How to prevent leaf functions from being automatically synced?

[inland pewter]

Hello!

Since v0.12.0 the CLI finds all functions without required arguments that return a scalar/primitive value, and prints them (https://github.com/dagger/dagger/pull/7479).

I really like the feature which is super useful most of the time. However, some of my modules have leaf functions that I do not want to automatically execute.

There are leaf functions that I do not want to execute every time as it might create artifacts somewhere, take a long time to execute, ...
e.g.

func (m *MyModule) Publish(ctx context.Context) (string, error) { ... }

There are functions that will fail the whole execution as they do not have required arguments defined when they actually need some.
e.g.

func (m *MyModule) Test(ctx context.Context,
    // SSH Auth Socket to access internal repositories
    // +optional
    sshSocket *dagger.Socket,
    // GitHub Token to access internal repositories
    // +optional
    ghToken *dagger.Secret,
) (string, error) { 
    if sshSocket == nil && ghToken == nil {
        return "", errors.New("either SSH or token authentication is required")
    }
    ...
}

Is there a way to exclude functions from being resolved automatically?

[heady totem]

I really like the feature which is super useful most of the time. However, some of my modules have leaf functions that I do not want to automatically execute.

👋 if you don't want your function to be callable through the CLI you can make it private and then nobody will be able to call it

[twin dove]

This is an open topic @inland pewter ... there's an issue somewhere for it, will try to find it.

One option we discussed is to change the selection filter to print the object state, in other words the actual fields in your object. There's typically a lot of overlap between the two possible selections, but it's not 100% overlap. One benefit of selecting fields is that your problem would go away (those problematic functions would not be selected). One possible downside is that you may have false negatives: accessor functions that are safe to print, but won't be selected.

What are your thoughts on this option @inland pewter ?

[inland pewter]

I still want the function to be callable there is a use-case for it, it's the automatic part that causes issues for me at the moment.

[inland pewter]

It seems like the leaf functions are more useful, I normally do not have anything in the module struct apart from a base container and a source directory.
For instance its super nice to have all kinds of tests run in parallel in my golang module and see this kind of output:

_type: Golang
check: ""
fmt: ""
test: |
    ?           github.com/vincent/example/cmd/hello    [no test files]
    ok          github.com/vincent/example/internal/reverse     (cached)
vet: ""

[inland pewter]

Some way to mark the function as "do not evaluate" would be nice, maybe a doc string decorator in Golang like we have for arguments, i.e.

// Publish artifacts
// +nosync
func (m *MyModule) Publish(ctx context.Context) (string, error) { ... }