Re: documentation around RE repetition metachars may need clarification

[Ed Morton]

On 5/22/2023 10:06 AM, Andrew J. Schorr wrote:
> Hi,
>
> On Sun, May 21, 2023 at 09:46:50AM -0500, Ed Morton wrote:
> > In the gawk manual 
> > underhttps://www.gnu.org/software/gawk/manual/html_node/Regexp-Operator-Details.html
> > we have this statement:
> >
> > > In POSIX |awk| and |gawk|, the ‘*’, ‘+’, and ‘?’ operators stand
> > > for themselves when there is nothing in the regexp that precedes
> > > them.
> > while in the POSIX spec 
> > underhttps://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap09.html#tag_09_04_03
> > we have this statement:
> >
> > > *+?{
> > >     The <asterisk>, <plus-sign>, <question-mark>, and <left-brace>
> > >     shall be special except when used in a bracket expression (see RE
> > >     Bracket Expression
> > >     
> > > <https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap09.html#tag_09_03_05>).
> > >     Any of the following uses produce undefined results:
> > >
> > >      *
> > >
> > >         If these characters appear first in an ERE
> > So the gawk manual statement says that /+foo/ in any POSIX awk will
> > match the literal string "+foo" while the POSIX spec statement says
> > it's undefined behavior.
> >
> > Should the gawk manual be tweaked to clarify/explain what it
> > currently says about POSIX awk since it apparently contradicts the
> > POSIX spec?
> Stupid question: when something says that the behavior is undefined, is
> it not the case that a given implementation is entitled to make its
> own choice about how to handle that situation? If so, why is gawk's
> choosing to match "+foo" at odds with POSIX? If it's "undefined", do
> you instead expect it to throw an error?
>
> Regards,
> Andy
Andy - it's not what the gawk manual says about gawks behavior that's 
the issue, it's what the gawk manual says about POSIX awk's behavior 
("`+` at the start of a regexp is a literal char") since it doesn't 
match what the POSIX spec says about POSIX awk's behavior ("`+` at the 
start of a regexp is undefined"). Yes, gawk or any other awk can do 
whatever it likes for behavior that's undefined by POSIX, as in this 
case, and no I don't expect gawk to throw an error (but a lint warning 
would be useful for people looking to write portable scripts to ensure 
they aren't doing anything undefined).

    Ed.