Re: regexp_replace grief

Thx, Albe. I tested both proposals, and I gravitate now more towards the
E'' nomenclature since it avoids something like this:

$pg_bs_char = ( $dbh->{pg_server_version} >= 90100 ) ? "\\" : "\\\\";

Hoping for a long half life of the E'' nomenclature ...

-ar

On Fri, Apr 12, 2013 at 5:16 AM, Albe Laurenz <laurenz(dot)albe(at)wien(dot)gv(dot)at>wrote:

> Armin Resch wrote:
> > Not sure this is the right list to vent about this but here you go:
> >
> > I) select regexp_replace('BEFORE.AFTER','(.*)\..*','\1','g') "Substring"
> > II) select regexp_replace('BEFORE.AFTER','(.*)\\..*','\\1','g')
> "Substring"
> >
> > Executing (II) against pg 8.4.4 or 9.0.4 yields 'BEFORE', but in order
> for 9.1.7 to yield the same one
> > has to execute (I) .. bummer
>
> To be immune against different settings of standard_conforming_strings,
> use the extended string literal syntax:
>
> select regexp_replace('BEFORE.AFTER',E'(.*)\\..*',E'\\1','g') "Substring";
>
> That will work in all versions.
>
> Yours,
> Laurenz Albe
>