What're Differences Between `typed/racket/unsafe` and `typed/racket/optional`?

Questions & Answers
,
1

I'm writing a module that adds types to identifiers imported from untyped code, and I want to avoid contract generation. I see two approaches:

Using typed/racket/unsafe:

(require typed/racket/unsafe)

Using typed/racket/optional with renamed imports:

(require (rename-in typed/racket/base/optional
                    [require/typed unsafe-require/typed]
                    [provide unsafe-provide]
                    [require/typed/provide unsafe-require/typed/provide]
                    [cast unsafe-cast]))

My goal: Type bindings from untyped modules, trusting that the values already match their types.

Both approaches seem to bypass contract generation. But I'm unsure about their exact differences. Could someone clarify:

  1. Are these approaches interchangeable, or do they have meaningful technical differences?

  2. When should each be preferred?

I'd appreciate any insights about these mechanisms or pointers to relevant documentation. Thanks!

2

On May 16, 2025, at 12:59 AM, Noah Ma via Racket Discourse notifications@racket.discoursemail.com says he wants "Type bindings from untyped modules, trusting that the values already match their types” and asks whether someone “could .. clarify”. Given your goal, you want #lang typed/racket/optional as your module level language. (Ben G. created the ‘shallow’ and ‘optional’ variants for his research on making the overhead of T/U interactions tolerable.)