Re: Leading substrings - alternatives with 8.1.3?

Back in the 7.4 era, to make leading substrings be indexed, I had to set
locale=C. See thread:

<http://archives.postgresql.org/pgsql-general/2005-02/msg00159.php>

I couldn't find any updates to this issue in the archives since then.

With 8.1.3, the default locale for my system is en_US.UTF8. Leading
substring searches result in a sequential search instead of an indexed
search. Do I still have to initdb to locale=C, or is there a better option
now?

Wes

On Mon, May 01, 2006 at 12:16:46PM -0500, Wes wrote:
> Back in the 7.4 era, to make leading substrings be indexed, I had to set
> locale=C. See thread:
>
> <http://archives.postgresql.org/pgsql-general/2005-02/msg00159.php>
>
> I couldn't find any updates to this issue in the archives since then.
>
> With 8.1.3, the default locale for my system is en_US.UTF8. Leading
> substring searches result in a sequential search instead of an indexed
> search. Do I still have to initdb to locale=C, or is there a better option
> now?

Check the documentation, but if you declare a index with (for example)
text_pettern_ops, then LIKE will work even if the rest of your DB is
UTF-8.

Have a nice day,
--
Martijn van Oosterhout <kleptog(at)svana(dot)org> http://svana.org/kleptog/
> From each according to his ability. To each according to his ability to litigate.

On 5/1/06 12:47 PM, "Martijn van Oosterhout" <kleptog(at)svana(dot)org> wrote:

> Check the documentation, but if you declare a index with (for example)
> text_pettern_ops, then LIKE will work even if the rest of your DB is
> UTF-8.

My understanding of using operator classes is that I'd have to create two
indexes for each column - one with and one without the operator class. That
is also what was indicated in the original thread. Defining multiple
indexes on a given column isn't feasible, due to the database size (100
million rows per day).

Wes

On Mon, May 01, 2006 at 01:45:35PM -0500, Wes wrote:
> On 5/1/06 12:47 PM, "Martijn van Oosterhout" <kleptog(at)svana(dot)org> wrote:
>
> > Check the documentation, but if you declare a index with (for example)
> > text_pettern_ops, then LIKE will work even if the rest of your DB is
> > UTF-8.
>
> My understanding of using operator classes is that I'd have to create two
> indexes for each column - one with and one without the operator class. That
> is also what was indicated in the original thread. Defining multiple
> indexes on a given column isn't feasible, due to the database size (100
> million rows per day).

Oh right. If you want ordinary <,=,> to work for utf-8 and be indexed,
you need to have both. OTOH, if you're happy with text_pattern_ops for
ordinary lookups, you can use ~<~, ~=~, ~>~ to use the that index for
matches..

Have a nice day,
--
Martijn van Oosterhout <kleptog(at)svana(dot)org> http://svana.org/kleptog/
> From each according to his ability. To each according to his ability to litigate.