dajac commented on code in PR #23732: URL: https://github.com/apache/kafka/pull/23732#discussion_r4248886181
########## group-coordinator/src/main/java/org/apache/kafka/coordinator/group/assignor/uniform2/GroupModel.java: ########## @@ -0,0 +1,429 @@ +/* + * 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.kafka.coordinator.group.assignor.uniform2; + +import org.apache.kafka.common.Uuid; +import org.apache.kafka.coordinator.group.api.assignor.GroupSpec; +import org.apache.kafka.coordinator.group.api.assignor.SubscribedTopicDescriber; +import org.apache.kafka.coordinator.group.api.assignor.SubscriptionType; +import org.apache.kafka.coordinator.group.assignor.uniform2.util.IntList; + +import java.util.ArrayList; +import java.util.Arrays; +import java.util.HashMap; +import java.util.List; +import java.util.Map; + +/** + * The members, the subscribed topics and the cohorts of a group, numbered by small integers. + * + * <p>Members are numbered in member id order and topics in topic id order, so that every + * decision taken on these numbers depends on the content of the input only, never on the + * iteration order of the collections of the spec. A cohort is a set of members having the same + * subscription; cohorts are numbered in the order of their first member. The topics are those + * subscribed by at least one member and which exist. + * + * <p>With a homogeneous subscription type, only the subscription of the first member is read, + * as the coordinator guarantees that every member has the same one, and the group is a single + * cohort. Otherwise every subscription is read once, and summarized by a hash of its topics + * that does not depend on their order, so that grouping the members into cohorts is linear in + * the size of the subscriptions. + */ +final class GroupModel { + /** + * Reads the subscriptions of the members, numbering their topics that exist in the order they + * are met, and remembering their partition counts. + */ + private static final class SubscriptionReader { + private final GroupSpec groupSpec; + private final SubscribedTopicDescriber describer; + + /** + * Per topic met, its number in the order met, or -1 if it does not exist, so that the + * describer is asked once per topic. + */ + private final Map<Uuid, Integer> indices = new HashMap<>(); + + /** + * The topics met that exist, in the order met. + */ + private final List<Uuid> ids = new ArrayList<>(); + + /** + * The partition counts of the topics met that exist, in the order met. + */ + private final IntList partitionCounts = new IntList(16); + + SubscriptionReader(GroupSpec groupSpec, SubscribedTopicDescriber describer) { + this.groupSpec = groupSpec; + this.describer = describer; + } + + /** + * Reads the subscription of the member. + * + * @param memberId The member id. + * @param topics Receives the numbers of its subscribed topics that exist. + */ + void subscribedTopics(String memberId, IntList topics) { + for (Uuid topicId : groupSpec.memberSubscription(memberId).subscribedTopicIds()) { + var index = indices.get(topicId); + if (index == null) { + // Met for the first time: numbered next, unless it does not exist. + int partitionCount = describer.numPartitions(topicId); + index = partitionCount < 0 ? -1 : ids.size(); + if (index >= 0) { + ids.add(topicId); + partitionCounts.add(partitionCount); + } + indices.put(topicId, index); + } + if (index >= 0) { + topics.add(index); + } + } + } + } + + /** + * The cohorts found so far, looked up by the topics of their subscription. + * + * <p>A subscription is summarized by a hash that does not depend on the order of its topics, + * and the cohorts with the same hash form a chain. A subscription equals a cohort of its + * chain when it has as many topics and its member subscribes to all the topics of the cohort, + * which the last subscriber of every topic tells without sorting the subscription nor copying + * it. + */ + private static final class CohortsByTopics { Review Comment: > This is actually a re-implementation of HashMap<subscription, int cohort> and could be swapped out, except Claude has written the class in a way that makes it very unclear. It is not quite the same as `HashMap<subscription, int cohort>`. Conceptually, yes, but performance matters a lot here. With `HashMap<subscription, int cohort>`, we would need to use `HashMap<Set<UUID>, int cohort>` or `HashMap<SortedList<UUID>, int cohort>` and using a Set and a sorted List has a higher cost than having our own hash. That being said, we can use `HashMap<Long, IntList>` and keep the performance boost. Let me know what you think about it. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
