`integer overflow` in CapacityByteArrayOutputStream

Inizia da CapacityByteArrayOutputStream.java, in particolare da addSlab alla riga 198 e write alla riga 220, quindi riproduci lo scenario di scrittura ARRAY segnalato usando Spark 3.5.0 e Parquet 1.15.2. Traccia il percorso con cui il calcolo della capacità raggiunge Math.addExact e conferma che l’errore di overflow degli interi viene impedito senza compromettere le scritture di grandi dimensioni.

Describe the bug, including details regarding any error messages, version, and platform.

The following exception was thrown when we read a column of ARRAY<STRING> in Spark 3.5.0 and Parquet 1.15.2

Caused by: java.lang.ArithmeticException: integer overflow
	at java.base/java.lang.Math.addExact(Math.java:883)
	at org.apache.parquet.bytes.CapacityByteArrayOutputStream.addSlab(CapacityByteArrayOutputStream.java:198)
	at org.apache.parquet.bytes.CapacityByteArrayOutputStream.write(CapacityByteArrayOutputStream.java:220)
	at org.apache.parquet.bytes.LittleEndianDataOutputStream.write(LittleEndianDataOutputStream.java:76)
	at java.base/java.io.OutputStream.write(OutputStream.java:127)
	at org.apache.parquet.io.api.Binary$ByteArrayBackedBinary.writeTo(Binary.java:319)
	at org.apache.parquet.column.values.plain.PlainValuesWriter.writeBytes(PlainValuesWriter.java:55)
	at org.apache.parquet.column.values.fallback.FallbackValuesWriter.writeBytes(FallbackValuesWriter.java:178)
	at org.apache.parquet.column.impl.ColumnWriterBase.write(ColumnWriterBase.java:196)
	at org.apache.parquet.io.MessageColumnIO$MessageColumnIORecordConsumer.addBinary(MessageColumnIO.java:473)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.$anonfun$makeWriter$9(ParquetWriteSupport.scala:212)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.$anonfun$makeWriter$9$adapted(ParquetWriteSupport.scala:210)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.$anonfun$makeArrayWriter$5(ParquetWriteSupport.scala:354)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.consumeField(ParquetWriteSupport.scala:490)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.$anonfun$makeArrayWriter$4(ParquetWriteSupport.scala:354)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.consumeGroup(ParquetWriteSupport.scala:484)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.$anonfun$makeArrayWriter$3(ParquetWriteSupport.scala:352)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.consumeField(ParquetWriteSupport.scala:490)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.$anonfun$makeArrayWriter$2(ParquetWriteSupport.scala:347)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.consumeGroup(ParquetWriteSupport.scala:484)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.$anonfun$makeArrayWriter$1(ParquetWriteSupport.scala:346)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.$anonfun$makeArrayWriter$1$adapted(ParquetWriteSupport.scala:342)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.$anonfun$writeFields$1(ParquetWriteSupport.scala:168)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.consumeField(ParquetWriteSupport.scala:490)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.writeFields(ParquetWriteSupport.scala:168)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.$anonfun$write$1(ParquetWriteSupport.scala:158)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.consumeMessage(ParquetWriteSupport.scala:478)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.write(ParquetWriteSupport.scala:158)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport.write(ParquetWriteSupport.scala:54)
	at org.apache.parquet.hadoop.InternalParquetRecordWriter.write(InternalParquetRecordWriter.java:152)
	at org.apache.parquet.hadoop.ParquetRecordWriter.write(ParquetRecordWriter.java:240)
	at org.apache.parquet.hadoop.ParquetRecordWriter.write(ParquetRecordWriter.java:41)
	at org.apache.spark.sql.execution.datasources.parquet.ParquetOutputWriter.write(ParquetOutputWriter.scala:39)

The issue can be worked around by increasing spark.sql.shuffle.partitions to divide data into smaller partitions.
Can it be solved at parquet side?

Component(s)

Core