[
https://issues.apache.org/jira/browse/SPARK-53982?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18043602#comment-18043602
]
André Souprayane commented on SPARK-53982:
------------------------------------------
There is no issue with spark 4.0.1 if that could help...
{code:java}
case class Pod(timestamp: String, id: String, value:Double)
val ds = Seq(Pod("2021-01-01T00:00:00.000+0000", "pod123",
99.95),Pod("2021-01-01T00:00:00.000+0000", "pod123",
99.95),Pod("2021-01-01T00:00:00.000+0000", "pod123",
99.95),Pod("2021-01-01T00:00:00.000+0000", "pod123",
99.95),Pod("2021-01-01T00:00:00.000+0000", "pod123",
99.95),Pod("2021-01-01T00:00:00.000+0000", "pod123",
99.95),Pod("2021-01-01T00:00:00.000+0000", "pod123",
99.95),Pod("2021-01-01T00:00:00.000+0000", "pod123",
99.95)).toDS()ds.printSchema()val res = ds.groupBy("id")
.agg(
avg("value").as("METADATA_COL_METRICVALUE"),
sum("value").as("METADATA_COL_SUM_VALUE")
);res.show()
``````
scala> res.show()
+------+------------------------+----------------------+
| id|METADATA_COL_METRICVALUE|METADATA_COL_SUM_VALUE|
+------+------------------------+----------------------+
|pod123| 99.95| 799.6|
+------+------------------------+----------------------+
``` {code}
> Spark aggregation is incorrect (floating point error)
> -----------------------------------------------------
>
> Key: SPARK-53982
> URL: https://issues.apache.org/jira/browse/SPARK-53982
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 3.5.6
> Reporter: Ian Manning
> Priority: Major
>
>
> {code:java}
> List<Row> data = Arrays.asList(
> RowFactory.create("2021-01-01T00:00:00.000+0000", "pod123", 99.95),
> RowFactory.create("2021-01-01T00:00:00.000+0000", "pod123", 99.95),
> RowFactory.create("2021-01-01T00:00:00.000+0000", "pod123", 99.95),
> RowFactory.create("2021-01-01T00:00:00.000+0000", "pod123", 99.95),
> RowFactory.create("2021-01-01T00:00:00.000+0000", "pod123", 99.95),
> RowFactory.create("2021-01-01T00:00:00.000+0000", "pod123", 99.95),
> RowFactory.create("2021-01-01T00:00:00.000+0000", "pod123", 99.95),
> RowFactory.create("2021-01-01T00:00:00.000+0000", "pod123", 99.95)
> );
>
> StructType schema = DataTypes.createStructType(new StructField[] {
> DataTypes.createStructField("timestamp", DataTypes.StringType, false),
> DataTypes.createStructField("id", DataTypes.StringType, false),
> DataTypes.createStructField("value", DataTypes.DoubleType, false)
> });
>
> Dataset<Row> df = spark.createDataFrame(data, schema);
>
> // Show the input data
> System.out.println("Input data:");
> df.show();
>
> // Perform the aggregation
> Dataset<Row> result = df.groupBy("id")
> .agg(
> avg("value").as(METADATA_COL_METRICVALUE),
> sum("value").as(METADATA_COL_SUM_VALUE)
> );
>
> // Show the results
> System.out.println("Aggregation results:");
> result.show();
>
> // Collect the results
> List<Row> results = result.collectAsList();
>
> // Print the results
> System.out.println("Number of results: " + results.size());
> for (Row row : results) {
> System.out.println("Metric value: " +
> row.getDouble(row.fieldIndex(METADATA_COL_METRICVALUE)));
> System.out.println("Sum value: " +
> row.getDouble(row.fieldIndex(METADATA_COL_SUM_VALUE)));
> }
>
> // Verify the results
> assertEquals(1, results.size(), "Expected 1 aggregated result");
>
> Row resultRow = results.get(0);
> doublesumValue =
> resultRow.getDouble(resultRow.fieldIndex(METADATA_COL_SUM_VALUE));
> doubleexpectedSum = 799.6; // 8 * 99.95
>
> System.out.println("Expected sum: " + expectedSum);
> System.out.println("Actual sum: " + sumValue);
> System.out.println("Difference: " + Math.abs(expectedSum - sumValue));
>
> // Check if the sum is close to the expected value
> assertTrue(Math.abs(expectedSum - sumValue) < 0.001,
> "Sum value should be close to " + expectedSum + " but was " + sumValue);
> {code}
> {code:java}
> Input data:
> +--------------------+------+-----+
> | timestamp| id|value|
> +--------------------+------+-----+
> |2021-01-01T00:00:...|pod123|99.95|
> |2021-01-01T00:00:...|pod123|99.95|
> |2021-01-01T00:00:...|pod123|99.95|
> |2021-01-01T00:00:...|pod123|99.95|
> |2021-01-01T00:00:...|pod123|99.95|
> |2021-01-01T00:00:...|pod123|99.95|
> |2021-01-01T00:00:...|pod123|99.95|
> |2021-01-01T00:00:...|pod123|99.95|
> +--------------------+------+-----+
> Aggregation results:
> +------+-----------------+-----------------+
> | id| metric_value| sum_value|
> +------+-----------------+-----------------+
> |pod123|99.95000000000002|799.6000000000001|
> +------+-----------------+-----------------+
> Number of results: 1
> Metric value: 99.95000000000002
> Sum value: 799.6000000000001
> Expected sum: 799.6
> Actual sum: 799.6000000000001
> Difference: 1.1368683772161603E-13{code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
