echauchot commented on code in PR #3: URL: https://github.com/apache/flink-connector-cassandra/pull/3#discussion_r1088816929
########## flink-connector-cassandra/src/main/java/org/apache/flink/connector/cassandra/source/reader/CassandraSplitReader.java: ########## @@ -0,0 +1,280 @@ +/* + * 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.flink.connector.cassandra.source.reader; + +import org.apache.flink.annotation.VisibleForTesting; +import org.apache.flink.connector.base.source.reader.RecordsBySplits; +import org.apache.flink.connector.base.source.reader.RecordsWithSplitIds; +import org.apache.flink.connector.base.source.reader.splitreader.SplitReader; +import org.apache.flink.connector.base.source.reader.splitreader.SplitsChange; +import org.apache.flink.connector.cassandra.source.split.CassandraSplit; +import org.apache.flink.connector.cassandra.source.split.CassandraSplitState; +import org.apache.flink.connector.cassandra.source.split.RingRange; +import org.apache.flink.streaming.connectors.cassandra.ClusterBuilder; + +import com.datastax.driver.core.Cluster; +import com.datastax.driver.core.ColumnMetadata; +import com.datastax.driver.core.Metadata; +import com.datastax.driver.core.PreparedStatement; +import com.datastax.driver.core.ResultSet; +import com.datastax.driver.core.Session; +import com.datastax.driver.core.Token; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import javax.annotation.Nullable; + +import java.math.BigInteger; +import java.util.ArrayList; +import java.util.Collection; +import java.util.HashMap; +import java.util.HashSet; +import java.util.Map; +import java.util.Set; +import java.util.concurrent.ConcurrentHashMap; +import java.util.concurrent.atomic.AtomicBoolean; +import java.util.regex.Matcher; +import java.util.regex.Pattern; +import java.util.stream.Collectors; + +/** + * {@link SplitReader} for Cassandra source. This class is responsible for fetching the records as + * {@link CassandraRow}. For that, it executes a range query (query that outputs records belonging + * to a {@link RingRange}) based on the user specified query. This class manages the Cassandra + * cluster and session. + */ +public class CassandraSplitReader implements SplitReader<CassandraRow, CassandraSplit> { + + private static final Logger LOG = LoggerFactory.getLogger(CassandraSplitReader.class); + public static final String SELECT_REGEXP = "(?i)select .+ from (\\w+)\\.(\\w+).*;$"; + + private final Cluster cluster; + private final Session session; + private final Set<CassandraSplitState> unprocessedSplits; + private final AtomicBoolean wakeup = new AtomicBoolean(false); + private final String query; + + public CassandraSplitReader(ClusterBuilder clusterBuilder, String query) { + // need a thread safe set + this.unprocessedSplits = ConcurrentHashMap.newKeySet(); + this.query = query; + cluster = clusterBuilder.getCluster(); + session = cluster.connect(); + } + + @Override + public RecordsWithSplitIds<CassandraRow> fetch() { + Map<String, Collection<CassandraRow>> recordsBySplit = new HashMap<>(); + Set<String> finishedSplits = new HashSet<>(); + Metadata clusterMetadata = cluster.getMetadata(); + + String partitionKey = getPartitionKey(clusterMetadata); + String finalQuery = generateRangeQuery(query, partitionKey); + PreparedStatement preparedStatement = session.prepare(finalQuery); + // Set wakeup to false to start consuming. + wakeup.compareAndSet(true, false); + for (CassandraSplitState cassandraSplitState : unprocessedSplits) { Review Comment: > from all splits into memory Well, in 99% of the cases, `unprocessedSplits` will contain only a single split as the enumerator now assigns only one split per reader. The remaining case I see is when the parallelism of the job is decreased (either by user or failover). In that case, as you know, upon restoration, a reader could be assigned the splits of another reader (as there are less readers after parallelism decrease). But still you're right, in case of parallelism=1 then, all the records would be in 1 split in 1 reader, there will be only a single query issued and a single resultset containing all the records. These records will indeed be stored in `recordsBySplit` to serve for constructing returned `RecordsBySplits`. That being said, as there will be a single resultset, we need to figure out a way of splitting it for that particular case. Luckily Cassandra `ResultSet` is paginated with results fetched only when all the records of the previous page were consumed. So, I could store in the `CassandraSplitState` a reference to the `ResultSet` to resume the output of remaining records on a later `fetch()`. My question is: what would be the condition to exit fetch() earlier? What about allowing the user to configure a `maxRecordPerSplit` and also provide a default value? It is true that fetch early exit should be mentioned in the docs. I'll open a ticket for that and do the PR later. -- 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]
