIMESTAMP". This forces Postgres to consider the constant a string
> type and then to convert the value to timestamp at runtime.
> 
> This appears to be untrue.  Is this a change since 6.4 or is there
> some cases where using CURRENT_TIMESTAMP will not do the expected thing?
> 
> Also, the title of the document (The PostgreSQL Development Team) seems
> to be incorrect.
> 
> -- 
> D'Arcy J.M. Cain <darcy@{druid|vex}.net>   |  Democracy is three wolves
> http://www.druid.net/darcy/                |  and a sheep voting on
> +1 416 424 2871     (DoD#0082)    (eNTP)   |  what's for dinner.
> 
> 


--  Bruce Momjian                        |  http://www.op.net/~candle maillist@candle.pha.pa.us            |  (610)
853-3000+  If your life is a hard drive,     |  830 Blythe Avenue +  Christ can be your backup.        |  Drexel Hill,
Pennsylvania19026
 


Re: [HACKERS] Niladic functions

От
Thomas Lockhart
Дата:
> Can someone comment on this?

Resolved.
                    - Tom

> > The "Integrated Document" on the web page has the following paragraph in
> > the section "CREATE TABLE" in chapter 20.
> >
> > In the current release (v6.4), Postgres evaluates all default 
> > expressions...

-- 
Thomas Lockhart                lockhart@alumni.caltech.edu
South Pasadena, California


Re: [HACKERS] Niladic functions

От
Bruce Momjian
Дата:
Was this addressed?


> The "Integrated Document" on the web page has the following paragraph in
> the section "CREATE TABLE" in chapter 20.
> 
> In the current release (v6.4), Postgres evaluates all default expressions at
> the time the table is defined. Hence, functions which are "non-cacheable"
> such as CURRENT_TIMESTAMP may not produce the desired effect. For the
> particular case of date/time types, one can work around this behavior by
> using "DEFAULT TEXT 'now'" instead of "DEFAULT 'now'" or "DEFAULT
> CURRENT_TIMESTAMP". This forces Postgres to consider the constant a string
> type and then to convert the value to timestamp at runtime.
> 
> This appears to be untrue.  Is this a change since 6.4 or is there
> some cases where using CURRENT_TIMESTAMP will not do the expected thing?
> 
> Also, the title of the document (The PostgreSQL Development Team) seems
> to be incorrect.
> 
> -- 
> D'Arcy J.M. Cain <darcy@{druid|vex}.net>   |  Democracy is three wolves
> http://www.druid.net/darcy/                |  and a sheep voting on
> +1 416 424 2871     (DoD#0082)    (eNTP)   |  what's for dinner.
> 
> 


--  Bruce Momjian                        |  http://www.op.net/~candle maillist@candle.pha.pa.us            |  (610)
853-3000+  If your life is a hard drive,     |  830 Blythe Avenue +  Christ can be your backup.        |  Drexel Hill,
Pennsylvania19026
 


Re: [HACKERS] Niladic functions

От
Thomas Lockhart
Дата:
> Was this addressed?

Yes.
                  - Thomas

> > The "Integrated Document" on the web page has the following paragraph in
> > the section "CREATE TABLE" in chapter 20.
> >
> > In the current release (v6.4), Postgres evaluates all default expressions at
> > the time the table is defined. Hence, functions which are "non-cacheable"
> > such as CURRENT_TIMESTAMP may not produce the desired effect. For the
> > particular case of date/time types, one can work around this behavior by
> > using "DEFAULT TEXT 'now'" instead of "DEFAULT 'now'" or "DEFAULT
> > CURRENT_TIMESTAMP". This forces Postgres to consider the constant a string
> > type and then to convert the value to timestamp at runtime.
> >
> > This appears to be untrue.  Is this a change since 6.4 or is there
> > some cases where using CURRENT_TIMESTAMP will not do the expected thing?
> >
> > Also, the title of the document (The PostgreSQL Development Team) seems
> > to be incorrect.


-- 
Thomas Lockhart                lockhart@alumni.caltech.edu
South Pasadena, California