ARN parsing is too strict on resources

[dacut]

I have a CloudWatch alarm named Production:LB:High4xx, which has an ARN in the form: arn:aws:cloudwatch:us-west-1:123456789012:alarm:Production:LB:High4xx. Note the multiple : after the account id.

This pattern doesn't satisfy either Resource::TypedId or Resource::QTypedId variants, so code like this fails:

ARN::from_str("arn:aws:cloudwatch:us-west-2:123456789012:alarm:Production:LB:High4xx").unwrap()

I'm not aware of any particular rules on the resource section of an ARN here; unfortunately, IIRC the original (AWS internal) Aspen spec left it up to the services teams, and CloudWatch made some particularly bad choices here with regard to separators and user-defined names. There are also cases (notably in S3 paths) where double slashes are allowed and needed (though this seems to work ok).

You might consider just letting the ':' codepath go similarly to the '/' codepath, or even abandoning validating the resource part entirely (beyond that "it's ASCII").

[johnstonskj]

Thanks! I'll take a look.

[johnstonskj]

Have undertaken a major refactor, removing the old Resource enum and adding new Identifier and ResourceIdentifier types. I specifically added a test case for your CloudWatch example as well.

rust-aws-arn/tests/test_arn.rs

Line 38 in 8f9d57f

fn test_github_issues_2() {

[dacut]

I read through the latest, and it definitely solves this issue. Thanks! I'll close this out here.