chenjunjiedada commented on a change in pull request #2841: URL: https://github.com/apache/iceberg/pull/2841#discussion_r672760398
########## File path: core/src/main/java/org/apache/iceberg/actions/RewriteDeleteStrategy.java ########## @@ -0,0 +1,51 @@ +/* + * 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.iceberg.actions; + +import java.util.Map; +import org.apache.iceberg.DeleteFile; +import org.apache.iceberg.Table; + +public interface RewriteDeleteStrategy { + Review comment: I didn't add `validOptions` because we haven't finalized what options are valid and it can be added late. But I'm fine to add them now. For `name`, It looks like the class name already includes the strategy name. Do we still need it? ########## File path: core/src/main/java/org/apache/iceberg/actions/RewriteDeleteStrategy.java ########## @@ -0,0 +1,51 @@ +/* + * 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.iceberg.actions; + +import java.util.Map; +import org.apache.iceberg.DeleteFile; +import org.apache.iceberg.Table; + +public interface RewriteDeleteStrategy { + + /** + * Returns the table being modified by this rewrite strategy + */ + Table table(); + + /** + * Select the deletes to rewrite. + * + * @return iterable of original delete file to be replaced. + */ + Iterable<DeleteFile> selectDeletes(); Review comment: The argument of `selectFilesToRewrite(Iterable<FileScanTask> dataFiles)` comes from `planFileGroups` and the filter logic. I think we could build similar logic by storing the iterable of `FileScanTask` as a local variable in the concrete `RewrtiteDeleteStrategy` class, and grouping them in `rewriteDeletes()` API. As for filter logic, I think we can add it later and apply it in `selectDeletes()` API. For strategies, I think it depends on different scenarios. In our cases, which mostly use HDFS as storage, we care more about the small files. they impact the planning, reading performance and also bring overhead to the name node. So the strategy we want badly is to merge deletes. This could be summarized to 1) merge all position deletes. 2) read all equality deletes under the partition and convert them into one position delete. I think the first one is also bin-packing, the second one is convert-and-bin-packing. ########## File path: core/src/main/java/org/apache/iceberg/actions/RewriteDeleteStrategy.java ########## @@ -0,0 +1,51 @@ +/* + * 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.iceberg.actions; + +import java.util.Map; +import org.apache.iceberg.DeleteFile; +import org.apache.iceberg.Table; + +public interface RewriteDeleteStrategy { + + /** + * Returns the table being modified by this rewrite strategy + */ + Table table(); + + /** + * Select the deletes to rewrite. + * + * @return iterable of original delete file to be replaced. + */ + Iterable<DeleteFile> selectDeletes(); Review comment: The argument of `selectFilesToRewrite(Iterable<FileScanTask> dataFiles)` comes from `planFileGroups` and the filter logic. I think we could build similar logic by storing the iterable of `FileScanTask` as a local variable in the concrete `RewrtiteDeleteStrategy` class, and grouping them in `rewriteDeletes()` API. As for filter logic, I think we can add it later and apply it in `selectDeletes()` API. For strategies, I think it depends on different scenarios. In our cases, which mostly use HDFS as storage, we care more about the small files. they impact the planning, reading performance and also bring overhead to the name node. So the strategy we want badly is to merge deletes. This could be summarized to 1) merge all position deletes. 2) read all equality deletes under the partition and convert them into one position delete. I think the first one is also bin-packing, the second one is convert-and-bin-packing. In the future, we might also want to sort the merged position delete by file name and break it into pieces so that executors can avoid reading useless delete contents. ########## File path: core/src/main/java/org/apache/iceberg/actions/RewriteDeleteStrategy.java ########## @@ -0,0 +1,51 @@ +/* + * 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.iceberg.actions; + +import java.util.Map; +import org.apache.iceberg.DeleteFile; +import org.apache.iceberg.Table; + +public interface RewriteDeleteStrategy { + + /** + * Returns the table being modified by this rewrite strategy + */ + Table table(); + + /** + * Select the deletes to rewrite. + * + * @return iterable of original delete file to be replaced. + */ + Iterable<DeleteFile> selectDeletes(); Review comment: The argument of `selectFilesToRewrite(Iterable<FileScanTask> dataFiles)` comes from `planFileGroups` and the filter logic. I think we could build similar logic by storing the iterable of `FileScanTask` as a local variable in the concrete `RewriteDeleteStrategy` class, and grouping them in `rewriteDeletes()` API. As for filter logic, we could add it later and apply it in `selectDeletes()` API. As for strategies, it depends on different scenarios. In our cases, which mostly use HDFS as storage, we care more about the small files. they could impact the planning, reading performance and also bring overhead to the name node. So the strategy we want badly is to merge deletes. So we are planning two strategies at first: 1) merge all position deletes. 2) read all equality deletes under the partition and convert them into one position delete. I think the first one is kind of bin-packing, the second one is convert-and-bin-packing. In the future, we might also want to sort the merged position delete by file name and break it into pieces so that executors could read only the delete contents that match the data files. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
