Bump v0.43.0

[ali-ramadhan]

In the spirit of getting closer to continuous delivery (see https://www.oxinabox.net/2019/09/28/Continuous-Delivery-For-Julia-Packages.html#what-if-i-dont-want-to-release-right-now--dev-versions) we should probably tag and release v0.43.0 soon since PRs #1070, #1057, and #1061 + couple of bug fix PRs would be useful to have.

Release notes:

• Fixes a bug in TwoDimensionalLeith (PR TendencyTermField (or something like it) for diagnosing exact tendency terms and fluxes #1073, issue Bug in TwoDimensionalLeith closure #1034). Previously tests were being skipped due to extreme slowness. Now we run tests on GPU (but not CPU, where the closure is much slower to compile).

• Rewrites the interface for "scheduling" output and diagnostics (PR AbstractSchedules for scheduling output and diagnostics #1070). Previously output and diagnostics were usually scheduled by specifying either time_interval or iteration_interval kwargs in the constrcutor for the object in question. Now, the relevant kwarg is called schedule and takes a callable AbstractSchedule object (or any user-defined function func that returns true or false depending on the single argument func(model)). This design is more flexible and extensible, and also simplifies underlying code. Four schedules are provided:

  • TimeInterval(interval)
  • IterationInterval(interval)
  • WallTimeInterval(interval)
  • AveragedTimeInterval(interval; window=interval, stride=1) (for time-averaging output).

Breaking changes:

• Output writers and diagnostics no longer have the keyword arguments time_interval or iteration_interval. The most commonly-used features that are affected are JLD2OutputWriter, NetCDFOutputWriter, and Checkpointer. JLD2OutputWriter and NetCDFOutputWriter no longer have the kwargs time_averaging_window and time_averaging_stride. The specific syntax changes are:

  • time_interval=T becomes schedule=TimeInterval(T)
  • iteration_interval=I becomes schedule=IterationInterval(I)
  • time_interval=T, time_averaging_window=W becomes schedule=AveragedTimeInterval(T, window=W).

[codecov]

Codecov Report

Merging #1077 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #1077   +/-   ##
=======================================
  Coverage   57.47%   57.47%           
=======================================
  Files         161      161           
  Lines        3800     3800           
=======================================
  Hits         2184     2184           
  Misses       1616     1616           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b11f047...688feaa. Read the comment docs.