Versioning and filenames

For a dataset that is downloaded only once

Filenames are written thus (major.minor.patch):

• 0.1.0.toml
• 0.1.0.csv
• 0.1.0.h5

Major changes only for the initial stable release (when you go from 0.y.z to 1.0.0), and when you have incompatible changes.
Minor changes when you're adding new functionality (e.g. a new component) or for the initial release that you think is probably right (when you go from 0.0.z to 0.1.0).
Patch changes for small bug fixes.

You can increase minor numbers a bit more freely than that if you're making a bigger change, or you want to make a new and improved script that definitely works better even though the output is technically the same, but those are the "official" rules. For example, you might want to start the disease data with 0.20200722.0, etc. until you're really confident that it's all good and you're happy to make a 1.2020mmdd.0 release, it's up to you.

For a dataset that is downloaded daily

Filenames are written thus (named after the download date):

• 1.20200716.0.csv
• 1.20200716.0.h5

If there's a bugfix to the dataset, then you'd go from 1.20200716.0.csv to 1.20200716.1.csv.

However if you have a completely new storage format, where you have different components and/or different formats of the components, then you go from 1.20200716.0.csv to 2.20200716.0.csv, because major number changes are supposed to not be backward compatible.