Participants - please go ahead and develop these Hack-a-thon topics, and use this page to provide notes
(If you need an ASTERICS wiki account, please email firstname.lastname@example.org)
Tuesday 26 th at 4:30 PM
I ) ObsCore and other VO standards in the context of EST and solar data
Participants Morten Franz, Thomas Hederer, Carl Schaffer, François Bonnarel
Discussion was about how Solar data can be described and discovered using an ObsCOre/EPNTAP strategy.
Spectral data at KIS (and probably also fo EST) are produced by combination of spatial/spectral measurements at different time. The time differentiation may be missed inside the datasets. The Polarization is also available These datasets may be made available as 4D FITS cubes. Sequences of observation at several times may also be gathered in the same cube, which may be seen as a TimeSeries of 3D or 4D cubes. Description of all this should fit with ObsCore but has to be prototyped. Spatial constraints would however require the possibility to use any of the coordiate systems attached to the SUN body. EPNTAP solutions have to be explored ( see: https://arxiv.org/abs/1407.5738 )
Extract data subset around a sunspot could be seen as some kind of extension of sODA functionality. Has to be checked
Links to useful documents:
II ) How should services and clients behave on the transition from ADQL-2.0 to ADQL-2.1 (in particular wrt TAP's LANG parameters)?
Wednesday 27 th at 2 PM : Tme Series + time Salle de réunion downstairs
17 attendants: Carlos Rodrigo, Markus Demleitner, Marco Molinaro, Mark Taylor, Stelios Voutsinas, Margarida Castro Neves, François-Xavier Pineau, Matthieu Baumann, Mireille Louys, Francois Bonnarel, Sebastien Derriere, Zheng Meyer-Zhao, Thomas Boch, Laurent Michel, Gilles Landais, Dave Morris, Ada Nebot
2 topics discussed: Discovery and Data Model 20190228-time-series-summary
1 topic need for further discussion and didn't have the time DM annotation and utypes (focusing on data part)
Wednesday 27 th at 4 PM : Provenance next steps Salle de Cours
Laurent, Mireille, François + Nicolas Renault Tinacci, Markus Nullmeier to discuss Provenance next steps.
implementation CTA at LUPM Montpellier shown by Nicolas Renault Tinacci ( LUTH)
Demo of a database instance in Jupyter Note book
Implementation test for an activity extracting Muons’ signal from a raw observation
All DM classes implemented. ParameterDescription This re-uses some provenance classes from CTA pipe library and adds some to be compliant to the model.
Mathieu Servillat &Catherine Boisson work in another part of the pipeline , on advanced analysis steps of gamma ray data gamma_py package
Some features : roles depend on the activity description e.g role = dl0.subevent could be published as a vocabulary with a name space cta:dl0.subevent
In this example execution config attributes are embedded into the instance of Activity
Parameter not implemented in this instance because not needed . question : is it using default values? then these can be discovered via the ParameterDescription links , but the Activity —> parameter link should tell it . e.g param_default =true in Activity
* config repository: All configs for activities are stored in a specific system, and re-used when necessary via copy or via reference comment by Laurent : how to store provenance objects Elements with diffrent life cycles , like parameters attached to an Activity should be stored in a different class / table than long lasting elements . They disapear with their related Activity.
* CTE functions for recursive queries in ADQL Markus Nullmeir will write some documentation for them and provide examples .
will be tried in Prov-Tap prototype.
Wednesday 27 th at 4 PM : Platforms Salle de réunion dowstairs
Zheng, Dave, Sara, Stelios … general discussion on science platforms, what they are, what our users need, what technologies are available and how we can make them interoperable. (Mark A.)
Wednesday 27 th after Provenance: after the Provenance hack-a-thon ~5pm
If needed : François + … to discuss Multi-D standards status an next steps. Planning for the radio meeting tomorrow —> Cancelled. No participants
Wednesday 27: Transition to ESCAPE:
Francoise (not available during Tuesday 26 Hack-a-Thon), Marco + …
Tuesday 16: Transition to ADQL 2.1
Participants: Markus D., Mark T., Grégory, Stelios, Dave + others
Topic: How should servers and clients behave with respect to minor version-sharp LANG parameters in TAP queries?
There is now ADQL 2.1, and services want to accept LANG=ADQL-2.1 (as per TAP). Should they accept LANG=ADQL-2.0, too. Should they declare version 2.0 in TAPRegExt?
Should we care about minor versions at all? Mark's use case: Syntax highlighting. There already is a language selector in TOPCAT, filled from language/version from TAPRegExt.
But then there's people with hacked scripts not looking at TAPRegExt, perhaps just passing in LANG=ADQL-2.0 as gleaned from TAP. Should this work on a 2.1 only service? Yes, because there's no way it could (regularly) fail (ADQL 2.0 is a strict subset of ADQL 2.1).
Should an ADQL-2.1 server declare support for 2.0 in TAPRegExt? Mark thinks he'd not turn off syntax highlighting just because he doesn't know a version. Mark would prefer no separate declaration if the behaviour is actually the same; users would worry about a meaningless decision. It's a different thing if there's actually a behavioural difference. So: No 2.0 declaration unless there's actually different parsers behind it.
Have a versioning explanation in the ADQL spec: If you're a client and a service only declares support for a newer minor version than you know, use the latest artefacts you know about and hope for the best. Also advse to leave out the version completely unless there's a strong reason to specify it (as in: when using new features in a cross-service query to fail early – perhaps)..
ADQL-2.2 to a ADQL-2.1 service: most likely it'd work because most people won't use the new versions. Even if it fails, perhaps “I don't support DISTINCT ON” might be more useful than “I don't support ADQL 2.2”. But then: people who do this kind of thing should be able to read “Just cut away the version number” error message, and this concrete information might be useful in other use cases. So: Fail with higher versions.