Erroneous “enum reference must resolve to a list”

[RemiBardon]

Description

Given the example Taskfile, when I:

task foo FOO=bar

then I get:

enum reference ".LIST" must resolve to a list

It usually works in other places, I don’t get why it doesn’t work here.

My use case is: allowing a set of possible values based on the result of a ls (i.e. “choosing a file”, more or less).

Version

3.52.0

Operating system

macOS 15.7.7 (24G720)

Experiments Enabled

No response

Example Taskfile

version: 3

vars:
  LIST_STR: { sh: printf '%s\n' aaa bbb ccc }
  LIST: { ref: splitLines .LIST_STR }

tasks:
  foo:
    requires:
      vars:
        - name: FOO
          enum: { ref: .LIST }

[trulede]

From google AI, this suggestion, it seems correct so give it a try.

version: '3'

vars:
  LIST_STR: { sh: printf '%s\n' aaa bbb ccc }
  LIST: '{{splitLines .LIST_STR}}'

tasks:
  foo:
    requires:
      vars:
        - name: FOO
          enum:
            ref: LIST
    cmds:
      - echo "You chose {{.FOO}}"

[RemiBardon]

$ task foo FOO=bar
template: resolver:1: function "LIST" not defined

I’m sorry but did you even try it?

[RemiBardon]

Even with ref: .LIST it goes back to enum reference ".LIST" must resolve to a list… I mean I know I’ve written a correct Taskfile, it’s just a bug I wanted to notify.

I’m sorry but I’m tired of AI just hallucinating solutions… it just doesn’t work.

[trulede]

No, I did not try it. AI is going to be more responsive for figuring this out ... its unlikely that this is going to be a bug. Not impossible, but unlikely.

You might try to locate LIST var to the task. Is possible that it tries to resolve LIST_STR before the sh is evaluated, relocating would address that.

See the documentation for examples (which I assume work), notice the location of the 'sh' variable (global) and enum variable (task).
https://taskfile.dev/docs/guide#using-variable-references-for-enum-values

[RemiBardon]

What do you mean by “AI is going to be more responsive for figuring this out”?

Also, this —taken from the docs (link you shared)— works:

version: '3'

vars:
  AVAILABLE_SERVICES:
    sh: ls services/

tasks:
  deploy:
    requires:
      vars:
        - name: SERVICE
          enum:
            ref: .AVAILABLE_SERVICES | splitLines | compact
    cmds:
      - echo "Deploying {{.SERVICE}}"

$ mkdir services
$ touch services/foo
$ task deploy SERVICE=foo
task: [deploy] echo "Deploying foo"
Deploying foo
$ task deploy SERVICE=bar
task: Task "deploy" cancelled because it is missing required variables:
  - SERVICE has an invalid value : 'bar' (allowed values : [foo])

I don’t see a meaningful difference from my example (which doesn’t work):

version: 3

vars:
  LIST_STR:
    sh: printf '%s\n' aaa bbb ccc

tasks:
  foo:
    requires:
      vars:
        - name: FOO
          enum:
            ref: splitLines .LIST_STR

(By the way I am using ls in my code; printf here was to make line separation explicit)

[RemiBardon]

version: '3'

vars:
  AVAILABLE_SERVICES:
    sh: ls services/

tasks:
  deploy:
    requires:
      vars:
        - name: SERVICE
          enum:
-           ref: .AVAILABLE_SERVICES | splitLines | compact
+           ref: .AVAILABLE_SERVICES | splitLines
    cmds:
      - echo "Deploying {{.SERVICE}}"

This is enough to make the docs example fail (enum reference ".AVAILABLE_SERVICES | splitLines" must resolve to a list). I’m sorry but I call this a bug. compact is just supposed to remove empty strings from the slice (https://sprig.taskfile.dev/lists.html#compact-mustcompact).

[pd93]

@RemiBardon Please update your version of Task to the latest.

> task --version
3.52.0
> task -d ./tmp/3020 foo FOO=bar
enum reference ".LIST" must resolve to a list

> task --version
3.53.1
> task -d ./tmp/3020 foo FOO=bar
task: Task "foo" cancelled because it is missing required variables:
  - FOO has an invalid value : 'bar' (allowed values : [aaa bbb ccc])

[RemiBardon]

Oh! That’s nice :) Hadn’t seen the fix being mentioned in the 3.53.1 release (thought I had a recent-enough version)

[trulede]

What do you mean by “AI is going to be more responsive for figuring this out”?

I mean, the AI has got spare time to help you. And its pretty good, even if you have to tell it "no, please try again, this time an example which actually works".

I have spent the whole day closing PR's submitted to the Task project which were written by AI which are poor quality.