Skip to content

MPIBGC-TEE/bgc_md2

Repository files navigation

test_binder_pinned test_binder_free

test_conda_developer_installation test_debian_pip_install test_windows_developer_installation

Binder

Installation

Please read carefully before you "Copy and paste", since some instructions only make sense for certain platforms.

  • For developers who work with CompartmentalSystems LAPM and testinfrastructure simultaneously:
    • Clone the repository and its submodules: Note: If you have forked the repository you probably want to put your forks ulr here. You can of course install the original but especially if you do so using setuptools develop mode to see the effects of your changes on the installed package you probably want to install from your fork.

      • If you do not have a bgc_md2 repo yet:

        git clone --recurse-submodules https://github.com/MPIBGC-TEE/bgc_md2.git
      • If you already have a bgc_md2 repo (and want to keep it):

        • Pull the changes in bgc_md2 and the submodules simultaneuously:
          git pull --recurse-submodules
          
        • Make sure that the submodule folders in src are not empty.
          git submodule init
          git submodule update
          
    • Update conda

      conda update --all
      
    • Create a conda environment and run the install script:

      conda create -y --name bgc_md2 python=3
      conda activate bgc_md2
      cd bgc_md2
      ./install_developer_conda.sh 

      (on MS-Windows replace the last line with)

      install_developer_conda.bat 

      This will install the dependencies and run python setup.py develop for every subpackage so that your code changes in one of these packages take immediate effect.

    • Run the tests.

      cd tests
      ./run_tests.py
      

      (on MS-Windows replace the last line with)

      python run_tests.py
      

      If you can run this script successfully, you have a working installation of bgc_md and can run all functions.

    • Troubleshooting:

      • We noticed that in MacOS, it is necessary to update packages in the conda environment before running the tests successfully. Try to update conda ( conda update --all) and run the tests again.
    • Working with the installation:

      • pulling: Since you will nearly always pull with the --recurse-submodules flag
        consider creating an alias

        git config alias.spull 'pull --recurse-submodules'
        

        which enables you to say git spull to achieve the same effect

      • Tips to work with git submodules:

Documentation

  • The latest build of the package documentation can be found here:.

Objectives

The package is supposed to assist in the creation of 'reports' in the form of jupyter notebooks. The notebooks will be of the following types.

  1. Investigations of a single model (or modelrun).
  2. Comparisons between models/modelruns.

In the first case the role of the bgc_md package is to guide the user (=author of a particular notebook concerned with a particular model, and simultaniously author of the source.py of that model) by using the computability graph (as represented by bgc_md/resolve/MvarsAndComputers.py) to either

  • show which addidional results can be computed, given the information already present in the models `source.py' or
  • show which additional information has to be provided in the models source.py to be able to obtain a desired result.

In the second case the same assistance is required for queries, which are best described by examples.

  • Create a table including all the models for which we can compute the compartmental matrix (from whatever Mvars are provided, in the different model files)
  • Compute the maximum set of Mvars we can compute for a given set of models
  • ...

Contribution

Green master

We try to keep the master green and develop new features or bug fixes in short lived branches that are then merged back into the master https://nvie.com/posts/a-successful-git-branching-model/ See also https://git-scm.com/book/en/v2/Git-Branching-Branches-in-a-Nutshell

  • Example workflow to work on a feature branch iss26-non-importable-models you are asked to review
    • git branch -a (shows all branches including remotes)
    • git checkout --track origin/iss26-non-importable-models (creates a local copy that you can test)
  • Example to create your own feature branch (here to fix an issue )
    • git checkout -b iss53

To merge into the master:

  • run the testsuites
  • merge into the branch test
  • merge test into master

The (github) workflows run the testsuites. This is an additional protection against forgotten files or other idiosycracies of your setup, that let the testsuites succeeed locally but brake for other users.

Green Binder

Another important branch is binder. It contains a smaller version of the repository to meet the binder requirements of total size<2GB which is deployed by mybinder.org to allow exploration without installation (the binder button on top). It relies on the smaller binder branch of CompartmentalSystems
(see https://git-scm.com/book/en/v2/Git-Tools-Submodules for the work on the dependencies) and therefore has a .gitmodules file that is permanently different from test and master branches. If you merge something to the binder branch the workflow is similar to merges to the master.

Before you try to merge into binder :

  • run the testsuites
  • merge into the branch test.
  • merge test into binder
  • make sure that your merge dosn't overwrite the branch specific files These are listed in the (version controlled) .gitattributes file with the special merge strategie "ours". You can automate this by configuring git to use the merge driver true (a posix command line tool that always exits with 0) by the command git config merge.ours.driver true
    https://git-scm.com/book/en/v2/Customizing-Git-Git-Attributes#_merge_strategies