Re: model inherited object

> Wrong; you need a (start_time, end_time) pair for a temporal fact.
> When the fact is not a completed event yet, use NULL for the end_time.
> This has to be in one database to work.

but can there be more than one null end date for a particular location? Because in what-if modelling my 2007 change does not complete my 2005 location.

> Procedures that build a new row with the proper time stamps, the
> changes and the "un-changes" in it.

Aaahh = but how do you record the "unchange" and then when I want a view of a location at a certain date, won't I have to do a massively complex select, maybe trying to analyze many "history" rows of the same location to see the correct values for all attributes?

Phil

-CELKO- wrote:
> >> I have many databases that want access to some relatively static data that is relevant to all of them i.e Location. Each of these database represents a period in time, 2005, 2006 etc and the locations will remain pretty much the same. But say LOC01 has 200 sq ft in 2005 but in 2007 (because of building work) it will have 400
> sq ft. However it also has 2 toilets (which have not changed because of
> the work). <<
>
> That is a temporal problem and not an OO problem. This is a job for
> Rick Snodgrass and SQL/MM proposals {play fanfare here} !!
>
> >> So I record tuple X as (ID,Sq Ft, Toilets, ValidFrom) as LOC01 - 200 - 2 - 2005 <<
>
> Wrong; you need a (start_time, end_time) pair for a temporal fact.
> When the fact is not a completed event yet, use NULL for the end_time.
> This has to be in one database to work.
>
> >> .. in fact those 2 toilets are inherited <<
>
> That is not quite what I had in mind when I my mother to leave me the
> family porcelain :) Sorry, too good a straight line not to use it.
> This is change to the same entity over time and not OO inheriting.
>
> >> So, how should this be done? <<
>
> Procedures that build a new row with the proper time stamps, the
> changes and the "un-changes" in it.
Received on Wed Jun 21 2006 - 22:59:12 CEST