Allow multiple commands in a task for loop

[JonZeolla]

I would like to be able to specify a list of commands inside of a task for loop.

This is useful if the second cmd relies on an output of the first cmd, but each iteration of the first cmd overwrites prior iterations. This is useful if, for instance, the first cmd loads the example:latest image of platform arm into the local docker daemon, and then the second command execs some tests against it. Because other image platforms would have the same name and tag (in the case of a multiplatform manifest), it would overwrite prior work locally and cause a mismatch when testing.

For example:

---
version: '3'

tasks:
  build:
    cmds:
      - echo 'hit {{.PLATFORM}}'

  example:
    cmds:
      - for:
          var: PLATFORM
          split: ','
          as: platform
        cmds:
          - task: build
            vars:
              PLATFORM: '{{.platform}}'
          - echo 'Do things with platform {{.platform}}'

Expected Output:

$ PLATFORM='arm,amd' task example
task: [build] echo 'hit arm'
hit arm
task: [example] echo 'Do things with platform arm'
Do things with platform arm
task: [build] echo 'hit amd'
hit amd
task: [example] echo 'Do things with platform amd'
Do things with platform amd

Instead you need to do something like:

---
version: '3'

tasks:
  build:
    cmds:
      - echo 'hit {{.PLATFORM}}'

  example:
    cmds:
      - for:
          var: PLATFORM
          split: ','
          as: platform
        task: build
        vars:
          PLATFORM: '{{.platform}}'
      - for:
          var: PLATFORM
          split: ','
          as: platform
        cmd: echo 'Do things with platform {{.platform}}'

Actual Output:

$ PLATFORM='arm,amd' task example
task: [build] echo 'hit arm'
hit arm
task: [build] echo 'hit amd'
hit amd
task: [example] echo 'Do things with platform arm'
Do things with platform arm
task: [example] echo 'Do things with platform amd'
Do things with platform amd

The way I read it this is similar to but different from #1299

[pd93]

I'm not a huge fan of the multi-nested cmds block (i.e. tasks[].cmds[].for.cmds[]). It makes the Taskfile hard to parse IMO (by both humans and our code). When designing this feature, there were a few ways we could have gone about doing it. The way it is implemented now keeps the YAML as simple and readable as possible and still offers maximum flexibility.

If you need to loop over multiple commands, why not just wrap up your functionality in a separate task? This means that if you ever need to run that set of commands for a single item or a different set of items, then your Taskfile will already support it. e.g.

version: '3'

tasks:
  do-thing-for-all-platforms:
    cmds:
      - for:
          var: PLATFORM
          split: ','
        task: do-thing-for-platform
        vars:
          PLATFORM: {{.ITEM}}
  
  do-thing-for-platform:
    cmds:
      - task: build
        vars:
          PLATFORM: '{{.PLATFORM}}'
      - echo 'Do things with platform {{.PLATFORM}}'

  build:
    cmds:
      - echo 'hit {{.PLATFORM}}'

It's also worth remembering that if you don't want certain tasks to be executable via the CLI, you can add internal: true to the definition.

  do-thing-for-platform:
    internal: true
    cmds:
      - ...

[JonZeolla]

While I still prefer the former, my workaround was essentially what you showed.

Currently our task files are getting pretty complex. We have a library of different task files for inclusion, indirection points (multiple includes), etc. I like my suggestion better because it reduces the amount of spaghetti - one less task - but this is not a critical item. Thanks!

[pd93]

I'm going to decline this for now for the reasons I gave above. for is still pretty new, so I'd like to let some time pass in its current state before we make any changes, but if we see more requests for this kind of feature then its something we can reconsider in the future.