Support stdin and stdout from piped process

[pecigonzalo]

Fixes: #32

Only format the content between ------ and output the rest, additionally allowing the user to pass input more clearly.

@dmlittle This a WIP implementation of the changes I requested on #32, I want to clean the code + add tests, but first wanted to check with you, is in scope of this to also format apply or only plan? because apply does not output ---.

Manual tests

Subproc and input

Test that a subproc and easily show and receive input. Previously, the input was passed, but because we could not see the output, it was hard to tell if we had to pass input, basically it was a bad unix pipe citizen.

./test/subproc.sh | ./scenery

New Output:

❯ ./test/subproc.sh | ./scenery
Enter some text:
asd
you typed: asd

Previous Output:

❯ ./test/subproc.sh | scenery

Format only between ----

Format only the code between ----, output the rest to the console, to make it easier to interact and see progress of commands, wrappers and even the initial part of the terraform plan.

pushd test
terraform init
terraform apply
# Modify test/main.tf actions
popd


./test/wrapper.sh | ./scenery

New Output:

❯ ./test/wrapper.sh | ./scenery
+ pushd test
+ trap popd EXIT
~/Workspace/src/scenery/test ~/Workspace/src/scenery
+ echo asdasd
asdasd
+ echo '!23123'
!23123
+ terraform plan -input=true -lock=false
Refreshing Terraform state in-memory prior to plan...
The refreshed state will be used to calculate this plan, but will not be
persisted to local or remote state storage.

data.aws_iam_policy_document.write: Refreshing state...
aws_iam_policy.write: Refreshing state... (ID: arn:aws:iam::155161382138:policy/testpolicy-deleteme)

------------------------------------------------------------------------
~ aws_iam_policy.write
    policy:  {
               "Statement": [
                 {
                   "Action": [
            -        "ecr:UploadLayerPart",
            +        "ecr:UploadLayer*",
                     "ecr:PutImage",
                     "ecr:InitiateLayerUpload",
                     "ecr:CompleteLayerUpload"
                   ],
                   "Effect": "Allow",
            

Plan: 0 to add, 1 to change, 0 to destroy.
------------------------------------------------------------------------

Note: You didn't specify an "-out" parameter to save this plan, so Terraform
can't guarantee that exactly these actions will be performed if
"terraform apply" is subsequently run.

+ popd
~/Workspace/src/scenery

Previous Output:

❯ ./test/wrapper.sh | scenery  
+ pushd test
+ trap popd EXIT
+ echo asdasd
+ echo '!23123'
+ terraform plan -input=true -lock=false
+ popd
~ aws_iam_policy.write
    policy:  {
               "Statement": [
                 {
                   "Action": [
            -        "ecr:UploadLayerPart",
            +        "ecr:UploadLayer*",
                     "ecr:PutImage",
                     "ecr:InitiateLayerUpload",
                     "ecr:CompleteLayerUpload"
                   ],
                   "Effect": "Allow",
            

Plan: 0 to add, 1 to change, 0 to destroy.

[dmlittle]

@pecigonzalo thanks for starting work on this issue!

is in scope of this to also format apply or only plan? because apply does not output ---.
I think there are 2 different issues currently open to add extra functionality:

• Feature: Show non-plan output #32 deals with adding the ability to show non-terraform plan output (things included before or after the plan). This doesn't need to add the ability to interact with prompts in the terraform commands. While I'm OK with adding this functionality I don't want this to be the default behavior of scenery and it should be enabled only when run with a special flag (maybe --full-output/-F).
• Add support for Terraform/Terragrunt Apply #30 deals with the ability to interact with the terraform apply command that asks for user-input.

I think we should tackle these 2 issues in different PRs. You do not need to add the ability to parse terraform apply output if you don't want to as this is currently not supported by scenery.

[pecigonzalo]

Indeed, im not trying to fix 30, only 32. I was just ensuring I was not removing any previous functionality.
I actually believe, it should not parse apply, only plan.

Regarding #32 and the -F option, I actually think it should be the opposite, right now scenery is not a good "unix" citizen, as its destroying the ability to combine or pipe other tools inline. I would prefer to make this default and have an option to remove it, as in its current behavior it prevents even terraform errors or prompts. Let me know your thoughts

[pecigonzalo]

@dmlittle Anything that holds merging this?

[dmlittle]

@pecigonzalo sorry for the delayed response, I've been busy these past few weeks and just had time to catch up on this issue. As I mentioned earlier I'm willing to include this functionality as an optional flag but not as the default behavior.

I don't necessarily agree with the Unix good citizen argument. The purpose of this tool is to make a terraform plan output easier to read by stripping away extraneous information and prettifying things when possible. If you have a script that calls terraform plan and the output of that script is being piped into scenery you can fix the stripping of information by scenery by calling scenery within the script instead of piping the entire script into scenery.

As for scenery preventing terraform errors and/or prompts that's a bug and should be fixed. I'll look more into it.

[pecigonzalo]

@dmlittle No worries, I wasnt sure how to proceed.

As for scenery preventing terraform errors and/or prompts that's a bug and should be fixed. I'll look more into it.

This is exactly what this change is doing, is covered by it.

AFAIK you cant detect other errors/prompts in a generic enough way to craft a "only these prompts" so you cant do this in a flag. You either pass stdin or not, I can make a flag that fixes the issue or not, but not fix the issue and also optionally strip other content from stdin.

Event in this case:

If you have a script that calls terraform plan and the output of that script is being piped into scenery you can fix the stripping of information by scenery by calling scenery within the script instead of piping the entire script into scenery.

Its not behaving and its stripping stderr/stdin content, preventing prompts and errors from going out.

[pecigonzalo]

@dmlittle any comments on this?