[Gambas-user] gb.pcre: RegExp.Replace() problem

[Benoît Minisini]

Le 21/12/2014 16:18, Tobias Boege a écrit :
> On Sun, 21 Dec 2014, Beno?t Minisini wrote:
>> Le 21/12/2014 14:33, Beno?t Minisini a ?crit :
>>> Le 20/12/2014 09:27, Tobias Boege a ?crit :
>>>> Hi,
>>>>
>>>> attached is a script which essentially executes
>>>>
>>>>     Print RegExp.Replace(" * * a *", "^[ ]*\\*", "'")
>>>>
>>>> On my system the output is
>>>>
>>>>     '' a *
>>>>
>>>> which I don't understand. As the regular expression indicates I want only
>>>> the one match at the beginning of the line to be replaced. Not the second
>>>> one behind it. The "a" acts like a border, to see if the "^" is
>>>> completely
>>>> ignored -- but it isn't.
>>>>
>>>> The error seems to come from
>>>>
>>>> --8<-- [ trunk/gb.pcre/src/regexp.c
>>>> ]---------------------------------------
>>>> 389                         r.subject = &subject[offset];
>>>> --8<------------------------------------------------------------------------
>>>>
>>>>
>>>> I guess that this should prevent recursive (indefinite) substitutions
>>>> in the
>>>> replacement string... But if you ask me, the result of my script is
>>>> wrong.
>>>>
>>>> Regards,
>>>> Tobi
>>>>
>>>>
>>>
>>> Mmm. I'm far from being a regexp user, but I will try to check that...
>>>
>>
>> I think you should not use Replace() for that. It tries to mimic the
>> behaviour of Replace$(), i.e. finding each match and replace it, by
>> going forward in the string.
>>
>> Maybe could I add a special case for pattern beginning with "^", which
>> then means "just replace once"?
>>
>
> Hmm, that would solve my problem and indeed "^" is a special case ("$"
> should not have this problem). I would say that this addition is worth it
> if one can document the new behaviour in just a few sentences that are not
> utterly complicated.
>
> But as I see you already committed. What is the concrete behaviour now? If
> the pattern starts with "^", then
>    (a) the rest of the pattern is replaced only once (not necessarily at the
>        beginning of the subject); or
>    (b) the pattern is replaced once at the beginning of the subject?
>
> In (a), the "^" acts merely as a flag and in (b) it acts as a flag AND
> retains its anchor meaning.
>
> Regards,
> Tobi
>

The (b) answer.

If the pattern starts with a "^", it just stops replacing as soon as we 
are not at the beginning of the string, as then the "^" cannot match.

Regards,

-- 
Benoît Minisini