release 1.2.3 or 1.3.0 in March

[newville]

Merging #941 and #942, I think we're approximately ready for a release -- it's been too long.

Are there other outstanding issues with code, documentation, or examples that we should try to squeeze in?

While reviewing the changes so far, and also #941, which is kind of a "behavior change", I think we can also ask whether this should be called 1.2.3 or 1.3.0? I have no strong opinion.

Maybe we can aim for March 15 or April 1?

[newville]

If I don't hear objections, I'll work on a ChangeLog to tag the current release (or, approximately so) as version 1.2.3.

Are there any other suggestions for changes to code, examples, project infrastructure, or docs?

[reneeotten]

if you haven't had the chance to work on this I might be able to do so sometime this week so that we can get a release out soon-ish? Let me know as I would rather not duplicate the work ;)

[newville]

@reneeotten Yeah, non-coding work has been taking up all my time at work the past few weeks, and the uncertainties project and other projects are causing a lot of traffic. Well, and it didn't seem like a huge rush ;).

But, I have this about half done -- I think I can get it done this week. FWIW, the rush at work should take a break in the next week or two..

[eendebakpt]

I tested lmfit-py locally with numpy2.0.0rc1 and all tests pass. See numpy/numpy#24300 (comment).

So if a release goes out before numpy 2.0 is released that would be great!

[newville]

@reneeotten Here is a draft changelog. Any suggestions? I guess we could always alter it later.

Changelog:

New features:

• add 'min_rel_change' as an optional variable in the calculation of confidence intervals with
  conf_interval(). (Update ConfidenceInterval class constructor parameters #937).

• Model.eval_uncertainty now takes an optional dscale parameter (default value of 0.01) to
  set the step size for calculating derivatives (Eval uncertainty dscale #933).

• add calculation of predicted_interval to Model.eval_uncertainty (Eval uncertainty dscale #933).

Bugfixes/enhancements:

• restore best-fit parameter values for high accuracy values of constrained values (PR restore best-fit parameter values for high accuracy values of constrained values #907)

• improvement to Model for the difference between "variable parameter", "independent variable", and "option".
  With this change, keyword arguments to model functions with non-numerice default values such as
  do_thing=True, or form='linear' has those arguments become clearly identified as independent variables,
  and use the provided values as default values. (More uniform independent vars #941)

• the dill package is now required. (Require dill #940)

• the numexpr is no longer required.

• better saving/loading saved states of Model now use dill, have several cleanups, and are now versioned for
  future-proofing. Also, propagate funcdets for Parameters when loading a Model. (propagate funcdefs to Parameters in model.loads #932, Saveload custom func #934)

• adding an example using basinhopping, can take other methods as command-line argument

• in the TNC method, maxfun is used instead of maxiter.

• fix broken links in the Examples gallery

• fix inter-sphinx mapping to scipy docs.

• updates for compatibility with numpy 2.0

• fix tests for Python 3.12 and Python 3.13-dev

• build swicthes to pyproject.toml (CI/BLD changes #928)

• drop support for Python 3.7, increase minimum numpy verstio to 1.23 and scipy to 1.8 (Drop support for Python 3.7, increase minimum versions for SciPy/NumPy #927)

• fix bug calculating r-squared for fits with weights (Model.fit() gives the wrong 'rsquared' when 'weights' is given #921, Rsquared with weights #923)

• fix for modelresult.eval_uncertainty() after load_modelresult (loaded ModelResult bug (eval_uncertainty() not working) #909)

• use StringIO for pandas.read_json.

• add test for MinimizerResult.uvars after successful fit (Minimize uvars test #913)

[reneeotten]

that looks good to me, without going through the PR list myself ;)

You can remove the line:
the numexpr is no longer required. - this was never a requirement and only (incorrectly) snug in at some point after the previous release, so this was never something the end-user observed.

Some aspects should probably be moved to the category Maintenance/Deprecations: as we usual have in the whatsnew.rst file. Other than that this looks fine and we can indeed always fine-tune it in the documentation if needed.

I would prefer to release this as version 1.3.0 given the changes, but I'll happily leave it up to you.

[newville]

@reneeotten Thanks - I agree with all of that, and will do.

[newville]

OK, I updated the whatsnew.rst, tagged 1.3.0, pushed to GitHub, and to PyPI. I rebuilt the docs, including PDF, EPUB, and zipped-html. I pushed that to gh-pages. I'm not sure why, but the https://lmfit.github.io/lmfit-py/ has not yet been updated from that yet (even after 2+ hours, and refreshing). I hope that happens soon!