ClimaDiagnostics.jl
provides a simple framework to add diagnostics to CliMA
simulations.
ClimaDiagnostics.jl
defines two important concepts:
DiagnosticVariable
: A recipe to compute a diagnostic from the integrator
alongside with names, units, comments.ScheduledDiagnostic
: When to compute and output the DiagnosticVariable
and
what type of accumulation to perform.To add the diagnostics to a simulation from a list of ScheduledDiagnostic
s,
just redefine your integrator with IntegratorWithDiagnostics(integrator, scheduled_diagnostics)
.
The documentation is rich and comprehensive, please find more information there.
These guidelines aim to ensure consistent code quality, maintainability, and a
smooth collaborative workflow for ClimaDiagnostics.jl
. Please, read these
guidelines even if you are familiar with other CliMA packages as there may be
some differences.
We prioritize well-tested code to guarantee ClimaDiagnostics.jl
functions
reliably. Here are some principles we follow:
test
folder and are exclusively thereThis means that:
Pkg.test()
;Pkg.test()
with MPI/GPUs.Manifest.toml
filesWhile checking in Manifest.toml
files ensures reproducibility, it also
introduces some nuisance, including:
In this repository, we have three environments:
The project environment defines the test dependencies in its extras
(to reduce
the number of environments and to avoid the "cannot merge projects" problem).
The buildkite environment, in .buildkite
, contains everything that is required
to test ClimaDiagnostics.jl
in its most general configuration (ie, including
CUDA.jl
and MPI.jl
). As the name suggests, this is the (only) environment
used by our Buildkite pipeline.
:note: Please, open an issue if you find workflow problems/friction with this system.
ClimaDiagnostics.jl
defines the test dependencies directly in the main
Project.toml
. This means that the package can be tested simply by running ] test
in a Julia REPL, as shown below:
Start a Julia session in the ClimaDignostics
directory:
julia --project
Enter Pkg
mode by typing ]
. This will change the prompt. Run test
.
When doing so, Julia
will start a new temporary environment where the tests
are run in isolation. Tests are running checking for in-bounds and for
deprecations, and this can result in code invalidation and new precompilation.
Note, the project environment does not contain the test dependencies. Therefore, you will find that some dependencies are missing if you try "manually" run the test in a REPL. There are two ways around this:
TestEnv
in your base environment (julia -e 'using Pkg; Pkg.add("TestEnv")'
). Then, when you want to use the test dependencies,
activate it from your REPL with using TestEnv; TestEnv.activate()
. This will
bump you to an environment where the test dependencies are available..buildkite
environment.:note: Please, open an issue if you find workflow problems/friction with this system.
JuliaFormatter.jl
One of the tests consists in checking that the code is uniformly formatted. We use JuliaFormatter.jl to achieve consistent formatting. Here's how to use it:
You can either install in your base environment with
julia -e 'using Pkg; Pkg.add("JuliaFormatter")'
or use it from within the TestEnv
or the .buildkite
environments (see previous section).
Then, you can format the package running:
using JuliaFormatter; format(".")
or just with format(".")
if the package is already imported.
The rules for formatting are defined in the .JuliaFormatter.toml
.
If you are used to formatting from the command line instead of the REPL, you can
install JuliaFormatter
in your base environment and call
julia -e 'using JuliaFormatter; format(".")'
You could also define a shell alias
alias julia_format_here="julia -e 'using JuliaFormatter; format(\".\")'"
:note: Please, open an issue if you find workflow problems/friction with this system.
Documentation is generated with Documenter.jl. We strive to have complete and up-to-date information.
To generate documentation, run
julia --project=docs docs/make.jl
Please, update the documentation if you add new features or change the behavior of existing ones.
Here's how to structure your contributions effectively:
Your pull request can contain one or multiple commits. In either cases, it is important that each commit is atomic (meaning that each commit represents a single logical change).
Please, squash commits that represent a single logical change (e.g., do not have two commits when the second just fixes the first).
Pull requests are not merged, but rebased, ensuring a linear history (this is handled automatically by GitHub).