sunchao commented on code in PR #45267: URL: https://github.com/apache/spark/pull/45267#discussion_r1536551512
########## sql/catalyst/src/main/java/org/apache/spark/sql/connector/catalog/functions/ReducibleFunction.java: ########## @@ -0,0 +1,106 @@ +/* + * 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.spark.sql.connector.catalog.functions; + +import org.apache.spark.annotation.Evolving; + +/** + * Base class for user-defined functions that can be 'reduced' on another function. + * + * A function f_source(x) is 'reducible' on another function f_target(x) if + * <ul> + * <li> There exists a reducer function r(x) such that r(f_source(x)) = f_target(x) + * for all input x, or </li> + * <li> More generally, there exists reducer functions r1(x) and r2(x) such that + * r1(f_source(x)) = r2(f_target(x)) for all input x. </li> + * </ul> + * <p> + * Examples: + * <ul> + * <li>Bucket functions where one side has reducer + * <ul> + * <li>f_source(x) = bucket(4, x)</li> + * <li>f_target(x) = bucket(2, x)</li> + * <li>r(x) = x % 2</li> + * </ul> + * + * <li>Bucket functions where both sides have reducer + * <ul> + * <li>f_source(x) = bucket(16, x)</li> + * <li>f_target(x) = bucket(12, x)</li> + * <li>r1(x) = x % 4</li> + * <li>r2(x) = x % 4</li> + * </ul> + * + * <li>Date functions + * <ul> + * <li>f_source(x) = days(x)</li> + * <li>f_target(x) = hours(x)</li> + * <li>r(x) = x / 24</li> + * </ul> + * </ul> + * @param <I> reducer function input type + * @param <O> reducer function output type + * @since 4.0.0 + */ +@Evolving +public interface ReducibleFunction<I, O> { + + /** + * This method is for parameterized functions. + * + * If this parameterized function is 'reducible' on another bucket function, + * return the {@link Reducer} function. + * <p> + * Example to return reducer for reducing f_source = bucket(4, x) on f_target = bucket(2, x) + * <ul> + * <li>thisFunction = bucket</li> + * <li>thisParam = Int(4)</li> + * <li>otherFunction = bucket</li> + * <li>otherParam = Int(2)</li> + * </ul> + * + * @param thisParam parameter for this function Review Comment: I'm thinking whether we can use `InternalRow` for this, which is consistent with the v2 function API (see `ScalarFunction` and `AggregateFunction`, and can support multiple parameters. We need to define a clear contract between Spark and the callers of this method though: **what should a caller expect to see from `thisParam` and `otherParam`**. For `bucket` it is quite clear, but I'm not sure how this will this work with geohash function that you pointed out. In Spark, we need a way for the data source provider to pass the `resolution` of the function to Spark via v2 catalog, and then pass it back to the data source via `reducer`. Any idea? -- 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]
