Re: What should /\08/ match and do?

On Thu, Jun 03, 2010 at 10:02:43AM -0600, karl williamson wrote:
> I'm still digesting what people have written.  But I
>
> What I'm hearing is that it's not ok to add warnings on long-standing  
> possible usages that don't make sense at least to the person proposing  
> the message, such as me with empty {}, and Ævar with the single  
> character sub names.  FWIW, I think that it's a mistake to not warn on  
> using a reserved term; I don't think that putting it into the docs alone  
> suffices.  And there would be more unhappy people when the construct  
> they were using suddenly goes away than there would be to get a  
> deprecation warning in advance that they are using a reserved term.
>
> But that's what I'm hearing.  But, Abigail originally said that it was  
> ok for \08 to warn; and Nicholas agreed; but then I hear in the latest  
> post on the subject from Abigail that "I'd say that pointing this  
> potential problem out in the docs should be enough."

I made that remark regarding to the more general concatenation problem:

    $a = '(.)\1';
    $p = qr /${a}1/;  # Equivalent to /(.)\011/

I do not think this should warn - it's not warning now, and, IMO, it's 
uncommon enough to not introduce a new warning. [*]

The warning on use of \18 already exists - regardless whether the \18
appears as is, or comes from a concatenation [+]. The only change I suggested
was the wording of the warning (as the '8' isn't ignored).

>                                                       I can go either  
> way.  The patch I had prepared simply silenced the warning; I can add  
> text to the documents, perhaps using his sly examples, as Yves  
> suggested.  And I think it is a good idea to change them to stress using  
> \g{} for backreferences.
>
> But I'm not sure we have consensus.


[*]  Unless one can manage to have a very accurate detection - that is, 
     concatenating a variable whose value ends in \\[0-7] with something
     that starts with [0-7]. My worry for new warnings for existing constructs
     is always false positives.

[+]  Although, when using C<< $a = '\1'; /${a}8/ >> Perl will warn, but
     is likely to do what the programmer intends. Eliminating that case 
     would be nice.

Abigail