Hi Graham,
On Wednesday 13 August 2003 22:01, Graham Barr wrote:
> On Wed, 2003-08-13 at 20:50, Kurt D. Zeilenga wrote:
> > I don't think the LDAP API should do any trancoding, the API should
> > be a simple conduit between the application and the wire. Any
> > transcoding desires should be done by the application (by directly
> > calling APIs specifically designed to do transcoding).
>
> I have been following this thread a bit. And while I would agree that
> the API (Net::LDAP in this case) should not do transcoding, I don't see
> any reason why it cannot provide hooks to make the application
> developers life easier.
I remember a thread this spring about perl-ldap, Convert::ASN1 and Unicode
support.
There were some issues with character semantics.
Are they still there or is it possible to feed strings with character
semantics into perl-ldap and get strings in character semantics back ?
Oops, wanting them back in character semantics is dangerous (as I wrote in
various previous mails in this thread ;-)), because it needs knowledge of the
data (schema, ...) But maybe here an option to get_value() can help.
The idea is to have a string e.g "Hägar" (Latin1) in character semantics
"Hägar" (UTF8, but also only 5 characters long because of the character
semantics: the "ä" is 2 bytes, but only one character) and accept this
regular Perl string as an input to operations in perl-ldap.
Of course even a string like "Any \x{0021} string \N{SMILEY FACE}" should
work ;-)
I know it works with byte semantics where "Hägar" is a string of length() six
and Perl has no idea that the "Ã" and the "¤" are actually the UTF8 encoding
for the Latin1 "ä".
I also know the character semantics will not work with versions < 5.8 because
there the Unicode support was not so complete (IIRC the utf8 flag was
lexically scoped and not an attribute of each variable).
For the checks whether a string is in byte or character semantics and the
appropriate conversion from character semantics to byte semantics the Encode
module (or Perl's 5.8.1 utf8 package) should do
e.g. with Encode:
# convert to byte semantics if string is in character semantics
$octets = encode("utf8", $string) if (is_utf8($string));
When reading an attribute's value, an additional option [e.g. chars => 1]
can tell get_value() to use character semantics instead of the default byte
semantics. This allows the user to get Perl strings from attributes he knows
to be encoded in UTF8.
e.g.
# get givenName as a string in character semantics
$string = $entry->get_value('givenname', chars => 1)
# get jpegPhoto as a sequence of bytes
$octets = $entry->get_value('jpegPhoto');
Peter
PS: Having written this I notice that the conversion from character semantics
strings to UTF8-encoded byte semantics in perl-ldap when writing might be
risky too since it assumes that this attribute is UTF8 encoded in LDAP.
Here, I think, the risk is tolerable as long as byte semantics is supported in
perl-ldap and the behaviour with character semantics is explanied in the man
page.
PPS: For me this is all pure theory since I only have Perl 5.6.0 at work (and
for compatibility's sake at home too).
--
Peter Marschall
eMail: [EMAIL PROTECTED]
