Notify argument?

On Tue, 2002-03-19 at 08:22, Jeff Davis wrote:

I am curious, why does notify not support a string argument of some kind, to
pass to the other connections? It seems it would be a little more useful.

You can pass data around using temp tables.

Is this due to oracle compatibility issues?

Actually, I think that Oracle's implementation of this feature actually
allows for a user-specified string argument.

Is it too difficult for it's usefulness?

AFAICT it shouldn't be too difficult. However, there is a note in the
TODO list referring to breaking backwards compatability with the
"pgNotify API". Exactly how backwards compatible do we need to be?

Cheers,

Neil

--
Neil Conway <neilconway@rogers.com>
PGP Key ID: DB3C29FC

Neil Conway wrote:

I am curious, why does notify not support a string argument of some kind, to
pass to the other connections? It seems it would be a little more useful.

Actually, I think that Oracle's implementation of this feature actually
allows for a user-specified string argument.

Commercial Ingres allowed one to specify a string also. I'm guessing
that the feature was not implemented in PostgreSQL *not* because there
is some good database design reason to leave it out, but rather because
someone did not bother to put it in.

AFAICT it shouldn't be too difficult. However, there is a note in the
TODO list referring to breaking backwards compatability with the
"pgNotify API". Exactly how backwards compatible do we need to be?

imho not much in this case (though of course we may find a way to be
very compatible when someone actually implements it). I had found the
equivalent feature very useful when building a large data handling
system a few years ago, and I'd think that it would be useful in
PostgreSQL also. Comments?

- Thomas

On Wed, 2002-03-20 at 10:24, Thomas Lockhart wrote:

AFAICT it shouldn't be too difficult. However, there is a note in the
TODO list referring to breaking backwards compatability with the
"pgNotify API". Exactly how backwards compatible do we need to be?

imho not much in this case (though of course we may find a way to be
very compatible when someone actually implements it). I had found the
equivalent feature very useful when building a large data handling
system a few years ago, and I'd think that it would be useful in
PostgreSQL also.

Okay, I'll implement this.

Cheers,

Neil

--
Neil Conway <neilconway@rogers.com>
PGP Key ID: DB3C29FC

Neil Conway wrote:

Is it too difficult for it's usefulness?

AFAICT it shouldn't be too difficult. However, there is a note in the
TODO list referring to breaking backwards compatability with the
"pgNotify API". Exactly how backwards compatible do we need to be?

The breakage will come when we lengthen NAMEDATALEN, which I plan to
tackle for 7.3. We will need to re-order the NOTIFY structure and put
the NAMEDATALEN string at the end of the struct so differing namedatalen
backend/clients will work. If you want to break it, 7.3 would probably
be the time to do it. :-) Users will need a recompile pre-7.3 to use
notify for 7.3 and later anyway.

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 853-3000
  +  If your life is a hard drive,     |  830 Blythe Avenue
  +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026

Bruce Momjian <pgman@candle.pha.pa.us> writes:

The breakage will come when we lengthen NAMEDATALEN, which I plan to
tackle for 7.3. We will need to re-order the NOTIFY structure and put
the NAMEDATALEN string at the end of the struct so differing namedatalen
backend/clients will work. If you want to break it, 7.3 would probably
be the time to do it. :-) Users will need a recompile pre-7.3 to use
notify for 7.3 and later anyway.

If we're going to change the structure anyway, let's fix it to be
independent of NAMEDATALEN. Instead of

char relname[NAMEDATALEN];
int be_pid;

let's do

char *relname;
int be_pid;

This should require no source-level changes in calling C code, thanks
to C's equivalence between pointers and arrays. We can preserve the
fact that freeing a PQnotifies result takes only one free() with a
little hacking to make the string be allocated in the same malloc call:

newNotify = (PGnotify *) malloc(sizeof(PGnotify) + strlen(str) + 1);
newNotify->relname = (char *) newNotify + sizeof(PGnotify);
strcpy(newNotify->relname, str);

Thus, with one line of extra ugliness inside the library, we solve the
problem permanently.

regards, tom lane

Bruce Momjian wrote:

Neil Conway wrote:

Is it too difficult for it's usefulness?

AFAICT it shouldn't be too difficult. However, there is a note in the
TODO list referring to breaking backwards compatability with the
"pgNotify API". Exactly how backwards compatible do we need to be?

The breakage will come when we lengthen NAMEDATALEN, which I plan to
tackle for 7.3. We will need to re-order the NOTIFY structure and put
the NAMEDATALEN string at the end of the struct so differing namedatalen
backend/clients will work. If you want to break it, 7.3 would probably
be the time to do it. :-) Users will need a recompile pre-7.3 to use
notify for 7.3 and later anyway.

Hmmm,

seems I have to get a little more familiar with the FE/BE
stuff again. Have been pretty good at that years ago.

IIRC, the FE/BE protocol itself does not limit any length or
depends on definitions like that. So that should be an
arbitrary (read bogus) usage in libpq. The TODO entry
therefore should read

Fix Notify API's usage of NAMEDATALEN.

Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== JanWieck@Yahoo.com #

_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com

On Wed, Mar 20, 2002 at 04:10:14PM -0500, Tom Lane wrote:

Bruce Momjian <pgman@candle.pha.pa.us> writes:

The breakage will come when we lengthen NAMEDATALEN, which I plan to
tackle for 7.3. We will need to re-order the NOTIFY structure and put
the NAMEDATALEN string at the end of the struct so differing namedatalen
backend/clients will work. If you want to break it, 7.3 would probably
be the time to do it. :-) Users will need a recompile pre-7.3 to use
notify for 7.3 and later anyway.

If we're going to change the structure anyway, let's fix it to be
independent of NAMEDATALEN.

Sounds good. If we're making other backwards-incompatible changes to
pgNotify, one thing that bugs me about the API is the use of "relname"
to refer to name of the NOTIFY/LISTEN condition. This is incorrect
because the name of a condition is _not_ the name of a relation -- there
is no necessary correspondence between these. The names of NOTIFY
conditions are arbitrary, and chosen by the application developer.

The same terminology is used in the backend NOTIFY/LISTEN code (e.g.
pg_listener), but the major source of incompatibility will be the change
to libpq. Would it be a good idea to rename "relname" to something more
sensible?

Cheers,

Neil

--
Neil Conway <neilconway@rogers.com>
PGP Key ID: DB3C29FC

Neil Conway wrote:

On Wed, Mar 20, 2002 at 04:10:14PM -0500, Tom Lane wrote:

Bruce Momjian <pgman@candle.pha.pa.us> writes:

The breakage will come when we lengthen NAMEDATALEN, which I plan to
tackle for 7.3. We will need to re-order the NOTIFY structure and put
the NAMEDATALEN string at the end of the struct so differing namedatalen
backend/clients will work. If you want to break it, 7.3 would probably
be the time to do it. :-) Users will need a recompile pre-7.3 to use
notify for 7.3 and later anyway.

If we're going to change the structure anyway, let's fix it to be
independent of NAMEDATALEN.

Sounds good. If we're making other backwards-incompatible changes to
pgNotify, one thing that bugs me about the API is the use of "relname"
to refer to name of the NOTIFY/LISTEN condition. This is incorrect
because the name of a condition is _not_ the name of a relation -- there
is no necessary correspondence between these. The names of NOTIFY
conditions are arbitrary, and chosen by the application developer.

The same terminology is used in the backend NOTIFY/LISTEN code (e.g.
pg_listener), but the major source of incompatibility will be the change
to libpq. Would it be a good idea to rename "relname" to something more
sensible?

Renaming the column would make an API change. I was talking only about
requiring a recompile.

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 853-3000
  +  If your life is a hard drive,     |  830 Blythe Avenue
  +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026

nconway@klamath.dyndns.org (Neil Conway) writes:

If we're going to change the structure anyway, let's fix it to be
independent of NAMEDATALEN.

Sounds good. If we're making other backwards-incompatible changes to
pgNotify, one thing that bugs me about the API is the use of "relname"
to refer to name of the NOTIFY/LISTEN condition.

I hear you ... but my proposal only requires a recompile, *not* any
source code changes. Renaming the field would break client source code.
I doubt it's worth that.

regards, tom lane

On Thu, 2002-03-21 at 00:16, Tom Lane wrote:

nconway@klamath.dyndns.org (Neil Conway) writes:

If we're going to change the structure anyway, let's fix it to be
independent of NAMEDATALEN.

Sounds good. If we're making other backwards-incompatible changes to
pgNotify, one thing that bugs me about the API is the use of "relname"
to refer to name of the NOTIFY/LISTEN condition.

I hear you ... but my proposal only requires a recompile, *not* any
source code changes. Renaming the field would break client source code.
I doubt it's worth that.

Okay, that's fair -- I'll leave it as it is.

Cheers,

Neil

--
Neil Conway <neilconway@rogers.com>
PGP Key ID: DB3C29FC

Here is a patch that implements Tom's suggestion of mallocing the
relation name string as part of PQnotify and not depending on
NAMEDATALEN. Nice trick.

---------------------------------------------------------------------------

Tom Lane wrote:

Bruce Momjian <pgman@candle.pha.pa.us> writes:

The breakage will come when we lengthen NAMEDATALEN, which I plan to
tackle for 7.3. We will need to re-order the NOTIFY structure and put
the NAMEDATALEN string at the end of the struct so differing namedatalen
backend/clients will work. If you want to break it, 7.3 would probably
be the time to do it. :-) Users will need a recompile pre-7.3 to use
notify for 7.3 and later anyway.

If we're going to change the structure anyway, let's fix it to be
independent of NAMEDATALEN. Instead of

char relname[NAMEDATALEN];
int be_pid;

let's do

char *relname;
int be_pid;

This should require no source-level changes in calling C code, thanks
to C's equivalence between pointers and arrays. We can preserve the
fact that freeing a PQnotifies result takes only one free() with a
little hacking to make the string be allocated in the same malloc call:

newNotify = (PGnotify *) malloc(sizeof(PGnotify) + strlen(str) + 1);
newNotify->relname = (char *) newNotify + sizeof(PGnotify);
strcpy(newNotify->relname, str);

Thus, with one line of extra ugliness inside the library, we solve the
problem permanently.

regards, tom lane

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 853-3000
  +  If your life is a hard drive,     |  830 Blythe Avenue
  +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026

Fix applied.

---------------------------------------------------------------------------

Tom Lane wrote:

Bruce Momjian <pgman@candle.pha.pa.us> writes:

The breakage will come when we lengthen NAMEDATALEN, which I plan to
tackle for 7.3. We will need to re-order the NOTIFY structure and put
the NAMEDATALEN string at the end of the struct so differing namedatalen
backend/clients will work. If you want to break it, 7.3 would probably
be the time to do it. :-) Users will need a recompile pre-7.3 to use
notify for 7.3 and later anyway.

If we're going to change the structure anyway, let's fix it to be
independent of NAMEDATALEN. Instead of

char relname[NAMEDATALEN];
int be_pid;

let's do

char *relname;
int be_pid;

This should require no source-level changes in calling C code, thanks
to C's equivalence between pointers and arrays. We can preserve the
fact that freeing a PQnotifies result takes only one free() with a
little hacking to make the string be allocated in the same malloc call:

newNotify = (PGnotify *) malloc(sizeof(PGnotify) + strlen(str) + 1);
newNotify->relname = (char *) newNotify + sizeof(PGnotify);
strcpy(newNotify->relname, str);

Thus, with one line of extra ugliness inside the library, we solve the
problem permanently.

regards, tom lane

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 853-3000
  +  If your life is a hard drive,     |  830 Blythe Avenue
  +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026