get resource specific version

[schmurfy]

I am not clear on how to specify the version for getstep in job, for example if I want to get two resources on the same version, can I do the following?

 "plan": [
    {
       "get": "res1",
       "trigger": true
    },
    {
       "get": "res2",
       "version": "res1"
    }
]

And if yes what can actually be passed as version, can it be produced by the output of a task like in this pseudo code ?

 "plan": [
    {
       "get": "res1",
       "trigger": true
    },
    {
      "task": "something",
      "config": {
        "inputs": [
          {"name": "res1"}
        ],
        
        "outputs": [
          {"name": "new_version"}
        ]
      }
    },
    {
       "get": "res2",
       "version": "new_version
    }
]

Sorry if I missed something but I read everything I found on this and I can't wrap my head on how it actually works.

[jamieklassen]

Can you be more explicit about these resources, perhaps giving a complete sample pipeline config? For example what type of resources are they? The docs give an example with a git resource, where you can use the version: field to configure a job to always fetch a static/fixed/constant version of a resource: https://concourse-ci.org/jobs.html#schema.step.get-step.version. This kind of "pinned/hard-coded" workflow is the use case for which the feature was designed.

When you ask

what can actually be passed as version

the answer is

An object with string keys and string values

https://concourse-ci.org/config-basics.html#schema.version, such as might appear in the stdout stream of the check, in or out script of a resource: https://concourse-ci.org/implementing-resource-types.html.

What you have in mind sounds like it is subtly more dynamic, and I'd be curious to understand what it means to have the same version of two different resources. Also potentially of interest is: whatever "getting two resources on the same version" means, why are you doing it/what is the purpose of this job?

[schmurfy]

I am playing with concourse to try building a pipeline for a golang repository which includes multiple applications, I wanted to provide a cache of dependencies so they are not fetched every time so I tried a few options including the npm cache way and finally found one I like:

For building the master branch everything is fine:

[git resource] => [cache task (pull the source and its depdencies and store them in s3 in a folder as tgz)
[s3 resource] => [build] => [release, ...]

For PR it also works the same way and works well:

[github-pr] => [cache task (uses the pr number as key in s3 and stores the version in the folder)]
[s3-resource] => [build]

The problem arise when I want to integrate in this flow updating the github-pr resource since it requires a version of the form: {ref: xxx, commit: xxx}, I have the data but not immediately in this form, this is the reason for the second plan in my first post.

I don't actually need the github-pr resource to update the pr or at least I hoped, but as it turns out there no exiting cli tool which does that and I am motivated enough yet to make my own cli tool, this is why I wanted to find a way to fetch both my cached folder and the associated github-pr.