amogh-jahagirdar commented on code in PR #12450: URL: https://github.com/apache/iceberg/pull/12450#discussion_r2152802240
########## api/src/main/java/org/apache/iceberg/actions/ComputePartitionStats.java: ########## @@ -0,0 +1,43 @@ +/* + * 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 org.apache.iceberg.PartitionStatisticsFile; + +/** + * An action that computes and writes the partition statistics of an Iceberg table. Current snapshot + * is used by default. + */ +public interface ComputePartitionStats + extends Action<ComputePartitionStats, ComputePartitionStats.Result> { Review Comment: I'm still skeptical about keeping these separate; imo there's a few good arguments for combining the procedures: 1. the examples @karuppayya provided for [Hive](https://cwiki.apache.org/confluence/display/hive/statsdev#StatsDev-Examples), [Spark](https://spark.apache.org/docs/latest/sql-ref-syntax-aux-analyze-table.html), [Trino](https://trino.io/docs/current/sql/analyze.html#examples) indicate that at least at the SQL level there is precedence for a common `ANALYZE` for both columnar and partition stats. I think it also makes sense at the Iceberg action API level to follow the same structure since users generally just think in terms of "generate these stats that will help my performance" (put another way "partitions are part of my table") and we can give the optionality to power users to choose which stats they want to collect. 3. I am not super concerned around the `columns` API not applying for partitions, by itself that doesn't seem like a strong justification to have 2 completely separate actions. Fundamentally, I think it'll be clear in the combined procedure that users express columns for the stats they want to collect and it's clear that partition stats are independent What I'm kind of envisioning: ``` // Generate NDV stats for "foo", "bar" and partition stats at snapshot 1234 action.computeTableStats(table).columns("foo", "bar").snapshot(1234L).withPartitionStats().commit() // Generate NDV stats for "foo", "bar" at snapshot 1234 action.computeTableStats(table).columns("foo", "bar").snapshot(1234L).commit() // Generate NDV stats for "foo", "bar" and partition stats at snapshot 1234 action.computeTableStats(table).snapshot(1234L).withPartitionStats().commit() // Generate NDV stats for all columns at snapshot 1234 action.computeTableStats(table).snapshot(1234L).commit() // Generate only partition stats for snapshot 1234 action.computeTableStats(table).withOnlyPartitionStats().commit() ``` The only additional slight awkwardness is in `withOnlyPartitionStats` (which if columns are specified along with this option we throw) but this is very minimal imo. It's also easier down the line to change behaviors of the different options and defaults of the procedure than potentially re-create entirely new actions imo. 4. Whether partition stats supports incremental computation today and table stats doesn't, seems like an implementation detail we can hide and is independent of the API imo. The implementation will choose the incremental computation if it's possible otherwise it won't. That's not a user opt in option. Besides, I think we probably could support incremental NDV computation down the line? -- 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: issues-unsubscr...@iceberg.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@iceberg.apache.org For additional commands, e-mail: issues-h...@iceberg.apache.org
