Handle anonymous environments when injecting it to the context

[glena]

• Splitting context.environment into rootEnvironment and currentEnvironment (potentially just environment)
• When evaluating an anonymous environment, the rootEnvironment will evaluate to the rootest non anonymous environment for the imported environments (wont propagate to the root)

Having these 3 envs:

# root
imports:
  - imported
values:
  root:
    currentEnvironment: ${context.currentEnvironment.name}
    rootEnvironment: ${context.rootEnvironment.name}

# imported
imports:
  - child
values:
  imported:
    currentEnvironment: ${context.currentEnvironment.name}
    rootEnvironment: ${context.rootEnvironment.name}

# child
values:
  child:
    currentEnvironment: ${context.currentEnvironment.name}
    rootEnvironment: ${context.rootEnvironment.name}

In the case the root env is anonymous, it resolves:

{
  "child": {
    "currentEnvironment": "child",
    "rootEnvironment": "imported"
  },
  "imported": {
    "currentEnvironment": "imported",
    "rootEnvironment": "imported"
  },
  "root": {
    "currentEnvironment": "<yaml>",
    "rootEnvironment": "<yaml>"
  }
}

In the case the root env is a named environment, it resolves:

{
  "child": {
    "currentEnvironment": "child",
    "rootEnvironment": "root"
  },
  "imported": {
    "currentEnvironment": "imported",
    "rootEnvironment": "root"
  },
  "root": {
    "currentEnvironment": "root",
    "rootEnvironment": "root"
  }
}

[glena]

I can merge this to reuse the copier between eval/val and this

https://github.com/glena/esc/compare/glena:esc:german/root-current-env-context...glena:esc:german/root-current-env-context-copy?expand=1

[pgavlin]

I think it's fine to have two implementations for now: using the generic copier feels a bit risky to me given the sensitivity of the *value graph to proper memoized copies.

FWIW, the *Values don't have the same restrictions, so we could also simplify this by removing the memoization.

[pgavlin]

one change re: not mutating execcontexts, but generally looks good.

[pgavlin]

I'd prefer not to mutate the values here. I'd prefer that we instead copy the context for each envName and set the values in Copy.

[glena]

done

[pgavlin]

nit: we generally use "reserved" for keys that the user is not allowed to set

Suggested change

	var ErrForbiddenContextkey = errors.New("forbidden context key")
	var ErrReservedContextkey = errors.New("reserved context key")

[glena]

done

[pgavlin]

I think it's fine to have two implementations for now: using the generic copier feels a bit risky to me given the sensitivity of the *value graph to proper memoized copies.

FWIW, the *Values don't have the same restrictions, so we could also simplify this by removing the memoization.

[pgavlin]

sorry, I think I wasn't clear here. this is what I was suggesting:

Suggested change

	func (ec *ExecContext) Copy() *ExecContext {
	return &ExecContext{
	values: ec.values,
	rootEvironment: ec.rootEvironment,
	}
	}
	func (ec *ExecContext) Values(envName string) map[string]Value {
	context := copyContext(ec.values)
	context["currentEnvironment"] = NewValue(map[string]Value{
	"name": NewValue(envName),
	})
	if ec.rootEvironment == AnonymousEnvironmentName || ec.rootEvironment == "" {
	ec.rootEvironment = envName
	}
	context["rootEnvironment"] = NewValue(map[string]Value{
	"name": NewValue(ec.rootEvironment),
	})
	return context
	}
	func (ec *ExecContext) Copy(envName string) *ExecContext {
	values := copyContext(ec.values)
	values["currentEnvironment"] = NewValue(map[string]Value{
	"name": NewValue(envName),
	})
	root := ec.rootEnvironment
	if ec.rootEvironment == AnonymousEnvironmentName || ec.rootEvironment == "" {
	root = envName
	}
	values["rootEnvironment"] = NewValue(map[string]Value{
	"name": NewValue(root),
	})
	return &ExecContext{
	values: values,
	rootEvironment: root,
	}
	}
	func (ec *ExecContext) Values() map[string]Value {
	return ec.values
	}

With this approach the contents of a context are never mutated once the context is created.

[glena]

ah I misunderstood. I thought the concern was mutating the value we have in memory, not the place were we mutate. Done

[pgavlin]

this is still mutating ec

[glena]

yes that is needed to propagate the root environment down the line

[glena]

done