|
| 1 | +package za.co.absa.standardization |
| 2 | + |
| 3 | + |
| 4 | +import org.scalatest.BeforeAndAfterAll |
| 5 | +import org.scalatest.funsuite.AnyFunSuite |
| 6 | +import org.apache.spark.sql.{Column, DataFrame, Row, SparkSession} |
| 7 | +import org.apache.spark.sql.types.{DataType, DateType, Metadata, StringType, StructField, StructType, TimestampType} |
| 8 | +import org.apache.spark.sql.functions.{col, lit, to_date, to_timestamp, when} |
| 9 | +import java.sql |
| 10 | +import java.sql.{Date, Timestamp} |
| 11 | +import java.text.{SimpleDateFormat,ParseException} |
| 12 | +import java.util.TimeZone |
| 13 | +import scala.util.Try |
| 14 | + |
| 15 | + |
| 16 | +class InfinitySupportIsoTest extends AnyFunSuite with BeforeAndAfterAll { |
| 17 | + var sparkSession: SparkSession = _ |
| 18 | + |
| 19 | + override def beforeAll(): Unit = { |
| 20 | + sparkSession = SparkSession.builder() |
| 21 | + .appName("InfinityISOTest") |
| 22 | + .master("local[*]") |
| 23 | + .config("spark.sql.legacy.parquet.datetimeRebaseModeInRead", "CORRECTED") |
| 24 | + .config("spark.sql.legacy.parquet.datetimeRebaseModeInWrite", "CORRECTED") |
| 25 | + .getOrCreate() |
| 26 | + } |
| 27 | + |
| 28 | + override def afterAll(): Unit ={ |
| 29 | + if (sparkSession != null) { |
| 30 | + sparkSession.stop() |
| 31 | + } |
| 32 | + } |
| 33 | + |
| 34 | + private def createTestDataFrame(data: Seq[String]): DataFrame = { |
| 35 | + sparkSession.createDataFrame( |
| 36 | + sparkSession.sparkContext.parallelize(data.map(Row(_))), |
| 37 | + StructType(Seq(StructField("value", StringType, nullable = false))) |
| 38 | + ) |
| 39 | + } |
| 40 | + |
| 41 | + private val configString = |
| 42 | + """ |
| 43 | + standardization.infinity { |
| 44 | + minus.symbol = "-inf" |
| 45 | + minus.value = "1970-01-01 00:00:00.000000" |
| 46 | + plus.symbol = "inf" |
| 47 | + plus.value ="9999-12-31 23:59:59.999999" |
| 48 | + } |
| 49 | + """ |
| 50 | + private def replaceInfinitySymbols(column: Column, dataType: DataType, pattern: Option[String], timezone:String, minusSymbol: String, minusValue: String, plusSymbol:String, plusValue:String): Column ={ |
| 51 | + def validateValue(value: String, patternOpt: Option[String], dataType: DataType) : Unit = { |
| 52 | + val isoPattern = dataType match { |
| 53 | + case TimestampType => "yyyy-MM-dd'T'HH:mm:ss.SSSSSS" |
| 54 | + case DateType => "yyyy-MM-dd" |
| 55 | + case _ => throw new IllegalArgumentException(s"Unsupported data type: $dataType") |
| 56 | + } |
| 57 | + |
| 58 | + val formatsToTry = patternOpt.toSeq ++ Seq(isoPattern) |
| 59 | + var lastException: Option[ParseException] = None |
| 60 | + |
| 61 | + for (fmt <- formatsToTry) { |
| 62 | + try { |
| 63 | + val sdf = new SimpleDateFormat(fmt) |
| 64 | + sdf.setTimeZone(TimeZone.getTimeZone(timezone)) |
| 65 | + sdf.setLenient(false) |
| 66 | + sdf.parse(value) |
| 67 | + return |
| 68 | + } catch { |
| 69 | + case e: ParseException => lastException = Some(e) |
| 70 | + } |
| 71 | + } |
| 72 | + |
| 73 | + val errorMsg = s"Invalid infinity value: '$value' for type: ${dataType.toString.toLowerCase} with pattern ${patternOpt.getOrElse("none")} and ISO fallback ($isoPattern)" |
| 74 | + throw new IllegalArgumentException(errorMsg,lastException.orNull) |
| 75 | + } |
| 76 | + |
| 77 | + validateValue(minusValue,pattern,dataType) |
| 78 | + validateValue(plusValue, pattern, dataType) |
| 79 | + |
| 80 | + |
| 81 | + dataType match { |
| 82 | + case TimestampType => |
| 83 | + when(col(column.toString) === minusSymbol, lit(minusValue)) |
| 84 | + .when(col(column.toString) === plusSymbol, lit(plusValue)) |
| 85 | + .otherwise( |
| 86 | + pattern.map(p => to_timestamp(col(column.toString),p)) |
| 87 | + .getOrElse(to_timestamp(col(column.toString))) |
| 88 | + ) |
| 89 | + case DateType => |
| 90 | + when(col(column.toString) === minusSymbol, lit(minusValue)) |
| 91 | + .when(col(column.toString) === plusSymbol, lit(plusValue)) |
| 92 | + .otherwise( |
| 93 | + pattern.map( p => to_date(col(column.toString), p)) |
| 94 | + .getOrElse(to_date(col(column.toString))) |
| 95 | + ) |
| 96 | + case _ => throw new IllegalArgumentException(s"Unsupported data type: $dataType") |
| 97 | + } |
| 98 | + } |
| 99 | + |
| 100 | + |
| 101 | + test("Replace infinity symbols for timestamp with valid pattern"){ |
| 102 | + val df = createTestDataFrame(Seq("-inf","inf", "2025-07-05 12:34:56", null)) |
| 103 | + val result = df.withColumn("result", replaceInfinitySymbols(col("value"), TimestampType,Some("yyyy-MM-dd HH:mm:ss"), "UTC","-inf", "1970-01-01 00:00:00","inf","9999-12-31 23:59:59")) |
| 104 | + .select("result") |
| 105 | + .collect() |
| 106 | + .map(_.getAs[TimestampType](0)) |
| 107 | + |
| 108 | + val expected = Seq( |
| 109 | + Timestamp.valueOf("1970-01-01 00:00:00"), |
| 110 | + Timestamp.valueOf("9999-12-31 23:59:59"), |
| 111 | + Timestamp.valueOf("2025-07-05 12:34:56"), |
| 112 | + null |
| 113 | + ) |
| 114 | + |
| 115 | + assert(result sameElements expected) |
| 116 | + } |
| 117 | + |
| 118 | + test("Convert invalid timestamp pattern to ISO"){ |
| 119 | + val df = createTestDataFrame(Seq("-inf","inf")) |
| 120 | + val result = df.withColumn("result", replaceInfinitySymbols( |
| 121 | + col("value"), |
| 122 | + TimestampType, |
| 123 | + Some("yyyy-MM-dd HH:mm:ss"), |
| 124 | + "UTC", |
| 125 | + "-inf", |
| 126 | + "1970-01-01 00:00:00", |
| 127 | + "inf", |
| 128 | + "9999-12-31 23:59:59")) |
| 129 | + .select("result") |
| 130 | + .collect() |
| 131 | + .map(_.getAs[TimestampType] (0)) |
| 132 | + |
| 133 | + |
| 134 | + val expected = Seq( |
| 135 | + Timestamp.valueOf("1970-01-01 00:00:00"), |
| 136 | + Timestamp.valueOf("9999-12-31 23:59:59") |
| 137 | + ) |
| 138 | + |
| 139 | + assert (result sameElements expected) |
| 140 | + } |
| 141 | + |
| 142 | + |
| 143 | + test("Replace infinity symbol for date with valid pattern"){ |
| 144 | + val df = createTestDataFrame(Seq("-inf", "inf", "20245-07-05",null)) |
| 145 | + val result = df.withColumn("result", replaceInfinitySymbols( |
| 146 | + col("value"), |
| 147 | + DateType, |
| 148 | + Some("yyyy-MM-dd"), |
| 149 | + "UTC", |
| 150 | + "-inf", |
| 151 | + "1970-01-01", |
| 152 | + "inf", |
| 153 | + "9999-12-31" |
| 154 | + )) |
| 155 | + .select("result") |
| 156 | + .collect() |
| 157 | + .map(_.getAs[Date](0)) |
| 158 | + |
| 159 | + val expected = Seq( |
| 160 | + Date.valueOf("1970-01-01"), |
| 161 | + Date.valueOf("9999-12-31"), |
| 162 | + Date.valueOf("2025-07-05"), |
| 163 | + null |
| 164 | + ) |
| 165 | + |
| 166 | + assert (result sameElements expected) |
| 167 | + } |
| 168 | + |
| 169 | + |
| 170 | + test("Throw error for unparseable infinity value"){ |
| 171 | + val exception = intercept[IllegalArgumentException] { |
| 172 | + replaceInfinitySymbols( |
| 173 | + col("value"), |
| 174 | + TimestampType, |
| 175 | + Some("yyyy-MM-dd HH:mm:ss"), |
| 176 | + "UTC", |
| 177 | + "-inf", |
| 178 | + "invalid_date", |
| 179 | + "inf", |
| 180 | + "9999-12-31 23:59:59" |
| 181 | + ) |
| 182 | + } |
| 183 | + |
| 184 | + assert(exception.getMessage.contains("Invalid infinity value: 'invalid_date' for type: timestamp")) |
| 185 | + assert(exception.getMessage.contains("pattern yyyy-MM-dd:mm:ss")) |
| 186 | + assert(exception.getMessage.contains("ISO fallback (yyyy-MM-dd'T'HH:mm:ss.SSSSSS")) |
| 187 | + } |
| 188 | + |
| 189 | + test("Handle missing pattern with ISO fallback"){ |
| 190 | + val df = createTestDataFrame(Seq("-inf","inf")) |
| 191 | + val result = df.withColumn("result", replaceInfinitySymbols( |
| 192 | + col("value"), |
| 193 | + DateType, |
| 194 | + None, |
| 195 | + "UTC", |
| 196 | + "-inf", |
| 197 | + "1970-01-01", |
| 198 | + "inf", |
| 199 | + "9999-12-31" |
| 200 | + )) |
| 201 | + .select("result") |
| 202 | + .collect() |
| 203 | + .map(_.getAs[Date](0)) |
| 204 | + |
| 205 | + |
| 206 | + val expected = Seq( |
| 207 | + Date.valueOf("1970-01-01"), |
| 208 | + Date.valueOf("9999-12-31") |
| 209 | + ) |
| 210 | + |
| 211 | + assert (result sameElements expected) |
| 212 | + } |
| 213 | +} |
0 commit comments