I don't think many will be caught-out, since it's an overwhelmingly
daft thing to do anyway!
If you've got a numeric type column then assign numeric values to, not
strings. Or explicitly convert.
I'm sure all occurrences of this "in the wild" are due to sloppy
SQL...
For the record:
lkind@coulin:~% sql -uingres iidbdb
INGRES TERMINAL MONITOR Copyright (c) 1981, 1997 Computer Associates Intl, Inc.
OpenIngres SPARC SOLARIS Version OI 2.0/9712 (su4.us5/00) login
Tue Dec 3 16:17:27 2002
continue
* create table test(f1 int);\g
Executing . . .
continue
* insert into test(f1) values('');\g
Executing . . .
E_US0B61 line 1, You cannot assign a value of type 'varchar' to a column
of type 'integer'. Explicitly convert the value to the required type.
(Tue Dec 3 16:18:22 2002)
continue
* select int4('');\g
Executing . . .
col1
0
(1 row)
continue
Lee.
Ben-Nes Michael writes:
> Then Why not set it to NULL
>
> Seems logic as there is nothing between ''
>
> What is the solution of other dbs ( oracle, db2 .. ) to:
> insert into table (num) value (''); ?
>
> Who knows how many application will suffer becouse of this.
>
> ----- Original Message -----
> From: "Bruce Momjian" <[EMAIL PROTECTED]>
> To: "Lee Kindness" <[EMAIL PROTECTED]>
> Cc: "Ben-Nes Michael" <[EMAIL PROTECTED]>; "postgresql"
> <[EMAIL PROTECTED]>; <[EMAIL PROTECTED]>
> Sent: Tuesday, December 03, 2002 6:08 PM
> Subject: Re: [GENERAL] 7.3 -> pg_atoi: zero-length string
>
> > The change was made to tighten up the code to catch errors sooner.
> > There isn't much logic to making '' be 0, and no one could make a case
> > for keeping such a mapping.
> >
> > --------------------------------------------------------------------------
> -
> >
> > Lee Kindness wrote:
> > > Was it necessary? No idea, you're welcome to search through the
> > > pgsql-hackers archives to determine the reasoning behind the change. I
> > > believe the change was made by Bruce Momjian (going by the release
> > > notes). I only remember reading the discussion in passing.
> > >
> > > This is also one of the reasons for beta releases - to allow people to
> > > test against the new version and pick up these sort of things. Either
> > > this gives them/you time to make changes, or to lobby to get the old
> > > behaviour back.
> > >
> > > If I were in your situation i'd probably hack back the old behaviour
> > > to 7.3, compile and run that while changes were made. Or stick with
> > > 7.2.x until changes were made to your applications, have you got a
> > > 'big carrot' for going with 7.3?
> > >
> > > Lee.
> > >
> > > Ben-Nes Michael writes:
> > > > That's indeed very nice but I don't see the logic in it.
> > > >
> > > > If I want to upgrade I need to go on all my projects and change
> thousands of
> > > > lines.
> > > >
> > > > And that's not all :(
> > > > I have other applications like phprojekt that was not developed by me
> and
> > > > became useless now as I cant insert.
> > > >
> > > > Was this step so necessary ?
> > > >
> > > > ----- Original Message -----
> > > > From: "Lee Kindness" <[EMAIL PROTECTED]>
> > > > To: "Ben-Nes Michael" <[EMAIL PROTECTED]>
> > > > Cc: "postgresql" <[EMAIL PROTECTED]>; "Lee Kindness"
> > > > <[EMAIL PROTECTED]>
> > > > Sent: Tuesday, December 03, 2002 12:37 PM
> > > > Subject: [GENERAL] 7.3 -> pg_atoi: zero-length string
> > > >
> > > > > Hi, one of the changes in 7.3 was to disallow '' being implicitly
> > > > > converted to 0. In your example below image_order is clearly not a
> > > > > text/char column, so what are you trying to set it too? If you want
> it
> > > > > be 0 then explicitly use 0, if you want it to be undefined then use
> > > > > NULL.
> > > > >
> > > > > Lee.
> > > > >
> > > > > Ben-Nes Michael writes:
> > > > > > Hi
> > > > > >
> > > > > > After I upgraded 7.2.3 to 7.3 I started to get the following
> errors:
> > > > > > pg_atoi: zero-length string
> > > > > >
> > > > > > its seems that i get it when not all field have content:
> > > > > >
> > > > > > this is one example that generate the error:
> > > > > > insert into images (section_id, pic_date, image_order) values
> ('8',
> > > > > > '2002-12-03', '')
> > >
> >
---------------------------(end of broadcast)---------------------------
TIP 4: Don't 'kill -9' the postmaster
