This is my "tracking effort" ticket for the SuperTask WG hack week, but I think it makes sense to use it for the formal code review of the prototype in obs_base, where I spent the vast majority of my time last week. I've created a tickets/DM-10600 branch that's just a copy of u/jbosch/repodb-initial, so the existing PR can be used:
https://github.com/lsst/obs_base/pull/45
Nate Pease [X], I think it probably makes sense to ask you to do the formal review. Hopefully with the classes now documented it'll be easier to parse than it was last week. I'd also of course love comments from anyone else who has time to look, especially Kian-Tat Lim, Andy Salnikov, Gregory Dubois-Felsmann, and Paul Price.
It should be emphasized that this prototype is very incomplete. In addition to the concerns mentioned in the RepoDatabase design notes in the docs, all it really supports right now are skymap-based datasets, which are the simplest since they don't involve any per-camera specialization. I have ideas on how to extend it to camera-based datasets, and I hope to find time to prototype that out a bit more before handing this work entirely off to someone else, but I have other work that needs to take priority right now.
This is my "tracking effort" ticket for the SuperTask WG hack week, but I think it makes sense to use it for the formal code review of the prototype in obs_base, where I spent the vast majority of my time last week. I've created a tickets/
DM-10600branch that's just a copy of u/jbosch/repodb-initial, so the existing PR can be used:https://github.com/lsst/obs_base/pull/45
Nate Pease [X], I think it probably makes sense to ask you to do the formal review. Hopefully with the classes now documented it'll be easier to parse than it was last week. I'd also of course love comments from anyone else who has time to look, especially Kian-Tat Lim, Andy Salnikov, Gregory Dubois-Felsmann, and Paul Price.
It should be emphasized that this prototype is very incomplete. In addition to the concerns mentioned in the RepoDatabase design notes in the docs, all it really supports right now are skymap-based datasets, which are the simplest since they don't involve any per-camera specialization. I have ideas on how to extend it to camera-based datasets, and I hope to find time to prototype that out a bit more before handing this work entirely off to someone else, but I have other work that needs to take priority right now.