Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Drop the in-range check at INT96 output path [databricks] #8824

Merged
merged 15 commits into from
Aug 4, 2023
Merged

Drop the in-range check at INT96 output path [databricks] #8824

Changes from 12 commits
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
063ae5d
wip
gerashegalov Jul 15, 2023
1f29224
wip
gerashegalov Jul 16, 2023
4b9d809
Revert CanonicalizeSuite.scala
gerashegalov Jul 19, 2023
7b96b35
undo gshegalov-dt
gerashegalov Jul 20, 2023
27255b8
int96-SNAPSHOT
gerashegalov Jul 22, 2023
97c1deb
debug
gerashegalov Jul 22, 2023
d1f2325
wip
gerashegalov Jul 23, 2023
e57ad3a
Merge remote-tracking branch 'origin/branch-23.10' into int96
gerashegalov Jul 26, 2023
e192895
Remove INT96 check for being in the Long.Min:Max range
gerashegalov Jul 27, 2023
36819cd
Merge remote-tracking branch 'origin/branch-23.10' into int96
gerashegalov Jul 27, 2023
07cdd41
cleanup
gerashegalov Jul 27, 2023
8f6766d
withResource
gerashegalov Jul 27, 2023
f266ea3
Update docs and tests
gerashegalov Jul 28, 2023
11b0159
Merge remote-tracking branch 'origin/branch-23.10' into int96
gerashegalov Aug 3, 2023
1eba8f9
Remove timestamp limits for parquet tests
gerashegalov Aug 4, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
30 changes: 1 addition & 29 deletions sql-plugin/src/main/scala/com/nvidia/spark/rapids/GpuParquetFileFormat.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@ import ai.rapids.cudf._
import com.nvidia.spark.RebaseHelper
import com.nvidia.spark.rapids.Arm.withResource
import com.nvidia.spark.rapids.RapidsPluginImplicits.AutoCloseableProducingArray
import com.nvidia.spark.rapids.shims.{ParquetFieldIdShims, ParquetTimestampNTZShims, SparkShimImpl}
import com.nvidia.spark.rapids.shims._
import org.apache.hadoop.mapreduce.{Job, OutputCommitter, TaskAttemptContext}
import org.apache.parquet.hadoop.{ParquetOutputCommitter, ParquetOutputFormat}
import org.apache.parquet.hadoop.ParquetOutputFormat.JobSummaryLevel
Expand Down Expand Up @@ -341,38 +341,10 @@ class GpuParquetWriter(
// included in Spark's `TimestampType`.
case (cv, _) if cv.getType.isTimestampType && cv.getType != DType.TIMESTAMP_DAYS =>
val typeMillis = ParquetOutputTimestampType.TIMESTAMP_MILLIS.toString
val typeInt96 = ParquetOutputTimestampType.INT96.toString

outputTimestampType match {
case `typeMillis` if cv.getType != DType.TIMESTAMP_MILLISECONDS =>
cv.castTo(DType.TIMESTAMP_MILLISECONDS)

case `typeInt96` =>
val inRange = withResource(Scalar.fromLong(Long.MaxValue / 1000)) { upper =>
withResource(Scalar.fromLong(Long.MinValue / 1000)) { lower =>
withResource(cv.bitCastTo(DType.INT64)) { int64 =>
withResource(int64.greaterOrEqualTo(upper)) { a =>
withResource(int64.lessOrEqualTo(lower)) { b =>
a.or(b)
}
}
}
}
}
val anyInRange = withResource(inRange)(_.any())
withResource(anyInRange) { _ =>
require(!(anyInRange.isValid && anyInRange.getBoolean),
// Its the writer's responsibility to close the input batch when this
// exception is thrown.
"INT96 column contains one " +
"or more values that can overflow and will result in data " +
"corruption. Please set " +
"`spark.rapids.sql.format.parquet.writer.int96.enabled` to false " +
"so we can fallback on CPU for writing parquet but still take " +
"advantage of parquet read on the GPU.")
}
cv.copyToColumnVector() /* the input is unchanged */

// Here the value of `outputTimestampType` should be `TIMESTAMP_MICROS`
case _ => cv.copyToColumnVector() /* the input is unchanged */
}
Expand Down