contrib features during beta period

"Christopher Kings-Lynne" <chriskl@familyhealth.com.au> writes:

Is it ok if Florian and I submit the improvements to the fulltextindex
contrib during beta?

What improvements are we talking about here? FTI is sufficiently widely
depended on that I think it ought to follow the same quality standard
as the main backend ... viz, "no new features during beta".

regards, tom lane

Well basically we have it all done, except neither of us has time to test at
the moment. Florian is free after sept 1st. Basically it adds full word
match support, stop words but keeps full backward compatibility. We are
getting a segfault in the new function (fti2) but the old is still working
fine.

I've also updated the docs and added a new WARNING file that strongly
discourages the use of contrib/fulltextindex in favour of contrib/tsearch.

Also, there's regression tests, etc.

I personally have switched to tsearch now exclusively but I promised Florian
I'd review his stuff so I want to get it in.

Maybe I'll try to figure out the bug today or tommorrow and commit it if I
can...

Chris

On Friday 30 August 2002 12:49 am, Tom Lane wrote:

"Christopher Kings-Lynne" <chriskl@familyhealth.com.au> writes:

Is it ok if Florian and I submit the improvements to the fulltextindex
contrib during beta?

What improvements are we talking about here? FTI is sufficiently widely
depended on that I think it ought to follow the same quality standard
as the main backend ... viz, "no new features during beta".

Does this mean we should be looking for a way to integrate it into the main
backend at this point? Isn't that what contrib is for?
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11

Lamar Owen <lamar.owen@wgcr.org> writes:

On Friday 30 August 2002 12:49 am, Tom Lane wrote:

What improvements are we talking about here? FTI is sufficiently widely
depended on that I think it ought to follow the same quality standard
as the main backend ... viz, "no new features during beta".

Does this mean we should be looking for a way to integrate it into the main
backend at this point? Isn't that what contrib is for?

Well, given that Chris also thinks that people should migrate to
tsearch, I'd guess that fulltextindex would not be the one to integrate.
But yeah, in the long run I'd like to see one of these packages become
mainstream, just because full-text search is such a widely used feature.

The more global point here is that during beta we need everyone to be
doing testing, bug fixing, and documentation, not new features ---
even if they're in contrib. Goodness knows there's not very much
organization to the Postgres project, but let's try to respect what
little process we do have...

regards, tom lane

On Friday 30 August 2002 09:29 am, Tom Lane wrote:

Lamar Owen <lamar.owen@wgcr.org> writes:

Does this mean we should be looking for a way to integrate [FTI] into the
main backend at this point? Isn't that what contrib is for?

Well, given that Chris also thinks that people should migrate to
tsearch, I'd guess that fulltextindex would not be the one to integrate.
But yeah, in the long run I'd like to see one of these packages become
mainstream, just because full-text search is such a widely used feature.

Agreed. We don't really have guidelines for this process; any idea what was
the last module moved over like that? There may be more modules to consider
-- earthdistance, for instance, which has been with us for a long time.

The more global point here is that during beta we need everyone to be
doing testing, bug fixing, and documentation, not new features ---
even if they're in contrib. Goodness knows there's not very much
organization to the Postgres project, but let's try to respect what
little process we do have...

ABSOLUTELY CORRECT. :-)
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11

Lamar Owen <lamar.owen@wgcr.org> writes:

On Friday 30 August 2002 09:29 am, Tom Lane wrote:

But yeah, in the long run I'd like to see one of these packages become
mainstream, just because full-text search is such a widely used feature.

Agreed. We don't really have guidelines for this process; any idea what was
the last module moved over like that?

It hasn't happened often, but I recall that the int8 code started life
as a contrib module. Let's see, looking at CVS I see
bit
dateformat
datetime
int8
ip_and_mac
metaphone
sequence
soundex
statmath
unixdate
as contrib subdirectories that aren't there anymore in current sources,
and that I think got merged, not dropped.

There may be more modules to consider
-- earthdistance, for instance, which has been with us for a long time.

Actually, that seems to have just disappeared completely. What the heck?
Marc, you're not physically removing stuff from CVS when you move it to
gborg, are you?

regards, tom lane

On Fri, 30 Aug 2002, Tom Lane wrote:

Lamar Owen <lamar.owen@wgcr.org> writes:

There may be more modules to consider
-- earthdistance, for instance, which has been with us for a long time.

Actually, that seems to have just disappeared completely. What the heck?
Marc, you're not physically removing stuff from CVS when you move it to
gborg, are you?

I think Marc put earthdistance in a separate CVS module, i.e. separate
from pgsql-server but included in the pgsql meta-module.

--
Alvaro Herrera (<alvherre[@]dcc.uchile.cl>)
"Estoy de acuerdo contigo en que la verdad absoluta no existe...
El problema es que tu est�s mintiendo y la mentira s� existe" (G. Lama)

Alvaro Herrera <alvherre@atentus.com> writes:

On Fri, 30 Aug 2002, Tom Lane wrote:
-- earthdistance, for instance, which has been with us for a long time.

Actually, that seems to have just disappeared completely. What the heck?
Marc, you're not physically removing stuff from CVS when you move it to
gborg, are you?

I think Marc put earthdistance in a separate CVS module, i.e. separate
from pgsql-server but included in the pgsql meta-module.

Hmm, you're right. It appears that that setup does not actually work
very well. I tried rm -rf contrib/earthdistance, after which I find
that "cvs update" does not bring back the earthdistance subdirectory;
it seems I have to do a full checkout to get back in sync. I suppose
this is a CVS bug. And it's one that worries me a lot more than the
annoyance that cvsweb won't show a merged view. I wonder whether cvs
update would even notice changes inside the earthdistance module?

regards, tom lane