Which decision have to be fulfilled for a 1.0 release?

[jan-janssen]

As @liamhuber correctly pointed out, 90% of unit test coverage, is a necessary but not sufficient condition for a 1.0 release. So I guess we should discuss which decisions have to be fulfilled to qualify for a 1.0 release.

From my perspective the majority of the decisions:

• 25-001-split-api.md
• 25-002-black-formatting.md
• 25-003-ruff-formatting.md

Are reasonable requirements for a 1.0 release. For others like 25-004-hatch-pyproject.md and 25-005-project-structure.md it is more a recommendation rather than a requirement. I believe it makes sense to follow the packaging standard but if anybody wants to publish a package using a different built system, for example for a package combining Python and C/C++ code, I could see a different packaging system be beneficial.

I guess from here we have two directions:

• Either we require all decisions to be fulfilled and adjust the decisions which are primarily recommendations to be more specific for example by restricting them to pure Python packages.
• Or we introduce a list of required decisions for 1.0 release and a second list of recommendations.

[liamhuber]

I fully agree with the "necessary" and "recommended" decision groups you suggest above.

From the most recent decision:

decisions/decisions/26-001-unit-test-coverage.md

Line 18 in 3f48296

For the 1.0 release which indicates a software is ready for production use, our users expect our software to be stable.

My main point is that all of these are "necessary but not sufficient", because we could have a fairly polished piece of software that we know is not "stable" because of some planned and imminent (<1/6/12 months?) change in the feature set or architecture that will feel unstable. Personal e.g., I expect pyiron_workflow to depend directly on flowrep recipes and python parsing.

Nonetheless, I think these being fulfilled is a very good hint that a v1 is viable and a good idea. Personal e.g. I should probably just have pyiron_workflow-1.0.0 be the last version not depending on flowrep, and go ahead and have a 2.0.0 once it does. We don't want new major versions every week, but even an annual cadence is not insane to me.