[Zope-CMF] Re: [dev] DateIndex and __CEILING_DATE

[Tres Seaver]

yuppie wrote:
> Hi!
> 
> 
> Tres Seaver wrote:
> 
>> yuppie wrote:
>>
>>> As proposed in <http://collector.zope.org/CMF/191> we would need to 
>>> support old and new indexes because Zope 2.5 doesn't have DateIndex / 
>>> DateRangeIndex.
>>>
>>> What about adding additional methods to the CatalogableDublinCore 
>>> interface that return the values needed by the new indexes? This way 
>>> they could co-exist with the old indexes that use the old methods.
>>
>>
>>
>> I would rather kill off the _CEILING_DATE / _FLOOR_DATE bit 
>> altogether;  perhaps what we really need is three indexes:
>>
>>   'effectiveRange' -- a DateRangeIndex against 'effective' / 'expires';
> 
> 
> Maybe we should better use a Dublin Core term: 'available' (or 'valid'?)
> <http://dublincore.org/documents/dcmi-terms/#available>
> <http://dublincore.org/documents/dcmi-terms/#valid>

'valid' is probably a closer match, I think.

>>   'effective' -- an as-yet notional DefaultedDateIndex against
>>    'effective', mapping 'None' as "earliest"
>>
>>   'expires' -- a DefaultedDateIndex against 'expires', mapping 'None'
>>     as "latest".
> 
> 
> I'm not sure if we still need these two at all. Are they used in any 
> place where we can't use the range index instead?

There are a number of different use cases which might need one or the 
other, but not the range.  E.g., archiving wants to find only content 
which has expired, in order to remove it.  DateRangeIndex is optimized 
for a different query.

>> This DefaultedDateIndex class would derive from DateIndex, special 
>> casing its handling of 'None', presumably via a radio button on the 
>> index UI itself, which would offer:
>>
>>   "earliest"  -- map 'None' or "missing" as earlier than all possible
>>     dates
>>
>>   "latest" -- map 'None' or "missing" as later than all possible dates
>>
>>   "ignore" -- don't index content with 'None' or missing values.
>>
>> We could propose adding this functionality to the Zope core (in the 
>> DateIndex package), but supply our own implementation in the 
>> meanwhile, using importability tests.
> 
> 
> That might make sense. Are there any volunteers for this work?

Hmm...  /me could be traveling a bit soon.

Tres.
-- 
===============================================================
Tres Seaver                                tseaver at zope.com
Zope Corporation      "Zope Dealers"       http://www.zope.com