Re: autocomplete - SELECT fx

Hello

I tested Peter's patch and it works well.

I don't see any performance problem on my notebook

last discussion
http://archives.postgresql.org/message-id/CAK3UJRHXgnChP9J5RuOEMEtFB0E9awPhNOV=RQuPwK6VX-ZaPA@mail.gmail.com

so we can do more - we can join this completation with "," too - but
it can add false compleations inside FROM clause.

Probably we are near of end of enhancing completation - for doing
usable completation we need syntax analysing :(

Regards

Pavel Stehule

On Mon, Jul 2, 2012 at 1:13 AM, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com> wrote:

> I tested Peter's patch and it works well.

I liked it as well. But I'm not sure what should happen with the patch
now. It seems like it'd be commit-ready with just a tweak or two to
the query as I noted in my last mail, but Tom did seem opposed[1] to
the idea in the first thread, and no one else has spoken up recently
in favor of the idea, so maybe it should just be marked Rejected or
Returned with Feedback?

Josh

[1] http://archives.postgresql.org/message-id/13612.1328887227%40sss.pgh.pa.us

2012/7/6 Josh Kupershmidt <schmiddy(at)gmail(dot)com>:
> On Mon, Jul 2, 2012 at 1:13 AM, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com> wrote:
>
>> I tested Peter's patch and it works well.
>
> I liked it as well. But I'm not sure what should happen with the patch
> now. It seems like it'd be commit-ready with just a tweak or two to
> the query as I noted in my last mail, but Tom did seem opposed[1] to
> the idea in the first thread, and no one else has spoken up recently
> in favor of the idea, so maybe it should just be marked Rejected or
> Returned with Feedback?

Hard to say - it is usable, but on second hand - some user can be
surprised, because it works only for first item in list.

Regards

Pavel

>
> Josh
>
> [1] http://archives.postgresql.org/message-id/13612.1328887227%40sss.pgh.pa.us

On Thu, Jul 05, 2012 at 06:43:09PM -0700, Josh Kupershmidt wrote:
> On Mon, Jul 2, 2012 at 1:13 AM, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com> wrote:
>
> > I tested Peter's patch and it works well.
>
> I liked it as well. But I'm not sure what should happen with the patch
> now. It seems like it'd be commit-ready with just a tweak or two to
> the query as I noted in my last mail, but Tom did seem opposed[1] to
> the idea in the first thread, and no one else has spoken up recently
> in favor of the idea, so maybe it should just be marked Rejected or
> Returned with Feedback?

I like the patch, as far as it goes. It's the natural addition to the
completions we already offer; compare the simplistic completion after WHERE.
Like Pavel and Robert, I think a delightful implementation of tab completion
for SELECT statements would require radical change.

Thanks,
nm

On Sat, Jul 7, 2012 at 5:43 PM, Noah Misch <noah(at)leadboat(dot)com> wrote:
> I like the patch, as far as it goes. It's the natural addition to the
> completions we already offer; compare the simplistic completion after WHERE.
> Like Pavel and Robert, I think a delightful implementation of tab completion
> for SELECT statements would require radical change.

Thanks for the feedback, Noah. Peter, are you interested in posting an
updated version of your patch? (The only problems I remember are
checking attisdropped and function visibility.)

Josh

On tis, 2012-07-10 at 07:29 -0700, Josh Kupershmidt wrote:
> On Sat, Jul 7, 2012 at 5:43 PM, Noah Misch <noah(at)leadboat(dot)com> wrote:
> > I like the patch, as far as it goes. It's the natural addition to the
> > completions we already offer; compare the simplistic completion after WHERE.
> > Like Pavel and Robert, I think a delightful implementation of tab completion
> > for SELECT statements would require radical change.
>
> Thanks for the feedback, Noah. Peter, are you interested in posting an
> updated version of your patch? (The only problems I remember are
> checking attisdropped and function visibility.)

Another problem was exluding functions that are not useful to call
directory, such as functions that used for type or operator definitions.
There is no simple solution for that.