Many Parquet datasets now default to Astropy instead of Pandas

Last night I merged DM-34875, which changes the butler “storage class” for many (but not all) of our Parquet datasets (including objectTable[_tract] and sourceTable[_visit]) from pandas.DataFrame to astropy.table.Table. This will change the Python type you get back from a butler only in new data repositories where those dataset types were not already registered.

I plan to rename many of these dataset types in the future, at which point they will get new dataset type registrations in our long-lived data repositories with the new storage class. In the meantime, please be aware that the Python type you get back for these may change from repo to repo.

1 Like

Another excellent reason for the DP1 repo to be created from scratch and not from an export from /repo/main (which will also be messing up our TaskMetadata serialization)