testbed/fairlearn__fairlearn/devops/pypi-release.yml

# Pipeline Definition for Releases to PyPI

# This Azure DevOps pipeline is for releasing fairlearn to PyPI. It performs the
# following:
# 1. Predeployment validation similar to the regular builds
# 2. Release to PyPI-Test
#    - Create package
#    - Upload
#    - Run tests against package on PyPI-Test
# 3. Release to PyPI itself, repeating the same stages

# Note that the pipeline MUST be queued with a variable DEV_VERSION set to an
# integer value. This is used to make the packages on PyPI-Test 'release
# candidates' so that minor fixes can be made without requiring a separate checkin
# to release to PyPI itself. See the logic in __init__.py

# TODO:
# Use Azure KeyVault so that uploads to PyPI actually occur

variables:
  poolImage: "ubuntu-latest"
  poolPythonVersion: 3.6
  FreezeFileStem: 'requirements-freeze-predeploy'

trigger: none # No CI build

pr: none # Not for pull requests


stages:
- stage: PreDeploymentValidation
  pool:
    vmImage: $(PoolImage)
    
  jobs:
  - template: templates/all-tests-job-template.yml
    parameters:
      name: Linux
      vmImage: 'ubuntu-16.04'
      pyVersions: [3.5, 3.6, 3.7]
      freezeArtifactStem: "PreDeployment"
      freezeFileStem: $(FreezeFileStem)

  - template: templates/all-tests-job-template.yml
    parameters:
      name: Windows
      vmImage:  'vs2017-win2016'
      pyVersions: [3.5, 3.6, 3.7]
      freezeArtifactStem: "PreDeployment"
      freezeFileStem: $(FreezeFileStem)
      
  - template: templates/all-tests-job-template.yml
    parameters:
      name: MacOS
      vmImage:  'macos-latest'
      pyVersions: [3.5, 3.6, 3.7] 
      freezeArtifactStem: "PreDeployment"
      freezeFileStem: $(FreezeFileStem)

  - template: templates/notebook-job-template.yml
    parameters:
      name: LinuxNotebooks
      vmImage: 'ubuntu-16.04'
      pyVersions: [3.5, 3.6, 3.7]

  - template: templates/notebook-job-template.yml
    parameters:
      name: WindowsNotebooks
      vmImage: 'vs2017-win2016'
      pyVersions: [3.5, 3.6, 3.7] 
  
  - template: templates/limited-installation-job-template.yml


# =================================================================================================================

# We rely on the implicit dependence on the previous stage
# No fan out/fan in of stages
- template: templates/pypi-deployment-stages-template.yml
  parameters:
    poolImage: $(PoolImage)
    targetType: 'Test'
    kvSubscription: "Fairness - Automation (cecafb73-04ae-4432-9f96-d96925d28058)" # This is actually the name of the service connection from the Project Settings page
    kvVaultName: fairlearndeploy
    kvUsername: 'usernametest'
    kvPassword: 'passwordtest'
    freezeArtifactStem: 'Freeze'
    freezeFileStem: 'requirements-freeze'

# =================================================================================================================

- template: templates/pypi-deployment-stages-template.yml
  parameters:
    poolImage: $(PoolImage)
    targetType: 'Prod'
    targetEnvironment: 'PyPI Deployment'
    kvSubscription: "Fairness - Automation (cecafb73-04ae-4432-9f96-d96925d28058)" # This is actually the name of the service connection from the Project Settings page
    kvVaultName: fairlearndeploy
    kvUsername: 'usernameprod'
    kvPassword: 'passwordprod'
    freezeArtifactStem: 'Freeze'
    freezeFileStem: 'requirements-freeze'