ray-chew / pybella Goto Github PK

Information on common parameters is currently duplicated across input files.

This issue moves common parameters to a common location, with attention to:

• Physical Constants
• Required parameters
• List of valid choices (e.g. bcs, ...)

From @tbnc:
@raychew What version of Sphinx should be used to compile the documentation? Running make docs (python 3.8.8) gives the following error:
Running Sphinx v6.1.3
Extension error:
Could not import extension sphinxcontrib.inkscapeconverter (exception: No module named 'sphinxcontrib.inkscapeconverter')

Currently, the readme file is missing some references and packages versions.

• #1
• #17

Not as easy as I thought it would be. I tried using some scripts, but they do not work as intended.

Gonna move them manually now.

Only the open issues will be moved!

I was using two deployment scripts test.yml and documentation.yml to run the pytest and build the Sphinx documentation. However, this is not ideal, as the dependencies have to be installed anew each time these scripts run.

Need to come up with maybe a cron job? Or something sensible for the deployment scripts.

Now I am just running documentation.yml.

For now, vertical slices with gravity are in the x-y plane, while horizontal slices with Coriolis are in the x-z plane. Simulations in the x-z plane are solved with a quasi-3D array, i.e. the full grid-size on the x-z plane and in the y-axis, 5 layers of the cellular grid and 6 layers of the nodal grid. These vertical layers comprise the 4 ghost layers.

Making Coriolis simulations work with a truly 2D array will decrease computation time significantly. Two possibilities are:

1. Make it possible for the code to solve x-z plane problems in 2D or
2. extend the x-y plane to support Coriolis.

Option 2. seems to be easier for now, but option 1. will simplify extension to 3D in the future.

Spoke everything from one "develop" branch.

A basic test suite is introduced to ensure the integrity of the code.

• #13
• #4
• #5
• #6

This issue aims to map out the strands of our discussions and provide an overview of the significant steps ahead.

[Infrastructure]

• Implement diagnostic module
• Implement tests for blending and DA
• Implement logger
• Implement profiler and coverage check into workflow
• Implement Christian's Baroclinic Instability test
• Add switch for IO precision
• Implement tests on code integrity via pytest @benacchio

[Documentation]

• Include write-up on the science
• Include write-up on the algorithmic implementation
• Do we want to include write-up on the discretisation? @RayChew thinks this would be too much work.

[Refactoring]

• Restructure code @RayChew
• Move DA and blending handling strictly to the respective modules @RayChew
• Clean up JSON/YAML driver scripts @RayChew
• Clean up clumsy hydrostatic initialisation implementation
• Make sure 3D simulations work again @RayChew

[Terrain-following coordinates (TFC)]

• Write out basic theory for TFC implementation @RayChew
• Decide on which implementation to go for (Prusa and Smolarkiewicz, 2003?) @RayChew
• Typeset Rupert's notes
• Develop new science on conservative discrete TFC formulations

The logger output looks a bit wonky when the text is coloured, e.g.,

�[33mNext tout = 600.000�[0m
�[32mStarting forecast...�[0m
�[33mFor ensemble member = 0...�[0m

This link might be useful to format the logger for coloured text:
https://stackoverflow.com/questions/13733552/logger-configuration-to-log-to-file-and-print-to-stdout

This should be the last step of the move repo milestone: package and upload pyBELLA to pip.

• #2
• #3

Ultimately, the data assimilation engine should be in a separate repository and imported as a submodule.

This will allow us to document that a tag/release corresponds to the code frozen for a particular publication.

Also make a new release after move to GitHub has been completed.

Now that the essentials of the code are working, it's time to clean-up and comment the code.

1. Remove all 'print debugs'.
2. Keep asserts and comment out outputs for debugging purposes.
3. Remove unnecessary variables in the data containers.
4. Remove unnecessary initial condtions in the userdata files.
5. Remove unnecessary comments..

The docs folder was removed during the move as it was massively deprecated.
So now, we have to restructure the documentation, make it work with modern Sphinx, and include it back in this repository.

The current restructuredText Readme has a bunch of broken links and references.

Fomat Readme as Markdown, as GitHub seems to work better with this format.

Mustn't import functions and classes directly...