aliehsaeedii commented on code in PR #22961: URL: https://github.com/apache/kafka/pull/22961#discussion_r3667910682
########## streams/src/main/java/org/apache/kafka/streams/state/internals/ChangeLoggingListValueBytesStoreWithHeaders.java: ########## @@ -0,0 +1,80 @@ +/* + * 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.streams.state.internals; + +import org.apache.kafka.common.header.Headers; +import org.apache.kafka.common.header.internals.RecordHeaders; +import org.apache.kafka.common.utils.Bytes; +import org.apache.kafka.streams.state.KeyValueStore; + +/** + * The HEADERS-mode changelog store for the outer-join {@link ListValueStore}. + * <p> + * The local store holds {@code [headersSize][headers][flag][value]} per list element, but that format + * must never reach the changelog: the changelog topic is the only durable copy of the state, so its + * value format is a permanent compatibility contract. If we logged the local bytes verbatim, an old + * PLAIN reader — after a version downgrade, or simply after flipping {@code dsl.store.format} back to + * PLAIN — would read each element's leading empty-headers {@code 0x00} as the {@code LeftOrRightValue} + * flag and silently mistake left values for right ones. + * <p> + * So this store does what every other KIP-1271 changelog store does (see + * {@link ChangeLoggingTimestampedKeyValueBytesStoreWithHeaders}, which logs + * {@link Utils#rawPlainValue(byte[])}): it keeps the headers out of the value and puts them in a record + * header instead. The list makes that a little more involved — one changelog record holds the whole + * list, so N sets of headers have to share one header field — which is why the stripped prefixes are + * concatenated into a single self-delimiting blob under + * {@link ListValueStoreUpgradeUtils#LIST_VALUE_HEADERS_HEADER_KEY} rather than unpacked into individual + * {@code RecordHeader}s. + * <p> + * Implements {@link HeadersAwareListValueStore} purely so {@code StateManagerUtil.converterForStore} + * selects {@link RecordConverters#rawListValueToHeadersListValue()}, which performs the inverse join on + * restore. + */ +public class ChangeLoggingListValueBytesStoreWithHeaders + extends ChangeLoggingListValueBytesStore + implements HeadersAwareListValueStore { + + ChangeLoggingListValueBytesStoreWithHeaders(final KeyValueStore<Bytes, byte[]> inner) { + super(inner); + } + + @Override + public void put(final Bytes key, final byte[] value) { + wrapped().put(key, value); + // As in the parent, a tombstone deletes the whole list, so there is nothing to read back and + // no per-element headers to carry. + if (value == null) { + log(key, null, internalContext.recordContext().timestamp(), changelogHeaders(null)); Review Comment: >do we actually need a copy No. What we needed was to never hand logChange an instance anyone else holds. A copy achieved that; starting empty achieves it too, and cheaper. So the copy is gone now. >and all other stores have a bug? `ChangeLoggingTimestampedKeyValueBytesStoreWithHeaders` passes the live `internalContext.recordContext().headers()` only when the value is null. every non-null path passes a freshly-deserialised headers(value). So the exposure is real but confined to tombstones, and only when consistencyEnabled is on. Do you agree with me? Sould I file a ticket? -- 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]
