feat: required arguments pre-filled when creating resource

[mazingerzzz]

It would be awesome to have pre-filled required arguments when created new ressource.

[radeksimko]

Agreed - this would be a useful feature to have.

This could be implemented as part of the completion method via snippets:

terraform-ls/internal/terraform/lang/schema.go

Lines 11 to 45 in b1c85b9

	func snippetForAttrType(placeholder int, attrType cty.Type) string {
	mapSnippet := func(aType cty.Type) string {
	return fmt.Sprintf("{\n"+` "${0:key}" = %s`+"\n}",
	snippetForAttrType(1, aType))
	}
	switch attrType {
	case cty.String:
	return fmt.Sprintf(`"${%d:value}"`, placeholder)
	case cty.List(cty.String), cty.Set(cty.String):
	return fmt.Sprintf(`["${%d:value}"]`, placeholder)
	case cty.Map(cty.String):
	return mapSnippet(cty.String)
	case cty.Bool:
	return fmt.Sprintf(`${%d:false}`, placeholder)
	case cty.List(cty.Bool), cty.Set(cty.Bool):
	return fmt.Sprintf(`[${%d:false}]`, placeholder)
	case cty.Map(cty.Bool):
	return mapSnippet(cty.Bool)
	case cty.Number:
	return fmt.Sprintf(`${%d:42}`, placeholder)
	case cty.List(cty.Number), cty.Set(cty.Number):
	return fmt.Sprintf(`[${%d:42}]`, placeholder)
	case cty.Map(cty.Number):
	return mapSnippet(cty.Number)
	}
	return ""
	}
	func snippetForNestedBlock(name string) string {
	return fmt.Sprintf("%s {\n ${0}\n}", name)
	}

just hooked from HCL labels, rather than attributes and blocks.

With that the required arguments would be pre-filled automatically after picking the type

There's just one edge which comes to my mind - what happens if there are already existing attributes within the block? Should we remove all of these and replace them with new ones which are relevant for the new picked type? I think it's important to understand why/when would this happen - it may be because the user made a mistake when picking the type at first and triggered completion again, or it could be because they are migrating away from a deprecated resource name (e.g. aws_lb -> aws_alb). I think the second use case could probably eventually be dealt with by Terraform's own CLI as it also has the power to move resources within state.

So I'm leaning more towards just replacing whatever there may be inside the block with new required arguments, but I'm keen to hear about other use cases where this may not be appropriate.

[FernandoMiguel]

replace only if file is saved is probably a good idea, so the user can revert if they dont want that.

and if possible, if auto-save is on (i do), only if committed

[danieldreier]

my impulse for the least-surprising or least-destructive would be to strictly add the required attributes/blocks without removing the old ones, and count on other validations to notify the user if they end up with unwanted attributes that no longer fit the resource.

For example, in the use case where I'm refactoring a resource from old name/attributes to the new, it would be annoying to lose everything, because I might want to reference the known-invalid syntax to copy/refactor an expression, reference, etc.

[radeksimko]

We have also discussed the possibility of pre-filling required attributes via code action, which may be the most flexible option here.

Example from gopls:

[radeksimko]

@jpogran implemented this as part of the following PRs which will be part of the upcoming LS / extension release:

• Add experimental Prefill Required Fields feature #657
• Add experimental Prefill Required Fields feature vscode-terraform#799

As mentioned within these issues, the functionality will be (at least initially) opt-in via prefillRequiredFields config option under terraform-ls.experimentalFeatures.

We also plan to add code action that would provide comparable functionality for existing blocks, which is tracked under hashicorp/vscode-terraform#801

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.