Release 0.5.0

[loli]

Pre:

• Write a pre-release checklist for future releases and add to repo as documentation

TODO

• Bump-up library version
• Supported formats
  • python3 tests/support.py > myformats.log fails due to random package in scipy
• Depreciation and error fixes
  • scipy. => numpy.*
  • list(slice) => tuple(slice)
• Run tests and fix errors:
  • switch from nose to pytest
  • add test requirements.txt: pytest-7.4.3 & hypothesis-6.92.0
  • fix errors
• Error in saving an image that is of type bool, at least for .nii format: fix
• Run all notebooks, to test the code
  • fix all error that occurr
• Check Readme.md
  • check consistency and update
• Update setup.py
• Check & improve documentation
  • check https://loli.github.io/medpy/#
  • generate and improve
  • Add graphcut help for Issue with graphcut support #103 CentOs
    • Add note that, under Debian / Ubuntu, lib can be found under /usr/lib/x86_64-linux-gnu/libboost_python* if linker error
    • Generally, write better error handling for GraphCut installation and link in open issues on the topic for people to find (keep open)
      • e.g. add install -v for more info to http://loli.github.io/medpy/installation/graphcutsupport.html
        [x]- Issue of broken links and notebooks: Broken links and incomplete notebooks #97
• Check badges in README.md
• Bioconda / conda
  • Now in conda-forge: https://anaconda.org/conda-forge/medpy
    • Add badge from here to Readme
  • test install and update docu (e.g. https://loli.github.io/medpy/installation/conda.html) accordingly
  • edit recipes to ppoint to 0.5.0?
    • maybe this: https://github.com/conda-forge/medpy-feedstock ?
    • seem like it, version is here: https://github.com/conda-forge/medpy-feedstock/blob/main/recipe/meta.yaml
  • fork & PR for file https://github.com/conda-forge/medpy-feedstock/blob/main/recipe/meta.yaml plus see Readme of that repo
• Remove travis build
  • move to github builds, if possible
• Check config files:
  • .bettercodehub
  • .gitmodules
• Try installation on different systems
  • with different library versions to figure out minimal
  • with graph cut support and without
  • use pip install -v -e . to see installation logs for graphcut
• check why they exists and remove/ edit requirements-dev.* files in root folder
• Dockerfiles: won't do anymore
  • remove linked repo
  • remove from documentation
• Others:
  • Maybe write a short docu / docstring comment on this: Anisotropic Diffusion Behaviour #76
• Auto-build and auto-tests: Introduce appropriate gituhub actions
  • test for pre-commit hook usage on commit
  • run install and test
    • linux
    • MacOs
    • Windows (won't do)
  • run graphcut install and test
    • ubuntu22.10 with p3.10 & ubuntu20.04 with p3.8
  • trigger build & upload to test.pypi on release
  • manually trigger release to pypi [needs: build]
  • manually trigger doc release [needs: build]

new minimal dependency versions

• old dtypes depreciates from numpy >= 1.24
• inspect.getfullargspec recommended since python >= 3.4
• dump minima dependencies in setup.py

Optional (rather for v0.6)

• Introduce new build system (e.g. poetry)
  • interesting, but would be quite some effort to get the c++ compiled extension running
  • maybe for the next release?
  • https://towardsdatascience.com/setuptools-python-571e7d5500f2
  • https://packaging.python.org/en/latest/guides/writing-pyproject-toml/
• Adapt GitHib actions to also build wheels
  • https://github.com/pypa/manylinux
  • for MacOs
  • for windows
• Harmonize imports, aka "import numpy" > "import numpy as np"
• Switch to more modern documentation builder

Release TODO

• Generate docs and place to upload
• Update setup.py
• Update CHANGES.txt
• Update github-pages
• create package and upload to pypi

Post-Release to do

• Update conda-forge recipe