CircleCI流水线跨Job传参后when条件不生效的解决方案咨询

我在CircleCI流水线里想根据前一个Job的执行结果，条件执行后续步骤。试了好几种方法把Step1的变量暴露成全局变量，在Step2里能读取到——能查看、打印这个变量，但when块的条件始终被判定为空。查了资料知道CircleCI的逻辑条件是在Job运行前就评估完的，现在求替代方案实现基于前Job的条件执行第二个Job里的步骤。

问题示例代码：

version: 2.1

orbs:

workflows:
  test-and-deploy:
    jobs:
      - set-data:
          context: my-context
      - read-data:
          context: my-context
          requires:
            - set-data      
   
definitions:
  node_image: &node-image
    docker:
      - image: cimg/node:14.15.5

executors:
  base-12-14-0:
    description: |
      Single Docker container with Node 12.14.0 and Cypress dependencies
      see https://github.com/cypress-io/cypress-docker-images/tree/master/base.
      Use example: `executor: cypress/base-12-14-0`.
    docker:
      - image: cypress/base:12.14.0

jobs:
  set-data:
    <<: *node-image
    description: Sets the data
    steps:
      - run: echo "VAR=app" > global-vars
      - persist_to_workspace: 
          root: .
          paths:
            - global-vars
  read-data: 
    <<: *node-image
    description: read the data
    steps:   
      - attach_workspace: 
          at: .
      - run: ls 
      - run: cat global-vars # 这里能看到global-vars里的正确变量
      - run: cat global-vars >> $BASH_ENV
      - run: echo "Test $VAR"  # 能成功打印变量
      - when: 
          condition: 
            matches: {
              pattern: "app",
              value: $VAR
            }
          steps:
            - run: echo "Condition Executed"

替代方案 #

方案1：用Shell脚本内置逻辑替代when块 #

因为CircleCI的when条件在Job启动前就完成评估，无法读取运行时生成的变量，直接在run步骤里用Shell的if判断即可实现条件执行：

jobs:
  read-data: 
    <<: *node-image
    description: read the data
    steps:   
      - attach_workspace: 
          at: .
      - run: ls 
      - run: cat global-vars >> $BASH_ENV
      - run: |
          if [[ "$VAR" == "app" ]]; then
            echo "Condition Executed"
            # 在这里添加需要条件执行的所有命令
          fi

方案2：通过流水线参数传递实现工作流层面的条件判断 #

将set-data的结果写入CircleCI流水线参数，这样read-data的条件可以在工作流层面评估：

1. 修改set-dataJob，把变量写入流水线参数：

jobs:
  set-data:
    <<: *node-image
    description: Sets the data
    steps:
      - run: echo "VAR=app" > global-vars
      - persist_to_workspace: 
          root: .
          paths:
            - global-vars
      - run: |
          export VAR_VALUE=$(cat global-vars | cut -d= -f2)
          circleci config set pipeline.parameters.VAR "$VAR_VALUE"

2. 在工作流中使用参数控制read-data的步骤或Job执行：

workflows:
  test-and-deploy:
    jobs:
      - set-data:
          context: my-context
      - read-data:
          context: my-context
          requires:
            - set-data
          # 可以控制整个Job是否执行
          when:
            matches:
              pattern: "app"
              value: << pipeline.parameters.VAR >>

如果不需要控制整个Job，也可以在read-data的run步骤中直接引用<< pipeline.parameters.VAR >>做判断。

方案3：拆分条件逻辑为独立Job #

如果条件执行的步骤可以独立，直接在工作流层面控制这个独立Job是否启动：

workflows:
  test-and-deploy:
    jobs:
      - set-data:
          context: my-context
      - read-data:
          context: my-context
          requires:
            - set-data
      - conditional-execution:
          context: my-context
          requires:
            - set-data
          when:
            matches:
              pattern: "app"
              value: << pipeline.parameters.VAR >>

jobs:
  conditional-execution:
    <<: *node-image
    steps:
      - run: echo "Condition Executed"
      # 添加其他需要条件执行的步骤

内容的提问来源于stack exchange，提问作者khaled Mohamed