feat: terramate generate: add overall tf code generation

[katcipis]

Introduces generate_hcl code generation on the cli, including outdated code detection. Here is an example to showcase how to use + expected results:

#!/bin/bash

set -o nounset
set -o errexit

basedir=$(mktemp -d)

cd "${basedir}"

# We need the project root config since it is not git
cat > terramate.tm.hcl <<- EOM
terramate {
  config {
  }
}
EOM

mkdir -p stacks/stack-{1,2}
terramate stacks init stacks/stack-{1,2}

# Defining some code for all stacks
# Globals are defined on root and also on stacks, to showcase possibilities
# Inspired on: https://github.com/mineiros-io/terramate-blueprint-iac-github/tree/main/stacks/repositories/teamA
cat > stacks/terramate.tm.hcl <<- EOM
globals {
    tf_version = "1.1.3"
    gh_owner = "the-awesome-mineiros"
    gh_provider_version = "4.19.1"
}

generate_hcl "provider.tf" {

    provider "github" {
      owner = global.gh_owner
    }

    terraform {
      required_providers {
        github = {
          source  = "integrations/github"
          version = global.gh_provider_version
        }
      }
    }

    terraform {
      required_version = global.tf_version
    }
}

generate_hcl "main.tf" {
  module "repositories" {
    source = "../../../modules/terramate-github-repositories"

    repositories        = global.repositories
    repository_defaults = global.repository_defaults
    member_map          = global.member_map
  }
}
EOM

# Now add some per stack globals

cat > stacks/stack-1/terramate.tm.hcl <<- EOM
globals {
    repositories = [
        {
            name = "stack-1"
            visibility = "private"
        }
    ]
    repository_defaults = {
        config = "stack-1"
    }
    member_map  = {
        "first.lastname" = "powerpuffGirls69"
    }
}
stack {}
EOM

cat > stacks/stack-2/terramate.tm.hcl <<- EOM
globals {
    repositories = [
        {
            name = "stack-2"
            visibility = "public"
        }
    ]
    repository_defaults = {
        config = "stack-2"
    }
    member_map  = {
        "first.lastname" = "onePunchMan666"
    }
}
stack {}
EOM


echo "initial tree"
tree

echo
echo "generating code for all stacks"
terramate generate

echo "tree after generating code"
tree

echo
echo "checking generated main.tf files"
terramate run -- cat main.tf

echo
echo "checking generated provider.tf files"
terramate run -- cat provider.tf

[codecov-commenter]

Codecov Report

Merging #193 (4f75d50) into main (4c69a40) will increase coverage by 0.91%.
The diff coverage is 92.73%.

❗ Current head 4f75d50 differs from pull request most recent head 74e1199. Consider uploading reports for the commit 74e1199 to get more accurate results

@@            Coverage Diff             @@
##             main     #193      +/-   ##
==========================================
+ Coverage   61.90%   62.82%   +0.91%     
==========================================
  Files          30       30              
  Lines        5064     5197     +133     
==========================================
+ Hits         3135     3265     +130     
- Misses       1755     1757       +2     
- Partials      174      175       +1     

Flag	Coverage Δ
tests	62.82% <92.73%> (+0.91%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
generate/generate.go	90.79% <91.54%> (+1.21%)	⬆️
generate/genhcl/genhcl.go	90.62% <95.00%> (ø)
hcl/hcl.go	80.02% <100.00%> (+0.40%)	⬆️
test/os.go	84.81% <100.00%> (+0.81%)	⬆️
test/sandbox/sandbox.go	86.85% <100.00%> (+0.48%)	⬆️
list.go	91.89% <0.00%> (+0.71%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4c69a40...74e1199. Read the comment docs.

[kassianh]

Lgtm but also a bit over my head.

[thiesen]

are we actually setting wantErr in any of the test cases? if this is not the case, shouldn't this assertion fail?

[katcipis]

If we never set wantErr, meaning that we are not expecting a failure, and no failure happens, this is the same as:

assert.IsError(t, nil, nil)

Which works, because nil == nil (they match). If we get an unexpected error, it fails, and if we define wantErr then it would fail if the got err doesn't match.

[katcipis]

But since we are not using it here we can remove it, I usually have some error handling tests but so far haven't added at this level. But here you can see how it works:

https://github.com/mineiros-io/terramate/blob/ad148a17de74f6f7b6e4594c69188f920f265483/generate/genhcl/genhcl_test.go#L577

When you set the error, it will check for a specific match with the sentinel wrapped inside the error, when you don't set it it will check if both errors are nil, since only a nil error "is" another nil error.

[katcipis]

Simpler example: https://go.dev/play/p/aMat5TWAU7t

[thiesen]

got it! thanks for taking the time to explain! ❤️

[thiesen]

dude, looks great to me! 🧙‍♂️

I added some questions but nothing should be blocking - just https://github.com/mineiros-io/terramate/pull/193/files#r798043896 but I assume I'm missing something since the tests are green)

[katcipis]

dude, looks great to me! mage_man

I added some questions but nothing should be blocking - just https://github.com/mineiros-io/terramate/pull/193/files#r798043896 but I assume I'm missing something since the tests are green)

Thanks for the review man ❤️

Tried to explain and provide some examples about errors.Is, the assert.IsError is a wrapper around errors.Is, so they behave the same way.

Also improved the code with all your suggestions, thanks again ❤️

[thiesen]

lgtm! 🚀