In discussing implementation of
RFC-449, it was noticed that obs_lsstSim and obs_sdss both import mysqlclient in python. The only reason for this is to support a class which can select images from a database based on a query. If we can remove the dependency on the python_mysqlclient package, the need for GnuTLS as a dependency on MacOS will go away, meaning we will not need to change the install process in any way as the DM stack dependence on mariadb will disappear.
I believe it is a reasonable request for several reasons:
1. This code is not used in any active pipelines, as far as I know.
2. That code path is not tested under CI since the test is disabled if there is no database connection which is the case for all VMs running inside Jenkins.
3. The mechanism for doing just this sort of data discovery is changing when the Gen3 data abstraction system lands (in the next few months).
This is not going to happen before v15 comes out (I hope), so I suggest we just test against v15 if that's needed for the Gen3 effort.
Is this code designed to select images from our own databases, or from the SDSS data release databases? Because astroquery exists for the latter purpose. We shouldn't have any database code in the obs packages anyway.
If the code is not being used and not being properly tested then it should go. It's not sophisticated and will be easy to reproduce for the gen3 butler.
It might be nice to keep this a little longer to ensure that Gen3 Butler produces the same results, but I would tend to agree that the countervailing support costs are excessive for the benefit derived.