Thread: psql wishlist: value completion

Wouldn't it be cool if someday psql could do value completion?

  # delete from person where lastname = 'Garam<tab>
  # delete from person where lastname = 'Garamond' _

  # delete from person where firstname = 'Da<tab><tab>
  # delete from person where firstname = 'Da
  Damian  Darren  Dave    David    Dawson

  # delete from person where firstname = 'D<tab><tab>
  Display all 215 possibilities? (y or n)

--
dave

On Friday 13 August 2004 12:18 pm, David Garamond wrote:
> Wouldn't it be cool if someday psql could do value completion?
>
>   # delete from person where lastname = 'Garam<tab>
>   # delete from person where lastname = 'Garamond' _
>
>   # delete from person where firstname = 'Da<tab><tab>
>   # delete from person where firstname = 'Da
>   Damian  Darren  Dave    David    Dawson
>
>   # delete from person where firstname = 'D<tab><tab>
>   Display all 215 possibilities? (y or n)

Sure it would be cool - but as far as I can tell totally impractical.

Determining all the possible expandable values for a particular column
could very well require a full-table-scan of a multi-million row
table each time you try to tab-expand. Not only would performance be
terrible but get a few users doing this and you could really hammer
the server.

Cheers,
Steve

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

El Vie 13 Ago 2004 16:27, Steve Crawford escribió:
> On Friday 13 August 2004 12:18 pm, David Garamond wrote:
> > Wouldn't it be cool if someday psql could do value completion?
> >
> >   # delete from person where lastname = 'Garam<tab>
> >   # delete from person where lastname = 'Garamond' _
> >
> >   # delete from person where firstname = 'Da<tab><tab>
> >   # delete from person where firstname = 'Da
> >   Damian  Darren  Dave    David    Dawson
> >
> >   # delete from person where firstname = 'D<tab><tab>
> >   Display all 215 possibilities? (y or n)
>
> Sure it would be cool - but as far as I can tell totally impractical.
>
> Determining all the possible expandable values for a particular column
> could very well require a full-table-scan of a multi-million row
> table each time you try to tab-expand. Not only would performance be
> terrible but get a few users doing this and you could really hammer
> the server.

This could be an extra cool-hack and not nesesarely be part of the core
distribution.

greets

j0rd1
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFBHUbuB5QJme5SU08RAhp4AJ433eG+glvfarErjdsWLXGZd1joxACdGPhr
l1TsTGx7ARJNCNen7sObTo4=
=rpFy
-----END PGP SIGNATURE-----

Steve Crawford wrote:

> On Friday 13 August 2004 12:18 pm, David Garamond wrote:
>
>>Wouldn't it be cool if someday psql could do value completion?
>>
>>  # delete from person where lastname = 'Garam<tab>
>>  # delete from person where lastname = 'Garamond' _
>>
>>  # delete from person where firstname = 'Da<tab><tab>
>>  # delete from person where firstname = 'Da
>>  Damian  Darren  Dave    David    Dawson
>>
>>  # delete from person where firstname = 'D<tab><tab>
>>  Display all 215 possibilities? (y or n)
>
>
> Sure it would be cool - but as far as I can tell totally impractical.
>
> Determining all the possible expandable values for a particular column
> could very well require a full-table-scan of a multi-million row
> table each time you try to tab-expand. Not only would performance be
> terrible but get a few users doing this and you could really hammer
> the server.

Why you need a full-table-scan ? The search will be an anchored search and
with the right operator class a index scan can occur.
However I have do admit that is nice but impratical.



Regards
Gaetano Mendola

Steve Crawford wrote:
>>  # delete from person where firstname = 'D<tab><tab>
>>  Display all 215 possibilities? (y or n)
>
> Sure it would be cool - but as far as I can tell totally impractical.
>
> Determining all the possible expandable values for a particular column
> could very well require a full-table-scan of a multi-million row
> table each time you try to tab-expand.

Well, psql could display "Warning: this would require a full-table-scan
of a multi-million row table..." or "Warning: there are more than 100
possibilities, continue?" if there are no indexes or the first SELECT
... LIMIT 101 returns 101 rows.

--
dave