Obscore 1.1 Erratum 3: Drop obs_id non-NULL requirement

[BONNAREL FRANCOIS]

Le 27/04/2022 à 15:04, Markus Demleitner a écrit :
> Hi Pat,
>
> On Tue, Apr 26, 2022 at 08:28:34AM -0700, Patrick Dowler wrote:
>> In the very last sentence with advice for query writers, "obs_id is not
>> null" will omit results which is likely not what people want. Instead, they
>> just need to treat NULL obs_id as not equal to another NULL obs_id, which
>> is normal database semantics for NULL values. Thus, NULL obs_id(s) always
>> indicates distinct observations.
> Fair enough.  I've changed the language to:
>
>    If this erratum is accepted authors of queries inspecting obs_id
>    are advised to not group observations with a NULL obs_id; NULL
>    obs_ids always indicate distinct observations.
>
> (where I'd be open to dropping the if clause in that statement).
>
> This perhaps also addresses Mireilles question...
>
> On Tue, Apr 26, 2022 at 07:54:27PM +0200, Mireille LOUYS wrote:
>> What does it mean to fill obs_id with NULL?
> ...somewhat more clearly.
>
> Still, I second her plea:
>
>> May be archive designers , providers can clarify this ?
> If you are running an obscore service or if you are running queries
> against obscore tables, please have a look at the impact assessment
> of https://wiki.ivoa.net/twiki/bin/view/IVOA/ObsCore-1_1-Erratum-3
> and see if I've missed something.
>
> Thanks,
>
>            Markus

Dear all,

As this may be discussed this evening at the TCG, I suddenly remember I 
had to check this before it happens ;-) !

With this definition the situation is better than with the simple 
relaxing. However .....

In the radio domain (see JIVE service for example) and also the High 
energy domain we often face the case where several dataproducts are 
produced from the same observation. But we can imagine services where 
some observations contain several dataproducts and some others only a 
single one (just by chance).

If you want to aggregate all the obs_publisher_did, or (s_ra, s_dec) or 
whatever property of the products belonging to the same observations I 
think the GROUP BY will fail if we relax "obs_id = null".

And it's easy to create obs_id from obs_publisher_did in the case of 
unique dataproduct in an observation

Cheers

François