Deprecating image and spectrum as registry servicetype-s

[msdemlei]

I'm also promising all-VO-searches in some other way in the docs (which perhaps I shouldn't).

This is supposed to address bug #429.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 80.10%. Comparing base (ded4360) to head (8c9c011).
Report is 14 commits behind head on main.

❗ Current head 8c9c011 differs from pull request most recent head e4e8c9f. Consider uploading reports for the commit e4e8c9f to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #449      +/-   ##
==========================================
- Coverage   80.38%   80.10%   -0.28%     
==========================================
  Files          52       52              
  Lines        6189     6027     -162     
==========================================
- Hits         4975     4828     -147     
+ Misses       1214     1199      -15     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[msdemlei]

On Fri, Aug 04, 2023 at 11:37:56AM -0700, Brigitta Sipőcz wrote: E.g. I strongly feel instead of this, one should be able to do one `image` search, that returns the ~90 services. ``` import pyvo image_services_v1 = pyvo.regsearch(servicetype='sia') image_services_v2 = pyvo.regsearch(servicetype='sia2') v2_irsa = [s for s in image_services_v2 if 'irsa' in s.ivoid] v1_irsa = [s for s in image_services_v1 if 'irsa' in s.ivoid] ```

[for the record: I'd much rather do the ivoid constraint on the server side; if what you're doing here is something people want to do, let's add a ``partial`` option to the registry.Ivoid constraint] Well, the trouble with this is: What would you do with such a resource list? I think almost always people would do this kind of thing to query all these services. This then would look like this: for rec in pyvo.regsearch(servicetype='image-meaning-any-sia'): svc = reg.get_service() # trouble if there's two sia caps: which # one do we return here? res = svc.query(????) # depending on what SIA version we have, # users have to pass different parameters for rec in res: ???? # depending on the SIA main version, you get different # sorts of records you need to treat in totally different # ways Frankly, I simply don't see a use case where this image-meaning-any-sia constraint would yield behaviour useful to any user – unless we abstracted the various versions away with a pyVO-specific ImageService class. And even with that, you still haven't started with obscore -- which can also contain images. My conclusion is that way the VO has worked out, normal astronomers have an almost zero chance of successfully doing an "All-VO image search" just using the pyvo.registry and pyvo.dal modules themselves. I'm now more convinced of that than ever, because I've just written a bit of code that I hope will evolve into an All-VO querier. See PR#470 to get an idea what's involved (and for a number of items where I'd be grateful for good ideas). And so I'd still like to warn people already in the next release that neither 'image' nor 'spectrum' does not do what they could think it does and that it likely never will. And I'd appreciate contributions to the my global-discovery branch.