Auto-naming produces invalid physical names when the Pulumi resource name contains characters the resource rejects (e.g. brackets)

[iwahbe]

Brackets are valid in Pulumi resource names, but auto-naming feeds the logical name straight into the physical name without sanitizing it against the resource's naming constraints.

Repro (aws 7.35.0):

import * as aws from "@pulumi/aws";
new aws.s3.Bucket("my-logs[0]", { forceDestroy: true });

error: sdk-v2/provider2.go:572: sdk.helper_schema: validating S3 Bucket (my-logs[0]-6c98083) name:
only alphanumeric characters, hyphens, periods, and underscores allowed in "my-logs[0]-6c98083"

Expected: the generated physical name is sanitized to the target resource's allowed charset (or at minimum the failure points at the logical name as the cause).

This surfaces naturally with runtime: hcl components, where a counted resource's address (logs[0]) becomes the Pulumi resource name — any name-validating resource under count without an explicit name fails to create.

[corymhall]

@iwahbe I'm not sure that this should be considered a bug in the provider and is probably not unique to this provider. Every resource has some naming limitations and it is on the user to make sure that their code follows those limitations. Otherwise we would have to maintain the naming limitation logic for every resource in every provider which is not work that I think we want to take on.

For hcl could we just convert [0] to -0?

[iwahbe]

It can't be done in HCL because the name keys from the actual name of the resource, as seen by the engine. We can express the same in typescript:

const index = 1;
const r = new aws.s3.Bucket(`bucket[{index}]`);

Given that Pulumi sees the above expression without issue, I think it's reasonable for providers to make a best effort attempt at getting this to work. Doing so is already policy in the provider:

pulumi-aws/provider/resources.go

Lines 3749 to 3758 in 2f7485c

	"aws_s3_bucket": {
	Tok: awsResource(s3Mod, "Bucket"),
	Fields: map[string]*tfbridge.SchemaInfo{
	"acl": {
	Type: "string",
	AltTypes: []tokens.Type{awsType(s3Mod, "CannedAcl", "CannedAcl")},
	},
	"bucket": tfbridge.AutoNameTransform("bucket", 63, func(name string) string {
	return strings.ToLower(name)
	}),

I'm simply saying that we need to expand that function to also handle other forbidden characters.