EventStoreOptions has no AddEventType — event type pre-registration is unreachable from the public API

[jeremydmiller]

Surfaced by the compliance-library adoption (#393). Small public-API parity gap.

Gap

StoreOptions.Events is an EventStoreOptions facade exposing RegisterTagType<TTag>() / RegisterTagType<TTag>(string) and a pile of flags — but no AddEventType. Marten has opts.Events.AddEventType<TEvent>() and AddEventTypes(IEnumerable<Type>), and the shared JasperFx.Events.IEventRegistry declares void AddEventType(Type).

So a Polecat user following Marten's documented pattern:

opts.Events.AddEventType<StudentEnrolled>();   // does not compile on Polecat

has to go around the facade:

opts.EventGraph.AddEventType(typeof(StudentEnrolled));

which works (StoreOptions.EventGraph is public and EventGraph implements IEventRegistry) but is not discoverable and is not what any Marten-derived docs or samples show.

This is exactly what the compliance fixture had to do — see PolecatComplianceFixture.PolecatComplianceRegistrar.AddEventType in #393, which carries a comment explaining the detour.

Why it matters beyond ergonomics

Pre-registration is not cosmetic. The shared registry doc for AddEventType calls out the async-daemon case: "can help with asynchronous projections where the daemon hasn't yet encountered the event type." A user who cannot reach it from the options facade cannot apply that mitigation without knowing about EventGraph.

Suggested fix

Additive, ~6 lines on EventStoreOptions, mirroring the existing RegisterTagType delegation pattern:

public void AddEventType<TEvent>() => EventGraph!.AddEventType(typeof(TEvent));
public void AddEventType(Type eventType) => EventGraph!.AddEventType(eventType);
public void AddEventTypes(IEnumerable<Type> types) { foreach (var t in types) EventGraph!.AddEventType(t); }

EventGraph is assigned in the StoreOptions constructor, so it is non-null at configure time — same lifetime assumption RegisterTagType already makes with its EventGraph!.

Note: deliberately not returning IEventStoreOptions for fluent chaining. Marten's AddEventType<T>() does, and that return type is precisely what makes marten#5114 (a generic DIM on IEventRegistry) impossible to land there without a source-breaking change. Keeping these void matches the shared interface and leaves the door open.

Related: marten#5114, marten#5119.

🤖 Generated with Claude Code