tpapp / DynamicHMC.jl

Implementation of robust dynamic Hamiltonian Monte Carlo methods (NUTS) in Julia.
Other
242 stars 21 forks source link

compat with LogDensityProblems=0.11 #154

Closed st-- closed 2 years ago

st-- commented 2 years ago

Otherwise our downstream code keeps getting CompatHelper PRs re LogDensityProblems, but we can't actually check that it works as DynamicHMC keeps back the version.

st-- commented 2 years ago

@tpapp would be great if you could merge this & register a new release!

tpapp commented 2 years ago

Apologies, I missed the first notification. Thanks for the PR and the friendly bump. Will merge and release now.

tpapp commented 2 years ago

I have to update some dependencies too, will finish this when things propagate through the registry, hopefully soon. Thanks for your patience.

tpapp commented 2 years ago

released, https://github.com/JuliaRegistries/General/pull/46261

st-- commented 2 years ago

Thanks! Just so I understand correctly, you are not following the ColPracs then? They seemed to advocate that for version>=1 each feature should lead up a minor version bump (patch bump exclusively for bugfixes) and compatibility with a new different version of another package counts is explicitly counted as a new feature the.

The other again I had is what does the ^ before the 0.9 add?

tpapp commented 2 years ago

IMO supporting another version of a dependency is not a feature addition in the API this package exposes. I never understood the reasoning behind that recommendation, but that said I am open to suggestions (and reading up on stuff if there is a discussion I did not follow).

The ^ is upward-compatibility, a no-op in this case since it is the default, just so that I could make change and trigger CI.

st-- commented 2 years ago

I agree that I would have tended to what you're saying, but e.g. the SciML ColPrac are very explicit about this... And SemVer also specifies that patch version for >1 is only for bugfixes. Might it be worth opening an issue on the ColPrac repo asking for clarification ?