RE: [VOTE] changing Octstr comparison behaviour (was: converting to and from HTML 4 entities)

My vote is to use strcmp only and let the underlying libc to do the proper
work concerning locale.

Angel Fradejas
Mediafusión España, S.A.
[email protected]
www.mediafusion.es
Tel. +34 91 252 32 00
Fax +34 91 572 27 08


-----Mensaje original-----
De: [email protected] [mailto:[email protected]]En
nombre de Oded Arbel
Enviado el: lunes 23 de septiembre de 2002 20:00
Para: Rene Kluwen / Chimit Software Solutions
CC: [email protected]
Asunto: [VOTE] changing Octstr comparison behaviour (was: converting to
and from HTML 4 entities)



> -----Original Message-----
> From: Rene Kluwen / Chimit Software Solutions

> MHO on A couple of things:
> - strcmp is (should be) Locale specific, according to POSIX.
> - The difference in octstr_compare and octstr_str_compare are at least
> inconsequent. They should both use either use strcmp or memcmp to
> avoid confusions (my vote is for strcmp).

I'd like to put a vote up on the list for changing the current behaviour of
the comparison functions in octstr.[ch] to be consistant, and whether they
should all use memcmp() or strcmp().

I'm voting to change the current behaviour so that it will be consistent and
use only strcmp() or stricmp() internally.

> - The discussion about whether entities should be compared to
> Locale or
> to byte order is more or less void. Entities do not contain
> locale-specific characters by definition. Because that is why entities
> are used for.

oh, right :-) thanks for pointing that out.

--
Oded Arbel
m-Wise mobile solutions
[email protected]

+972-9-9581711 (116)
+972-67-340014

::..
If they wrote error messages in Haiku ?
  The ten thousand things
  How long do any persist?
  Netscape, too, has gone.





> OA> -----Original Message-----
> OA> From: Andreas Fink [mailto:[email protected]]
> OA> Sent: Monday, September 23, 2002 4:13 PM
> OA> To: Oded Arbel
> OA> Cc: [email protected]
> OA> Subject: Re: Octstr support for converting to and from
> HTML 4 entities.
>
>
>
>
>
> OA> Oh. that was me - sorry. its no hard rule, but basicly
> what I meant is - don't use c strings for doing complicated
> stuff that the Octstr sub system handles better and more
> safely. do use c-string
> OA> library calls for simple stuff that the library call
> would do better, especially if the call doesnot modify the
> content of the string (hence - no risk for buffer overflows).
> for example, when you
> OA> want to compare octstrings you can use
> strcmp(octstr_get_cstr(.... or you can use octstr_compare().
> this is the border line case IMO, where in this case I would
> have used the octstr_compare() as
> OA> it involves fewer calls. you have already demonstrated
> that you can choose correctly between complexity and safety,
> and you can make the same decisions here. most importantly -
> don't take
> OA> anything said by Kannel developers , especially me, as
> hard truth - those are simply recomendations. if we hate your
> code so much (not that I do) - we can change when it is in
> the CVS :-)
>
>
>
> OA> strcmp(octstr_get_cstr(... and octstr_compare() behave
> DIFFERENTLY.
> OA> if I remember correctly octstr_compare is case sensitive
> where strcmp is more or less simply comparing the bytes.
>
> OA> strcmp should be locale sensitive as per the POSIX
> standard, IIRC. it has nothing to do with case sensitivity.
>
> OA>   I wouldnt count on it to compare on upper/lowercase and
> even diacritical characters.
>
> OA> Well - you should. on the contrary - while strcmp is (or
> at least should be) locale aware, octstr_compare() uses
> memcmp() internally, which does simple byte comparison. OTOH,
> octstr_str_compare()
> OA> which IMO should behave no different then
> octstr_compare() uses strcmp() internally.
>
>
> OA> as you can see comparing strings is not always obvious.
> but it should be at least consistent. So if you can write a
> parameter in a config file in upper or lower case and it
> doesn't really matter,
> OA> suddendly using strcmp would make the behaviour really strange.
>
> OA> if you want to be case insensitive, you can use stricmp()
> or octstr_case_compare(). since Rene uses
> octstr_str_compare() which internally uses the locale
> sensitive strcmp(), maybe we should
> OA> change it to use octstr_compare() somehow so that
> comparison would be byte ordered (locale insensitive) and
> sort the entity list to match ?
>
> OA> octstr.c is an abstraction layer. Thats what it is there
> for. So use it for that purpose.
> OA> and frankly, the memory checking stuff is really useful
> in it. I'm fixing a few dozen memory leaks by using it. It
> points with the finger to it all the time. Very easy to fix this way.
>
> OA> stdc sting.h stuff doesn't give you any of that.
>
> OA> That's why its always a good idea to use it. sometimes
> though, if you are careful, the std lib calls provide extra
> performance which is also important.
>
> OA> --
> OA> Oded Arbel
> OA> m-Wise mobile solutions
> OA> [email protected]
>
> OA> +972-9-9581711 (116)
> OA> +972-67-340014
>
> OA> ::..
> OA> If you don't care where you are, then you ain't lost.
>
>
>
>
> --
> Best regards,
>  Rene                            mailto:[email protected]
>
>