-
.If(pred)(branch) -
.ElseIf(pred)(branch) -
.Else(branch) -
If[A](pred)(branch) -
IfCtx[C](pred)(branch) -
.IfCtx(pred)(branch) -
.ElseIfCtx(pred)(branch)
Conditional branching
Send data down different pipelines with If, ElseIf and Else. You can branch on the data itself, or on config.
Branch on data
The condition runs on the input, and only the matching branch runs:
With an .Else every input is covered. Without one, anything that doesn't match passes straight through:
Start with a branch
If[A] starts a pipeline with the branch itself, no upstream node needed:
Use whole pipelines as branches
A branch can be any pipeline, fan-outs included:
val offers = extractUser
.If(_.tier == "premium") (validate ~> enrich ~> premiumOffer)
.ElseIf(_.tier == "standard") (validate ~> standardOffer)
.Else (guestNotice)
val profile = loadUser
.If(_.wantsDetails) ((self & loadMetrics & loadHistory) ~> fullProfile)
.Else (simpleProfile)
Swap & for &> to fan out concurrently under an effect.
Branch on config and data
When the decision needs config too, write the condition as (cfg: Config) => (data: A) => Boolean:
val route = source
.If((cfg: Config) => (n: Int) => n < cfg.threshold) (formatBelow)
.Else (formatAbove)
route.provide(Config(10)).unsafeRun(5) // "below:5"
Branch on config alone
When only the config matters, use IfCtx / ElseIfCtx, where the condition is just Config => Boolean.
IfCtx[Config] can start a pipeline on its own, no source node needed:
val ingest =
IfCtx[Config](_.isBackfill)(readSnapshot ~> replay ~> load)
.ElseIfCtx(_.isDryRun) (readStream ~> validate ~> logOnly)
.Else (readStream ~> validate ~> load)
ingest.provide(Config(isBackfill = true, isDryRun = false)).unsafeRun(batch)
If you already have an upstream node, call .IfCtx on it instead:
val ingest = source
.IfCtx(_.isBackfill) (readSnapshot ~> replay ~> load)
.ElseIfCtx(_.isDryRun)(readStream ~> validate ~> logOnly)
.Else (readStream ~> validate ~> load)
Scala 2 vs Scala 3
The API is the same on both, but Scala 3 is a bit more flexible.
Scala 3: branches can return different types (you get a union):
val router = score
.If(_ > 0) (toLabel) // String
.ElseIf(_ < 0) (negate) // Int
.Else (toDouble) // Double
The result type is Node[Int, String | Int | Double].
Scala 3: branches can need different configs. Mix one that needs DbConfig with one that needs CacheConfig
and you provide DbConfig & CacheConfig.
etl4s