Recovering from failures — catchAll & catchTag

Effect v3.12.0

Typed failures only pay off when you can answer them, and that is recovery. catchAll is the broadest answer. You give it any error and a value to fall back to, and once nothing can fail, the error channel collapses to never. The type now promises a value that cannot fail.

Handling every failure the same way is rarely what you want. Usually you answer one and let the rest carry on. catchTag matches a single tag by name and leaves the others in place. Watch the type follow along. E narrows from two failures to just NotFound. You closed one case and the compiler still holds you to the rest.

To clear several failures at once, catchTags takes one handler per tag, a named branch for each. Cover them all and the channel is back to never. The compiler acts as the checklist, so a failure you forgot is a type error instead of a surprise in production.

import { Data, Effect } from "effect";

class MissingId extends Data.TaggedError("MissingId")<{}> {}
class NotFound extends Data.TaggedError("NotFound")<{}> {}

const lookup = (id: string): Effect.Effect<string, MissingId | NotFound> =>
  id === "" ? Effect.fail(new MissingId()) : Effect.succeed("Ada");

// catchAll handles EVERY failure with a recovery effect. Once nothing can
// fail, the error channel collapses to never.
const safe: Effect.Effect<string> = lookup("").pipe(
  Effect.catchAll(() => Effect.succeed("anonymous")),
);

console.log(Effect.runSync(safe)); // anonymous

❚❚ beat