ENH: siav2 implementation doesn't allow "contain coordinate" queries

[bsipocz]

The first use case in the SIAv2 document is a simple contain or not question

find data that includes specified coordinates (e.g. for some object)

Yet, with POS it's not possible to query for this, one always have to pass on a search radius or define a range/polygon.

Would you think there could be/should be a way to work around this? E.g. in pyvo could default to very small radius when only a skycoord is passed along for a coordinate?

[molinaro-m]

I am no pyvo expert at all, however I think that defaulting to a "small radius" on the client side (i.e. when preparing the query) should be fine.

It might be the case to alert DAL of this because POS=CIRCLE... needing a radius > 0 (from DALI, unless I read it wrong) is, in a sense, a contradiction with the reported use case.

Also, SIAv2 inherited only the discovery ("atlas" and "pointed") service types of SIAv1, and the latter allowed for SIZE=0 exactly for those cases. So, maybe, there's something to clarify on the specification itself (i.e. take it into account for the new DAP proposal).

[msdemlei]

On Wed, Jul 30, 2025 at 12:21:09PM -0700, Brigitta Sipőcz wrote: bsipocz created an issue (astropy/pyvo#692) Would you think there could be/should be a way to work around this? E.g. in pyvo could default to very small radius when only a skycoord is passed along for a coordinate?

That sounds like a good idea. SIA2 says: A POS constraint is satisfied if the specified shape intersects the bounds (s_region in the ObsCore [7] data model) of the observation. That's nice because in that way using a 1 arcsec radius will result in pretty much what a sane user expects: Datasets that contain the point they were looking for. Very strictly speaking, we may fetch metadata for images that contain a patch 1 arcsec next to the point; but since that's probably an optimistic guess for the WCS accuracy of many images in their outer regions, I'd count that as a feature rather than a bug.

[bsipocz]

Also, SIAv2 inherited only the discovery ("atlas" and "pointed") service types of SIAv1, and the latter allowed for SIZE=0 exactly for those cases. So, maybe, there's something to clarify on the specification itself (i.e. take it into account for the new DAP proposal).

@molinaro-m - what would be the workflow to do this? A while back we run into something similar with Inf ranges, too, but afaik that hasn't been propagated to the specs either.

[molinaro-m]

An email to dal@ivoa.net would be nice, but maybe an issue on https://github.com/ivoa-std/DAP could also work.
Then document editors, authors or group coordination will possibly act upon and, if needed, bring in the TCG.