This message was deleted.

The

json.dumps

should be within the apply

cfg = pulumi.Config()
    ssm.Parameter(
        f"{prefix}-app-dn-credentials",
        name=f"/{prefix}",
        value=cfg.require_secret('secret').apply(lambda x: json.dumps({"user": "user", "password": x}))
    )

ah, I will not get used to this weirdness 😞

Use Output.all

thanks I looked there

🤷🏽

my_dict = {"user": "user", "password": x, "another": another_secret}
Output.all(**my_dict).apply(lambda args: json.dumps(args))

Thanks that works! I should go to bed, I've been trying too hard past week

To be fair, it isn't very pythonic @great-sunset-355. Pulumi's Output model still trips me up regularly. I'd love to see the api move to asyncio, but I believe there are some fundamental architectural differences which make this unlikely.

I'm simply unable to grasp it I just need to create complex dicts that can be serialized to JSON with this apply magic it's simply just too much

pulumi stack output

?

that won't help me, I have a complex dict of items which I want to json.dump

@great-sunset-355 I understand your frustration, this trips a lot of users up and often comes up when users adopt Pulumi. We’re continuously working on improving it where we can Having said that, your frustration is starting to become difficult to interact with. If you pose concrete questions we can try to answer them, but belligerence and complaints make it difficult for us and other community members to help. Please try and consider the way you’re interacting with the community

Thank you, I'm really sorry for my behaviour, I'm being pushed in the corner and have no results and sleep deprivation. I really appreciate you are trying to help me. I had to clean up the code because it's pretty messy right now. Are there any examples of recommended patterns? I studied the example you shared before

ol-infrastructure

but I could not find enough with outputs. I'm trying to create a

container_definitions

https://www.pulumi.com/docs/reference/pkg/aws/ecs/taskdefinition/#example-using-container_definitions-and-inference_accelerator using

json.dumps

app_env_vars = {
        "EMAIL_HOST_USER": ses_user_access_key.id,
        "EMAIL_HOST_PASSWORD": ses_user_access_key.ses_smtp_password_v4,
        "a": "b",
    }

    # container_name = container_name

    container_definition = {
        "name": container_name,  # is output Config.require
        "image": image_name,  # just str
        "portMappings": [{"containerPort": 80, "hostPort": 80, "protocol": "tcp"}],
        "environment": [{"name": k, "value": v} for k, v in app_env_vars.items()],
    }

    expected_result = container_definition.apply(json.dumps)

@great-sunset-355 every time you want to go from a dictionary of outputs to an output dictionary, you’ll need to use

apply

. So for your code, I think something like the following should work:

app_env_vars_dict = {
        "EMAIL_HOST_USER": ses_user_access_key.id,
        "EMAIL_HOST_PASSWORD": ses_user_access_key.ses_smtp_password_v4,
        "a": "b",
    }

app_env_vars = Output.all(**app_env_vars_dict).apply(args: [{"name": k, "value": v} for k, v in args.items()])

container_definition = {
        "name": container_name,
        "image": image_name,
        "portMappings": [{"containerPort": 80, "hostPort": 80, "protocol": "tcp"}],
        "environment": app_env_vars,
    }

expected_result = Output.all(**container_definition).apply(json.dumps)

Note that you have multiple outputs that you are trying to convert to json. Output.all accepts keyword args, so when you do

**container_definition

it spreads the keys/values as keyword args.

I’ll also replug my own blog post: https://www.leebriggs.co.uk/blog/2021/05/09/pulumi-apply.html

yeah the blogpost was great entry point, helped for a little until I got stuck. I'm not great with absorbing theory, so more complex examples would be awesome

We are happy to help if you provide concrete examples of what you need

@billowy-army-68599 I've been thinking about this, do you think that custom json Encoder passed to

json.dump

could handle the outputs?

yes, if it can handle the async nature of the output. it needs to await the result, but otherwise this would work

Are there any examples of creating a custom output so I could play with this without calling real resources? I'd like to create isolated environment where I could control and simulate Outputs. If I understand it correctly I'd use component resource and register the outputs. Correct?

yes that sounds right, you can also implement a dynamic provider

Are there any examples of creating a custom output so I could play with this without calling real resources?

You can use

Output.from_input()

as described in https://www.pulumi.com/docs/intro/concepts/inputs-outputs/#convert-input-to-output-through-interpolation

@red-match-15116 @billowy-army-68599 I finally got the expected structure! After trying

Output.from_input()

I found out that it can also take the Output or anything else. Also reviewing the ECS Task docs one more time. It turned that

ContainerDefinitions

is not a JSON object but JSON array of objects! from there I made this code: That yields into the desired structure!

container_definition = {
            "name": config.container_name,  # is output Config.require
            "image": config.image_name,  # just str
            "portMappings": [{"containerPort": 80, "hostPort": 80, "protocol": "tcp"}],
            "environment": [{"name": k, "value": v} for k, v in config.env_vars.items()],  # config.env_vars is Dict[str,Any]
        }
        container_definitions = Output.from_input([container_definition]).apply(json.dumps)  # List[Dict[str,str]]

Thank you so much for baring with me, in case you'd like to have this in docs or any examples, let me know and I can put something more informative together. Because I believe others will find this helpful.