Akshay Mankar d93df4f70a helpers.addTags: Add tags to job hooks too (#6) | 3 éve | |
---|---|---|
ci | 3 éve | |
defaults | 4 éve | |
extractors | 4 éve | |
helpers | 3 éve | |
lib | 4 éve | |
nix | 3 éve | |
render | 4 éve | |
schemas | 3 éve | |
scripts | 3 éve | |
types | 4 éve | |
utils | 4 éve | |
.envrc | 3 éve | |
.gitignore | 3 éve | |
README.md | 3 éve | |
direnv.nix | 3 éve | |
package.dhall | 3 éve |
Concourse types and helpers in dhall.
Dhall Concourse provides Dhall bindings for Concourse, so you can generate concourse pipelines from Dhall expressions. This lets the pipelines be easily typechecked, templated and modularized.
There are a lot of issues one could face while building any non-trivial pipeline. Few of them could be:
Most common ways to deal with these have been to use a templating language like erb or tools like spruce. But this gets very messy very fast. We can do a lot better with a typed total language. I'll let dhall speak for itself.
One way to translate pipelines written in dhall-concourse into yaml is using dhall-fly.
To use native rendering to render a list of jobs in a file called jobs.dhall
, you'd have to write a dhall expression like this:
let Concourse =
https://raw.githubusercontent.com/akshaymankar/dhall-concourse/0.6.0/package.dhall
let jobs = ./jobs.dhall
in Concourse.render.pipeline jobs
Now you can render this using dhall-to-json and jq like this:
dhall-to-json <<< './pipeline.dhall' \
| jq '.resources = (.resources|unique)' \
| jq '.resource_types = (.resource_types|unique)'
Similarly, to render a list of GroupedJob
s in a filed called grouped-jobs.dhall
, this would be the expression to render:
let Concourse =
https://raw.githubusercontent.com/akshaymankar/dhall-concourse/0.6.0/package.dhall
let groupedJobs = ./grouped-jobs.dhall
in Concourse.render.groupedJobs groupedJobs
Now you can render this using dhall-to-json and jq like this:
dhall-to-json <<< './pipeline.dhall' \
| jq '.resources = (.resources|unique)' \
| jq '.resource_types = (.resource_types|unique)' \
| jq '.groups = (.groups | group_by(.name) | map({name: .[0].name, jobs: (map(.jobs) | flatten) }))'
This dhall expression will create a pipeline with one job, which would have one task. The task would run in a busybox container and echo "Hello Dhall".
let Concourse =
https://raw.githubusercontent.com/akshaymankar/dhall-concourse/0.5.0/package.dhall
let Prelude =
https://prelude.dhall-lang.org/v11.1.0/package.dhall sha256:99462c205117931c0919f155a6046aec140c70fb8876d208c7c77027ab19c2fa
let busyboxImage =
Concourse.schemas.ImageResource::{
, type = "docker-image"
, source = Some (toMap { repository = Prelude.JSON.string "busybox" })
}
let job =
Concourse.schemas.Job::{
, name = "hello"
, plan =
[ Concourse.helpers.taskStep
Concourse.schemas.TaskStep::{
, task = "hello"
, config =
Concourse.Types.TaskSpec.Config
Concourse.schemas.TaskConfig::{
, image_resource = Some busyboxImage
, run =
Concourse.schemas.TaskRunConfig::{
, path = "bash"
, args = Some [ "-c", "echo Hello Dhall" ]
}
}
}
]
}
in [ job ]
To set the pipeline, run this command:
fly -t <TARGET> set-pipeline -p hello-dhall -c <(dhall-fly <example1.dhall)
let Concourse =
https://raw.githubusercontent.com/akshaymankar/dhall-concourse/0.5.0/package.dhall
let Prelude =
https://prelude.dhall-lang.org/v11.1.0/package.dhall sha256:99462c205117931c0919f155a6046aec140c70fb8876d208c7c77027ab19c2fa
-- Assuming above file is here
let jobs = ./example1.dhall
in Prelude.List.map
Concourse.Types.Job
Concourse.Types.GroupedJob
(λ(j : Concourse.Types.Job) → { job = j, groups = [ "hello-world" ] })
jobs
To notify dhall-fly
that we'd be passing it a List Concourse.Types.GroupedJob
instead of List Concourse.Types.Job
, we have to call it with --pipeline-type grouped-jobs
like this:
fly -t <TARGET> set-pipeline -p hello-dhall -c <(dhall-fly --pipeline-type grouped-jobs <example1.dhall)
The pipeline is defined in ./ci/pipeline.dhall.
The script ./ci/set-pipeline.sh is used to set the pipeline.
The pipeline can be viewed here.
In concourse, a step can be one of get
, put
, task
, in_parallel
, aggregate
, do
or try
. The in_parallel
, do
and aggregate
are list of steps, try
represents one step. This makes definition of the Step
type recursive, as in to define a Step
we'd already need definition of a Step. In total languages like dhall, this is a little tricky to do. There is an explanation in dhall docs about how to do this. An example of this can be found in dhall prelude's definition of the JSON Type
But there is more!: Each step can also have 4 types of hooks
: on_failure
, on_success
, on_abort
and ensure
. Each of these is optional and you guessed it, is of type Step.
So, I decided the type definition for Step
would look like this:
let StepHooks =
λ(Step : Type)
→ { on_failure : Optional Step
, on_success : Optional Step
, on_abort : Optional Step
, ensure : Optional Step
}
let StepConstructors =
λ(Step : Type)
→ { get : ./GetStep.dhall → StepHooks Step → Step
, put : ./PutStep.dhall → StepHooks Step → Step
, task : ./TaskStep.dhall → StepHooks Step → Step
, aggregate : List Step → StepHooks Step → Step
, in_parallel : ./InParallelStep.dhall → StepHooks Step → Step
, do : Step → StepHooks Step → Step
, try : Step → StepHooks Step → Step
}
in ∀(Step : Type) → StepConstructors Step → Step
The way I read it is, a Step is defined for any type for which one can provide constructors for get, put, task, aggregate, etc.
This is the reason this repository includes helpers to construct steps with or without hooks.