Should BoutDataset.save() record the version of xBOUT used to produce the files?

[johnomotani]

Might be useful for provenance tracking and reproducibility to know what version of xBOUT was used to produce some file. Similar to boutproject/boutdata#79.

[ZedThree]

Oh, I perhaps should've thought about this on that other PR, it would be good to have this set of metadata as attributes:

• title
• software_name
• software_version
• netcdf_version
• date_created
• id

I've cobbled this set together from various conventions, including netCDF and the Climate Forcasting conventions, and I've been trying to get it into various output files.

I think we might already have some of these?

[bendudson]

Quick bikeshedding question: Should software_version be a semantic or at least sequential version number, or a git checksum, or a combination?

[ZedThree]

Yes :)

Personally, I'm a big fan of setuptools_scm's scheme, which is SemVer compatible. It's basically:

• no distance and clean:
  {tag}
• distance and clean:
  {next_version}.dev{distance}+{scm letter}{revision hash}

So a release would be 1.2.3 say, and the first commit after that would get version 1.2.3.dev1+gdeadbeef

[johnomotani]

👍 I was gonna say what @ZedThree said.

[dschwoerer]

Offtopic: Isn't the next one 1.2.4.dev1+gdeadbeef?

[ZedThree]

Good catch, yes. I think there's some setting to make it bump the minor version instead too.