Re: New substring matching rule (ITS#3041)

[alexandre.pauzies@linagora.com]

Le vendredi 26 Mars 2004 17:55, Kurt D. Zeilenga a écrit :
> Without a formal technical specification (e.g., an RFC), your
> new rule addition is likely will not be accepted.
>
> I suggest instead you look at ways to improve the approximate
> matching rules and submit patches in that area.  Note that
> the technical specification for these rules allows a great
> deal of freedom on how to implement approximate matching,
> that is, it's basically a local matter.
>
> Kurt
>
> At 02:37 AM 3/26/2004, alexandre.pauzies@linagora.com wrote:
> >Full_Name: Alexandre PAUZIES
> >Version: 2.1.27
> >OS: Gentoo GNU/Linux
> >URL:
> > http://etudiant.epita.fr/~pauzie_a/openldap/subapprox_matchingrule.patch
> > Submission from: (NULL) (62.4.23.8)
> >
> >
> >I added a new matching rule which allow substring searchs without taking
> > care of accents (this could be useful for latin laguages because actually
> > we can't really use approx with phonetic).
> >
> >Please change the OID of the rule and maybe it's name.

Hi,

maybe the name of the patch causes confusion.

Let me explain :

- this patch is not a new approximation system ;

- this patch is an improvment of the actual substring search mechanism (a new 
matching rule), which allow (like you done it with caseExact/caseIgnore) 
latin languages to choose if they want, or not, to take care of accents.

As i said, it works like caseExact/caseIgnore, it meens that no change is 
required to the substring match function, only the normalization flag is 
changed by the matching rule.

I know that this kind of matching rule is not described on an RFC but it could 
be very useful for all latin languages.

Also, people intersted to do this must change the substring matching rule of 
some attributes in their schema files, it meens they must understand what 
they're doing.


I followed your advise and looked at approximate matching rule, but i don't 
think it will be a better way to use approx for doing a substring search.

Here, it's really a (more flexible) substring search that is requested, not an 
approximative one, approx could be use for something else.

Of course i could copy the substring search function and paste it in approx 
but it doesn't makes sence and makes me lose the possibility to use approx. 
An other problem by doing this in approx, is that wildcards are not allowed 
on approx filter by the ldapsearch client so it require more modifications to 
get it work.

Maybe you could tell me an other and proper way to do this.

I have looked at lang attribute option but i don't know how i can change the 
normalization flags corresponding.

Thanks.

Alex
-- 
Alexandre Pauziès <alexandre.pauzies@linagora.com>
LINAGORA - Société de services en Logiciels Libres
http://www.linagora.com/ - Tél: 01.58.18.68.28