I believe XMATCH is just a built-in function.
If we add a spatial library for footprints and overlaps and point in
polygon and ... then that will be a function library as well.
That's the rationale for including table-valued functions in the language.
So, Bob Mann can define his own function that for each object O, gives
the Mann-Type1-Cross-Match (or type 2 ...) objects.
This function has as parameters, object O and any other params Bob
wants to add.
It returns a table of objects.
Then one can use CROSS APPLY to invoke that function and get the answer
set.
The point is that astronomers need a way to do these table-valued
functions and do cross-apply in order to do their particular kind of
science.
We are going to get to this point eventually.
Having the current spec is a good start in that direction.
Of course these functions have to be programmed in some language (Java, C#, Fortran, Python) and have to be downloadable to any skynode. That is a lot of mechanism, but... What alternative is there?
-----Original Message-----
From: owner-voql-at-eso.org [mailto:owner-voql-at-eso.org] On Behalf Of
Masatoshi OHISHI
Sent: Sunday, July 24, 2005 3:37 AM
To: voql-at-ivoa.net
Subject: Re: comments on ADQL v1.01 & SNI v1.01
Bob,
My apologies for my very delayed response. m(_ _)m
In our Kyoto meeting we made extensive discussion to adopt currently proposed XMATCH method. Sevral of our colleagues proposed to adopt other method as well. And our German collegues reported their wonderful experience to include errors into the XMATCH functions.
It seemed to me that people realized that it would take very long to investigate how our XMATCH should be, and people agreed to adopt currntly proposed XMATCH scheme, as a temporaty solution, and also people agreed that we need to work more on detailed aspects of the XMATCH functions. The upgraded XMATCH function will be incorporated into ADQL / SKyNodeInterface version 2.
The arguments contained in the current XMATCH specification was made
based on discussion at CDS to guarantee extensibility of the XMATCH
function.
If you have serious difficulty on this format, and if you have much
better format, please don not hesitate to propose it.
Cheers,
Masatoshi Received on 2005-07-25Z05:46:44