Allow to exclude nested namespaces

[ghostbuster91]

When generating scala classes from smithy model we often need to exclude certain namespaces due to various reasons.

For that we were using --exclude-ns option in smithy4s-cli. However, recently we did some small refactoring that moved some of the shapes from the top namespace that was excluded (e.g. a.b) into sub-namespaces (e.g. a.b.c and a.b.d).

This cause issues in the generated code as those nested namespaces were not excluded.

Would it be possible to allow excluding namespaces in a more convenient way? e.g. --exclude-ns a.b.* would exclude .a.b as well as any other namespace that starts with a.b

[kubukoz]

I'm all for it, if it's simple to implement then it shouldn't be a big deal to @Baccata either I assume

[Baccata]

PR away :)

[kubukoz]

@Baccata how flexible do you think this should be? Accept arbitrary regex, or just wildcards (if so, only trailing/leading or everywhere?)

[Baccata]

just wildcards (glob patterns) would be fine. Whatever you can find a concise implementation for.

[msosnicki]

I can have a look into it. But first I would like to clarify the behavior.
Going with glob patterns, it would mean that we have:

• a.b.** - matches all namespaces below a.b (but not a.b itself)
• a.b.*.d - matches only a single level on a namespace

Exact namespaces can be excluded with a.b and the combination of * and ** gives a lot flexibility in other scenarios.

But the a.b.* could be treated more like regex (in which case dots would have to be escaped internally to mean dot literals), which would mean that it matches all of the below:

• a.b
• a.b.c
• a.b.c.d

Not sure which is more preferable here. I like the first solution more, as it feels more precise. But can be cumbersome too, if I want to exclude all the shapes from a.b and nested namespaces: I would have to include two patterns: a.b and a.b.**.

[kubukoz]

IMO

• a.b.* matches everything below a.b but not a.b itself
• a.b* matches a.b and everything below

I don't think we have a need for * in the middle, it could be just a distraction at this point

so I guess I'm just proposing prefix-based exclusion/inclusion, with the special case of "no asterisk" which is an exact match

[msosnicki]

Had a brief look, looks like alloy will also have to be involved https://github.com/disneystreaming/alloy/blob/6ac81f362362ea48165c07589021181e484f10ec/modules/openapi/src/alloy/openapi/package.scala#L33

I wonder why it only takes the allowedNS param, and the excludedNS is ignored in openapi context?

[Baccata]

No particular reason, probably just an oversight