Re: Regex query not using index

Tom,

I was looking for another approach but didn't come across that array
syntax in my searches (perhaps because it's newer. Thanks for a
solution.

Now to end my fixation, one last item. What about the case of a null
or empty param value- is there a way to assign a condition value that
Postgres will ignore when processing the query?

This syntax results in a seq scan: WHERE fielda = Coalesce(param, fielda)
because it applies only to non-nulls

Is there another way to write this- perhaps using your array syntax on
an empty array? Basically I'd PG to ignore the condition just as it
ignores WHERE 1 = 1

On Wed, Feb 20, 2008 at 8:31 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> "Postgres User" <postgres(dot)developer(at)gmail(dot)com> writes:
>
> > My users are developers and the goal was to accept a simple
> > comma-delimited list of string values as a function's input parameter.
> > The function would then parse this input param into a valid regex
> > expression.
>
> Why are you fixated on this being a regex? If you aren't actually
> trying to expose regex capabilities to the users, you'll just be having
> to suppress a bunch of strange behaviors for special characters.
>
> ISTM that the best solution is to use an array-of-text parameter,
> along the lines of
>
> where name = any (array['Smith', 'Jones', ...])
>
> For what you're doing, you'd not actually want the array[] syntax,
> it would look more like
>
> where name = any ('{Smith,Jones}'::text[])
>
> This should optimize into an indexscan in 8.2 or later.
>
> regards, tom lane
>