Release 1.0.6 - Githubissues

lewisfogden / heavylight

A lightweight actuarial modelling framework for Python

https://lewisfogden.github.io/heavylight/

MIT License

5 stars 2 forks source link

Release 1.0.6 #23

Closed lewisfogden closed 5 months ago

lewisfogden commented 5 months ago

@MatthewCaseres Think we're ready for releasing v1.0.6 to pypi soon, capturing all the latest changes.

[ ] I'm adding a template/example generator 'make_example' which can set up a few examples for users. (I'm going to use them for demos/getting started)
[ ] I've added you as an author in pyproject.toml - please set email if you want it :)

Anything else to add? (this is in branch dev_core_docs, I'll merge soon)

MatthewCaseres commented 5 months ago

I'd like to close out #9 and start using proper release notes but I can't upload to pypi so it seems like you'll have to do that one?

Another note is that we are not abiding by the rules of semantic versioning. If you had not released v1.0.0 then we would not be breaking the rules. And we still might ship more breaking changes (unless we are certain we won't) so it is a funny spot to be in https://semver.org/. This is more of a comment, I have no idea what we are supposed to do except keep iterating on the design until we are very confident and a 2.0.0 is released and we promise API stability.

lewisfogden commented 5 months ago

I've still to read (and understand) the automated building, I've got hatch set up which does it all on my own machine for now.

SemVer: Agree - I wasn't aware of semantic versioning when I first released. I think we're pretty safe to go for it as 1.0.6 and then run semver on that going forward:

not many people using the library yet (I suspect just me & you)
breaking changes aren't really breaking, e.g. I've got a Warning on do_run, and .ToDataFrame() isn't removed yet in favour of .df property.
can caveat in readme.

API wise, it feels like I've got what I want in heavylight.Model, we probably need to look at consistency between that and the optimised versions though. heavylight.Table still needs a bit of thinking on API - for now I'm leaving it as opinionated on having metadata encoded in the column names (|int, etc.)

(I could yank the current versions from pypi, and put in at 0.2.0 but it might cause other problems.)

lewisfogden commented 5 months ago

I've put in PR but tests fail for 3.8, 3.9 due to path: str|pathlib.Path not being recognised. I'll probably just drop the type hints here for now. Need to sleep now, so probably easier to merge and then fix in main.

MatthewCaseres commented 5 months ago

No rush lets wait on merging for another day, type signatures in user facing functions is usually nice. gives me some time to review

MatthewCaseres commented 5 months ago

I went ahead and merged it after writing a test and changing some code style but not the substance

lewisfogden commented 5 months ago

Thank you! Would you like to add your email address to pyproject.toml (https://github.com/lewisfogden/heavylight/blob/main/pyproject.toml) and then I'll publish to pypi?

MatthewCaseres commented 5 months ago

its changed

lewisfogden commented 5 months ago

Fab - we're live with 1.0.6: https://pypi.org/project/heavylight/1.0.6/ (Just running post-live checks now using a new colab environment)

lewisfogden commented 5 months ago

All good!