Skip to content

[SPARK-14267] [SQL] [PYSPARK] execute multiple Python UDFs within single batch #12057

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

Closed
davies wants to merge 7 commits into from
Closed

[SPARK-14267] [SQL] [PYSPARK] execute multiple Python UDFs within single batch #12057

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
f6b7373
multiple Python UDFs in single batch
Mar 30, 2016
8e6e5bc
fast path for single UDF
Mar 30, 2016
8dc1adf
improve performance, address comments
Mar 31, 2016
dd71ba9
be safe
Mar 31, 2016
8597bba
fix udt with udf
Mar 31, 2016
72a5ec0
avoid unnecessary conversion
Mar 31, 2016
876f9f9
udpate error message
Mar 31, 2016
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
Prev Previous commit
Next Next commit
be safe
  • Loading branch information
Davies Liu committed Mar 31, 2016
commit dd71ba90cd41722c8fd3811d2c698da344aeb690
4 changes: 2 additions & 2 deletions core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -86,7 +86,7 @@ private[spark] case class ChainedPythonFunctions(funcs: Seq[PythonFunction])
private[spark] object PythonRunner {
def apply(func: PythonFunction, bufferSize: Int, reuse_worker: Boolean): PythonRunner = {
new PythonRunner(
Seq(ChainedPythonFunctions(Seq(func))), bufferSize, reuse_worker, false, Seq(Seq(0)))
Seq(ChainedPythonFunctions(Seq(func))), bufferSize, reuse_worker, false, Array(Array(0)))
}
}

Expand All @@ -101,7 +101,7 @@ private[spark] class PythonRunner(
bufferSize: Int,
reuse_worker: Boolean,
isUDF: Boolean,
Copy link
Contributor

@JoshRosen JoshRosen Mar 30, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Similarly, do you mind adding scaldoc for these two new parameters?

argOffsets: Seq[Seq[Int]])
argOffsets: Array[Array[Int]])
extends Logging {

require(funcs.length == argOffsets.length, "numArgs should have the same length as funcs")
Copy link
Contributor

@JoshRosen JoshRosen Mar 31, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

numArgs -> argOffsets

Expand Down
4 changes: 2 additions & 2 deletions python/pyspark/worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@
from pyspark.broadcast import Broadcast, _broadcastRegistry
from pyspark.files import SparkFiles
from pyspark.serializers import write_with_length, write_int, read_long, \
write_long, read_int, SpecialLengths, UTF8Deserializer, PickleSerializer, AutoBatchedSerializer
write_long, read_int, SpecialLengths, UTF8Deserializer, PickleSerializer, BatchedSerializer
from pyspark import shuffle

pickleSer = PickleSerializer()
Expand Down Expand Up @@ -101,7 +101,7 @@ def read_udfs(pickleSer, infile):
mapper = eval(mapper_str, udfs)

func = lambda _, it: map(mapper, it)
ser = AutoBatchedSerializer(PickleSerializer())
ser = BatchedSerializer(PickleSerializer(), 100)
Copy link
Contributor

@JoshRosen JoshRosen Mar 31, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What serializer did we use before? 100 seems arbitrary here.

Copy link
Contributor Author

@davies davies Mar 31, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Before this patch, we use AutoBatchedSerialzier, which could hold thousands of rows (holding more rows in JVM, may cause OOM).
The 100 is used on Java side.

# profiling is not supported for UDF
return func, None, ser, ser

Expand Down
6 changes: 3 additions & 3 deletions sql/core/src/main/scala/org/apache/spark/sql/execution/python/BatchPythonEvaluation.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -86,13 +86,13 @@ case class BatchPythonEvaluation(udfs: Seq[PythonUDF], output: Seq[Attribute], c
dataTypes += e.dataType
allInputs.length - 1
}
}
}
}.toArray
}.toArray
val projection = newMutableProjection(allInputs, child.output)()

// Input iterator to Python: input rows are grouped so we send them in batches to Python.
// For each row, add it to the queue.
val inputIterator = iter.grouped(1024).map { inputRows =>
val inputIterator = iter.grouped(100).map { inputRows =>
val toBePickled = inputRows.map { inputRow =>
queue.add(inputRow)
val row = projection(inputRow)
Expand Down