Thread: BUG #13310: Please update docs of regexp_replace()

The following bug has been logged on the website:

Bug reference:      13310
Logged by:          Thomas Güttler
Email address:      guettli.postgres@thomas-guettler.de
PostgreSQL version: 9.4.1
Operating system:   All
Description:

Please improve the example of regexp_replace():

I am missing a "group insert" in the example.

I would replace this example:

    regexp_replace('Thomas', '.[mN]a.', 'M')     ThM

With this one:

    regexp_replace('fooooooobar', 'f(o*)bar', '\\1')

It can be any other example, but a "group insert" (\\1) would be very nice.

I talk about this:

 Table 9-6. SQL String Functions and Operators

 http://www.postgresql.org/docs/9.4/static/functions-string.html

On Mon, May 18, 2015 at 2:11 AM, <guettli.postgres@thomas-guettler.de>
wrote:

> The following bug has been logged on the website:
>
> Bug reference:      13310
> Logged by:          Thomas G=C3=BCttler
> Email address:      guettli.postgres@thomas-guettler.de
> PostgreSQL version: 9.4.1
> Operating system:   All
> Description:
>
> Please improve the example of regexp_replace():
>
> I am missing a "group insert" in the example.
>
>     regexp_replace('fooooooobar', 'f(o*)bar', '\\1')
>
>  http://www.postgresql.org/docs/9.4/static/functions-string.html


=E2=80=8BThis example assumes non-standard conforming strings.  Standard co=
nforming
strings, which are what should be documented, would give two possibilities:

regexp_replace('','', E'\\1')
or
regexp_replace('','','\1')

My personal preference is that if you do not intend to have the string
literal actually escape something the second form, without the 'E'scape
specifier, should be used.=E2=80=8B However, it seems that many people take=
 the
option requiring less thought and always prepend the E''.  The
documentation elsewhere, I think, tends to take that approach as
well...though I haven't gone a tried to figure out statistics.

David J.