support for Future-based predict-method, see apache/predictionio#495

Author: Chris Wewerka

build.sbt

@@ -14,7 +14,7 @@ scalaVersion in ThisBuild := "2.11.11"
 
 val mahoutVersion = "0.13.0"
 
-val pioVersion = "0.12.0-incubating"
+val pioVersion = "0.14.0-SNAPSHOT"
 
 val elasticsearchVersion = "5.5.2"
 

src/main/scala/EsClient.scala

@@ -22,13 +22,13 @@ import java.util
 
 import grizzled.slf4j.Logger
 import org.apache.http.util.EntityUtils
-import org.apache.predictionio.data.storage.{ DataMap, Storage, StorageClientConfig }
+import org.apache.predictionio.data.storage.{DataMap, Storage, StorageClientConfig}
 import org.apache.predictionio.workflow.CleanupFunctions
 import org.apache.spark.SparkContext
 import org.apache.spark.rdd.RDD
 import org.elasticsearch.client.RestClient
 import org.apache.http.HttpHost
-import org.apache.http.auth.{ AuthScope, UsernamePasswordCredentials }
+import org.apache.http.auth.{AuthScope, UsernamePasswordCredentials}
 import org.apache.http.entity.ContentType
 import org.apache.http.entity.StringEntity
 import org.apache.http.impl.client.BasicCredentialsProvider
@@ -42,6 +42,9 @@ import org.elasticsearch.spark._
 import org.json4s.JValue
 import org.json4s.DefaultFormats
 import org.json4s.JsonAST.JString
+import ScalaRestClient.ExtendedScalaRestClient
+
+import scala.concurrent.{ExecutionContext, Future}
 // import org.json4s.native.Serialization.writePretty
 import com.actionml.helpers.{ ItemID, ItemProps }
 
@@ -367,20 +370,23 @@ object EsClient {
    *  @param indexName the index to search
    *  @return a [PredictedResults] collection
    */
-  def search(query: String, indexName: String): Option[JValue] = {
+  def search(query: String, indexName: String)(implicit ec: ExecutionContext): Future[Option[JValue]] = {
     logger.info(s"Query:\n${query}")
-    val response = client.performRequest(
+    val responseFuture = client.performRequestFuture(
       "POST",
       s"/$indexName/_search",
-      Map.empty[String, String].asJava,
+      Map.empty[String, String],
       new StringEntity(query, ContentType.APPLICATION_JSON))
-    response.getStatusLine.getStatusCode match {
-      case 200 =>
-        logger.info(s"Got source from query: ${query}")
-        Some(parse(EntityUtils.toString(response.getEntity)))
-      case _ =>
-        logger.info(s"Query: ${query}\nproduced status code: ${response.getStatusLine.getStatusCode}")
-        None
+    responseFuture.map {
+      response =>
+        response.getStatusLine.getStatusCode match {
+          case 200 =>
+            logger.info(s"Got source from query: ${query}")
+            Some(parse(EntityUtils.toString(response.getEntity)))
+          case _ =>
+            logger.info(s"Query: ${query}\nproduced status code: ${response.getStatusLine.getStatusCode}")
+            None
+        }
     }
   }
 

src/main/scala/ScalaRestClient.scala

@@ -0,0 +1,23 @@
+package com.actionml
+
+import org.apache.http.{Header, HttpEntity}
+import org.elasticsearch.client.{Response, ResponseListener, RestClient}
+import scala.collection.JavaConverters._
+import scala.concurrent.{Future, Promise}
+
+object ScalaRestClient {
+
+  implicit class ExtendedScalaRestClient(restClient: RestClient) {
+
+    def performRequestFuture(method: String, endpoint: String, params: Map[String, String],
+                             entity: HttpEntity, headers: Header*): Future[Response] = {
+      val promise: Promise[Response] = Promise()
+      val responseListener = new ResponseListener {
+        override def onSuccess(response: Response): Unit = promise.success(response)
+        override def onFailure(exception: Exception): Unit = promise.failure(exception)
+      }
+      restClient.performRequestAsync(method, endpoint, params.asJava, entity, responseListener, headers: _*)
+      promise.future
+    }
+  }
+}

src/main/scala/URAlgorithm.scala

@@ -20,10 +20,10 @@ package com.actionml
 import java.util
 
 import grizzled.slf4j.Logger
-import org.apache.predictionio.controller.{ P2LAlgorithm, Params }
-import org.apache.predictionio.data.storage.{ DataMap, Event, NullModel, PropertyMap }
+import org.apache.predictionio.controller.{P2LAlgorithm, Params}
+import org.apache.predictionio.data.storage.{DataMap, Event, NullModel, PropertyMap}
 import org.apache.predictionio.data.store.LEventStore
-import org.apache.mahout.math.cf.{ DownsamplableCrossOccurrenceDataset, SimilarityAnalysis }
+import org.apache.mahout.math.cf.{DownsamplableCrossOccurrenceDataset, SimilarityAnalysis}
 import org.apache.mahout.sparkbindings.indexeddataset.IndexedDatasetSpark
 import org.apache.spark.SparkContext
 import org.apache.spark.rdd.RDD
@@ -34,10 +34,12 @@ import org.json4s.JsonAST._
 import org.json4s.JsonDSL._
 import org.json4s.jackson.JsonMethods._
 import com.actionml.helpers._
-
+import scala.concurrent.ExecutionContext
 import scala.collection.JavaConverters._
+import scala.concurrent.Future
 import scala.concurrent.duration.Duration
-import scala.language.{ implicitConversions, postfixOps }
+import scala.language.{implicitConversions, postfixOps}
+import ScalaRestClient.ExtendedScalaRestClient
 
 /** Available value for algorithm param "RecsModel" */
 object RecsModels { // todo: replace this with rankings
@@ -481,51 +483,58 @@ class URAlgorithm(val ap: URAlgorithmParams)
    *  @todo Need to prune that query to minimum required for data include, for instance no need for the popularity
    *       ranking if no PopModel is being used, same for "must" clause and dates.
    */
-  def predict(model: NullModel, query: Query): PredictedResult = {
+  def predict(model: NullModel, query: Query)(implicit ec: ExecutionContext): Future[PredictedResult] = {
 
     queryEventNames = query.eventNames.getOrElse(modelEventNames) // eventNames in query take precedence
 
-    val (queryStr, blacklist) = buildQuery(ap, query, rankingFieldNames)
-    // old es1 query
-    // val searchHitsOpt = EsClient.search(queryStr, esIndex, queryEventNames)
-    val searchHitsOpt = EsClient.search(queryStr, esIndex)
-
-    val withRanks = query.withRanks.getOrElse(false)
-    val predictedResults = searchHitsOpt match {
-      case Some(searchHits) =>
-        val hits = (searchHits \ "hits" \ "hits").extract[Seq[JValue]]
-        val recs = hits.map { hit =>
-          if (withRanks) {
-            val source = hit \ "source"
-            val ranks: Map[String, Double] = rankingsParams map { backfillParams =>
-              val backfillType = backfillParams.`type`.getOrElse(DefaultURAlgoParams.BackfillType)
-              val backfillFieldName = backfillParams.name.getOrElse(PopModel.nameByType(backfillType))
-              backfillFieldName -> (source \ backfillFieldName).extract[Double]
-            } toMap
-
-            ItemScore((hit \ "_id").extract[String], (hit \ "_score").extract[Double],
-              ranks = if (ranks.nonEmpty) Some(ranks) else None)
-          } else {
-            ItemScore((hit \ "_id").extract[String], (hit \ "_score").extract[Double])
-          }
-        }.toArray
-        logger.info(s"Results: ${hits.length} retrieved of a possible ${(searchHits \ "hits" \ "total").extract[Long]}")
-        PredictedResult(recs)
-
-      case _ =>
-        logger.info(s"No results for query ${parse(queryStr)}")
-        PredictedResult(Array.empty[ItemScore])
+    val queryStrBlacklistFuture = buildQuery(ap, query, rankingFieldNames)
+
+    queryStrBlacklistFuture.flatMap {
+      case (queryStr, blacklist) =>
+        // old es1 query
+        // val searchHitsOpt = EsClient.search(queryStr, esIndex, queryEventNames)
+        val searchHitsOptFuture = EsClient.search(queryStr, esIndex)
+
+        val withRanks = query.withRanks.getOrElse(false)
+        searchHitsOptFuture.map {
+          searchHitsOpt =>
+            val predictedResults = searchHitsOpt match {
+              case Some(searchHits) =>
+                val hits = (searchHits \ "hits" \ "hits").extract[Seq[JValue]]
+                val recs = hits.map { hit =>
+                  if (withRanks) {
+                    val source = hit \ "source"
+                    val ranks: Map[String, Double] = rankingsParams map { backfillParams =>
+                      val backfillType = backfillParams.`type`.getOrElse(DefaultURAlgoParams.BackfillType)
+                      val backfillFieldName = backfillParams.name.getOrElse(PopModel.nameByType(backfillType))
+                      backfillFieldName -> (source \ backfillFieldName).extract[Double]
+                    } toMap
+
+                    ItemScore((hit \ "_id").extract[String], (hit \ "_score").extract[Double],
+                      ranks = if (ranks.nonEmpty) Some(ranks) else None)
+                  } else {
+                    ItemScore((hit \ "_id").extract[String], (hit \ "_score").extract[Double])
+                  }
+                }.toArray
+                logger.info(s"Results: ${hits.length} retrieved of a possible ${(searchHits \ "hits" \ "total").extract[Long]}")
+                PredictedResult(recs)
+
+              case _ =>
+                logger.info(s"No results for query ${parse(queryStr)}")
+                PredictedResult(Array.empty[ItemScore])
+            }
+
+            // todo: is this needed to remove ranked items from recs?
+            //if (recsModel == RecsModels.CF) {
+            //  PredictedResult(predictedResults.filter(_.score != 0.0))
+            //} else PredictedResult(predictedResults)
+
+            // should have all blacklisted items excluded
+            // todo: need to add dithering, mean, sigma, seed required, make a seed that only changes on some fixed time
+            // period so the recs ordering stays fixed for that time period.
+          predictedResults
+        }
     }
-
-    // todo: is this needed to remove ranked items from recs?
-    //if (recsModel == RecsModels.CF) {
-    //  PredictedResult(predictedResults.filter(_.score != 0.0))
-    //} else PredictedResult(predictedResults)
-
-    // should have all blacklisted items excluded
-    // todo: need to add dithering, mean, sigma, seed required, make a seed that only changes on some fixed time
-    // period so the recs ordering stays fixed for that time period.
-    predictedResults
   }
 
   /** Calculate all fields and items needed for ranking.
@@ -563,56 +572,60 @@ class URAlgorithm(val ap: URAlgorithmParams)
   def buildQuery(
     ap: URAlgorithmParams,
     query: Query,
-    backfillFieldNames: Seq[String] = Seq.empty): (String, Seq[Event]) = {
+    backfillFieldNames: Seq[String] = Seq.empty)(implicit ec: ExecutionContext): Future[(String, Seq[Event])] = {
 
     logger.info(s"Got query: \n${query}")
 
     val startPos = query.from.getOrElse(0)
     logger.info(s"from: ${startPos}")
 
-    try {
-      // create a list of all query correlators that can have a bias (boost or filter) attached
-      val (boostable, events) = getBiasedRecentUserActions(query)
-      logger.info(s"getBiasedRecentUserActions returned boostable: ${boostable} and events: ${events}")
-
-      // since users have action history and items have correlators and both correspond to the same "actions" like
-      // purchase or view, we'll pass both to the query if the user history or items correlators are empty
-      // then metadata or backfill must be relied on to return results.
-      val numRecs = if (query.num.isDefined) query.num.get else limit // num in query orerrides num in config
-      logger.info(s"UR query num = ${query.num}")
-      logger.info(s"query.num.getOrElse returned numRecs: ${numRecs}")
-
-      val should = buildQueryShould(query, boostable)
-      logger.info(s"buildQueryShould returned should: ${should}")
-      val must = buildQueryMust(query, boostable)
-      logger.info(s"buildQueryMust returned must: ${must}")
-      val mustNot = buildQueryMustNot(query, events)
-      logger.info(s"buildQueryMustNot returned mustNot: ${mustNot}")
-      val sort = buildQuerySort()
-      logger.info(s"buildQuerySort returned sort: ${sort}")
-
-      val json =
-        ("from" -> startPos) ~
-          ("size" -> numRecs) ~
-          ("query" ->
-            ("bool" ->
-              ("should" -> should) ~
-              ("must" -> must) ~
-              ("must_not" -> mustNot) ~
-              ("minimum_should_match" -> 1))) ~
-              ("sort" -> sort)
-
-      logger.info(s"json is: ${json}")
-      val compactJson = compact(render(json))
-      logger.info(s"compact json is: ${compactJson}")
-
-      //logger.info(s"Query:\n$compactJson")
-      (compactJson, events)
-    } catch {
-      case e: IllegalArgumentException => {
-        logger.warn("whoops, IllegalArgumentException for something in buildQuery.")
-        ("", Seq.empty[Event])
-      }
+    // create a list of all query correlators that can have a bias (boost or filter) attached
+    val biasedRecentUserActionsFuture = getBiasedRecentUserActions(query)
+
+    biasedRecentUserActionsFuture.map {
+      case (boostable, events) =>
+        try {
+          logger.info(s"getBiasedRecentUserActions returned boostable: ${boostable} and events: ${events}")
+
+          // since users have action history and items have correlators and both correspond to the same "actions" like
+          // purchase or view, we'll pass both to the query if the user history or items correlators are empty
+          // then metadata or backfill must be relied on to return results.
+          val numRecs = if (query.num.isDefined) query.num.get else limit // num in query orerrides num in config
+          logger.info(s"UR query num = ${query.num}")
+          logger.info(s"query.num.getOrElse returned numRecs: ${numRecs}")
+
+          val should = buildQueryShould(query, boostable)
+          logger.info(s"buildQueryShould returned should: ${should}")
+          val must = buildQueryMust(query, boostable)
+          logger.info(s"buildQueryMust returned must: ${must}")
+          val mustNot = buildQueryMustNot(query, events)
+          logger.info(s"buildQueryMustNot returned mustNot: ${mustNot}")
+          val sort = buildQuerySort()
+          logger.info(s"buildQuerySort returned sort: ${sort}")
+
+          val json =
+            ("from" -> startPos) ~
+              ("size" -> numRecs) ~
+              ("query" ->
+                ("bool" ->
+                  ("should" -> should) ~
+                  ("must" -> must) ~
+                  ("must_not" -> mustNot) ~
+                  ("minimum_should_match" -> 1))) ~
+                  ("sort" -> sort)
+
+          logger.info(s"json is: ${json}")
+          val compactJson = compact(render(json))
+          logger.info(s"compact json is: ${compactJson}")
+
+          //logger.info(s"Query:\n$compactJson")
+          (compactJson, events)
+        } catch {
+          case e: IllegalArgumentException => {
+            logger.warn("whoops, IllegalArgumentException for something in buildQuery.")
+            ("", Seq.empty[Event])
+          }
+        }
     }
   }
 
@@ -792,10 +805,10 @@ class URAlgorithm(val ap: URAlgorithmParams)
   }
 
   /** Get recent events of the user on items to create the recommendations query from */
-  def getBiasedRecentUserActions(query: Query): (Seq[BoostableCorrelators], Seq[Event]) = {
+  def getBiasedRecentUserActions(query: Query)(implicit ec: ExecutionContext): Future[(Seq[BoostableCorrelators], Seq[Event])] = {
 
-    val recentEvents = try {
-      LEventStore.findByEntity(
+    val recentEventsFuture =
+      LEventStore.findByEntityAsync(
         appName = appName,
         // entityType and entityId is specified for fast lookup
         entityType = "user",
@@ -806,13 +819,9 @@ class URAlgorithm(val ap: URAlgorithmParams)
         // targetEntityType = None,
         // limit = Some(maxQueryEvents), // this will get all history then each action can be limited before using in
         // the query
-        latest = true,
-        // set time limit to avoid super long DB access
-        timeout = Duration(200, "millis")).toSeq
-    } catch {
-      case e: scala.concurrent.TimeoutException =>
-        logger.error(s"Timeout when reading recent events. Empty list is used. $e")
-        Seq.empty[Event]
+        latest = true).map(_.toSeq)
+
+    val recoveredRecentEventsFuture = recentEventsFuture.recover {
       case e: NoSuchElementException =>
         logger.info("No user id for recs, returning item-based recs if an item is specified in the query.")
         Seq.empty[Event]
@@ -821,21 +830,24 @@ class URAlgorithm(val ap: URAlgorithmParams)
         Seq.empty[Event]
     }
 
-    val userEventBias = query.userBias.getOrElse(userBias)
-    val userEventsBoost = if (userEventBias > 0 && userEventBias != 1) Some(userEventBias) else None
-    val rActions = queryEventNames.map { action =>
-      var items = Seq.empty[String]
-
-      for (event <- recentEvents) { // todo: use indidatorParams for each indicator type
-        if (event.event == action && items.size < indicatorParams(action).maxItemsPerUser) {
-          items = event.targetEntityId.get +: items
-          // todo: may throw exception and we should ignore the event instead of crashing
+    recoveredRecentEventsFuture.map {
+      recentEvents =>
+        val userEventBias = query.userBias.getOrElse(userBias)
+        val userEventsBoost = if (userEventBias > 0 && userEventBias != 1) Some(userEventBias) else None
+        val rActions = queryEventNames.map { action =>
+          var items = Seq.empty[String]
+
+          for (event <- recentEvents) { // todo: use indidatorParams for each indicator type
+            if (event.event == action && items.size < indicatorParams(action).maxItemsPerUser) {
+              items = event.targetEntityId.get +: items
+              // todo: may throw exception and we should ignore the event instead of crashing
+            }
+            // userBias may be None, which will cause no JSON output for this
+          }
+          BoostableCorrelators(action, items.distinct, userEventsBoost)
         }
-        // userBias may be None, which will cause no JSON output for this
-      }
-      BoostableCorrelators(action, items.distinct, userEventsBoost)
+        (rActions, recentEvents)
     }
-    (rActions, recentEvents)
   }
 
   /** get all metadata fields that potentially have boosts (not filters) */