PQL requirements [Was Re: SSA-1.1]

Douglas Tody dtody at nrao.edu
Fri Apr 29 15:46:59 PDT 2011


On Fri, 29 Apr 2011, Juande Santander Vela wrote:

> On 27.04.2011, at 19:19, Patrick Dowler wrote:
>
>>> I can, however, very well see a strong need for lenient matching.
>>
>> This sounds exactly like material that should be considered carefully enough
>> to cover the general case, covered in a PQL specification, and common to all
>> services that accept PQL queries. Other query languages (eg ADQL) have their
>> own way of handling such things.
>
> I was thinking exactly that when I started reading this discussion.
>
> Is there a place in the IVOA wiki where PQL is discussed? Because I see PQL has a lot of advantages for a TAP-technology based archive backend/frontend, with SQL/ADQL not being sent, and parameters allowing for ranges, transformations...

PQL has not been worked on for some time now as it was set aside while
ObsTAP was being done.  Although there have already been prototype
implementations of PQL, it needs a stable Observation data model (the
basis of most of the query parameters) hence getting ObsTAP to a stable
state is needed before we can take PQL further.  As Pat points out it is
about time to restart work on PQL; PQL and ObsTAP prototyping with real
archives (using both ADQL and PQL) is what we need to do next with
ObsTAP now wrapping up.

All there is for PQL currently is the draft and related papers on the
main table access page:

     http://www.ivoa.net/cgi-bin/twiki/bin/view/IVOA/TableAccess

We tried to discuss this in Nara but never got to it.  Hopefully there
will be time in Naples.

The big thing likely needed for PQL in the next phase is to flesh out
the query parameters using the same ObsDM as in ObsTAP (PQL in fact
should be an alternative query mechanism for ObsTAP, as well as for
general tables/catalogs a la cone search).  But other things like
TARGETNAME need to be considered as well.  While parameter semantics may
need to be specialized for the "OO/typed" interfaces (SSA, SIA etc.) PQL
can define the generic mechanism which is assumed to be followed unless
otherwise specified in a specific object interface.

 	- Doug



More information about the dal mailing list