azure-pipelines.yml

# Azure Pipelines configuration file for Continuous Integration
# for building the package and running the tests under Windows.
variables:
  PIP_CACHE_DIR: $(Pipeline.Workspace)/.pip
  # data cache dir determined via: print(pooch.os_cache('scikit-image'))
  SKIMAGE_DATA_CACHE_FOLDER: C:\Users\VssAdministrator\AppData\Local\scikit-image\scikit-image\Cache

jobs:
  - job: "Default"
    pool:
      vmImage: "windows-latest"
    strategy:
      maxParallel: 10
      matrix:
        Python311-x64:
          PYTHON_VERSION: "3.11"
          ARCH: "x64"
          PIP_FLAGS: ""
          BUILD_DOCS: "true"
        Python311-x64-pre:
          PYTHON_VERSION: "3.11"
          ARCH: "x64"
          PIP_FLAGS: "--pre"
        Python310:
          PYTHON_VERSION: "3.10"
          ARCH: "x64"
          PIP_FLAGS: ""
        Python310-x64:
          PYTHON_VERSION: "3.10"
          ARCH: "x64"
          PIP_FLAGS: ""
          TEST_EXAMPLES: "true"
    continueOnError: false
    timeoutInMinutes: 60

    steps:
      - task: UsePythonVersion@0
        inputs:
          versionSpec: "$(PYTHON_VERSION)"
          architecture: "$(ARCH)"
        name: python

      # https://docs.microsoft.com/en-us/azure/devops/pipelines/release/caching?view=azure-devops
      - task: Cache@2
        inputs:
          key: '"skimage_data" | skimage/data/_registry.py'
          path: $(SKIMAGE_DATA_CACHE_FOLDER)
        displayName: Cache scikit-image data files
        # the line below (continueOnError: true) can be removed when the following issue is addressed:
        # https://github.com/scikit-image/scikit-image/issues/6048
        continueOnError: true

      # https://docs.microsoft.com/en-us/azure/devops/pipelines/release/caching?view=azure-devops#pythonpip
      - task: Cache@2
        inputs:
          key: '"Python$(PYTHON_VERSION)-$(ARCH)" | "$(Agent.OS)" | requirements/build.txt | requirements/default.txt | requirements/test.txt | requirements/docs.txt'
          restoreKeys: |
            "Python$(PYTHON_VERSION)-$(ARCH)" | "$(Agent.OS)"
            "Python$(PYTHON_VERSION)-$(ARCH)"
          path: $(PIP_CACHE_DIR)
        displayName: Cache pip packages
        # the line below (continueOnError: true) can be removed when the following issue is addressed:
        # https://github.com/scikit-image/scikit-image/issues/6048
        continueOnError: true

      - bash: |
          set -ex
          PYTHON="$(python.pythonLocation)\\python.exe"

          # Update pip
          $PYTHON -m pip install -U pip

          # Check that we have the expected version and architecture for Python
          $PYTHON --version
          $PYTHON -m pip --version
          $PYTHON -c "import struct; print('Void pointer width is', struct.calcsize('P') * 8)"

          # print out the pip cache dir followed by the variable defined above
          # to confirm that they match
          $PYTHON -m pip cache dir
          echo "PIP_CACHE_DIR: $PIP_CACHE_DIR"

          # Install the build and runtime dependencies of the project
          $PYTHON -m pip install ${PIP_FLAGS} -r requirements/build.txt
          $PYTHON -m pip list

          # Disable C99 complex if PyWavelets needs to be built from source.
          # The compiler used will be MSVC, but C99 may be detected improperly
          USE_C99_COMPLEX=0 $PYTHON -m pip install ${PIP_FLAGS} -r requirements/default.txt

          $PYTHON -m pip list
        displayName: "Pre-installation"

      - bash: |
          set -ex
          PYTHON="$(python.pythonLocation)\\python.exe"
          if [ "$(ARCH)" = "x86" ] ; then
              export CXXFLAGS="-m32"
          fi

          ## To list all installed packages:
          choco list

          # Remove other gcc compilers

          ## This would be the best way to uninstall, but is SLOW:
          # choco uninstall llvm mingw rtools strawberryperl -x

          # Remove mingw gcc
          #rm /c/ProgramData/Chocolatey/bin/gcc
          #rm /c/ProgramData/Chocolatey/bin/g++
          #rm /c/ProgramData/Chocolatey/bin/c++
          # Remove perl gcc
          rm -rf /c/Strawberry

          which ccache || true
          which c++ || true

          ## Uncomment the next line, and you'll remain with only MSVC.
          # rm -rf "/c/Program Files/LLVM"
          ## We're compiling with clang for now because MSVC 19 and
          ## Pythran don't play well together.

          which clang || true

          # Compile the package and build the wheel
          $PYTHON -m build --no-isolation --skip-dependency-check --wheel

          # Install the generated wheel package
          ls dist
          $PYTHON -m pip install ${PIP_FLAGS} --no-index --find-links dist/ scikit-image
        displayName: "Installation"

      - bash: |
          set -ex
          PYTHON="$(python.pythonLocation)\\python.exe"

          # Install the test dependencies
          $PYTHON -m pip install ${PIP_FLAGS} -r requirements/test.txt
          $PYTHON -m pip list

          # Set non-UI Matplotlib backend
          cd ${AGENT_BUILDDIRECTORY}  # D:\a\1
          echo "backend : Agg" > matplotlibrc
        displayName: "Pre-testing"

      - bash: |
          set -ex
          PYTHON="$(python.pythonLocation)\\python.exe"

          # Change the working directory in order to run the tests
          # on the installed version of skimage
          cd ${AGENT_BUILDDIRECTORY}  # D:\a\1

          # Show the info about the installed scikit-image
          $PYTHON -c "import skimage; print(skimage.__path__)"

          # Show the path pooch is using for the data cache
          # (this should match the SKIMAGE_DATA_CACHE_FOLDER variable above)
          $PYTHON -c "import pooch; print(pooch.os_cache('scikit-image'))"
          echo "SKIMAGE_DATA_CACHE_FOLDER: $SKIMAGE_DATA_CACHE_FOLDER"

          # Force matplotlib to use the prepared config
          export MATPLOTLIBRC=${AGENT_BUILDDIRECTORY}

          # Run unit tests with pytest
          # We don't test docstring examples (--doctest-modules) on
          # Windows due to inconsistent ndarray formatting in `numpy`.
          # For more details, see https://github.com/numpy/numpy/issues/13468
          export TEST_ARGS="-v --cov=skimage"
          $PYTHON -m pytest ${TEST_ARGS} --pyargs skimage
        displayName: "Package testing"

      - bash: |
          set -ex
          PYTHON="$(python.pythonLocation)\\python.exe"
          $PYTHON -m asv check -v -E existing
        displayName: Check benchmarks

      - bash: |
          set -ex
          export PYTHON="$(python.pythonLocation)\\python.exe"

          # Install the doc dependencies
          $PYTHON -m pip install ${PIP_FLAGS} -r requirements/docs.txt
          $PYTHON -m pip list

          # Build the documentation
          choco install optipng
          export SPHINXCACHE=${AGENT_BUILDDIRECTORY}/.cache/sphinx
          export SPHINXOPTS=-W
          make -C doc html
        condition: eq(variables['BUILD_DOCS'], 'true')
        displayName: "Documentation testing"

      - bash: |
          set -ex
          PYTHON="$(python.pythonLocation)\\python.exe"

          # Install the doc dependencies
          $PYTHON -m pip install ${PIP_FLAGS} -r requirements/docs.txt
          $PYTHON -m pip list

          # Force matplotlib to use the prepared config
          export MATPLOTLIBRC=${AGENT_BUILDDIRECTORY}

          # Run example applications
          for f in doc/examples/*/*.py; do
            # Comment out any plotly.io.show() calls before running the example.
            # Plotly opens a web browser, which often seemed to cause the CI to
            # hang and timeout.
            sed -i 's/plotly.io.show/# plotly.io.show/g' ${f}

            $PYTHON -W ignore:Matplotlib:UserWarning "${f}"
            if [ $? -ne 0 ]; then
              exit 1
            fi
          done
        condition: eq(variables['TEST_EXAMPLES'], 'true')
        displayName: "Gallery testing"
#  - bash: |
#      # -- Publish the .whl artifacts
#      # -- Upload the content of dist/*.whl to a public wheelhouse
#    displayName: 'Further consideration'