szehon-ho commented on code in PR #7105: URL: https://github.com/apache/iceberg/pull/7105#discussion_r1318005645
########## format/partition-statistics-spec.md: ########## @@ -0,0 +1,43 @@ +--- +title: "Partition Statistics Spec" +url: partition-statistics-spec +toc: true +disableSidebar: true +--- +<!-- + - 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. + --> + +# Partition Statistics file format + +This is a specification for partition statistics files. It is designed to store statistics information +for every partition value as a row in the **table default format** sorted based on the `partition`. + +The schema of the file is as follows: + +| Field Id | Field Name | Field Type | isOptional | Doc | +|----------|------------------------------|--------------|------------|-----------------------------------------------------------------------------------------------------------------------------| +| 1 | partition | StructType | false | See [PartitionData](https://github.com/apache/iceberg/blob/master/core/src/main/java/org/apache/iceberg/PartitionData.java) | +| 2 | spec_id | IntegerType | false | Partition spec id | +| 3 | data_record_count | LongType | false | Count of records in data files | +| 4 | data_file_count | IntegerType | false | Count of data files | +| 5 | position_delete_record_count | LongType | true | Count of records in position delete files | +| 6 | position_delete_file_count | IntegerType | true | Count of position delete files | +| 7 | equality_delete_record_count | LongType | true | Count of records in equality delete files | +| 8 | equality_delete_file_count | IntegerType | true | Count of equality delete files | Review Comment: > Partitions metadata table is also an async call w.r.t write. But we still don't apply the actual delete files. @szehon-ho: Do you have any suggestions? I still not very convinced that Partitions metadata table should have that column, as it would really reduce the performance of that table. It's a metadata table, so its sync in the sense that it needs to be calculated when the user queries it. I slightly prefer having more stats than less. ie total_record_count, data_file_record_count , eq_delete_record_count, pos_delete_record_count. Internally we can definitely make use of delete record count in planning, for example, like potentially whether to cache deletes, etc. Users can also use them as @flyrain mentioned. It is true we already can access all file record counts via manifest entries, so maybe an argument can be made against having them here. But I think as they are cheap to get, don't see a strong reason why to hide this from the user. The total_record_count seems the most valuable but expensive to compute. Also agree with @flyrain that total count may not be available in writer, for example row-level operation writers , not sure how the writer will know how many total data records are in partition after the write..? -- 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]
