Add versions to packages in environment.yml

[abostroem]

The CAC discussed the fact that we've had instances of package updates that break this lesson. We recommend that the maintainers add version numbers to the environment.yml file that are updated (after testing is done on the new versions) periodically (quarterly? semi-annually?)

[rudyphd]

We would probably also want to add a version note to setup.html for version and update schedule of the environment file?

Is there any indication of the scope of the warnings?

[abostroem]

I think the fact that the setup asks you to use the environment file means we don't need much detail in the setup. At most we could add a sentence saying that specific versions are given in the environment file if you plan on building/using a personal environment.

The issues we've encountered:

• When using describe on the best neighbor multiplicity column of the panstarrs table, we get an numpy warning 'partition' will ignore the 'mask' of the MaskedColumn. We added a call out to the lesson that this can be ignored
• When creating a pandas data frame, we used to create the data frame and then add new columns on the fly to the data frame. A pandas update caused this to produce an error when describe was called on the data frame. We fixed this by adding the columns on the fly to the astropy table before it was converted to a pandas dataframe
• (Note - this would not be fixed with versioning): An ADQL update in the Gaia database caused the POLYGON function to not work on the proper motion column. We fixed this by switching to using BETWEEN instead of POLYGON

[rudyphd]

Will test this with my clean set-up on my new laptop I will check the most recent versions work and produce a YAML from there if it works and version backwards if needed.