How to make Pattern and alert for the following - scala

Question

I've this code which gives locationID and temp, I want a pattern that spits alert whenever the temp > THRESHOLD_TEMPERATURE
I've Tried :-
val pattern1: Pattern[Event,_] = Pattern.begin[Event]("first")
.subtype(Event.getClass)
.where( (evt -> evt.getTemp()) >= TEMPERATURE_THRESHOLD)
.within(Time.seconds(5))
val patternStream: PatternStream[Event] = CEP.pattern(f,pattern1)
val alerts: DataStream[String] = patternStream.flatSelect(
(in: Map[String,String], out: Collector[String]) => {
var first: String = in.get("first")
if (first >= TEMPERATURE_THRESHOLD){
out.collect("Temperature above danger zone")
}
}
)
This is the code for which alert is to be made :-
case class Event(locationID: String, temp: Double)
val TEMPERATURE_THRESHOLD: Double = 50.00
val see: StreamExecutionEnvironment = StreamExecutionEnvironment.getExecutionEnvironment
val properties = new Properties()
properties.setProperty("zookeeper.connect", "localhost:2181")
properties.setProperty("bootstrap.servers", "localhost:9092")
val src = see.addSource(new FlinkKafkaConsumer010[ObjectNode]("broadcast",
new JSONKeyValueDeserializationSchema(false), properties))
var ask = src.map{
r => r.get("value")
}
var data = ask.map { v => {
val loc = v.get("locationID").asInstanceOf[String]
val temperature = v.get("temp").asDouble()
(loc, temperature)
}}
// data.print()
var f = data.keyBy(
v => v._2
)
f.print()
see.execute()
pattern is getting overloaded, also flatSelect

Answer 1

CEP is for detecting Patterns in sequences of events, which doesn't really fit this particular problem. Finding events where the temp > THRESHOLD doesn't require pattern matching -- a simple filter or flatmap will do the job. E.g., you could have a flatmap that transforms every event where the temp is too high to an alert, and ignores all other events.
As for your CEP-based solution, I don't understand what you are saying is going wrong with it. But a couple of things look wrong. The within clause won't do anything: since your patterns are only one event long, they don't have any duration. And you are keying the stream by the temperature, which seems odd.