Uploaded image for project: 'Request For Comments'
  1. Request For Comments
  2. RFC-433

Update LoadAstrometryNetObjectsTask to use config instead of env variables

    Details

    • Type: RFC
    • Status: Withdrawn
    • Resolution: Done
    • Component/s: DM
    • Labels:
      None

      Description

      The dataset name and data directory for LoadAstrometryNetObjectsTask are defined via environment variables, whereas for LoadIndexedReferenceObjectsTask they are configurables. Thus, there is no common api for identifying the dataset name for Colorterms, and switching between the catalog types may require setting those environment variables.

      If we want to keep A.Net refcats viable, we need to at least somewhat modernize the Task and keep the APIs consistent. This RFC proposes to remove the SETUP_ASTROMETRY_NET_DATA and ASTROMETRY_NET_DATA_DIR env vars and replace them with Config parameters, including a ref_dataset_name in the base class, to match LoadIndexedReferenceObjectsTask.

      An alternate option for this RFC would be to write an a.net->Indexed refcat converter, and remove LoadAstrometryNetObjectsTask.

        Attachments

          Issue Links

            Activity

            Hide
            ctslater Colin Slater added a comment -

            I suspect your alternate option—converting the refcats to the new style—is a more effective use of time.

            Show
            ctslater Colin Slater added a comment - I suspect your alternate option—converting the refcats to the new style—is a more effective use of time.
            Hide
            jbosch Jim Bosch added a comment - - edited

            Colin Slater's point that conversion may be easier is strengthed by the fact that the space of "best reference catalogs" has shrunk considerably in the past few years, with a public Pan-STARRS data release and now Gaia. It's even easier to remove an old refcat than it would be to update, and I think it's possible that all the refcats we could imagine using are already in the new format.

            The other argument for keeping the old astrometry_net loader around is I think the possibility that we'd want to use the astrometry.net matcher to do some blind solving in the future. I'm personally pretty skeptical that that possibility is enough to merit keeping astrometry.net code in the stack.

            Show
            jbosch Jim Bosch added a comment - - edited Colin Slater 's point that conversion may be easier is strengthed by the fact that the space of "best reference catalogs" has shrunk considerably in the past few years, with a public Pan-STARRS data release and now Gaia. It's even easier to remove an old refcat than it would be to update, and I think it's possible that all the refcats we could imagine using are already in the new format. The other argument for keeping the old astrometry_net loader around is I think the possibility that we'd want to use the astrometry.net matcher to do some blind solving in the future. I'm personally pretty skeptical that that possibility is enough to merit keeping astrometry.net code in the stack.
            Hide
            krughoff Simon Krughoff added a comment -

            I agree conversion, or dropping a.net style references is probably a better use of time.  My biggest worry, and one that was brought up on slack, is that most (all?) of our test data use a.net style references.  That being said, I don't think it would be much work to replace those with new style reference catalogs.

            Show
            krughoff Simon Krughoff added a comment - I agree conversion, or dropping a.net style references is probably a better use of time.  My biggest worry, and one that was brought up on slack, is that most (all?) of our test data use a.net style references.  That being said, I don't think it would be much work to replace those with new style reference catalogs.
            Hide
            Parejkoj John Parejko added a comment -

            There seems to be a consensus that replacing existing testdata a.net refcats with Gaia/PS1 indexed refcats is the preferred solution. I'll see about scheduling some work for that and doing something to help automate it.

            Watch for a future RFC about removing LoadAstrometryNetObjectsTask...

            Show
            Parejkoj John Parejko added a comment - There seems to be a consensus that replacing existing testdata a.net refcats with Gaia/PS1 indexed refcats is the preferred solution. I'll see about scheduling some work for that and doing something to help automate it. Watch for a future RFC about removing LoadAstrometryNetObjectsTask...

              People

              • Assignee:
                Parejkoj John Parejko
                Reporter:
                Parejkoj John Parejko
                Watchers:
                Colin Slater, Jim Bosch, John Parejko, John Swinbank, Kian-Tat Lim, Lauren MacArthur, Paul Price, Simon Krughoff
              • Votes:
                1 Vote for this issue
                Watchers:
                8 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved:
                  Planned End:

                  Summary Panel