=== Checklist for MDP release === - Check that new nodes have been explicitly imported in nodes/__init__.py and that they are listed in __all__: # create a list of defined nodes with: git grep 'class .*Node(' mdp/nodes | grep -v test | grep -v Scikits | cut -d ':' -f 2 | cut -d ' ' -f 2 | cut -d '(' -f 1 | sort > /tmp/list_defined # create a list of nodes imported in mdp.nodes with: python3 -c "import sys, mdp; [sys.stdout.write(i+'\n') for i in sorted([obj for obj in mdp.nodes.__dict__ if obj.endswith('Node') and not obj.endswith('ScikitsLearnNode')])]" > /tmp/list_in_dict # create a list of nodes in __all__ with: python3 -c "import sys, mdp; [sys.stdout.write(i+'\n') for i in sorted([obj for obj in mdp.nodes.__all__ if obj.endswith('Node') and not obj.endswith('ScikitsLearnNode')])]" > /tmp/list_in_all # compare those listsm for example with meld /tmp/list_defined /tmp/list_in_dict /tmp/list_in_all # These lists should be almost identical. Some nodes defined or found in the dict could be used only internally, # and thus not available through __all__, as of 2020.04.18 the only nodes that do not need to be in all three # lists are: - ICANode -> this is an internal node to be subclassed by nodes implementing some form of ICA - ProjectionNode -> this is an internal node used by XSFA - _ExpansionNode -> this is used internally by many SFA nodes) - Make sure that mdp/__init__.py has the right version number - Update date in variable __copyright__ in file mdp/__init__.py and in COPYRIGHT - Update CHANGES: you can generate a new bunch of CHANGES with: git log --no-color --pretty="format:%w(79,0,12)%ad: %s%+b" --date=short --no-merges --since=$LASTRELEASE where LASTRELEASE is the tag of the last release, e.g. MDP-3.6 You can then prepend the output of this command to the original CHANGES file, but *even better* would be to edit the result to only keep the changes that are relevant for the user like incompatibilities, new features, etc.. Not every commit is interesting in CHANGES: the exact commits are available in git. - Test that MDP can be installed with pip locally: # create a wheel python3 setup.py sdist bdist_wheel # create a new Python3 virtualenv python3 -m venv /tmp/mdp_test # activate the virtualenv source /tmp/mdp_test/bin/activate # install the wheel in the virtualenv pip install dist/XXX.whl # this should also install MDP hard-dependencies (numpy, future) # manually install MDP soft dependencies pip install scipy scikit-learn joblib libsvm pytest # change directory to be sure we are importing the pip-installed MDP cd ~ # now we should be able to import python -c 'import mdp; print(mdp.__file__, mdp.__version__)' # -> in the above verify that the file is in the clone of the MDP repo, # and that the version is correct # now we can run the tests python -c 'import mdp; mdp.test(); import bimdp; bimdp.test()' # -> verify that all soft depenencies are correctly detected by MDP, for example ================================================== test session starts ============= platform linux -- Python 3.8.2, pytest-5.4.1, py-1.8.1, pluggy-0.13.1 python: 3.8.2.final.0 mdp: 3.6 parallel python: NOT AVAILABLE: No module named 'pp' shogun: NOT AVAILABLE: No module named 'shogun' libsvm: 3.23.0.4 joblib: 0.14.1 sklearn: 0.22.2.post1 numx: scipy 1.4.1 symeig: scipy.linalg.eigh Random Seed: 1343069334 rootdir: /tmp/mdp_test/lib/python3.8/site-packages/mdp/test # -> also, of course check that all tests are passing - create a release notes file (this is an edit version of the CHANGES file, it will be used as annotation for the git tag, and will be shown on GitHub in the release package) - tag release in git and use the release notes file as annotation for the tag git tag -a MDP-3.6 -F RELEASE_NOTES - push the tag git push --tags ###### The stuff below is obsolete and needs to be updated ###### - generate tutorial, website, and API documentation [make website] - change homepage colors - short/long description should go: on SF.net description, tutorial, home page, modules __init__, software.incf.net. - update the package on PyPI: # first test on the test PyPI instance (you have to register there first) # - create a ~/.pypirc first: cat > ~/.pypirc < ~/.pypirc <