JulianJaffePinterest commented on a change in pull request #10920: URL: https://github.com/apache/druid/pull/10920#discussion_r614545490
########## File path: extensions-core/spark-extensions/src/main/scala/org/apache/druid/spark/clients/DruidClient.scala ########## @@ -0,0 +1,180 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.spark.clients + +import com.fasterxml.jackson.core.`type`.TypeReference +import com.fasterxml.jackson.databind.ObjectMapper +import com.google.common.net.HostAndPort +import org.apache.druid.java.util.common.{ISE, StringUtils} +import org.apache.druid.java.util.http.client.response.{StringFullResponseHandler, + StringFullResponseHolder} +import org.apache.druid.java.util.http.client.{HttpClient, Request} +import org.apache.druid.query.Druids +import org.apache.druid.query.metadata.metadata.{ColumnAnalysis, SegmentAnalysis, + SegmentMetadataQuery} +import org.apache.druid.spark.MAPPER +import org.apache.druid.spark.configuration.{Configuration, DruidConfigurationKeys} +import org.apache.druid.spark.mixins.Logging +import org.jboss.netty.handler.codec.http.{HttpMethod, HttpResponseStatus} +import org.joda.time.{Duration, Interval} + +import java.net.URL +import java.util.{List => JList} +import javax.ws.rs.core.MediaType +import scala.collection.JavaConverters.{asScalaBufferConverter, mapAsJavaMapConverter, + mapAsScalaMapConverter, seqAsJavaListConverter} + +/** + * Going with SegmentMetadataQueries despite the significant overhead because there's no other way + * to get accurate information about Druid columns. + * + * This is pulled pretty directly from the druid-hadoop-input-format project on GitHub. There is + * likely substantial room for improvement. + * + * @param httpClient + * @param objectMapper + * @param hostAndPort + */ +class DruidClient( + val httpClient: HttpClient, + val objectMapper: ObjectMapper, + val hostAndPort: HostAndPort + ) extends Logging { + private val RETRY_COUNT = 5 + private val RETRY_WAIT_SECONDS = 5 + private val TIMEOUT_MILLISECONDS = 300000 + private val druidBaseQueryURL: HostAndPort => String = + (hostAndPort: HostAndPort) => s"http://$hostAndPort/druid/v2/"; + + /** + * The SQL system catalog tables are incorrect for multivalue columns and don't have accurate + * type info, so we need to fall back to segmentMetadataQueries. Given a DATASOURCE and a range + * of INTERVALS to query over, return a map from column name to a tuple of + * (columnType, hasMultipleValues). Note that this is a very expensive operation over large + * numbers of segments. If possible, this method should only be called with the most granular + * starting and ending intervals instead of over a larger interval. + * + * @param dataSource The Druid dataSource to fetch the schema for. + * @param intervals The intervals to return the schema for. + * @return A map from column name to data type and whether or not the column is multi-value + * for the schema of DATASOURCE. + */ + def getSchema(dataSource: String, intervals: List[Interval]): Map[String, (String, Boolean)] = { Review comment: Intervals are required because the default behavior is unlikely to be what Spark users would expect (e.g. if no time bounds are provided, I would expect to read the whole table in a batch context, not just the last week.) For now, this is only called in [one place](https://github.com/apache/druid/pull/10920/files/331e0c9d088cad3778f3a57432680a1cc607c084#diff-fdc08076d81ecaaf0431bfa58ec51cd8ea8c9efd6aa2ba72faf0eea9928c0ffaR76), which uses `JodaTime.MIN_INSTANT` and `JodaTime.MAX_INSTANT` as the interval bounds if no time filters are provided. This will be an expensive query for large datasources, but until #9707 is implemented I don't know of a better way. I suppose it would make sense to move that default handling into the `getSchema` method here instead though, so I'll make the change. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
