|
| 1 | +// |
| 2 | +// Licensed to the Apache Software Foundation (ASF) under one |
| 3 | +// or more contributor license agreements. See the NOTICE file |
| 4 | +// distributed with this work for additional information |
| 5 | +// regarding copyright ownership. The ASF licenses this file |
| 6 | +// to you under the Apache License, Version 2.0 (the |
| 7 | +// "License"); you may not use this file except in compliance |
| 8 | +// with the License. You may obtain a copy of the License at |
| 9 | +// |
| 10 | +// http://www.apache.org/licenses/LICENSE-2.0 |
| 11 | +// |
| 12 | +// Unless required by applicable law or agreed to in writing, |
| 13 | +// software distributed under the License is distributed on an |
| 14 | +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY |
| 15 | +// KIND, either express or implied. See the License for the |
| 16 | +// specific language governing permissions and limitations |
| 17 | +// under the License. |
| 18 | +// |
| 19 | +import Atomics |
| 20 | +import Foundation |
| 21 | +import GRPCCore |
| 22 | +import GRPCNIOTransportHTTP2 |
| 23 | +import GRPCProtobuf |
| 24 | +import NIOCore |
| 25 | +import SwiftyTextTable |
| 26 | +import Synchronization |
| 27 | + |
| 28 | +/// A DataFrame which supports only SQL queries |
| 29 | +public actor DataFrame: Sendable { |
| 30 | + var spark: SparkSession |
| 31 | + var plan: Plan |
| 32 | + var schema: DataType? = nil |
| 33 | + private var batches: [RecordBatch] = [RecordBatch]() |
| 34 | + |
| 35 | + /// Create a new `DataFrame`instance with the given Spark session and plan. |
| 36 | + /// - Parameters: |
| 37 | + /// - spark: A ``SparkSession`` instance to use. |
| 38 | + /// - plan: A plan to execute. |
| 39 | + init(spark: SparkSession, plan: Plan) async throws { |
| 40 | + self.spark = spark |
| 41 | + self.plan = plan |
| 42 | + } |
| 43 | + |
| 44 | + /// Create a new `DataFrame` instance with the given SparkSession and a SQL statement. |
| 45 | + /// - Parameters: |
| 46 | + /// - spark: A `SparkSession` instance to use. |
| 47 | + /// - sqlText: A SQL statement. |
| 48 | + init(spark: SparkSession, sqlText: String) async throws { |
| 49 | + self.spark = spark |
| 50 | + self.plan = sqlText.toSparkConnectPlan |
| 51 | + } |
| 52 | + |
| 53 | + /// Set the schema. This is used to store the analized schema response from `Spark Connect` server. |
| 54 | + /// - Parameter schema: <#schema description#> |
| 55 | + private func setSchema(_ schema: DataType) { |
| 56 | + self.schema = schema |
| 57 | + } |
| 58 | + |
| 59 | + /// Add `Apache Arrow`'s `RecordBatch`s to the intenal array. |
| 60 | + /// - Parameter batches: A ``RecordBatch`` instance. |
| 61 | + private func addBathes(_ batches: [RecordBatch]) { |
| 62 | + self.batches.append(contentsOf: batches) |
| 63 | + } |
| 64 | + |
| 65 | + /// A method to access the underlying Spark's `RDD`. |
| 66 | + /// In `Spark Connect`, this feature is not allowed by design. |
| 67 | + public func rdd() throws { |
| 68 | + // SQLSTATE: 0A000 |
| 69 | + // [UNSUPPORTED_CONNECT_FEATURE.RDD] |
| 70 | + // Feature is not supported in Spark Connect: Resilient Distributed Datasets (RDDs). |
| 71 | + throw SparkConnectError.UnsupportedOperationException |
| 72 | + } |
| 73 | + |
| 74 | + /// Return a `JSON` string of data type because we cannot expose the internal type ``DataType``. |
| 75 | + /// - Returns: a `JSON` string. |
| 76 | + public func schema() async throws -> String { |
| 77 | + var dataType: String? = nil |
| 78 | + |
| 79 | + try await withGRPCClient( |
| 80 | + transport: .http2NIOPosix( |
| 81 | + target: .dns(host: spark.client.host, port: spark.client.port), |
| 82 | + transportSecurity: .plaintext |
| 83 | + ) |
| 84 | + ) { client in |
| 85 | + let service = Spark_Connect_SparkConnectService.Client(wrapping: client) |
| 86 | + let response = try await service.analyzePlan( |
| 87 | + spark.client.getAnalyzePlanRequest(spark.sessionID, plan)) |
| 88 | + dataType = try response.schema.schema.jsonString() |
| 89 | + } |
| 90 | + return dataType! |
| 91 | + } |
| 92 | + |
| 93 | + /// Return the total number of rows. |
| 94 | + /// - Returns: a `Int64` value. |
| 95 | + public func count() async throws -> Int64 { |
| 96 | + let counter = Atomic(Int64(0)) |
| 97 | + |
| 98 | + try await withGRPCClient( |
| 99 | + transport: .http2NIOPosix( |
| 100 | + target: .dns(host: spark.client.host, port: spark.client.port), |
| 101 | + transportSecurity: .plaintext |
| 102 | + ) |
| 103 | + ) { client in |
| 104 | + let service = Spark_Connect_SparkConnectService.Client(wrapping: client) |
| 105 | + try await service.executePlan(spark.client.getExecutePlanRequest(spark.sessionID, plan)) { |
| 106 | + response in |
| 107 | + for try await m in response.messages { |
| 108 | + counter.add(m.arrowBatch.rowCount, ordering: .relaxed) |
| 109 | + } |
| 110 | + } |
| 111 | + } |
| 112 | + return counter.load(ordering: .relaxed) |
| 113 | + } |
| 114 | + |
| 115 | + /// Execute the plan and try to fill `schema` and `batches`. |
| 116 | + private func execute() async throws { |
| 117 | + try await withGRPCClient( |
| 118 | + transport: .http2NIOPosix( |
| 119 | + target: .dns(host: spark.client.host, port: spark.client.port), |
| 120 | + transportSecurity: .plaintext |
| 121 | + ) |
| 122 | + ) { client in |
| 123 | + let service = Spark_Connect_SparkConnectService.Client(wrapping: client) |
| 124 | + try await service.executePlan(spark.client.getExecutePlanRequest(spark.sessionID, plan)) { |
| 125 | + response in |
| 126 | + for try await m in response.messages { |
| 127 | + if m.hasSchema { |
| 128 | + // The original schema should arrive before ArrowBatches |
| 129 | + await self.setSchema(m.schema) |
| 130 | + } |
| 131 | + let ipcStreamBytes = m.arrowBatch.data |
| 132 | + if !ipcStreamBytes.isEmpty && m.arrowBatch.rowCount > 0 { |
| 133 | + let IPC_CONTINUATION_TOKEN = Int32(-1) |
| 134 | + // Schema |
| 135 | + assert(ipcStreamBytes[0..<4].int32 == IPC_CONTINUATION_TOKEN) |
| 136 | + let schemaSize = Int64(ipcStreamBytes[4..<8].int32) |
| 137 | + let schema = Data(ipcStreamBytes[8..<(8 + schemaSize)]) |
| 138 | + |
| 139 | + // Arrow IPC Data |
| 140 | + assert( |
| 141 | + ipcStreamBytes[(8 + schemaSize)..<(8 + schemaSize + 4)].int32 |
| 142 | + == IPC_CONTINUATION_TOKEN) |
| 143 | + var pos: Int64 = 8 + schemaSize + 4 |
| 144 | + let dataHeaderSize = Int64(ipcStreamBytes[pos..<(pos + 4)].int32) |
| 145 | + pos += 4 |
| 146 | + let dataHeader = Data(ipcStreamBytes[pos..<(pos + dataHeaderSize)]) |
| 147 | + pos += dataHeaderSize |
| 148 | + let dataBodySize = Int64(ipcStreamBytes.count) - pos - 8 |
| 149 | + let dataBody = Data(ipcStreamBytes[pos..<(pos + dataBodySize)]) |
| 150 | + |
| 151 | + // Read ArrowBatches |
| 152 | + let reader = ArrowReader() |
| 153 | + let arrowResult = ArrowReader.makeArrowReaderResult() |
| 154 | + _ = reader.fromMessage(schema, dataBody: Data(), result: arrowResult) |
| 155 | + _ = reader.fromMessage(dataHeader, dataBody: dataBody, result: arrowResult) |
| 156 | + await self.addBathes(arrowResult.batches) |
| 157 | + } |
| 158 | + } |
| 159 | + } |
| 160 | + } |
| 161 | + } |
| 162 | + |
| 163 | + /// This is designed not to support this feature in order to simplify the Swift client. |
| 164 | + public func collect() async throws { |
| 165 | + throw SparkConnectError.UnsupportedOperationException |
| 166 | + } |
| 167 | + |
| 168 | + /// Execute the plan and show the result. |
| 169 | + public func show() async throws { |
| 170 | + try await execute() |
| 171 | + |
| 172 | + if let schema = self.schema { |
| 173 | + var columns: [TextTableColumn] = [] |
| 174 | + for f in schema.struct.fields { |
| 175 | + columns.append(TextTableColumn(header: f.name)) |
| 176 | + } |
| 177 | + var table = TextTable(columns: columns) |
| 178 | + for batch in self.batches { |
| 179 | + for i in 0..<batch.length { |
| 180 | + var values: [String] = [] |
| 181 | + for column in batch.columns { |
| 182 | + let str = column.array as! AsString |
| 183 | + if column.data.isNull(i) { |
| 184 | + values.append("NULL") |
| 185 | + } else { |
| 186 | + values.append(str.asString(i)) |
| 187 | + } |
| 188 | + } |
| 189 | + table.addRow(values: values) |
| 190 | + } |
| 191 | + } |
| 192 | + print(table.render()) |
| 193 | + } |
| 194 | + } |
| 195 | +} |
0 commit comments