[python-ldap] O/R mapper

[Michael Ströder]

Florian Friesdorf wrote:
> On Fri, 22 Apr 2011 17:15:25 +0200, Michael Ströder <michael at stroeder.com> wrote:
>> Florian Friesdorf wrote:
>>> For me it looks like the attribute conversions could be handled by
>>> subclassing and overwriting _ldap_call [3], in pseudo-code:
>>>
>>> def _ldap_call(self, func, *args, **kwargs):
>>>     args = deep_utf8_copy(args)
>>>     kwargs = deep_utf8_copy(kwargs)
>>>     res = super(...)...
>>>     if not self.convmap:
>>>        return res
>>>     return check_for_things_to_convert(res, self.convmap)
>>
>> http://www.freeipa.org is doing something like this AFAICT with decorators.
>> Check their code.
>>
>> But as repeated on this mailing list several times: In general this won't work
>> without having knowledge of the subschema. E.g. your automatic conversion
>> won't work for attributes like jpegPhoto, userCertificate, userPKCS12 etc.
> 
>>> Attribute conversion maps could be generated from automatically parsing
>>> an ldap's schemas (if supported) or be provided by the user.
> 
> For fields were automatic conversion does not work, a conversion needs
> to be specified by the user of the library. 

Yes. So converting attribute values in LDAPObject._ldap_call() is the wrong place.

>> So you could have a look how web2ldap deals with all that based on
>> schema-aware classes. Basically there are classes for each LDAP syntax
>> registered by syntax OID with several possibilities to override that for
>> certain attribute types. web2ldap itself is only interested in (Unicode)
>> strings. One could extend the concept to convert to native Python data types.
> 
> Thank you for the pointer, will definitely have a look at it.
> 
> Why do you keep this functionality in web2ldap and not in python-ldap?

The schema-ware classes are tightly coupled to web2ldap's HTML generating
needs and e.g. depend on pyweblib. You can use them as inspiration but not more.

Ciao, Michael.