grundprinzip commented on code in PR #37075: URL: https://github.com/apache/spark/pull/37075#discussion_r915749533
########## sql/core/src/main/protobuf/spark/connect/relations.proto: ########## @@ -0,0 +1,191 @@ +/* + * 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. + */ + +syntax = 'proto3'; + +package spark.connect; + +import "spark/connect/expressions.proto"; + +option java_multiple_files = true; +option java_package = "org.apache.spark.connect.protobuf"; + +/* + The main [[Relation]] type. Fundamentally, a relation is a typed container + that has exactly one explicit relation type set. + + When adding new relation types, they have to be registered here. + */ +message Relation { Review Comment: add ID for incremental plan building. ########## sql/core/src/main/protobuf/spark/connect/base.proto: ########## @@ -0,0 +1,110 @@ +/* + * 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. + */ + +syntax = 'proto3'; + +package spark.connect; + +import "spark/connect/commands.proto"; +import "spark/connect/relations.proto"; + +option java_multiple_files = true; +option java_package = "org.apache.spark.connect.protobuf"; + +// A [[Plan]] is the structure that carries the runtime information for the execution from the +// client to the server. A [[Plan]] can either be of the type [[Relation]] which is a reference +// to the underlying logical plan or it can be of the [[Command]] type that is used to execute +// commands on the server. +message Plan { + oneof op_type { + Relation root = 1; + Command command = 2; + } +} + +// A request to be executed by the service. +message Request { Review Comment: Add option to pass spark conf values as part of the request? ########## sql/core/src/main/protobuf/spark/connect/base.proto: ########## @@ -0,0 +1,110 @@ +/* + * 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. + */ + +syntax = 'proto3'; + +package spark.connect; + +import "spark/connect/commands.proto"; +import "spark/connect/relations.proto"; + +option java_multiple_files = true; +option java_package = "org.apache.spark.connect.protobuf"; + +// A [[Plan]] is the structure that carries the runtime information for the execution from the +// client to the server. A [[Plan]] can either be of the type [[Relation]] which is a reference +// to the underlying logical plan or it can be of the [[Command]] type that is used to execute +// commands on the server. +message Plan { + oneof op_type { + Relation root = 1; + Command command = 2; + } +} + +// A request to be executed by the service. +message Request { + // The client_id is set by the client to be able to collate streaming responses from + // different queries. + string client_id = 1; + // User context + UserContext user_context = 2; + // The logical plan to be executed / analyzed. + Plan plan = 3; + + // User Context is used to refer to one particular user session that is executing + // queries in the backend. + message UserContext { + string user_id = 1; + string user_name = 2; + } +} + +// The response of a query, can be one or more for each request. Responses belonging to the +// same input query, carry the same `client_id`. +message Response { + string client_id = 1; + ArrowBatch batch = 2; + + // Metrics for the query execution. Typically, this field is only present in the last + // batch of results and then represent the overall state of the query execution. + Metrics metrics = 3; + + // Batch results of metrics. + message ArrowBatch { + int64 row_count = 1; + int64 uncompressed_bytes = 2; + int64 compressed_bytes = 3; + bytes data = 4; + bytes schema = 5; + } + + message Metrics { + + repeated MetricObject metrics = 1; + + message MetricObject { + string name = 1; + int64 plan_id = 2; + int64 parent = 3; + map<string, MetricValue> execution_metrics = 4; + } + + message MetricValue { + string name = 1; + int64 value = 2; + string metric_type = 3; + } + } +} + +message AnalyzeResponse { + string client_id = 1; + repeated string column_names = 2; + repeated string column_types = 3; +} + +// Main interface for the SparkConnect service. +service SparkConnectService { + + // Executes a request that contains the query and returns a stream of [[Response]]. + rpc ExecutePlan(Request) returns (stream Response) {} Review Comment: I suggest renaming this to something that avoids having Plan in the request as it can be confusing. ########## sql/core/src/main/protobuf/spark/connect/base.proto: ########## @@ -0,0 +1,110 @@ +/* + * 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. + */ + +syntax = 'proto3'; + +package spark.connect; + +import "spark/connect/commands.proto"; +import "spark/connect/relations.proto"; + +option java_multiple_files = true; +option java_package = "org.apache.spark.connect.protobuf"; + +// A [[Plan]] is the structure that carries the runtime information for the execution from the +// client to the server. A [[Plan]] can either be of the type [[Relation]] which is a reference +// to the underlying logical plan or it can be of the [[Command]] type that is used to execute +// commands on the server. +message Plan { + oneof op_type { + Relation root = 1; + Command command = 2; + } +} + +// A request to be executed by the service. +message Request { + // The client_id is set by the client to be able to collate streaming responses from + // different queries. + string client_id = 1; + // User context + UserContext user_context = 2; + // The logical plan to be executed / analyzed. + Plan plan = 3; + + // User Context is used to refer to one particular user session that is executing + // queries in the backend. + message UserContext { + string user_id = 1; + string user_name = 2; + } +} + +// The response of a query, can be one or more for each request. Responses belonging to the +// same input query, carry the same `client_id`. +message Response { + string client_id = 1; + ArrowBatch batch = 2; + + // Metrics for the query execution. Typically, this field is only present in the last + // batch of results and then represent the overall state of the query execution. + Metrics metrics = 3; + + // Batch results of metrics. + message ArrowBatch { Review Comment: Yes, the desired format is actually Arrow IPC Streams which include the schema directly. I will address this in this PR. ########## sql/core/src/main/protobuf/spark/connect/relations.proto: ########## @@ -0,0 +1,191 @@ +/* + * 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. + */ + +syntax = 'proto3'; + +package spark.connect; + +import "spark/connect/expressions.proto"; + +option java_multiple_files = true; +option java_package = "org.apache.spark.connect.protobuf"; + +/* + The main [[Relation]] type. Fundamentally, a relation is a typed container + that has exactly one explicit relation type set. + + When adding new relation types, they have to be registered here. + */ +message Relation { + RelationCommon common = 1; + oneof rel_type { + Read read = 2; + Project project = 3; + Filter filter = 4; + Join join = 5; + Union union = 6; + Sort sort = 7; + Fetch fetch = 8; + Aggregate aggregate = 9; + Sql sql = 10; + } +} +/* + Common metadata of all relations. + */ +message RelationCommon { + string source_info = 1; + string alias = 2; +} + +/* + Relation that uses a SQL query to generate the output. + */ +message Sql { + string query = 1; +} + +/* + Relation that reads from a file / table or other data source. Does not have additional + inputs. + */ +message Read { + oneof read_type { + NamedTable named_table = 1; + } + + message NamedTable { + repeated string parts = 1; + } +} + +/* + Projection of a bag of expressions for a given input relation. + + The input relation must be specified. + The projected expression can be an arbitrary expression. + */ +message Project { + Relation input = 1; + repeated Expression expressions = 3; +} + +/* + Relation that applies a boolean expression `condition` on each row of `input` to produce the output result. + */ +message Filter { + Relation input = 1; + Expression condition = 2; +} + +/* + Relation of type [[Join]]. + + `left` and `right` must be present. + */ +message Join { + Relation left = 1; + Relation right = 2; + Expression on = 3; + JoinType how = 4; + + enum JoinType { + JOIN_TYPE_UNSPECIFIED = 0; + JOIN_TYPE_INNER = 1; + JOIN_TYPE_OUTER = 2; + JOIN_TYPE_LEFT_OUTER = 3; + JOIN_TYPE_RIGHT_OUTER = 4; + JOIN_TYPE_ANTI = 5; + } +} + +/* + Relation of type [[Union]], at least one input must be set. + */ +message Union { + repeated Relation inputs = 1; + UnionType union_type = 2; + + enum UnionType { + UNION_TYPE_UNSPECIFIED = 0; + UNION_TYPE_DISTINCT = 1; + UNION_TYPE_ALL = 2; + } +} + +/* + Relation of type [[Fetch]] that is used to read `limit` / `offset` rows from the input relation. + */ +message Fetch { + Relation input = 1; + int32 limit = 2; + int32 offset = 3; +} + +/* + Relation of type [[Aggregate]]. + */ +message Aggregate { + Relation input = 1; + + // Grouping sets are used in rollups + repeated GroupingSet grouping_sets = 2; + + // Measures + repeated Measure measures = 3; + + message GroupingSet { + repeated Expression aggregate_expressions = 1; + } + + message Measure { + AggregateFunction function = 1; + // Conditional filter for SUM(x FILTER WHERE x < 10) + Expression filter = 2; + } + + message AggregateFunction { + string name = 1; + repeated Expression arguments = 2; + } +} + +/* + Relation of type [[Sort]]. + */ +message Sort { Review Comment: For now, I don't think we need to differentiate between local and global sort here as this will always treat the sort on the full relation. Pushing a local sort, FWIR, is a physical optimization. ########## sql/core/src/main/protobuf/spark/connect/relations.proto: ########## @@ -0,0 +1,191 @@ +/* + * 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. + */ + +syntax = 'proto3'; + +package spark.connect; + +import "spark/connect/expressions.proto"; + +option java_multiple_files = true; +option java_package = "org.apache.spark.connect.protobuf"; + +/* + The main [[Relation]] type. Fundamentally, a relation is a typed container + that has exactly one explicit relation type set. + + When adding new relation types, they have to be registered here. + */ +message Relation { + RelationCommon common = 1; + oneof rel_type { + Read read = 2; + Project project = 3; + Filter filter = 4; + Join join = 5; + Union union = 6; + Sort sort = 7; + Fetch fetch = 8; + Aggregate aggregate = 9; + Sql sql = 10; + } +} +/* + Common metadata of all relations. + */ +message RelationCommon { + string source_info = 1; + string alias = 2; +} + +/* + Relation that uses a SQL query to generate the output. + */ +message Sql { + string query = 1; +} + +/* + Relation that reads from a file / table or other data source. Does not have additional + inputs. + */ +message Read { + oneof read_type { + NamedTable named_table = 1; + } + + message NamedTable { + repeated string parts = 1; + } +} + +/* + Projection of a bag of expressions for a given input relation. + + The input relation must be specified. + The projected expression can be an arbitrary expression. + */ +message Project { + Relation input = 1; + repeated Expression expressions = 3; +} + +/* + Relation that applies a boolean expression `condition` on each row of `input` to produce the output result. + */ +message Filter { + Relation input = 1; + Expression condition = 2; +} + +/* + Relation of type [[Join]]. + + `left` and `right` must be present. + */ +message Join { + Relation left = 1; + Relation right = 2; + Expression on = 3; + JoinType how = 4; + + enum JoinType { + JOIN_TYPE_UNSPECIFIED = 0; + JOIN_TYPE_INNER = 1; + JOIN_TYPE_OUTER = 2; + JOIN_TYPE_LEFT_OUTER = 3; + JOIN_TYPE_RIGHT_OUTER = 4; + JOIN_TYPE_ANTI = 5; + } +} + +/* + Relation of type [[Union]], at least one input must be set. + */ +message Union { + repeated Relation inputs = 1; + UnionType union_type = 2; + + enum UnionType { + UNION_TYPE_UNSPECIFIED = 0; + UNION_TYPE_DISTINCT = 1; + UNION_TYPE_ALL = 2; + } +} + +/* + Relation of type [[Fetch]] that is used to read `limit` / `offset` rows from the input relation. + */ +message Fetch { + Relation input = 1; + int32 limit = 2; + int32 offset = 3; +} + +/* + Relation of type [[Aggregate]]. + */ +message Aggregate { + Relation input = 1; + + // Grouping sets are used in rollups + repeated GroupingSet grouping_sets = 2; + + // Measures + repeated Measure measures = 3; + + message GroupingSet { + repeated Expression aggregate_expressions = 1; + } + + message Measure { + AggregateFunction function = 1; + // Conditional filter for SUM(x FILTER WHERE x < 10) + Expression filter = 2; + } + + message AggregateFunction { + string name = 1; + repeated Expression arguments = 2; Review Comment: From my experiments UDAF registered by name should just work, as the `AggregateFunction` is "unresolved". I might make it clearer by calling it `UnresolvedAggregateFunction` -- 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]
