Awesome
Dhall AWS CloudFormation
dhall-aws-cloudformation
contains Dhall bindings to AWS CloudFormation, so you can generate CloudFormation template from Dhall expressions. This will let you easily typecheck, template and modularize your CloudFormation definitions.
:mag: References
:bulb: Examples
:book: Usage
Use resource schema
AWS Cloudformation has massive amount of specifications, to load all dhall remotely will be very slow and impractical.
One simply way to make import faster is by only importing just each resource you need
Remote import resource
let Function =
-- import Lambda Function type definition
https://raw.githubusercontent.com/jcouyang/dhall-aws-cloudformation/0.10.82/cloudformation/AWS::Lambda::Function.dhall
sha256:c2e1876f87edb0e7c08058a8a0c90c6a56be530b6c4c7ba3ef3c897feb75cde1
let Fn =
-- Intrinsic functions
https://raw.githubusercontent.com/jcouyang/dhall-aws-cloudformation/0.10.82/Fn.dhall
sha256:b2cf7212998902c44ba1bf1670a8e0bc40562542b9b525587cd044f317644e47
let S =
{-
Each AWS String field can be either a String or a Intrinsic function, we can use `Fn.renderText "abc"` to create static string
Or `Fn.render (Ref "abc")` to create a function that ref to a string
-} Fn.renderText
let render =
-- function can be nested `render (Fn.Ref (Fn.GetAtt (Fn.String "abc.property")))`
Fn.render
let example0 =
{ Resources.HelloWorldFunction
= Function.Resources::{
, Properties = Function.Properties::{
, Handler = Some (S "index.handler")
, Code = Function.Code::{
, S3Bucket = Some (S "lambda-functions")
, S3Key = Some (S "amilookup.zip")
}
, Runtime = Some (S "nodejs12.x")
, Role = render (Fn.Ref "role logical id")
, Timeout = Some +25
, TracingConfig = Some { Mode = Some (S "Active") }
}
}
}
in example0
to convert to CloudFormation JSON file just
dhall-to-json < ./template.dhall > ./template.json
which generates
{
"Resources": {
"HelloWorldFunction": {
"Properties": {
"Code": {
"S3Bucket": "lambda-functions",
"S3Key": "amilookup.zip"
},
"Handler": "index.handler",
"Role": {
"Ref": "role logical id"
},
"Runtime": "nodejs12.x",
"Timeout": 25,
"TracingConfig": {
"Mode": "Active"
}
},
"Type": "AWS::Lambda::Function"
}
}
}
Other way around is build the binary of subset of the resources using nix
Build and load package.dhall binary to local cache
Have something like ./examples/example0.nix
, and the dhall file you want to compile ./examples/example0.dhall
let aws =
missing
sha256:a04e4db67b092e40987639cca5cd845f452b3984ee7ec77172f815a31e830325
let Function = aws.Cloudformation.`AWS::Lambda::Function`
let Fn = aws.Fn
let S =
{-
Each AWS String field can be either a String or a Intrinsic function, we can use `Fn.renderText "abc"` to create static string
Or `Fn.render (Ref "abc")` to create a function that ref to a string
-} Fn.renderText
let render =
-- function can be nested `render (Fn.Ref (Fn.GetAtt (Fn.String "abc.property")))`
Fn.render
let example0 =
{ Resources.HelloWorldFunction
= Function.Resources::{
, Properties = Function.Properties::{
, Handler = Some (S "index.handler")
, Code = Function.Code::{
, S3Bucket = Some (S "lambda-functions")
, S3Key = Some (S "amilookup.zip")
}
, Runtime = Some (S "nodejs12.x")
, Role = render (Fn.Ref "role logical id")
, Timeout = Some +25
, TracingConfig = Some { Mode = Some (S "Active") }
}
}
}
in example0
Add all the resources you need to cf-preset
, run nix-build, if the subset is not too large it will be very quick, and you will see something like:
⚠ ─ If error occured, you may need to update the sha256 in your dhall file e.g.
│ let aws = missing sha256:a04e4db67b092e40987639cca5cd845f452b3984ee7ec77172f815a31e830325
Actually the first time it will fail since you can't guess the correct sha of the subset binary, now if you update example0.dhall with the correct sha,
it should then be able to compile to ./result/example0.yaml
let aws =
missing
sha256:a04e4db67b092e40987639cca5cd845f452b3984ee7ec77172f815a31e830325
let Function = aws.Cloudformation.`AWS::Lambda::Function`
Intrinsic Function
The following intrinsic functions are implemented, you can find examples of using intrinsic function in Fn.dhall document
- Fn::Base64
- Fn::Cidr
- Condition functions
- Fn::FindInMap
- Fn::GetAtt
- Fn::GetAZs
- Fn::ImportValue
- Fn::Join
- Fn::Select
- Fn::Split
- Fn::Sub
- Fn::Transform
- Ref
Type Safe Fn::GetAttr
Instead of manually looking for AWS documents to make sure the resource has what attributes, we can just use <Resource>.GetAttr.<attribute name>
:
render (Role.GetAttr.Arn "HelloWorldFunctionRole")
So the compiler can just help you find the correct attributes available.
Sam Policy Templates
Cloudformation's Policy document is loosy type as just JSON, it is hard to get the policy right and too many boilerplates to create a Dhall JSON data thanks to AWS SAM there are some common policy documents we can laverage
these templates are translated into Dhall functions, so you don't need to use SAM to be able to use these policy documents.
let Policy = https://github.com/jcouyang/dhall-aws-cloudformation/raw/0.10.82/cloudformation/AWS::IAM::Role/Policy.dhall
let Sam/Policy = https://github.com/jcouyang/dhall-aws-cloudformation/raw/0.10.82/sam/policy-template/package.dhall
...
Policies = Some [Policy::{
, PolicyDocument = Sam/Policy.DynamoDBReadPolicy (Fn.String "DBName")
, PolicyName = s "dynamo read only"
}]
...
will generates
{
"Policies": [
{
"PolicyDocument": {
"Statement": [
{
"Action": [
"dynamodb:GetItem",
"dynamodb:Scan",
"dynamodb:Query",
"dynamodb:BatchGetItem",
"dynamodb:DescribeTable"
],
"Effect": "Allow",
"Resource": [
{
"Fn::Sub": [
"arn:${AWS::Partition}:dynamodb:${AWS::Region}:${AWS::AccountId}:table/${tableName}",
{
"tableName": "DBName"
}
]
},
{
"Fn::Sub": [
"arn:${AWS::Partition}:dynamodb:${AWS::Region}:${AWS::AccountId}:table/${tableName}/index/*",
{
"tableName": "DBName"
}
]
}
]
}
]
},
"PolicyName": "dynamo read only"
}
]
}
:coffee: Contribute
Build and Test
> nix-shell
$ cabal build
$ cabal test
Generate Type Definitions
e definitions are generated from config file ./config.dhall
which contains specifications used by AWS CDK as well:
regenerate types definition files, simply run
$ cabal run
if you just want to regenerate dhall files without setting up haskell dev environment, just
docker run --rm -v $(pwd):/data -w /data ghcr.io/jcouyang/dhall-aws-cloudformation
:warning: Known Issue
The following CloudFormation definitions will raise assertion error due to invalid type definition such as empty type or cyclic import
AWS::EMR::Cluster
AWS::EMR::InstanceGroupConfig
AWS::EMR::InstanceFleetConfig
AWS::Macie::FindingsFilter
AWS::Connect::EvaluationForm
AWS::IoTTwinMaker::ComponentType
AWS::IoTTwinMaker::Entity
AWS::Lex::Bot
AWS::DataBrew::Recipe
AWS::FIS::ExperimentTemplate
AWS::SageMaker::ModelBiasJobDefinition
AWS::SageMaker::ModelQualityJobDefinition
AWS::SageMaker::MonitoringSchedule
AWS::SageMaker::DataQualityJobDefinition
AWS::SageMaker::ModelExplainabilityJobDefinition
AWS::S3::StorageLens
AWS::StepFunctions::StateMachine
AWS::MWAA::Environment
AWS::WAFv2::RuleGroup
AWS::WAFv2::WebACL
AWS::ServiceDiscovery::PrivateDnsNamespace
AWS::ServiceDiscovery::PublicDnsNamespace
AWS::AmplifyUIBuilder::Component
AWS::AmplifyUIBuilder::Theme
AWS::EMRServerless::Application